blob: b48b8ada3b3d1ef6ed394bb18e20f47c01d07feb [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
Willy Tarreau8317b282014-04-23 01:49:41 +02002 HAProxy
Willy Tarreau6a06a402007-07-15 20:15:28 +02003 Configuration Manual
4 ----------------------
Willy Tarreau1db55792020-11-05 17:20:35 +01005 version 2.4
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau095275f2022-02-16 16:29:03 +01007 2022/02/16
Willy Tarreau6a06a402007-07-15 20:15:28 +02008
9
10This document covers the configuration language as implemented in the version
Davor Ocelice9ed2812017-12-25 17:49:28 +010011specified above. It does not provide any hints, examples, or advice. For such
Willy Tarreau0ba27502007-12-24 16:55:16 +010012documentation, please refer to the Reference Manual or the Architecture Manual.
Davor Ocelice9ed2812017-12-25 17:49:28 +010013The summary below is meant to help you find sections by name and navigate
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014through the document.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016Note to documentation contributors :
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017 This document is formatted with 80 columns per line, with even number of
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018 spaces for indentation and without tabs. Please follow these rules strictly
19 so that it remains easily printable everywhere. If a line needs to be
20 printed verbatim and does not fit, please end each line with a backslash
Willy Tarreau62a36c42010-08-17 15:53:10 +020021 ('\') and continue on next line, indented by two characters. It is also
Davor Ocelice9ed2812017-12-25 17:49:28 +010022 sometimes useful to prefix all output lines (logs, console outputs) with 3
23 closing angle brackets ('>>>') in order to emphasize the difference between
24 inputs and outputs when they may be ambiguous. If you add sections,
Willy Tarreau62a36c42010-08-17 15:53:10 +020025 please update the summary below for easier searching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020026
27
28Summary
29-------
30
311. Quick reminder about HTTP
321.1. The HTTP transaction model
331.2. HTTP request
Davor Ocelice9ed2812017-12-25 17:49:28 +0100341.2.1. The request line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200351.2.2. The request headers
361.3. HTTP response
Davor Ocelice9ed2812017-12-25 17:49:28 +0100371.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200381.3.2. The response headers
39
402. Configuring HAProxy
412.1. Configuration file format
William Lallemandf9873ba2015-05-05 17:37:14 +0200422.2. Quoting and escaping
William Lallemandb2f07452015-05-12 14:27:13 +0200432.3. Environment variables
Willy Tarreau4b103022021-02-12 17:59:10 +0100442.4. Conditional blocks
452.5. Time format
462.6. Examples
Willy Tarreauc57f0e22009-05-10 13:12:33 +020047
483. Global parameters
493.1. Process management and security
503.2. Performance tuning
513.3. Debugging
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100523.4. Userlists
Cyril Bontédc4d9032012-04-08 21:57:39 +0200533.5. Peers
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200543.6. Mailers
William Lallemandc9515522019-06-12 16:32:11 +0200553.7. Programs
Christopher Faulet76edc0f2020-01-13 15:52:01 +0100563.8. HTTP-errors
Emeric Brun99c453d2020-05-25 15:01:04 +0200573.9. Rings
William Lallemand0217b7b2020-11-18 10:41:24 +0100583.10. Log forwarding
Willy Tarreauc57f0e22009-05-10 13:12:33 +020059
604. Proxies
614.1. Proxy keywords matrix
624.2. Alphabetically sorted keywords reference
63
Davor Ocelice9ed2812017-12-25 17:49:28 +0100645. Bind and server options
Willy Tarreau086fbf52012-09-24 20:34:51 +0200655.1. Bind options
665.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200675.3. Server DNS resolution
685.3.1. Global overview
695.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020070
Julien Pivotto6ccee412019-11-27 15:49:54 +0100716. Cache
726.1. Limitation
736.2. Setup
746.2.1. Cache section
756.2.2. Proxy section
76
Willy Tarreau74ca5042013-06-11 23:12:07 +0200777. Using ACLs and fetching samples
787.1. ACL basics
797.1.1. Matching booleans
807.1.2. Matching integers
817.1.3. Matching strings
827.1.4. Matching regular expressions (regexes)
837.1.5. Matching arbitrary data blocks
847.1.6. Matching IPv4 and IPv6 addresses
857.2. Using ACLs to form conditions
867.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200877.3.1. Converters
887.3.2. Fetching samples from internal states
897.3.3. Fetching samples at Layer 4
907.3.4. Fetching samples at Layer 5
917.3.5. Fetching samples from buffer contents (Layer 6)
927.3.6. Fetching HTTP samples (Layer 7)
Christopher Faulete596d182020-05-05 17:46:34 +0200937.3.7. Fetching samples for developers
Willy Tarreau74ca5042013-06-11 23:12:07 +0200947.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020095
968. Logging
978.1. Log levels
988.2. Log formats
998.2.1. Default log format
1008.2.2. TCP log format
1018.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +01001028.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +01001038.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001048.3. Advanced logging options
1058.3.1. Disabling logging of external tests
1068.3.2. Logging before waiting for the session to terminate
1078.3.3. Raising log level upon errors
1088.3.4. Disabling logging of successful connections
1098.4. Timing events
1108.5. Session state at disconnection
1118.6. Non-printable characters
1128.7. Capturing HTTP cookies
1138.8. Capturing HTTP headers
1148.9. Examples of logs
115
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001169. Supported filters
1179.1. Trace
1189.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001199.3. Stream Processing Offload Engine (SPOE)
Christopher Faulet99a17a22018-12-11 09:18:27 +01001209.4. Cache
Christopher Fauletb30b3102019-09-12 23:03:09 +02001219.5. fcgi-app
Miroslav Zagoracdc32cd92020-12-13 18:32:57 +01001229.6. OpenTracing
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200123
Christopher Fauletb30b3102019-09-12 23:03:09 +020012410. FastCGI applications
12510.1. Setup
12610.1.1. Fcgi-app section
12710.1.2. Proxy section
12810.1.3. Example
12910.2. Default parameters
13010.3. Limitations
131
Emeric Brunce325c42021-04-02 17:05:09 +020013211. Address formats
13311.1. Address family prefixes
13411.2. Socket type prefixes
13511.3. Protocol prefixes
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200136
1371. Quick reminder about HTTP
138----------------------------
139
Davor Ocelice9ed2812017-12-25 17:49:28 +0100140When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200141fully analyzed and indexed, thus it becomes possible to build matching criteria
142on almost anything found in the contents.
143
144However, it is important to understand how HTTP requests and responses are
145formed, and how HAProxy decomposes them. It will then become easier to write
146correct rules and to debug existing configurations.
147
148
1491.1. The HTTP transaction model
150-------------------------------
151
152The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100153to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100154from the client to the server, a request is sent by the client through the
155connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200156will involve a new connection :
157
158 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
159
160In this mode, called the "HTTP close" mode, there are as many connection
161establishments as there are HTTP transactions. Since the connection is closed
162by the server after the response, the client does not need to know the content
163length.
164
165Due to the transactional nature of the protocol, it was possible to improve it
166to avoid closing a connection between two subsequent transactions. In this mode
167however, it is mandatory that the server indicates the content length for each
168response so that the client does not wait indefinitely. For this, a special
169header is used: "Content-length". This mode is called the "keep-alive" mode :
170
171 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
172
173Its advantages are a reduced latency between transactions, and less processing
174power required on the server side. It is generally better than the close mode,
175but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200176a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177
Willy Tarreau95c4e142017-11-26 12:18:55 +0100178Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179keep-alive, but the client does not wait for the first response to send the
180second request. This is useful for fetching large number of images composing a
181page :
182
183 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
184
185This can obviously have a tremendous benefit on performance because the network
186latency is eliminated between subsequent requests. Many HTTP agents do not
187correctly support pipelining since there is no way to associate a response with
188the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100189server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200190
Willy Tarreau95c4e142017-11-26 12:18:55 +0100191The next improvement is the multiplexed mode, as implemented in HTTP/2. This
192time, each transaction is assigned a single stream identifier, and all streams
193are multiplexed over an existing connection. Many requests can be sent in
194parallel by the client, and responses can arrive in any order since they also
195carry the stream identifier.
196
Willy Tarreau70dffda2014-01-30 03:07:23 +0100197By default HAProxy operates in keep-alive mode with regards to persistent
198connections: for each connection it processes each request and response, and
199leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100200start of a new request. When it receives HTTP/2 connections from a client, it
201processes all the requests in parallel and leaves the connection idling,
202waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200203
Christopher Faulet315b39c2018-09-21 16:26:19 +0200204HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +0100205 - keep alive : all requests and responses are processed (default)
206 - tunnel : only the first request and response are processed,
Christopher Faulet6c9bbb22019-03-26 21:37:23 +0100207 everything else is forwarded with no analysis (deprecated).
Willy Tarreau70dffda2014-01-30 03:07:23 +0100208 - server close : the server-facing connection is closed after the response.
Christopher Faulet315b39c2018-09-21 16:26:19 +0200209 - close : the connection is actively closed after end of response.
Willy Tarreau70dffda2014-01-30 03:07:23 +0100210
Willy Tarreau95c4e142017-11-26 12:18:55 +0100211
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200212
2131.2. HTTP request
214-----------------
215
216First, let's consider this HTTP request :
217
218 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100219 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200220 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
221 2 Host: www.mydomain.com
222 3 User-agent: my small browser
223 4 Accept: image/jpeg, image/gif
224 5 Accept: image/png
225
226
2271.2.1. The Request line
228-----------------------
229
230Line 1 is the "request line". It is always composed of 3 fields :
231
232 - a METHOD : GET
233 - a URI : /serv/login.php?lang=en&profile=2
234 - a version tag : HTTP/1.1
235
236All of them are delimited by what the standard calls LWS (linear white spaces),
237which are commonly spaces, but can also be tabs or line feeds/carriage returns
238followed by spaces/tabs. The method itself cannot contain any colon (':') and
239is limited to alphabetic letters. All those various combinations make it
240desirable that HAProxy performs the splitting itself rather than leaving it to
241the user to write a complex or inaccurate regular expression.
242
243The URI itself can have several forms :
244
245 - A "relative URI" :
246
247 /serv/login.php?lang=en&profile=2
248
249 It is a complete URL without the host part. This is generally what is
250 received by servers, reverse proxies and transparent proxies.
251
252 - An "absolute URI", also called a "URL" :
253
254 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
255
256 It is composed of a "scheme" (the protocol name followed by '://'), a host
257 name or address, optionally a colon (':') followed by a port number, then
258 a relative URI beginning at the first slash ('/') after the address part.
259 This is generally what proxies receive, but a server supporting HTTP/1.1
260 must accept this form too.
261
262 - a star ('*') : this form is only accepted in association with the OPTIONS
263 method and is not relayable. It is used to inquiry a next hop's
264 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100265
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200266 - an address:port combination : 192.168.0.12:80
267 This is used with the CONNECT method, which is used to establish TCP
268 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
269 other protocols too.
270
271In a relative URI, two sub-parts are identified. The part before the question
272mark is called the "path". It is typically the relative path to static objects
273on the server. The part after the question mark is called the "query string".
274It is mostly used with GET requests sent to dynamic scripts and is very
275specific to the language, framework or application in use.
276
Willy Tarreau95c4e142017-11-26 12:18:55 +0100277HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100278assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100279
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200280
2811.2.2. The request headers
282--------------------------
283
284The headers start at the second line. They are composed of a name at the
285beginning of the line, immediately followed by a colon (':'). Traditionally,
286an LWS is added after the colon but that's not required. Then come the values.
287Multiple identical headers may be folded into one single line, delimiting the
288values with commas, provided that their order is respected. This is commonly
289encountered in the "Cookie:" field. A header may span over multiple lines if
290the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
291define a total of 3 values for the "Accept:" header.
292
Davor Ocelice9ed2812017-12-25 17:49:28 +0100293Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200294their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100295"Connection:" header). In HTTP/2, header names are always sent in lower case,
Willy Tarreau253c2512020-07-07 15:55:23 +0200296as can be seen when running in debug mode. Internally, all header names are
297normalized to lower case so that HTTP/1.x and HTTP/2 use the exact same
298representation, and they are sent as-is on the other side. This explains why an
299HTTP/1.x request typed with camel case is delivered in lower case.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200300
301The end of the headers is indicated by the first empty line. People often say
302that it's a double line feed, which is not exact, even if a double line feed
303is one valid form of empty line.
304
305Fortunately, HAProxy takes care of all these complex combinations when indexing
306headers, checking values and counting them, so there is no reason to worry
307about the way they could be written, but it is important not to accuse an
308application of being buggy if it does unusual, valid things.
309
310Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000311 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200312 in the middle of headers by LWS in order to join multi-line headers. This
313 is necessary for proper analysis and helps less capable HTTP parsers to work
314 correctly and not to be fooled by such complex constructs.
315
316
3171.3. HTTP response
318------------------
319
320An HTTP response looks very much like an HTTP request. Both are called HTTP
321messages. Let's consider this HTTP response :
322
323 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100324 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200325 1 HTTP/1.1 200 OK
326 2 Content-length: 350
327 3 Content-Type: text/html
328
Willy Tarreau816b9792009-09-15 21:25:21 +0200329As a special case, HTTP supports so called "Informational responses" as status
330codes 1xx. These messages are special in that they don't convey any part of the
331response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100332continue to post its request for instance. In the case of a status 100 response
333the requested information will be carried by the next non-100 response message
334following the informational one. This implies that multiple responses may be
335sent to a single request, and that this only works when keep-alive is enabled
336(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
337correctly forward and skip them, and only process the next non-100 response. As
338such, these messages are neither logged nor transformed, unless explicitly
339state otherwise. Status 101 messages indicate that the protocol is changing
Daniel Corbett9f0843f2021-05-08 10:50:37 -0400340over the same connection and that HAProxy must switch to tunnel mode, just as
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100341if a CONNECT had occurred. Then the Upgrade header would contain additional
342information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200343
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200344
Davor Ocelice9ed2812017-12-25 17:49:28 +01003451.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200346------------------------
347
348Line 1 is the "response line". It is always composed of 3 fields :
349
350 - a version tag : HTTP/1.1
351 - a status code : 200
352 - a reason : OK
353
354The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100355 - 1xx = informational message to be skipped (e.g. 100, 101)
356 - 2xx = OK, content is following (e.g. 200, 206)
357 - 3xx = OK, no content following (e.g. 302, 304)
358 - 4xx = error caused by the client (e.g. 401, 403, 404)
359 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200360
Lukas Tribus23953682017-04-28 13:24:30 +0000361Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100362"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200363found there, but it's a common practice to respect the well-established
364messages. It can be composed of one or multiple words, such as "OK", "Found",
365or "Authentication Required".
366
Davor Ocelice9ed2812017-12-25 17:49:28 +0100367HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200368
369 Code When / reason
370 200 access to stats page, and when replying to monitoring requests
371 301 when performing a redirection, depending on the configured code
372 302 when performing a redirection, depending on the configured code
373 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100374 307 when performing a redirection, depending on the configured code
375 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200376 400 for an invalid or too large request
377 401 when an authentication is required to perform the action (when
378 accessing the stats page)
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200379 403 when a request is forbidden by a "http-request deny" rule
Florian Tham9205fea2020-01-08 13:35:30 +0100380 404 when the requested resource could not be found
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200381 408 when the request timeout strikes before the request is complete
Florian Tham272e29b2020-01-08 10:19:05 +0100382 410 when the requested resource is no longer available and will not
383 be available again
Daniel Corbett9f0843f2021-05-08 10:50:37 -0400384 500 when HAProxy encounters an unrecoverable internal error, such as a
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200385 memory allocation failure, which should never happen
Daniel Corbett9f0843f2021-05-08 10:50:37 -0400386 501 when HAProxy is unable to satisfy a client request because of an
Christopher Faulete095f312020-12-07 11:22:24 +0100387 unsupported feature
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200388 502 when the server returns an empty, invalid or incomplete response, or
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200389 when an "http-response deny" rule blocks the response.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200390 503 when no server was available to handle the request, or in response to
391 monitoring requests which match the "monitor fail" condition
392 504 when the response timeout strikes before the server responds
393
394The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3954.2).
396
397
3981.3.2. The response headers
399---------------------------
400
401Response headers work exactly like request headers, and as such, HAProxy uses
402the same parsing function for both. Please refer to paragraph 1.2.2 for more
403details.
404
405
4062. Configuring HAProxy
407----------------------
408
4092.1. Configuration file format
410------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200411
412HAProxy's configuration process involves 3 major sources of parameters :
413
414 - the arguments from the command-line, which always take precedence
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100415 - the configuration file(s), whose format is described here
Thayne McCombscdbcca92021-01-07 21:24:41 -0700416 - the running process's environment, in case some environment variables are
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100417 explicitly referenced
Willy Tarreau6a06a402007-07-15 20:15:28 +0200418
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100419The configuration file follows a fairly simple hierarchical format which obey
420a few basic rules:
Willy Tarreau0ba27502007-12-24 16:55:16 +0100421
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100422 1. a configuration file is an ordered sequence of statements
423
424 2. a statement is a single non-empty line before any unprotected "#" (hash)
425
426 3. a line is a series of tokens or "words" delimited by unprotected spaces or
427 tab characters
428
429 4. the first word or sequence of words of a line is one of the keywords or
430 keyword sequences listed in this document
431
432 5. all other words are all arguments of the first one, some being well-known
433 keywords listed in this document, others being values, references to other
434 parts of the configuration, or expressions
435
436 6. certain keywords delimit a section inside which only a subset of keywords
437 are supported
438
439 7. a section ends at the end of a file or on a special keyword starting a new
440 section
441
442This is all that is needed to know to write a simple but reliable configuration
443generator, but this is not enough to reliably parse any configuration nor to
444figure how to deal with certain corner cases.
445
446First, there are a few consequences of the rules above. Rule 6 and 7 imply that
447the keywords used to define a new section are valid everywhere and cannot have
448a different meaning in a specific section. These keywords are always a single
449word (as opposed to a sequence of words), and traditionally the section that
450follows them is designated using the same name. For example when speaking about
451the "global section", it designates the section of configuration that follows
452the "global" keyword. This usage is used a lot in error messages to help locate
453the parts that need to be addressed.
454
455A number of sections create an internal object or configuration space, which
456requires to be distinguished from other ones. In this case they will take an
457extra word which will set the name of this particular section. For some of them
458the section name is mandatory. For example "frontend foo" will create a new
459section of type "frontend" named "foo". Usually a name is specific to its
460section and two sections of different types may use the same name, but this is
461not recommended as it tends to complexify configuration management.
462
463A direct consequence of rule 7 is that when multiple files are read at once,
464each of them must start with a new section, and the end of each file will end
465a section. A file cannot contain sub-sections nor end an existing section and
466start a new one.
467
468Rule 1 mentioned that ordering matters. Indeed, some keywords create directives
469that can be repeated multiple times to create ordered sequences of rules to be
470applied in a certain order. For example "tcp-request" can be used to alternate
471"accept" and "reject" rules on varying criteria. As such, a configuration file
472processor must always preserve a section's ordering when editing a file. The
473ordering of sections usually does not matter except for the global section
474which must be placed before other sections, but it may be repeated if needed.
475In addition, some automatic identifiers may automatically be assigned to some
476of the created objects (e.g. proxies), and by reordering sections, their
477identifiers will change. These ones appear in the statistics for example. As
478such, the configuration below will assign "foo" ID number 1 and "bar" ID number
4792, which will be swapped if the two sections are reversed:
480
481 listen foo
482 bind :80
483
484 listen bar
485 bind :81
486
487Another important point is that according to rules 2 and 3 above, empty lines,
488spaces, tabs, and comments following and unprotected "#" character are not part
489of the configuration as they are just used as delimiters. This implies that the
490following configurations are strictly equivalent:
491
492 global#this is the global section
493 daemon#daemonize
494 frontend foo
495 mode http # or tcp
496
497and:
498
499 global
500 daemon
501
502 # this is the public web frontend
503 frontend foo
504 mode http
505
506The common practice is to align to the left only the keyword that initiates a
507new section, and indent (i.e. prepend a tab character or a few spaces) all
508other keywords so that it's instantly visible that they belong to the same
509section (as done in the second example above). Placing comments before a new
510section helps the reader decide if it's the desired one. Leaving a blank line
511at the end of a section also visually helps spotting the end when editing it.
512
513Tabs are very convenient for indent but they do not copy-paste well. If spaces
514are used instead, it is recommended to avoid placing too many (2 to 4) so that
515editing in field doesn't become a burden with limited editors that do not
516support automatic indent.
517
518In the early days it used to be common to see arguments split at fixed tab
519positions because most keywords would not take more than two arguments. With
520modern versions featuring complex expressions this practice does not stand
521anymore, and is not recommended.
522
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200523
William Lallemandf9873ba2015-05-05 17:37:14 +02005242.2. Quoting and escaping
525-------------------------
526
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100527In modern configurations, some arguments require the use of some characters
528that were previously considered as pure delimiters. In order to make this
529possible, HAProxy supports character escaping by prepending a backslash ('\')
530in front of the character to be escaped, weak quoting within double quotes
531('"') and strong quoting within single quotes ("'").
William Lallemandf9873ba2015-05-05 17:37:14 +0200532
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100533This is pretty similar to what is done in a number of programming languages and
534very close to what is commonly encountered in Bourne shell. The principle is
535the following: while the configuration parser cuts the lines into words, it
536also takes care of quotes and backslashes to decide whether a character is a
537delimiter or is the raw representation of this character within the current
538word. The escape character is then removed, the quotes are removed, and the
539remaining word is used as-is as a keyword or argument for example.
William Lallemandf9873ba2015-05-05 17:37:14 +0200540
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100541If a backslash is needed in a word, it must either be escaped using itself
542(i.e. double backslash) or be strongly quoted.
543
544Escaping outside quotes is achieved by preceding a special character by a
545backslash ('\'):
William Lallemandf9873ba2015-05-05 17:37:14 +0200546
547 \ to mark a space and differentiate it from a delimiter
548 \# to mark a hash and differentiate it from a comment
549 \\ to use a backslash
550 \' to use a single quote and differentiate it from strong quoting
551 \" to use a double quote and differentiate it from weak quoting
552
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100553In addition, a few non-printable characters may be emitted using their usual
554C-language representation:
555
556 \n to insert a line feed (LF, character \x0a or ASCII 10 decimal)
557 \r to insert a carriage return (CR, character \x0d or ASCII 13 decimal)
558 \t to insert a tab (character \x09 or ASCII 9 decimal)
559 \xNN to insert character having ASCII code hex NN (e.g \x0a for LF).
560
561Weak quoting is achieved by surrounding double quotes ("") around the character
562or sequence of characters to protect. Weak quoting prevents the interpretation
563of:
William Lallemandf9873ba2015-05-05 17:37:14 +0200564
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100565 space or tab as a word separator
William Lallemandf9873ba2015-05-05 17:37:14 +0200566 ' single quote as a strong quoting delimiter
567 # hash as a comment start
568
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100569Weak quoting permits the interpretation of environment variables (which are not
570evaluated outside of quotes) by preceding them with a dollar sign ('$'). If a
571dollar character is needed inside double quotes, it must be escaped using a
572backslash.
William Lallemandb2f07452015-05-12 14:27:13 +0200573
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100574Strong quoting is achieved by surrounding single quotes ('') around the
575character or sequence of characters to protect. Inside single quotes, nothing
576is interpreted, it's the efficient way to quote regular expressions.
William Lallemandf9873ba2015-05-05 17:37:14 +0200577
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100578As a result, here is the matrix indicating how special characters can be
579entered in different contexts (unprintable characters are replaced with their
580name within angle brackets). Note that some characters that may only be
581represented escaped have no possible representation inside single quotes,
582hence the '-' there:
William Lallemandf9873ba2015-05-05 17:37:14 +0200583
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100584 Character | Unquoted | Weakly quoted | Strongly quoted
585 -----------+---------------+-----------------------------+-----------------
586 <TAB> | \<TAB>, \x09 | "<TAB>", "\<TAB>", "\x09" | '<TAB>'
587 <LF> | \n, \x0a | "\n", "\x0a" | -
588 <CR> | \r, \x0d | "\r", "\x0d" | -
589 <SPC> | \<SPC>, \x20 | "<SPC>", "\<SPC>", "\x20" | '<SPC>'
590 " | \", \x22 | "\"", "\x22" | '"'
591 # | \#, \x23 | "#", "\#", "\x23" | '#'
592 $ | $, \$, \x24 | "\$", "\x24" | '$'
593 ' | \', \x27 | "'", "\'", "\x27" | -
594 \ | \\, \x5c | "\\", "\x5c" | '\'
William Lallemandf9873ba2015-05-05 17:37:14 +0200595
596 Example:
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100597 # those are all strictly equivalent:
William Lallemandf9873ba2015-05-05 17:37:14 +0200598 log-format %{+Q}o\ %t\ %s\ %{-Q}r
599 log-format "%{+Q}o %t %s %{-Q}r"
600 log-format '%{+Q}o %t %s %{-Q}r'
601 log-format "%{+Q}o %t"' %s %{-Q}r'
602 log-format "%{+Q}o %t"' %s'\ %{-Q}r
603
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100604There is one particular case where a second level of quoting or escaping may be
605necessary. Some keywords take arguments within parenthesis, sometimes delimited
606by commas. These arguments are commonly integers or predefined words, but when
607they are arbitrary strings, it may be required to perform a separate level of
608escaping to disambiguate the characters that belong to the argument from the
609characters that are used to delimit the arguments themselves. A pretty common
610case is the "regsub" converter. It takes a regular expression in argument, and
611if a closing parenthesis is needed inside, this one will require to have its
612own quotes.
613
614The keyword argument parser is exactly the same as the top-level one regarding
Thayne McCombsf87e4f72021-10-04 01:02:58 -0600615quotes, except that the \#, \$, and \xNN escapes are not processed. But what is
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +0500616not always obvious is that the delimiters used inside must first be escaped or
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100617quoted so that they are not resolved at the top level.
618
619Let's take this example making use of the "regsub" converter which takes 3
620arguments, one regular expression, one replacement string and one set of flags:
621
622 # replace all occurrences of "foo" with "blah" in the path:
623 http-request set-path %[path,regsub(foo,blah,g)]
624
625Here no special quoting was necessary. But if now we want to replace either
626"foo" or "bar" with "blah", we'll need the regular expression "(foo|bar)". We
627cannot write:
628
629 http-request set-path %[path,regsub((foo|bar),blah,g)]
630
631because we would like the string to cut like this:
632
633 http-request set-path %[path,regsub((foo|bar),blah,g)]
634 |---------|----|-|
635 arg1 _/ / /
636 arg2 __________/ /
637 arg3 ______________/
638
639but actually what is passed is a string between the opening and closing
640parenthesis then garbage:
641
642 http-request set-path %[path,regsub((foo|bar),blah,g)]
643 |--------|--------|
644 arg1=(foo|bar _/ /
645 trailing garbage _________/
646
647The obvious solution here seems to be that the closing parenthesis needs to be
648quoted, but alone this will not work, because as mentioned above, quotes are
649processed by the top-level parser which will resolve them before processing
650this word:
651
652 http-request set-path %[path,regsub("(foo|bar)",blah,g)]
653 ------------ -------- ----------------------------------
654 word1 word2 word3=%[path,regsub((foo|bar),blah,g)]
655
656So we didn't change anything for the argument parser at the second level which
657still sees a truncated regular expression as the only argument, and garbage at
658the end of the string. By escaping the quotes they will be passed unmodified to
659the second level:
660
661 http-request set-path %[path,regsub(\"(foo|bar)\",blah,g)]
662 ------------ -------- ------------------------------------
663 word1 word2 word3=%[path,regsub("(foo|bar)",blah,g)]
664 |---------||----|-|
665 arg1=(foo|bar) _/ / /
666 arg2=blah ___________/ /
667 arg3=g _______________/
668
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +0500669Another approach consists in using single quotes outside the whole string and
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100670double quotes inside (so that the double quotes are not stripped again):
671
672 http-request set-path '%[path,regsub("(foo|bar)",blah,g)]'
673 ------------ -------- ----------------------------------
674 word1 word2 word3=%[path,regsub("(foo|bar)",blah,g)]
675 |---------||----|-|
676 arg1=(foo|bar) _/ / /
677 arg2 ___________/ /
678 arg3 _______________/
679
680When using regular expressions, it can happen that the dollar ('$') character
681appears in the expression or that a backslash ('\') is used in the replacement
682string. In this case these ones will also be processed inside the double quotes
683thus single quotes are preferred (or double escaping). Example:
684
685 http-request set-path '%[path,regsub("^/(here)(/|$)","my/\1",g)]'
686 ------------ -------- -----------------------------------------
687 word1 word2 word3=%[path,regsub("^/(here)(/|$)","my/\1",g)]
688 |-------------| |-----||-|
689 arg1=(here)(/|$) _/ / /
690 arg2=my/\1 ________________/ /
691 arg3 ______________________/
692
Daniel Corbett9f0843f2021-05-08 10:50:37 -0400693Remember that backslashes are not escape characters within single quotes and
Thayne McCombsf87e4f72021-10-04 01:02:58 -0600694that the whole word above is already protected against them using the single
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100695quotes. Conversely, if double quotes had been used around the whole expression,
696single the dollar character and the backslashes would have been resolved at top
697level, breaking the argument contents at the second level.
698
Thayne McCombsf87e4f72021-10-04 01:02:58 -0600699Unfortunately, since single quotes can't be escaped inside of strong quoting,
700if you need to include single quotes in your argument, you will need to escape
701or quote them twice. There are a few ways to do this:
702
703 http-request set-var(txn.foo) str("\\'foo\\'")
704 http-request set-var(txn.foo) str(\"\'foo\'\")
705 http-request set-var(txn.foo) str(\\\'foo\\\')
706
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100707When in doubt, simply do not use quotes anywhere, and start to place single or
708double quotes around arguments that require a comma or a closing parenthesis,
Thayne McCombsf87e4f72021-10-04 01:02:58 -0600709and think about escaping these quotes using a backslash if the string contains
Willy Tarreau6f1129d2020-11-25 19:58:20 +0100710a dollar or a backslash. Again, this is pretty similar to what is used under
711a Bourne shell when double-escaping a command passed to "eval". For API writers
712the best is probably to place escaped quotes around each and every argument,
713regardless of their contents. Users will probably find that using single quotes
714around the whole expression and double quotes around each argument provides
715more readable configurations.
William Lallemandf9873ba2015-05-05 17:37:14 +0200716
717
William Lallemandb2f07452015-05-12 14:27:13 +02007182.3. Environment variables
719--------------------------
720
721HAProxy's configuration supports environment variables. Those variables are
722interpreted only within double quotes. Variables are expanded during the
723configuration parsing. Variable names must be preceded by a dollar ("$") and
724optionally enclosed with braces ("{}") similarly to what is done in Bourne
725shell. Variable names can contain alphanumerical characters or the character
Amaury Denoyellefa41cb62020-10-01 14:32:35 +0200726underscore ("_") but should not start with a digit. If the variable contains a
727list of several values separated by spaces, it can be expanded as individual
728arguments by enclosing the variable with braces and appending the suffix '[*]'
729before the closing brace.
William Lallemandb2f07452015-05-12 14:27:13 +0200730
731 Example:
732
733 bind "fd@${FD_APP1}"
734
735 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
736
737 user "$HAPROXY_USER"
738
William Lallemand4d03e432019-06-14 15:35:37 +0200739Some variables are defined by HAProxy, they can be used in the configuration
740file, or could be inherited by a program (See 3.7. Programs):
William Lallemanddaf4cd22018-04-17 16:46:13 +0200741
William Lallemand4d03e432019-06-14 15:35:37 +0200742* HAPROXY_LOCALPEER: defined at the startup of the process which contains the
743 name of the local peer. (See "-L" in the management guide.)
744
745* HAPROXY_CFGFILES: list of the configuration files loaded by HAProxy,
746 separated by semicolons. Can be useful in the case you specified a
747 directory.
748
749* HAPROXY_MWORKER: In master-worker mode, this variable is set to 1.
750
John Roeslerfb2fce12019-07-10 15:45:51 -0500751* HAPROXY_CLI: configured listeners addresses of the stats socket for every
William Lallemand4d03e432019-06-14 15:35:37 +0200752 processes, separated by semicolons.
753
John Roeslerfb2fce12019-07-10 15:45:51 -0500754* HAPROXY_MASTER_CLI: In master-worker mode, listeners addresses of the master
William Lallemand4d03e432019-06-14 15:35:37 +0200755 CLI, separated by semicolons.
756
Willy Tarreaua46f1af2021-05-06 10:25:11 +0200757In addition, some pseudo-variables are internally resolved and may be used as
758regular variables. Pseudo-variables always start with a dot ('.'), and are the
759only ones where the dot is permitted. The current list of pseudo-variables is:
760
761* .FILE: the name of the configuration file currently being parsed.
762
763* .LINE: the line number of the configuration file currently being parsed,
764 starting at one.
765
766* .SECTION: the name of the section currently being parsed, or its type if the
767 section doesn't have a name (e.g. "global"), or an empty string before the
768 first section.
769
770These variables are resolved at the location where they are parsed. For example
771if a ".LINE" variable is used in a "log-format" directive located in a defaults
772section, its line number will be resolved before parsing and compiling the
773"log-format" directive, so this same line number will be reused by subsequent
774proxies.
775
776This way it is possible to emit information to help locate a rule in variables,
777logs, error statuses, health checks, header values, or even to use line numbers
778to name some config objects like servers for example.
779
William Lallemand4d03e432019-06-14 15:35:37 +0200780See also "external-check command" for other variables.
William Lallemandb2f07452015-05-12 14:27:13 +0200781
Willy Tarreau4b103022021-02-12 17:59:10 +0100782
7832.4. Conditional blocks
784-----------------------
785
786It may sometimes be convenient to be able to conditionally enable or disable
787some arbitrary parts of the configuration, for example to enable/disable SSL or
788ciphers, enable or disable some pre-production listeners without modifying the
789configuration, or adjust the configuration's syntax to support two distinct
790versions of HAProxy during a migration.. HAProxy brings a set of nestable
791preprocessor-like directives which allow to integrate or ignore some blocks of
792text. These directives must be placed on their own line and they act on the
793lines that follow them. Two of them support an expression, the other ones only
794switch to an alternate block or end a current level. The 4 following directives
795are defined to form conditional blocks:
796
797 - .if <condition>
798 - .elif <condition>
799 - .else
800 - .endif
801
802The ".if" directive nests a new level, ".elif" stays at the same level, ".else"
803as well, and ".endif" closes a level. Each ".if" must be terminated by a
804matching ".endif". The ".elif" may only be placed after ".if" or ".elif", and
805there is no limit to the number of ".elif" that may be chained. There may be
806only one ".else" per ".if" and it must always be after the ".if" or the last
807".elif" of a block.
808
809Comments may be placed on the same line if needed after a '#', they will be
810ignored. The directives are tokenized like other configuration directives, and
811as such it is possible to use environment variables in conditions.
812
813The conditions are currently limited to:
814
815 - an empty string, always returns "false"
816 - the integer zero ('0'), always returns "false"
817 - a non-nul integer (e.g. '1'), always returns "true".
Willy Tarreau42ed14b2021-05-06 15:55:14 +0200818 - a predicate optionally followed by argument(s) in parenthesis.
819
820The list of currently supported predicates is the following:
821
822 - defined(<name>) : returns true if an environment variable <name>
823 exists, regardless of its contents
824
Willy Tarreau58ca7062021-05-06 16:34:23 +0200825 - feature(<name>) : returns true if feature <name> is listed as present
826 in the features list reported by "haproxy -vv"
827 (which means a <name> appears after a '+')
828
Willy Tarreau6492e872021-05-06 16:10:09 +0200829 - streq(<str1>,<str2>) : returns true only if the two strings are equal
830 - strneq(<str1>,<str2>) : returns true only if the two strings differ
831
Willy Tarreau0b7c78a2021-05-06 16:53:26 +0200832 - version_atleast(<ver>): returns true if the current haproxy version is
833 at least as recent as <ver> otherwise false. The
834 version syntax is the same as shown by "haproxy -v"
835 and missing components are assumed as being zero.
836
837 - version_before(<ver>) : returns true if the current haproxy version is
838 strictly older than <ver> otherwise false. The
839 version syntax is the same as shown by "haproxy -v"
840 and missing components are assumed as being zero.
841
Willy Tarreau42ed14b2021-05-06 15:55:14 +0200842Example:
Willy Tarreau4b103022021-02-12 17:59:10 +0100843
Willy Tarreau42ed14b2021-05-06 15:55:14 +0200844 .if defined(HAPROXY_MWORKER)
845 listen mwcli_px
846 bind :1111
847 ...
848 .endif
Willy Tarreau4b103022021-02-12 17:59:10 +0100849
Willy Tarreau6492e872021-05-06 16:10:09 +0200850 .if strneq("$SSL_ONLY",yes)
851 bind :80
852 .endif
853
854 .if streq("$WITH_SSL",yes)
Willy Tarreau58ca7062021-05-06 16:34:23 +0200855 .if feature(OPENSSL)
Willy Tarreau6492e872021-05-06 16:10:09 +0200856 bind :443 ssl crt ...
Willy Tarreau58ca7062021-05-06 16:34:23 +0200857 .endif
Willy Tarreau6492e872021-05-06 16:10:09 +0200858 .endif
859
Willy Tarreau0b7c78a2021-05-06 16:53:26 +0200860 .if version_atleast(2.4-dev19)
861 profiling.memory on
862 .endif
863
Willy Tarreau7190b982021-05-07 08:59:50 +0200864Four other directives are provided to report some status:
Willy Tarreau4b103022021-02-12 17:59:10 +0100865
Willy Tarreau7190b982021-05-07 08:59:50 +0200866 - .diag "message" : emit this message only when in diagnostic mode (-dD)
Willy Tarreau4b103022021-02-12 17:59:10 +0100867 - .notice "message" : emit this message at level NOTICE
868 - .warning "message" : emit this message at level WARNING
869 - .alert "message" : emit this message at level ALERT
870
871Messages emitted at level WARNING may cause the process to fail to start if the
872"strict-mode" is enabled. Messages emitted at level ALERT will always cause a
873fatal error. These can be used to detect some inappropriate conditions and
874provide advice to the user.
875
876Example:
877
878 .if "${A}"
879 .if "${B}"
880 .notice "A=1, B=1"
881 .elif "${C}"
882 .notice "A=1, B=0, C=1"
883 .elif "${D}"
884 .warning "A=1, B=0, C=0, D=1"
885 .else
886 .alert "A=1, B=0, C=0, D=0"
887 .endif
888 .else
889 .notice "A=0"
890 .endif
891
Willy Tarreau7190b982021-05-07 08:59:50 +0200892 .diag "WTA/2021-05-07: replace 'redirect' with 'return' after switch to 2.4"
893 http-request redirect location /goaway if ABUSE
894
Willy Tarreau4b103022021-02-12 17:59:10 +0100895
8962.5. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200897----------------
898
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100899Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100900values are generally expressed in milliseconds (unless explicitly stated
901otherwise) but may be expressed in any other unit by suffixing the unit to the
902numeric value. It is important to consider this because it will not be repeated
903for every keyword. Supported units are :
904
905 - us : microseconds. 1 microsecond = 1/1000000 second
906 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
907 - s : seconds. 1s = 1000ms
908 - m : minutes. 1m = 60s = 60000ms
909 - h : hours. 1h = 60m = 3600s = 3600000ms
910 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
911
912
Willy Tarreau4b103022021-02-12 17:59:10 +01009132.6. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200914-------------
915
916 # Simple configuration for an HTTP proxy listening on port 80 on all
917 # interfaces and forwarding requests to a single backend "servers" with a
918 # single server "server1" listening on 127.0.0.1:8000
919 global
920 daemon
921 maxconn 256
922
923 defaults
924 mode http
925 timeout connect 5000ms
926 timeout client 50000ms
927 timeout server 50000ms
928
929 frontend http-in
930 bind *:80
931 default_backend servers
932
933 backend servers
934 server server1 127.0.0.1:8000 maxconn 32
935
936
937 # The same configuration defined with a single listen block. Shorter but
938 # less expressive, especially in HTTP mode.
939 global
940 daemon
941 maxconn 256
942
943 defaults
944 mode http
945 timeout connect 5000ms
946 timeout client 50000ms
947 timeout server 50000ms
948
949 listen http-in
950 bind *:80
951 server server1 127.0.0.1:8000 maxconn 32
952
953
954Assuming haproxy is in $PATH, test these configurations in a shell with:
955
Willy Tarreauccb289d2010-12-11 20:19:38 +0100956 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200957
958
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009593. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200960--------------------
961
962Parameters in the "global" section are process-wide and often OS-specific. They
963are generally set once for all and do not need being changed once correct. Some
964of them have command-line equivalents.
965
966The following keywords are supported in the "global" section :
967
968 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200969 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200970 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200971 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200972 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200973 - daemon
Willy Tarreau8a022d52021-04-27 20:29:11 +0200974 - default-path
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200975 - description
976 - deviceatlas-json-file
977 - deviceatlas-log-level
978 - deviceatlas-separator
979 - deviceatlas-properties-cookie
Amaury Denoyelled2e53cd2021-05-06 16:21:39 +0200980 - expose-experimental-directives
Simon Horman98637e52014-06-20 12:30:16 +0900981 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200982 - gid
983 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100984 - hard-stop-after
Christopher Faulet98fbe952019-07-22 16:18:24 +0200985 - h1-case-adjust
986 - h1-case-adjust-file
Willy Tarreaud96f1122019-12-03 07:07:36 +0100987 - insecure-fork-wanted
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100988 - insecure-setuid-wanted
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +0100989 - issuers-chain-path
Amaury Denoyelle0ea2c4f2021-07-09 17:14:30 +0200990 - h2-workaround-bogus-websocket-clients
Dragan Dosen13cd54c2020-06-18 18:24:05 +0200991 - localpeer
Willy Tarreau6a06a402007-07-15 20:15:28 +0200992 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200993 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100994 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200995 - lua-load
Thierry Fournier59f11be2020-11-29 00:37:41 +0100996 - lua-load-per-thread
Tim Duesterhusdd74b5f2020-01-12 13:55:40 +0100997 - lua-prepend-path
William Lallemand27edc4b2019-05-07 17:49:33 +0200998 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200999 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001000 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001001 - node
Amaury Denoyelle0f50cb92021-03-26 18:50:33 +01001002 - numa-cpu-mapping
Willy Tarreau6a06a402007-07-15 20:15:28 +02001003 - pidfile
Willy Tarreau119e50e2020-05-22 13:53:29 +02001004 - pp2-never-send-local
Willy Tarreau1d549722016-02-16 12:41:57 +01001005 - presetenv
1006 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +02001007 - uid
1008 - ulimit-n
1009 - user
Willy Tarreau636848a2019-04-15 19:38:50 +02001010 - set-dumpable
Willy Tarreau13d2ba22021-03-26 11:38:08 +01001011 - set-var
Willy Tarreau1d549722016-02-16 12:41:57 +01001012 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +02001013 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001014 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001015 - ssl-default-bind-ciphersuites
Jerome Magninb203ff62020-04-03 15:28:22 +02001016 - ssl-default-bind-curves
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001017 - ssl-default-bind-options
1018 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001019 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001020 - ssl-default-server-options
1021 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +01001022 - ssl-server-verify
Emmanuel Hocdetc3b7e742020-04-22 11:06:19 +02001023 - ssl-skip-self-issued-ca
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001024 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +01001025 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +01001026 - 51degrees-data-file
1027 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +02001028 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001029 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001030 - wurfl-data-file
1031 - wurfl-information-list
1032 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001033 - wurfl-cache-size
William Dauchy0fec3ab2019-10-27 20:08:11 +01001034 - strict-limits
Willy Tarreaud72758d2010-01-12 10:42:19 +01001035
Willy Tarreau6a06a402007-07-15 20:15:28 +02001036 * Performance tuning
William Dauchy0a8824f2019-10-27 20:08:09 +01001037 - busy-polling
Willy Tarreau1746eec2014-04-25 10:46:47 +02001038 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +02001039 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +02001040 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +01001041 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +01001042 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001043 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +02001044 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +02001045 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +02001046 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001047 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +02001048 - noepoll
1049 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001050 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +02001051 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001052 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001053 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001054 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +01001055 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +02001056 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +02001057 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001058 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +00001059 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001060 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +02001061 - tune.buffers.limit
1062 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +02001063 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +02001064 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +01001065 - tune.comp.maxlevel
Willy Tarreaubc52bec2020-06-18 08:58:47 +02001066 - tune.fd.edge-triggered
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001067 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +02001068 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +02001069 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +01001070 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001071 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +02001072 - tune.http.maxhdr
Willy Tarreau76cc6992020-07-01 18:49:24 +02001073 - tune.idle-pool.shared
Willy Tarreau7e312732014-02-12 16:35:14 +01001074 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001075 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +01001076 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001077 - tune.lua.session-timeout
1078 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001079 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001080 - tune.maxaccept
1081 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +02001082 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +02001083 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001084 - tune.pipesize
Willy Tarreaua8e2d972020-07-01 18:27:16 +02001085 - tune.pool-high-fd-ratio
1086 - tune.pool-low-fd-ratio
Willy Tarreaue803de22010-01-21 17:43:04 +01001087 - tune.rcvbuf.client
1088 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +01001089 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +02001090 - tune.runqueue-depth
Willy Tarreaue7723bd2020-06-24 11:11:02 +02001091 - tune.sched.low-latency
Willy Tarreaue803de22010-01-21 17:43:04 +01001092 - tune.sndbuf.client
1093 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001094 - tune.ssl.cachesize
William Lallemand7d42ef52020-07-06 11:41:30 +02001095 - tune.ssl.keylog
Willy Tarreaubfd59462013-02-21 07:46:09 +01001096 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +02001097 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +01001098 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001099 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +02001100 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01001101 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001102 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +01001103 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001104 - tune.vars.reqres-max-size
1105 - tune.vars.sess-max-size
1106 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +01001107 - tune.zlib.memlevel
1108 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +01001109
Willy Tarreau6a06a402007-07-15 20:15:28 +02001110 * Debugging
Willy Tarreau6a06a402007-07-15 20:15:28 +02001111 - quiet
Willy Tarreau3eb10b82020-04-15 16:42:39 +02001112 - zero-warning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001113
1114
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011153.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +02001116------------------------------------
1117
Emeric Brunc8e8d122012-10-02 18:42:10 +02001118ca-base <dir>
1119 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emmanuel Hocdet842e94e2019-12-16 16:39:17 +01001120 relative path is used with "ca-file", "ca-verify-file" or "crl-file"
1121 directives. Absolute locations specified in "ca-file", "ca-verify-file" and
1122 "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +02001123
Willy Tarreau6a06a402007-07-15 20:15:28 +02001124chroot <jail dir>
1125 Changes current directory to <jail dir> and performs a chroot() there before
1126 dropping privileges. This increases the security level in case an unknown
1127 vulnerability would be exploited, since it would make it very hard for the
1128 attacker to exploit the system. This only works when the process is started
1129 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +01001130 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +01001131
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001132cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
1133 On Linux 2.6 and above, it is possible to bind a process or a thread to a
1134 specific CPU set. This means that the process or the thread will never run on
1135 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
1136 sets. The first argument is a process set, eventually followed by a thread
1137 set. These sets have the format
1138
1139 all | odd | even | number[-[number]]
1140
1141 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +01001142 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001143 ignored. It is possible to specify a range with two such number delimited by
1144 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +01001145 "all", only odd numbers using "odd" or even numbers using "even", just like
1146 with the "bind-process" directive. The second and forthcoming arguments are
Amaury Denoyelle982fb532021-04-21 18:39:58 +02001147 CPU sets. Each CPU set is either a unique number starting at 0 for the first
1148 CPU or a range with two such numbers delimited by a dash ('-'). Outside of
1149 Linux and BSDs, there may be a limitation on the maximum CPU index to either
1150 31 or 63. Multiple CPU numbers or ranges may be specified, and the processes
1151 or threads will be allowed to bind to all of them. Obviously, multiple
1152 "cpu-map" directives may be specified. Each "cpu-map" directive will replace
1153 the previous ones when they overlap. A thread will be bound on the
1154 intersection of its mapping and the one of the process on which it is
1155 attached. If the intersection is null, no specific binding will be set for
1156 the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +01001157
Christopher Fauletff4121f2017-11-22 16:38:49 +01001158 Ranges can be partially defined. The higher bound can be omitted. In such
1159 case, it is replaced by the corresponding maximum value, 32 or 64 depending
1160 on the machine's word size.
1161
Christopher Faulet26028f62017-11-22 15:01:51 +01001162 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001163 automatically bind a process or a thread to a CPU by incrementing
1164 process/thread and CPU sets. To be valid, both sets must have the same
1165 size. No matter the declaration order of the CPU sets, it will be bound from
1166 the lowest to the highest bound. Having a process and a thread range with the
1167 "auto:" prefix is not supported. Only one range is supported, the other one
1168 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +01001169
1170 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001171 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
1172
1173 cpu-map 1/all 0-3 # bind all threads of the first process on the
1174 # first 4 CPUs
1175
1176 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
1177 # or "cpu-map 1-32 0-31" depending on the machine's
1178 # word size.
1179
Christopher Faulet26028f62017-11-22 15:01:51 +01001180 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001181 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +01001182 cpu-map auto:1-4 0-3
1183 cpu-map auto:1-4 0-1 2-3
1184 cpu-map auto:1-4 3 2 1 0
1185
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001186 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
1187 # and so on.
1188 cpu-map auto:1/1-4 0-3
1189 cpu-map auto:1/1-4 0-1 2-3
1190 cpu-map auto:1/1-4 3 2 1 0
1191
Davor Ocelice9ed2812017-12-25 17:49:28 +01001192 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +01001193 cpu-map auto:all 0-63
1194 cpu-map auto:even 0-31
1195 cpu-map auto:odd 32-63
1196
1197 # invalid cpu-map because process and CPU sets have different sizes.
1198 cpu-map auto:1-4 0 # invalid
1199 cpu-map auto:1 0-3 # invalid
1200
Christopher Fauletcb6a9452017-11-22 16:50:41 +01001201 # invalid cpu-map because automatic binding is used with a process range
1202 # and a thread range.
1203 cpu-map auto:all/all 0 # invalid
1204 cpu-map auto:all/1-4 0 # invalid
1205 cpu-map auto:1-4/all 0 # invalid
1206
Emeric Brunc8e8d122012-10-02 18:42:10 +02001207crt-base <dir>
1208 Assigns a default directory to fetch SSL certificates from when a relative
William Dauchy238ea3b2020-01-11 13:09:12 +01001209 path is used with "crtfile" or "crt" directives. Absolute locations specified
1210 prevail and ignore "crt-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +02001211
Willy Tarreau6a06a402007-07-15 20:15:28 +02001212daemon
1213 Makes the process fork into background. This is the recommended mode of
1214 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +01001215 disabled by the command line "-db" argument. This option is ignored in
1216 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001217
Willy Tarreau8a022d52021-04-27 20:29:11 +02001218default-path { current | config | parent | origin <path> }
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001219 By default HAProxy loads all files designated by a relative path from the
Willy Tarreau8a022d52021-04-27 20:29:11 +02001220 location the process is started in. In some circumstances it might be
1221 desirable to force all relative paths to start from a different location
1222 just as if the process was started from such locations. This is what this
1223 directive is made for. Technically it will perform a temporary chdir() to
1224 the designated location while processing each configuration file, and will
1225 return to the original directory after processing each file. It takes an
1226 argument indicating the policy to use when loading files whose path does
1227 not start with a slash ('/'):
1228 - "current" indicates that all relative files are to be loaded from the
1229 directory the process is started in ; this is the default.
1230
1231 - "config" indicates that all relative files should be loaded from the
1232 directory containing the configuration file. More specifically, if the
1233 configuration file contains a slash ('/'), the longest part up to the
1234 last slash is used as the directory to change to, otherwise the current
1235 directory is used. This mode is convenient to bundle maps, errorfiles,
1236 certificates and Lua scripts together as relocatable packages. When
1237 multiple configuration files are loaded, the directory is updated for
1238 each of them.
1239
1240 - "parent" indicates that all relative files should be loaded from the
1241 parent of the directory containing the configuration file. More
1242 specifically, if the configuration file contains a slash ('/'), ".."
1243 is appended to the longest part up to the last slash is used as the
1244 directory to change to, otherwise the directory is "..". This mode is
1245 convenient to bundle maps, errorfiles, certificates and Lua scripts
1246 together as relocatable packages, but where each part is located in a
1247 different subdirectory (e.g. "config/", "certs/", "maps/", ...).
1248
1249 - "origin" indicates that all relative files should be loaded from the
1250 designated (mandatory) path. This may be used to ease management of
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001251 different HAProxy instances running in parallel on a system, where each
Willy Tarreau8a022d52021-04-27 20:29:11 +02001252 instance uses a different prefix but where the rest of the sections are
1253 made easily relocatable.
1254
1255 Each "default-path" directive instantly replaces any previous one and will
1256 possibly result in switching to a different directory. While this should
1257 always result in the desired behavior, it is really not a good practice to
1258 use multiple default-path directives, and if used, the policy ought to remain
1259 consistent across all configuration files.
1260
1261 Warning: some configuration elements such as maps or certificates are
1262 uniquely identified by their configured path. By using a relocatable layout,
1263 it becomes possible for several of them to end up with the same unique name,
1264 making it difficult to update them at run time, especially when multiple
1265 configuration files are loaded from different directories. It is essential to
1266 observe a strict collision-free file naming scheme before adopting relative
1267 paths. A robust approach could consist in prefixing all files names with
1268 their respective site name, or in doing so at the directory level.
1269
David Carlier8167f302015-06-01 13:50:06 +02001270deviceatlas-json-file <path>
1271 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001272 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +02001273
1274deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001275 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +02001276 optional and set to 0 by default if not set.
1277
1278deviceatlas-separator <char>
1279 Sets the character separator for the API properties results. This directive
1280 is optional and set to | by default if not set.
1281
Cyril Bonté0306c4a2015-10-26 22:37:38 +01001282deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001283 Sets the client cookie's name used for the detection if the DeviceAtlas
1284 Client-side component was used during the request. This directive is optional
1285 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +01001286
Amaury Denoyelled2e53cd2021-05-06 16:21:39 +02001287expose-experimental-directives
1288 This statement must appear before using directives tagged as experimental or
1289 the config file will be rejected.
1290
Simon Horman98637e52014-06-20 12:30:16 +09001291external-check
Willy Tarreaud96f1122019-12-03 07:07:36 +01001292 Allows the use of an external agent to perform health checks. This is
1293 disabled by default as a security precaution, and even when enabled, checks
Willy Tarreaua45a8b52019-12-06 16:31:45 +01001294 may still fail unless "insecure-fork-wanted" is enabled as well. If the
1295 program launched makes use of a setuid executable (it should really not),
1296 you may also need to set "insecure-setuid-wanted" in the global section.
1297 See "option external-check", and "insecure-fork-wanted", and
1298 "insecure-setuid-wanted".
Simon Horman98637e52014-06-20 12:30:16 +09001299
Willy Tarreau6a06a402007-07-15 20:15:28 +02001300gid <number>
Thayne McCombscdbcca92021-01-07 21:24:41 -07001301 Changes the process's group ID to <number>. It is recommended that the group
Willy Tarreau6a06a402007-07-15 20:15:28 +02001302 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1303 be started with a user belonging to this group, or with superuser privileges.
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001304 Note that if HAProxy is started from a user having supplementary groups, it
Michael Schererab012dd2013-01-12 18:35:19 +01001305 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001306 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +01001307
Willy Tarreau11770ce2019-12-03 08:29:22 +01001308group <group name>
1309 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
1310 See also "gid" and "user".
1311
Cyril Bonté203ec5a2017-03-23 22:44:13 +01001312hard-stop-after <time>
1313 Defines the maximum time allowed to perform a clean soft-stop.
1314
1315 Arguments :
1316 <time> is the maximum time (by default in milliseconds) for which the
1317 instance will remain alive when a soft-stop is received via the
1318 SIGUSR1 signal.
1319
1320 This may be used to ensure that the instance will quit even if connections
1321 remain opened during a soft-stop (for example with long timeouts for a proxy
1322 in tcp mode). It applies both in TCP and HTTP mode.
1323
1324 Example:
1325 global
1326 hard-stop-after 30s
1327
Christopher Faulet98fbe952019-07-22 16:18:24 +02001328h1-case-adjust <from> <to>
1329 Defines the case adjustment to apply, when enabled, to the header name
1330 <from>, to change it to <to> before sending it to HTTP/1 clients or
1331 servers. <from> must be in lower case, and <from> and <to> must not differ
1332 except for their case. It may be repeated if several header names need to be
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05001333 adjusted. Duplicate entries are not allowed. If a lot of header names have to
Christopher Faulet98fbe952019-07-22 16:18:24 +02001334 be adjusted, it might be more convenient to use "h1-case-adjust-file".
1335 Please note that no transformation will be applied unless "option
1336 h1-case-adjust-bogus-client" or "option h1-case-adjust-bogus-server" is
1337 specified in a proxy.
1338
1339 There is no standard case for header names because, as stated in RFC7230,
1340 they are case-insensitive. So applications must handle them in a case-
1341 insensitive manner. But some bogus applications violate the standards and
1342 erroneously rely on the cases most commonly used by browsers. This problem
1343 becomes critical with HTTP/2 because all header names must be exchanged in
1344 lower case, and HAProxy follows the same convention. All header names are
1345 sent in lower case to clients and servers, regardless of the HTTP version.
1346
1347 Applications which fail to properly process requests or responses may require
1348 to temporarily use such workarounds to adjust header names sent to them for
1349 the time it takes the application to be fixed. Please note that an
1350 application which requires such workarounds might be vulnerable to content
1351 smuggling attacks and must absolutely be fixed.
1352
1353 Example:
1354 global
1355 h1-case-adjust content-length Content-Length
1356
1357 See "h1-case-adjust-file", "option h1-case-adjust-bogus-client" and
1358 "option h1-case-adjust-bogus-server".
1359
1360h1-case-adjust-file <hdrs-file>
1361 Defines a file containing a list of key/value pairs used to adjust the case
1362 of some header names before sending them to HTTP/1 clients or servers. The
1363 file <hdrs-file> must contain 2 header names per line. The first one must be
1364 in lower case and both must not differ except for their case. Lines which
1365 start with '#' are ignored, just like empty lines. Leading and trailing tabs
1366 and spaces are stripped. Duplicate entries are not allowed. Please note that
1367 no transformation will be applied unless "option h1-case-adjust-bogus-client"
1368 or "option h1-case-adjust-bogus-server" is specified in a proxy.
1369
1370 If this directive is repeated, only the last one will be processed. It is an
1371 alternative to the directive "h1-case-adjust" if a lot of header names need
1372 to be adjusted. Please read the risks associated with using this.
1373
1374 See "h1-case-adjust", "option h1-case-adjust-bogus-client" and
1375 "option h1-case-adjust-bogus-server".
1376
Willy Tarreaud96f1122019-12-03 07:07:36 +01001377insecure-fork-wanted
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001378 By default HAProxy tries hard to prevent any thread and process creation
Willy Tarreaud96f1122019-12-03 07:07:36 +01001379 after it starts. Doing so is particularly important when using Lua files of
1380 uncertain origin, and when experimenting with development versions which may
1381 still contain bugs whose exploitability is uncertain. And generally speaking
1382 it's good hygiene to make sure that no unexpected background activity can be
1383 triggered by traffic. But this prevents external checks from working, and may
1384 break some very specific Lua scripts which actively rely on the ability to
1385 fork. This option is there to disable this protection. Note that it is a bad
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001386 idea to disable it, as a vulnerability in a library or within HAProxy itself
Willy Tarreaud96f1122019-12-03 07:07:36 +01001387 will be easier to exploit once disabled. In addition, forking from Lua or
1388 anywhere else is not reliable as the forked process may randomly embed a lock
1389 set by another thread and never manage to finish an operation. As such it is
1390 highly recommended that this option is never used and that any workload
1391 requiring such a fork be reconsidered and moved to a safer solution (such as
1392 agents instead of external checks). This option supports the "no" prefix to
1393 disable it.
1394
Willy Tarreaua45a8b52019-12-06 16:31:45 +01001395insecure-setuid-wanted
1396 HAProxy doesn't need to call executables at run time (except when using
1397 external checks which are strongly recommended against), and is even expected
1398 to isolate itself into an empty chroot. As such, there basically is no valid
1399 reason to allow a setuid executable to be called without the user being fully
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001400 aware of the risks. In a situation where HAProxy would need to call external
Willy Tarreaua45a8b52019-12-06 16:31:45 +01001401 checks and/or disable chroot, exploiting a vulnerability in a library or in
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001402 HAProxy itself could lead to the execution of an external program. On Linux
Willy Tarreaua45a8b52019-12-06 16:31:45 +01001403 it is possible to lock the process so that any setuid bit present on such an
1404 executable is ignored. This significantly reduces the risk of privilege
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001405 escalation in such a situation. This is what HAProxy does by default. In case
Willy Tarreaua45a8b52019-12-06 16:31:45 +01001406 this causes a problem to an external check (for example one which would need
1407 the "ping" command), then it is possible to disable this protection by
1408 explicitly adding this directive in the global section. If enabled, it is
1409 possible to turn it back off by prefixing it with the "no" keyword.
1410
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +01001411issuers-chain-path <dir>
1412 Assigns a directory to load certificate chain for issuer completion. All
1413 files must be in PEM format. For certificates loaded with "crt" or "crt-list",
1414 if certificate chain is not included in PEM (also commonly known as
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001415 intermediate certificate), HAProxy will complete chain if the issuer of the
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +01001416 certificate corresponds to the first certificate of the chain loaded with
1417 "issuers-chain-path".
1418 A "crt" file with PrivateKey+Certificate+IntermediateCA2+IntermediateCA1
1419 could be replaced with PrivateKey+Certificate. HAProxy will complete the
1420 chain if a file with IntermediateCA2+IntermediateCA1 is present in
1421 "issuers-chain-path" directory. All other certificates with the same issuer
1422 will share the chain in memory.
1423
Amaury Denoyelle0ea2c4f2021-07-09 17:14:30 +02001424h2-workaround-bogus-websocket-clients
1425 This disables the announcement of the support for h2 websockets to clients.
1426 This can be use to overcome clients which have issues when implementing the
1427 relatively fresh RFC8441, such as Firefox 88. To allow clients to
1428 automatically downgrade to http/1.1 for the websocket tunnel, specify h2
1429 support on the bind line using "alpn" without an explicit "proto" keyword. If
1430 this statement was previously activated, this can be disabled by prefixing
1431 the keyword with "no'.
1432
Dragan Dosen13cd54c2020-06-18 18:24:05 +02001433localpeer <name>
1434 Sets the local instance's peer name. It will be ignored if the "-L"
1435 command line argument is specified or if used after "peers" section
1436 definitions. In such cases, a warning message will be emitted during
1437 the configuration parsing.
1438
1439 This option will also set the HAPROXY_LOCALPEER environment variable.
1440 See also "-L" in the management guide and "peers" section below.
1441
Jan Wagner3e678602020-12-17 22:22:32 +01001442log <address> [len <length>] [format <format>] [sample <ranges>:<sample_size>]
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02001443 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +01001444 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +01001445 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +01001446 configured with "log global".
1447
1448 <address> can be one of:
1449
Willy Tarreau2769aa02007-12-27 18:26:09 +01001450 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +01001451 no port is specified, 514 is used by default (the standard syslog
1452 port).
1453
David du Colombier24bb5f52011-03-17 10:40:23 +01001454 - An IPv6 address followed by a colon and optionally a UDP port. If
1455 no port is specified, 514 is used by default (the standard syslog
1456 port).
1457
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01001458 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +01001459 considerations for chroot (be sure the path is accessible inside
1460 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +01001461 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +01001462
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01001463 - A file descriptor number in the form "fd@<number>", which may point
1464 to a pipe, terminal, or socket. In this case unbuffered logs are used
1465 and one writev() call per log is performed. This is a bit expensive
1466 but acceptable for most workloads. Messages sent this way will not be
1467 truncated but may be dropped, in which case the DroppedLogs counter
1468 will be incremented. The writev() call is atomic even on pipes for
1469 messages up to PIPE_BUF size, which POSIX recommends to be at least
1470 512 and which is 4096 bytes on most modern operating systems. Any
1471 larger message may be interleaved with messages from other processes.
1472 Exceptionally for debugging purposes the file descriptor may also be
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001473 directed to a file, but doing so will significantly slow HAProxy down
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01001474 as non-blocking calls will be ignored. Also there will be no way to
1475 purge nor rotate this file without restarting the process. Note that
1476 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +01001477 for use with a TCP syslog server. See also the "short" and "raw"
1478 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01001479
1480 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
1481 "fd@2", see above.
1482
Willy Tarreauc046d162019-08-30 15:24:59 +02001483 - A ring buffer in the form "ring@<name>", which will correspond to an
1484 in-memory ring buffer accessible over the CLI using the "show events"
1485 command, which will also list existing rings and their sizes. Such
1486 buffers are lost on reload or restart but when used as a complement
1487 this can help troubleshooting by having the logs instantly available.
1488
William Lallemandb2f07452015-05-12 14:27:13 +02001489 You may want to reference some environment variables in the address
1490 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001491
Willy Tarreau18324f52014-06-27 18:10:07 +02001492 <length> is an optional maximum line length. Log lines larger than this value
1493 will be truncated before being sent. The reason is that syslog
1494 servers act differently on log line length. All servers support the
1495 default value of 1024, but some servers simply drop larger lines
1496 while others do log them. If a server supports long lines, it may
1497 make sense to set this value here in order to avoid truncating long
1498 lines. Similarly, if a server drops long lines, it is preferable to
1499 truncate them before sending them. Accepted values are 80 to 65535
1500 inclusive. The default value of 1024 is generally fine for all
1501 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +01001502 JSON-formatted logs may require larger values. You may also need to
1503 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +02001504
Dragan Dosen7ad31542015-09-28 17:16:47 +02001505 <format> is the log format used when generating syslog messages. It may be
1506 one of the following :
1507
Emeric Brun0237c4e2020-11-27 16:24:34 +01001508 local Analog to rfc3164 syslog message format except that hostname
1509 field is stripped. This is the default.
1510 Note: option "log-send-hostname" switches the default to
1511 rfc3164.
1512
1513 rfc3164 The RFC3164 syslog message format.
Dragan Dosen7ad31542015-09-28 17:16:47 +02001514 (https://tools.ietf.org/html/rfc3164)
1515
1516 rfc5424 The RFC5424 syslog message format.
1517 (https://tools.ietf.org/html/rfc5424)
1518
Emeric Brun54648852020-07-06 15:54:06 +02001519 priority A message containing only a level plus syslog facility between
1520 angle brackets such as '<63>', followed by the text. The PID,
1521 date, time, process name and system name are omitted. This is
1522 designed to be used with a local log server.
1523
Willy Tarreaue8746a02018-11-12 08:45:00 +01001524 short A message containing only a level between angle brackets such as
1525 '<3>', followed by the text. The PID, date, time, process name
1526 and system name are omitted. This is designed to be used with a
1527 local log server. This format is compatible with what the systemd
1528 logger consumes.
1529
Emeric Brun54648852020-07-06 15:54:06 +02001530 timed A message containing only a level between angle brackets such as
1531 '<3>', followed by ISO date and by the text. The PID, process
1532 name and system name are omitted. This is designed to be
1533 used with a local log server.
1534
1535 iso A message containing only the ISO date, followed by the text.
1536 The PID, process name and system name are omitted. This is
1537 designed to be used with a local log server.
1538
Willy Tarreauc1b06452018-11-12 11:57:56 +01001539 raw A message containing only the text. The level, PID, date, time,
1540 process name and system name are omitted. This is designed to be
1541 used in containers or during development, where the severity only
1542 depends on the file descriptor used (stdout/stderr).
1543
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02001544 <ranges> A list of comma-separated ranges to identify the logs to sample.
1545 This is used to balance the load of the logs to send to the log
1546 server. The limits of the ranges cannot be null. They are numbered
1547 from 1. The size or period (in number of logs) of the sample must be
1548 set with <sample_size> parameter.
1549
1550 <sample_size>
1551 The size of the sample in number of logs to consider when balancing
1552 their logging loads. It is used to balance the load of the logs to
1553 send to the syslog server. This size must be greater or equal to the
1554 maximum of the high limits of the ranges.
1555 (see also <ranges> parameter).
1556
Robert Tsai81ae1952007-12-05 10:47:29 +01001557 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001558
Willy Tarreaue8746a02018-11-12 08:45:00 +01001559 kern user mail daemon auth syslog lpr news
1560 uucp cron auth2 ftp ntp audit alert cron2
1561 local0 local1 local2 local3 local4 local5 local6 local7
1562
Willy Tarreauc1b06452018-11-12 11:57:56 +01001563 Note that the facility is ignored for the "short" and "raw"
1564 formats, but still required as a positional field. It is
1565 recommended to use "daemon" in this case to make it clear that
1566 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001567
1568 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +02001569 all messages are sent. If a maximum level is specified, only messages with a
1570 severity at least as important as this level will be sent. An optional minimum
1571 level can be specified. If it is set, logs emitted with a more severe level
1572 than this one will be capped to this level. This is used to avoid sending
1573 "emerg" messages on all terminals on some default syslog configurations.
1574 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001575
Cyril Bontédc4d9032012-04-08 21:57:39 +02001576 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +02001577
Joe Williamsdf5b38f2010-12-29 17:05:48 +01001578log-send-hostname [<string>]
1579 Sets the hostname field in the syslog header. If optional "string" parameter
1580 is set the header is set to the string contents, otherwise uses the hostname
1581 of the system. Generally used if one is not relaying logs through an
1582 intermediate syslog server or for simply customizing the hostname printed in
1583 the logs.
1584
Kevinm48936af2010-12-22 16:08:21 +00001585log-tag <string>
1586 Sets the tag field in the syslog header to this string. It defaults to the
1587 program name as launched from the command line, which usually is "haproxy".
1588 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +01001589 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +00001590
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001591lua-load <file>
Thierry Fournier59f11be2020-11-29 00:37:41 +01001592 This global directive loads and executes a Lua file in the shared context
1593 that is visible to all threads. Any variable set in such a context is visible
1594 from any thread. This is the easiest and recommended way to load Lua programs
1595 but it will not scale well if a lot of Lua calls are performed, as only one
1596 thread may be running on the global state at a time. A program loaded this
1597 way will always see 0 in the "core.thread" variable. This directive can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001598 used multiple times.
1599
Thierry Fournier59f11be2020-11-29 00:37:41 +01001600lua-load-per-thread <file>
1601 This global directive loads and executes a Lua file into each started thread.
1602 Any global variable has a thread-local visibility so that each thread could
1603 see a different value. As such it is strongly recommended not to use global
1604 variables in programs loaded this way. An independent copy is loaded and
1605 initialized for each thread, everything is done sequentially and in the
1606 thread's numeric order from 1 to nbthread. If some operations need to be
1607 performed only once, the program should check the "core.thread" variable to
1608 figure what thread is being initialized. Programs loaded this way will run
1609 concurrently on all threads and will be highly scalable. This is the
1610 recommended way to load simple functions that register sample-fetches,
1611 converters, actions or services once it is certain the program doesn't depend
1612 on global variables. For the sake of simplicity, the directive is available
1613 even if only one thread is used and even if threads are disabled (in which
1614 case it will be equivalent to lua-load). This directive can be used multiple
1615 times.
1616
Tim Duesterhusdd74b5f2020-01-12 13:55:40 +01001617lua-prepend-path <string> [<type>]
1618 Prepends the given string followed by a semicolon to Lua's package.<type>
1619 variable.
1620 <type> must either be "path" or "cpath". If <type> is not given it defaults
1621 to "path".
1622
1623 Lua's paths are semicolon delimited lists of patterns that specify how the
1624 `require` function attempts to find the source file of a library. Question
1625 marks (?) within a pattern will be replaced by module name. The path is
1626 evaluated left to right. This implies that paths that are prepended later
1627 will be checked earlier.
1628
1629 As an example by specifying the following path:
1630
1631 lua-prepend-path /usr/share/haproxy-lua/?/init.lua
1632 lua-prepend-path /usr/share/haproxy-lua/?.lua
1633
1634 When `require "example"` is being called Lua will first attempt to load the
1635 /usr/share/haproxy-lua/example.lua script, if that does not exist the
1636 /usr/share/haproxy-lua/example/init.lua will be attempted and the default
1637 paths if that does not exist either.
1638
1639 See https://www.lua.org/pil/8.1.html for the details within the Lua
1640 documentation.
1641
William Lallemand4cfede82017-11-24 22:02:34 +01001642master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +02001643 Master-worker mode. It is equivalent to the command line "-W" argument.
1644 This mode will launch a "master" which will monitor the "workers". Using
1645 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001646 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +02001647 or daemon mode. It is recommended to use this mode with multiprocess and
1648 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +01001649 By default, if a worker exits with a bad return code, in the case of a
1650 segfault for example, all workers will be killed, and the master will leave.
1651 It is convenient to combine this behavior with Restart=on-failure in a
1652 systemd unit file in order to relaunch the whole process. If you don't want
1653 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +02001654
William Lallemand4cfede82017-11-24 22:02:34 +01001655 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +02001656
William Lallemand27edc4b2019-05-07 17:49:33 +02001657mworker-max-reloads <number>
1658 In master-worker mode, this option limits the number of time a worker can
John Roeslerfb2fce12019-07-10 15:45:51 -05001659 survive to a reload. If the worker did not leave after a reload, once its
William Lallemand27edc4b2019-05-07 17:49:33 +02001660 number of reloads is greater than this number, the worker will receive a
1661 SIGTERM. This option helps to keep under control the number of workers.
1662 See also "show proc" in the Management Guide.
1663
Willy Tarreauf42d7942020-10-20 11:54:49 +02001664nbproc <number> (deprecated)
Willy Tarreau6a06a402007-07-15 20:15:28 +02001665 Creates <number> processes when going daemon. This requires the "daemon"
1666 mode. By default, only one process is created, which is the recommended mode
1667 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +01001668 process, it may be needed to fork multiple daemons. When set to a value
1669 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreauf42d7942020-10-20 11:54:49 +02001670 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. This directive is deprecated
1671 and scheduled for removal in 2.5. Please use "nbthread" instead. See also
1672 "daemon" and "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001673
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001674nbthread <number>
1675 This setting is only available when support for threads was built in. It
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001676 makes HAProxy run on <number> threads. This is exclusive with "nbproc". While
Willy Tarreau26f6ae12019-02-02 12:56:15 +01001677 "nbproc" historically used to be the only way to use multiple processors, it
1678 also involved a number of shortcomings related to the lack of synchronization
1679 between processes (health-checks, peers, stick-tables, stats, ...) which do
1680 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +01001681 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
1682 HAProxy is started in foreground. On some platforms supporting CPU affinity,
1683 when nbproc is not used, the default "nbthread" value is automatically set to
1684 the number of CPUs the process is bound to upon startup. This means that the
1685 thread count can easily be adjusted from the calling process using commands
1686 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1687 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001688
Amaury Denoyelle0f50cb92021-03-26 18:50:33 +01001689numa-cpu-mapping
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001690 By default, if running on Linux, HAProxy inspects on startup the CPU topology
Amaury Denoyelle0f50cb92021-03-26 18:50:33 +01001691 of the machine. If a multi-socket machine is detected, the affinity is
1692 automatically calculated to run on the CPUs of a single node. This is done in
1693 order to not suffer from the performance penalties caused by the inter-socket
1694 bus latency. However, if the applied binding is non optimal on a particular
1695 architecture, it can be disabled with the statement 'no numa-cpu-mapping'.
1696 This automatic binding is also not applied if a nbthread statement is present
1697 in the configuration, or the affinity of the process is already specified,
1698 for example via the 'cpu-map' directive or the taskset utility.
1699
Willy Tarreau6a06a402007-07-15 20:15:28 +02001700pidfile <pidfile>
MIZUTA Takeshic32f3942020-08-26 13:46:19 +09001701 Writes PIDs of all daemons into file <pidfile> when daemon mode or writes PID
1702 of master process into file <pidfile> when master-worker mode. This option is
1703 equivalent to the "-p" command line argument. The file must be accessible to
1704 the user starting the process. See also "daemon" and "master-worker".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001705
Willy Tarreau119e50e2020-05-22 13:53:29 +02001706pp2-never-send-local
1707 A bug in the PROXY protocol v2 implementation was present in HAProxy up to
1708 version 2.1, causing it to emit a PROXY command instead of a LOCAL command
1709 for health checks. This is particularly minor but confuses some servers'
1710 logs. Sadly, the bug was discovered very late and revealed that some servers
1711 which possibly only tested their PROXY protocol implementation against
1712 HAProxy fail to properly handle the LOCAL command, and permanently remain in
1713 the "down" state when HAProxy checks them. When this happens, it is possible
1714 to enable this global option to revert to the older (bogus) behavior for the
1715 time it takes to contact the affected components' vendors and get them fixed.
1716 This option is disabled by default and acts on all servers having the
1717 "send-proxy-v2" statement.
1718
Willy Tarreau1d549722016-02-16 12:41:57 +01001719presetenv <name> <value>
1720 Sets environment variable <name> to value <value>. If the variable exists, it
1721 is NOT overwritten. The changes immediately take effect so that the next line
1722 in the configuration file sees the new value. See also "setenv", "resetenv",
1723 and "unsetenv".
1724
1725resetenv [<name> ...]
1726 Removes all environment variables except the ones specified in argument. It
1727 allows to use a clean controlled environment before setting new values with
1728 setenv or unsetenv. Please note that some internal functions may make use of
1729 some environment variables, such as time manipulation functions, but also
1730 OpenSSL or even external checks. This must be used with extreme care and only
1731 after complete validation. The changes immediately take effect so that the
1732 next line in the configuration file sees the new environment. See also
1733 "setenv", "presetenv", and "unsetenv".
1734
Christopher Fauletff4121f2017-11-22 16:38:49 +01001735stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001736 Limits the stats socket to a certain set of processes numbers. By default the
1737 stats socket is bound to all processes, causing a warning to be emitted when
1738 nbproc is greater than 1 because there is no way to select the target process
1739 when connecting. However, by using this setting, it becomes possible to pin
1740 the stats socket to a specific set of processes, typically the first one. The
1741 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001742 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001743 word size (32 or 64). Ranges can be partially defined. The higher bound can
1744 be omitted. In such case, it is replaced by the corresponding maximum
1745 value. A better option consists in using the "process" setting of the "stats
1746 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001747
Baptiste Assmann5626f482015-08-23 10:00:10 +02001748server-state-base <directory>
1749 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001750 file names which do not start with a '/'. See also "server-state-file",
1751 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001752
1753server-state-file <file>
1754 Specifies the path to the file containing state of servers. If the path starts
1755 with a slash ('/'), it is considered absolute, otherwise it is considered
1756 relative to the directory specified using "server-state-base" (if set) or to
1757 the current directory. Before reloading HAProxy, it is possible to save the
1758 servers' current state using the stats command "show servers state". The
1759 output of this command must be written in the file pointed by <file>. When
1760 starting up, before handling traffic, HAProxy will read, load and apply state
1761 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001762 configuration. See also "server-state-base" and "show servers state",
1763 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001764
Willy Tarreau13d2ba22021-03-26 11:38:08 +01001765set-var <var-name> <expr>
1766 Sets the process-wide variable '<var-name>' to the result of the evaluation
1767 of the sample expression <expr>. The variable '<var-name>' may only be a
1768 process-wide variable (using the 'proc.' prefix). It works exactly like the
1769 'set-var' action in TCP or HTTP rules except that the expression is evaluated
1770 at configuration parsing time and that the variable is instantly set. The
1771 sample fetch functions and converters permitted in the expression are only
1772 those using internal data, typically 'int(value)' or 'str(value)'. It's is
1773 possible to reference previously allocated variables as well. These variables
1774 will then be readable (and modifiable) from the regular rule sets.
1775
1776 Example:
1777 global
1778 set-var proc.current_state str(primary)
1779 set-var proc.prio int(100)
1780 set-var proc.threshold int(200),sub(proc.prio)
1781
Willy Tarreau1d549722016-02-16 12:41:57 +01001782setenv <name> <value>
1783 Sets environment variable <name> to value <value>. If the variable exists, it
1784 is overwritten. The changes immediately take effect so that the next line in
1785 the configuration file sees the new value. See also "presetenv", "resetenv",
1786 and "unsetenv".
1787
Willy Tarreau636848a2019-04-15 19:38:50 +02001788set-dumpable
1789 This option is better left disabled by default and enabled only upon a
William Dauchyec730982019-10-27 20:08:10 +01001790 developer's request. If it has been enabled, it may still be forcibly
1791 disabled by prefixing it with the "no" keyword. It has no impact on
1792 performance nor stability but will try hard to re-enable core dumps that were
1793 possibly disabled by file size limitations (ulimit -f), core size limitations
1794 (ulimit -c), or "dumpability" of a process after changing its UID/GID (such
1795 as /proc/sys/fs/suid_dumpable on Linux). Core dumps might still be limited by
1796 the current directory's permissions (check what directory the file is started
1797 from), the chroot directory's permission (it may be needed to temporarily
1798 disable the chroot directive or to move it to a dedicated writable location),
1799 or any other system-specific constraint. For example, some Linux flavours are
1800 notorious for replacing the default core file with a path to an executable
1801 not even installed on the system (check /proc/sys/kernel/core_pattern). Often,
1802 simply writing "core", "core.%p" or "/var/log/core/core.%p" addresses the
1803 issue. When trying to enable this option waiting for a rare issue to
1804 re-appear, it's often a good idea to first try to obtain such a dump by
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001805 issuing, for example, "kill -11" to the "haproxy" process and verify that it
William Dauchyec730982019-10-27 20:08:10 +01001806 leaves a core where expected when dying.
Willy Tarreau636848a2019-04-15 19:38:50 +02001807
Willy Tarreau610f04b2014-02-13 11:36:41 +01001808ssl-default-bind-ciphers <ciphers>
1809 This setting is only available when support for OpenSSL was built in. It sets
1810 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001811 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001812 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001813 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1814 information and recommendations see e.g.
1815 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1816 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1817 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1818 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001819
1820ssl-default-bind-ciphersuites <ciphersuites>
1821 This setting is only available when support for OpenSSL was built in and
1822 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1823 describing the list of cipher algorithms ("cipher suite") that are negotiated
1824 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1825 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001826 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1827 cipher configuration for TLSv1.2 and earlier, please check the
1828 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001829 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001830
Jerome Magninb203ff62020-04-03 15:28:22 +02001831ssl-default-bind-curves <curves>
1832 This setting is only available when support for OpenSSL was built in. It sets
1833 the default string describing the list of elliptic curves algorithms ("curve
1834 suite") that are negotiated during the SSL/TLS handshake with ECDHE. The format
1835 of the string is a colon-delimited list of curve name.
1836 Please check the "bind" keyword for more information.
1837
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001838ssl-default-bind-options [<option>]...
1839 This setting is only available when support for OpenSSL was built in. It sets
1840 default ssl-options to force on all "bind" lines. Please check the "bind"
1841 keyword to see available options.
1842
1843 Example:
1844 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001845 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001846
Willy Tarreau610f04b2014-02-13 11:36:41 +01001847ssl-default-server-ciphers <ciphers>
1848 This setting is only available when support for OpenSSL was built in. It
1849 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001850 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001851 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001852 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1853 information and recommendations see e.g.
1854 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1855 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1856 For TLSv1.3 cipher configuration, please check the
1857 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1858 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001859
1860ssl-default-server-ciphersuites <ciphersuites>
1861 This setting is only available when support for OpenSSL was built in and
1862 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1863 string describing the list of cipher algorithms that are negotiated during
1864 the TLSv1.3 handshake with the server, for all "server" lines which do not
1865 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001866 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1867 cipher configuration for TLSv1.2 and earlier, please check the
1868 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1869 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001870
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001871ssl-default-server-options [<option>]...
1872 This setting is only available when support for OpenSSL was built in. It sets
1873 default ssl-options to force on all "server" lines. Please check the "server"
1874 keyword to see available options.
1875
Remi Gacogne47783ef2015-05-29 15:53:22 +02001876ssl-dh-param-file <file>
1877 This setting is only available when support for OpenSSL was built in. It sets
1878 the default DH parameters that are used during the SSL/TLS handshake when
1879 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001880 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001881 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001882 are not specified either by using ssl-dh-param-file or by setting them
1883 directly in the certificate file, pre-generated DH parameters of the size
1884 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1885 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001886 Custom DH parameters may be generated by using the OpenSSL command
1887 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1888 parameters should not be considered secure anymore.
1889
William Lallemand8e8581e2020-10-20 17:36:46 +02001890ssl-load-extra-del-ext
1891 This setting allows to configure the way HAProxy does the lookup for the
1892 extra SSL files. By default HAProxy adds a new extension to the filename.
William Lallemand089c1382020-10-23 17:35:12 +02001893 (ex: with "foobar.crt" load "foobar.crt.key"). With this option enabled,
William Lallemand8e8581e2020-10-20 17:36:46 +02001894 HAProxy removes the extension before adding the new one (ex: with
William Lallemand089c1382020-10-23 17:35:12 +02001895 "foobar.crt" load "foobar.key").
1896
1897 Your crt file must have a ".crt" extension for this option to work.
William Lallemand8e8581e2020-10-20 17:36:46 +02001898
1899 This option is not compatible with bundle extensions (.ecdsa, .rsa. .dsa)
1900 and won't try to remove them.
1901
1902 This option is disabled by default. See also "ssl-load-extra-files".
1903
William Lallemand4c5adbf2020-02-24 14:23:22 +01001904ssl-load-extra-files <none|all|bundle|sctl|ocsp|issuer|key>*
William Lallemand3af48e72020-02-03 17:15:52 +01001905 This setting alters the way HAProxy will look for unspecified files during
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001906 the loading of the SSL certificates. This option applies to certificates
1907 associated to "bind" lines as well as "server" lines but some of the extra
1908 files will not have any functional impact for "server" line certificates.
William Lallemand3af48e72020-02-03 17:15:52 +01001909
1910 By default, HAProxy discovers automatically a lot of files not specified in
1911 the configuration, and you may want to disable this behavior if you want to
1912 optimize the startup time.
1913
1914 "none": Only load the files specified in the configuration. Don't try to load
1915 a certificate bundle if the file does not exist. In the case of a directory,
1916 it won't try to bundle the certificates if they have the same basename.
1917
1918 "all": This is the default behavior, it will try to load everything,
William Lallemand4c5adbf2020-02-24 14:23:22 +01001919 bundles, sctl, ocsp, issuer, key.
William Lallemand3af48e72020-02-03 17:15:52 +01001920
1921 "bundle": When a file specified in the configuration does not exist, HAProxy
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001922 will try to load a "cert bundle". Certificate bundles are only managed on the
1923 frontend side and will not work for backend certificates.
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001924
1925 Starting from HAProxy 2.3, the bundles are not loaded in the same OpenSSL
1926 certificate store, instead it will loads each certificate in a separate
1927 store which is equivalent to declaring multiple "crt". OpenSSL 1.1.1 is
1928 required to achieve this. Which means that bundles are now used only for
1929 backward compatibility and are not mandatory anymore to do an hybrid RSA/ECC
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001930 bind configuration.
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001931
1932 To associate these PEM files into a "cert bundle" that is recognized by
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001933 HAProxy, they must be named in the following way: All PEM files that are to
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001934 be bundled must have the same base name, with a suffix indicating the key
1935 type. Currently, three suffixes are supported: rsa, dsa and ecdsa. For
1936 example, if www.example.com has two PEM files, an RSA file and an ECDSA
1937 file, they must be named: "example.pem.rsa" and "example.pem.ecdsa". The
1938 first part of the filename is arbitrary; only the suffix matters. To load
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001939 this bundle into HAProxy, specify the base name only:
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001940
1941 Example : bind :8443 ssl crt example.pem
1942
Daniel Corbett9f0843f2021-05-08 10:50:37 -04001943 Note that the suffix is not given to HAProxy; this tells HAProxy to look for
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001944 a cert bundle.
1945
1946 HAProxy will load all PEM files in the bundle as if they were configured
1947 separately in several "crt".
1948
1949 The bundle loading does not have an impact anymore on the directory loading
1950 since files are loading separately.
1951
1952 On the CLI, bundles are seen as separate files, and the bundle extension is
1953 required to commit them.
1954
William Dauchy57dd6f12020-10-06 15:22:37 +02001955 OCSP files (.ocsp), issuer files (.issuer), Certificate Transparency (.sctl)
William Lallemandf9ff3ec2020-10-02 17:57:44 +02001956 as well as private keys (.key) are supported with multi-cert bundling.
William Lallemand3af48e72020-02-03 17:15:52 +01001957
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001958 "sctl": Try to load "<basename>.sctl" for each crt keyword. If provided for
1959 a backend certificate, it will be loaded but will not have any functional
1960 impact.
William Lallemand3af48e72020-02-03 17:15:52 +01001961
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001962 "ocsp": Try to load "<basename>.ocsp" for each crt keyword. If provided for
1963 a backend certificate, it will be loaded but will not have any functional
1964 impact.
William Lallemand3af48e72020-02-03 17:15:52 +01001965
1966 "issuer": Try to load "<basename>.issuer" if the issuer of the OCSP file is
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001967 not provided in the PEM file. If provided for a backend certificate, it will
1968 be loaded but will not have any functional impact.
William Lallemand3af48e72020-02-03 17:15:52 +01001969
William Lallemand4c5adbf2020-02-24 14:23:22 +01001970 "key": If the private key was not provided by the PEM file, try to load a
1971 file "<basename>.key" containing a private key.
1972
William Lallemand3af48e72020-02-03 17:15:52 +01001973 The default behavior is "all".
1974
1975 Example:
1976 ssl-load-extra-files bundle sctl
1977 ssl-load-extra-files sctl ocsp issuer
1978 ssl-load-extra-files none
1979
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +02001980 See also: "crt", section 5.1 about bind options and section 5.2 about server
1981 options.
William Lallemand3af48e72020-02-03 17:15:52 +01001982
Emeric Brun850efd52014-01-29 12:24:34 +01001983ssl-server-verify [none|required]
1984 The default behavior for SSL verify on servers side. If specified to 'none',
1985 servers certificates are not verified. The default is 'required' except if
1986 forced using cmdline option '-dV'.
1987
Emmanuel Hocdetc3b7e742020-04-22 11:06:19 +02001988ssl-skip-self-issued-ca
Daniel Corbett67a82712020-07-06 23:01:19 -04001989 Self issued CA, aka x509 root CA, is the anchor for chain validation: as a
Emmanuel Hocdetc3b7e742020-04-22 11:06:19 +02001990 server is useless to send it, client must have it. Standard configuration
1991 need to not include such CA in PEM file. This option allows you to keep such
1992 CA in PEM file without sending it to the client. Use case is to provide
1993 issuer for ocsp without the need for '.issuer' file and be able to share it
1994 with 'issuers-chain-path'. This concerns all certificates without intermediate
1995 certificates. It's useless for BoringSSL, .issuer is ignored because ocsp
William Lallemand9a1d8392020-08-10 17:28:23 +02001996 bits does not need it. Requires at least OpenSSL 1.0.2.
Emmanuel Hocdetc3b7e742020-04-22 11:06:19 +02001997
Willy Tarreauabb175f2012-09-24 12:43:26 +02001998stats socket [<address:port>|<path>] [param*]
1999 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
2000 Connections to this socket will return various statistics outputs and even
2001 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02002002 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02002003 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02002004
Willy Tarreauabb175f2012-09-24 12:43:26 +02002005 All parameters supported by "bind" lines are supported, for instance to
2006 restrict access to some users or their access rights. Please consult
2007 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02002008
2009stats timeout <timeout, in milliseconds>
2010 The default timeout on the stats socket is set to 10 seconds. It is possible
2011 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01002012 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02002013
2014stats maxconn <connections>
2015 By default, the stats socket is limited to 10 concurrent connections. It is
2016 possible to change this value with "stats maxconn".
2017
Willy Tarreau6a06a402007-07-15 20:15:28 +02002018uid <number>
Thayne McCombscdbcca92021-01-07 21:24:41 -07002019 Changes the process's user ID to <number>. It is recommended that the user ID
Willy Tarreau6a06a402007-07-15 20:15:28 +02002020 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
2021 be started with superuser privileges in order to be able to switch to another
2022 one. See also "gid" and "user".
2023
2024ulimit-n <number>
2025 Sets the maximum number of per-process file-descriptors to <number>. By
2026 default, it is automatically computed, so it is recommended not to use this
2027 option.
2028
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002029unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
2030 [ group <group> ] [ gid <gid> ]
2031
2032 Fixes common settings to UNIX listening sockets declared in "bind" statements.
2033 This is mainly used to simplify declaration of those UNIX sockets and reduce
2034 the risk of errors, since those settings are most commonly required but are
2035 also process-specific. The <prefix> setting can be used to force all socket
2036 path to be relative to that directory. This might be needed to access another
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002037 component's chroot. Note that those paths are resolved before HAProxy chroots
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002038 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
2039 all have the same meaning as their homonyms used by the "bind" statement. If
2040 both are specified, the "bind" statement has priority, meaning that the
2041 "unix-bind" settings may be seen as process-wide default settings.
2042
Willy Tarreau1d549722016-02-16 12:41:57 +01002043unsetenv [<name> ...]
2044 Removes environment variables specified in arguments. This can be useful to
2045 hide some sensitive information that are occasionally inherited from the
2046 user's environment during some operations. Variables which did not exist are
2047 silently ignored so that after the operation, it is certain that none of
2048 these variables remain. The changes immediately take effect so that the next
2049 line in the configuration file will not see these variables. See also
2050 "setenv", "presetenv", and "resetenv".
2051
Willy Tarreau6a06a402007-07-15 20:15:28 +02002052user <user name>
2053 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
2054 See also "uid" and "group".
2055
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02002056node <name>
2057 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
2058
2059 This statement is useful in HA configurations where two or more processes or
2060 servers share the same IP address. By setting a different node-name on all
2061 nodes, it becomes easy to immediately spot what server is handling the
2062 traffic.
2063
2064description <text>
2065 Add a text that describes the instance.
2066
2067 Please note that it is required to escape certain characters (# for example)
2068 and this text is inserted into a html page so you should avoid using
2069 "<" and ">" characters.
2070
Thomas Holmesdb04f192015-05-18 13:21:39 +0100207151degrees-data-file <file path>
2072 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01002073 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01002074
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002075 Please note that this option is only available when HAProxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01002076 compiled with USE_51DEGREES.
2077
Ben Shillitof25e8e52016-12-02 14:25:37 +0000207851degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01002079 A list of 51Degrees property names to be load from the dataset. A full list
2080 of names is available on the 51Degrees website:
2081 https://51degrees.com/resources/property-dictionary
2082
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002083 Please note that this option is only available when HAProxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01002084 compiled with USE_51DEGREES.
2085
Dragan Dosen93b38d92015-06-29 16:43:25 +0200208651degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01002087 A char that will be appended to every property value in a response header
2088 containing 51Degrees results. If not set that will be set as ','.
2089
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002090 Please note that this option is only available when HAProxy has been
Dragan Dosenae6d39a2015-06-29 16:43:27 +02002091 compiled with USE_51DEGREES.
2092
209351degrees-cache-size <number>
2094 Sets the size of the 51Degrees converter cache to <number> entries. This
2095 is an LRU cache which reminds previous device detections and their results.
2096 By default, this cache is disabled.
2097
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002098 Please note that this option is only available when HAProxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01002099 compiled with USE_51DEGREES.
2100
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002101wurfl-data-file <file path>
2102 The path of the WURFL data file to provide device detection services. The
2103 file should be accessible by HAProxy with relevant permissions.
2104
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002105 Please note that this option is only available when HAProxy has been compiled
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002106 with USE_WURFL=1.
2107
2108wurfl-information-list [<capability>]*
2109 A space-delimited list of WURFL capabilities, virtual capabilities, property
2110 names we plan to use in injected headers. A full list of capability and
2111 virtual capability names is available on the Scientiamobile website :
2112
2113 https://www.scientiamobile.com/wurflCapability
2114
2115 Valid WURFL properties are:
2116 - wurfl_id Contains the device ID of the matched device.
2117
2118 - wurfl_root_id Contains the device root ID of the matched
2119 device.
2120
2121 - wurfl_isdevroot Tells if the matched device is a root device.
2122 Possible values are "TRUE" or "FALSE".
2123
2124 - wurfl_useragent The original useragent coming with this
2125 particular web request.
2126
2127 - wurfl_api_version Contains a string representing the currently
2128 used Libwurfl API version.
2129
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002130 - wurfl_info A string containing information on the parsed
2131 wurfl.xml and its full path.
2132
2133 - wurfl_last_load_time Contains the UNIX timestamp of the last time
2134 WURFL has been loaded successfully.
2135
2136 - wurfl_normalized_useragent The normalized useragent.
2137
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002138 Please note that this option is only available when HAProxy has been compiled
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002139 with USE_WURFL=1.
2140
2141wurfl-information-list-separator <char>
2142 A char that will be used to separate values in a response header containing
2143 WURFL results. If not set that a comma (',') will be used by default.
2144
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002145 Please note that this option is only available when HAProxy has been compiled
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002146 with USE_WURFL=1.
2147
2148wurfl-patch-file [<file path>]
2149 A list of WURFL patch file paths. Note that patches are loaded during startup
2150 thus before the chroot.
2151
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002152 Please note that this option is only available when HAProxy has been compiled
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002153 with USE_WURFL=1.
2154
paulborilebad132c2019-04-18 11:57:04 +02002155wurfl-cache-size <size>
2156 Sets the WURFL Useragent cache size. For faster lookups, already processed user
2157 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002158 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02002159 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002160
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002161 Please note that this option is only available when HAProxy has been compiled
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02002162 with USE_WURFL=1.
2163
William Dauchy0fec3ab2019-10-27 20:08:11 +01002164strict-limits
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002165 Makes process fail at startup when a setrlimit fails. HAProxy tries to set the
William Dauchya5194602020-03-28 19:29:58 +01002166 best setrlimit according to what has been calculated. If it fails, it will
2167 emit a warning. This option is here to guarantee an explicit failure of
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002168 HAProxy when those limits fail. It is enabled by default. It may still be
William Dauchya5194602020-03-28 19:29:58 +01002169 forcibly disabled by prefixing it with the "no" keyword.
William Dauchy0fec3ab2019-10-27 20:08:11 +01002170
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021713.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02002172-----------------------
2173
Willy Tarreaubeb859a2018-11-22 18:07:59 +01002174busy-polling
2175 In some situations, especially when dealing with low latency on processors
2176 supporting a variable frequency or when running inside virtual machines, each
2177 time the process waits for an I/O using the poller, the processor goes back
2178 to sleep or is offered to another VM for a long time, and it causes
2179 excessively high latencies. This option provides a solution preventing the
2180 processor from sleeping by always using a null timeout on the pollers. This
2181 results in a significant latency reduction (30 to 100 microseconds observed)
2182 at the expense of a risk to overheat the processor. It may even be used with
2183 threads, in which case improperly bound threads may heavily conflict,
2184 resulting in a worse performance and high values for the CPU stolen fields
2185 in "show info" output, indicating which threads are misconfigured. It is
2186 important not to let the process run on the same processor as the network
2187 interrupts when this option is used. It is also better to avoid using it on
2188 multiple CPU threads sharing the same core. This option is disabled by
2189 default. If it has been enabled, it may still be forcibly disabled by
2190 prefixing it with the "no" keyword. It is ignored by the "select" and
2191 "poll" pollers.
2192
William Dauchy3894d972019-12-28 15:36:02 +01002193 This option is automatically disabled on old processes in the context of
2194 seamless reload; it avoids too much cpu conflicts when multiple processes
2195 stay around for some time waiting for the end of their current connections.
2196
Willy Tarreau1746eec2014-04-25 10:46:47 +02002197max-spread-checks <delay in milliseconds>
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002198 By default, HAProxy tries to spread the start of health checks across the
Willy Tarreau1746eec2014-04-25 10:46:47 +02002199 smallest health check interval of all the servers in a farm. The principle is
2200 to avoid hammering services running on the same server. But when using large
2201 check intervals (10 seconds or more), the last servers in the farm take some
2202 time before starting to be tested, which can be a problem. This parameter is
2203 used to enforce an upper bound on delay between the first and the last check,
2204 even if the servers' check intervals are larger. When servers run with
2205 shorter intervals, their intervals will be respected though.
2206
Willy Tarreau6a06a402007-07-15 20:15:28 +02002207maxconn <number>
2208 Sets the maximum per-process number of concurrent connections to <number>. It
2209 is equivalent to the command-line argument "-n". Proxies will stop accepting
2210 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02002211 automatically adjusted according to this value. See also "ulimit-n". Note:
2212 the "select" poller cannot reliably use more than 1024 file descriptors on
2213 some platforms. If your platform only supports select and reports "select
2214 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01002215 below 500 in general). If this value is not set, it will automatically be
2216 calculated based on the current file descriptors limit reported by the
2217 "ulimit -n" command, possibly reduced to a lower value if a memory limit
2218 is enforced, based on the buffer size, memory allocated to compression, SSL
2219 cache size, and use or not of SSL and the associated maxsslconn (which can
2220 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02002221
Willy Tarreau81c25d02011-09-07 15:17:21 +02002222maxconnrate <number>
2223 Sets the maximum per-process number of connections per second to <number>.
2224 Proxies will stop accepting connections when this limit is reached. It can be
2225 used to limit the global capacity regardless of each frontend capacity. It is
2226 important to note that this can only be used as a service protection measure,
2227 as there will not necessarily be a fair share between frontends when the
2228 limit is reached, so it's a good idea to also limit each frontend to some
2229 value close to its expected share. Also, lowering tune.maxaccept can improve
2230 fairness.
2231
William Lallemandd85f9172012-11-09 17:05:39 +01002232maxcomprate <number>
2233 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01002234 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01002235 level will be decreased during the session. If the maximum is reached at the
2236 beginning of a session, the session will not compress at all. If the maximum
2237 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01002238 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01002239 default value.
2240
William Lallemand072a2bf2012-11-20 17:01:01 +01002241maxcompcpuusage <number>
2242 Sets the maximum CPU usage HAProxy can reach before stopping the compression
2243 for new requests or decreasing the compression level of current requests.
2244 It works like 'maxcomprate' but measures CPU usage instead of incoming data
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002245 bandwidth. The value is expressed in percent of the CPU used by HAProxy. In
William Lallemand072a2bf2012-11-20 17:01:01 +01002246 case of multiple processes (nbproc > 1), each process manages its individual
2247 usage. A value of 100 disable the limit. The default value is 100. Setting
2248 a lower value will prevent the compression work from slowing the whole
2249 process down and from introducing high latencies.
2250
Willy Tarreauff4f82d2009-02-06 11:28:13 +01002251maxpipes <number>
2252 Sets the maximum per-process number of pipes to <number>. Currently, pipes
2253 are only used by kernel-based tcp splicing. Since a pipe contains two file
2254 descriptors, the "ulimit-n" value will be increased accordingly. The default
2255 value is maxconn/4, which seems to be more than enough for most heavy usages.
2256 The splice code dynamically allocates and releases pipes, and can fall back
2257 to standard copy, so setting this value too low may only impact performance.
2258
Willy Tarreau93e7c002013-10-07 18:51:07 +02002259maxsessrate <number>
2260 Sets the maximum per-process number of sessions per second to <number>.
2261 Proxies will stop accepting connections when this limit is reached. It can be
2262 used to limit the global capacity regardless of each frontend capacity. It is
2263 important to note that this can only be used as a service protection measure,
2264 as there will not necessarily be a fair share between frontends when the
2265 limit is reached, so it's a good idea to also limit each frontend to some
2266 value close to its expected share. Also, lowering tune.maxaccept can improve
2267 fairness.
2268
Willy Tarreau403edff2012-09-06 11:58:37 +02002269maxsslconn <number>
2270 Sets the maximum per-process number of concurrent SSL connections to
2271 <number>. By default there is no SSL-specific limit, which means that the
2272 global maxconn setting will apply to all connections. Setting this limit
2273 avoids having openssl use too much memory and crash when malloc returns NULL
2274 (since it unfortunately does not reliably check for such conditions). Note
2275 that the limit applies both to incoming and outgoing connections, so one
2276 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01002277 If this value is not set, but a memory limit is enforced, this value will be
2278 automatically computed based on the memory limit, maxconn, the buffer size,
2279 memory allocated to compression, SSL cache size, and use of SSL in either
2280 frontends, backends or both. If neither maxconn nor maxsslconn are specified
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002281 when there is a memory limit, HAProxy will automatically adjust these values
Willy Tarreaud0256482015-01-15 21:45:22 +01002282 so that 100% of the connections can be made over SSL with no risk, and will
2283 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02002284
Willy Tarreaue43d5322013-10-07 20:01:52 +02002285maxsslrate <number>
2286 Sets the maximum per-process number of SSL sessions per second to <number>.
2287 SSL listeners will stop accepting connections when this limit is reached. It
2288 can be used to limit the global SSL CPU usage regardless of each frontend
2289 capacity. It is important to note that this can only be used as a service
2290 protection measure, as there will not necessarily be a fair share between
2291 frontends when the limit is reached, so it's a good idea to also limit each
2292 frontend to some value close to its expected share. It is also important to
2293 note that the sessions are accounted before they enter the SSL stack and not
2294 after, which also protects the stack against bad handshakes. Also, lowering
2295 tune.maxaccept can improve fairness.
2296
William Lallemand9d5f5482012-11-07 16:12:57 +01002297maxzlibmem <number>
2298 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
2299 When the maximum amount is reached, future sessions will not compress as long
2300 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01002301 The default value is 0. The value is available in bytes on the UNIX socket
2302 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
2303 "ZlibMemUsage" in bytes.
2304
Willy Tarreau6a06a402007-07-15 20:15:28 +02002305noepoll
2306 Disables the use of the "epoll" event polling system on Linux. It is
2307 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01002308 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02002309
2310nokqueue
2311 Disables the use of the "kqueue" event polling system on BSD. It is
2312 equivalent to the command-line argument "-dk". The next polling system
2313 used will generally be "poll". See also "nopoll".
2314
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00002315noevports
2316 Disables the use of the event ports event polling system on SunOS systems
2317 derived from Solaris 10 and later. It is equivalent to the command-line
2318 argument "-dv". The next polling system used will generally be "poll". See
2319 also "nopoll".
2320
Willy Tarreau6a06a402007-07-15 20:15:28 +02002321nopoll
2322 Disables the use of the "poll" event polling system. It is equivalent to the
2323 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002324 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00002325 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
2326 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02002327
Willy Tarreauff4f82d2009-02-06 11:28:13 +01002328nosplice
2329 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002330 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01002331 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002332 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01002333 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
2334 be used. This option makes it easier to globally disable kernel splicing in
2335 case of doubt. See also "option splice-auto", "option splice-request" and
2336 "option splice-response".
2337
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002338nogetaddrinfo
2339 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
2340 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
2341
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00002342noreuseport
2343 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
2344 command line argument "-dR".
2345
Willy Tarreauca3afc22021-05-05 18:33:19 +02002346profiling.memory { on | off }
2347 Enables ('on') or disables ('off') per-function memory profiling. This will
2348 keep usage statistics of malloc/calloc/realloc/free calls anywhere in the
2349 process (including libraries) which will be reported on the CLI using the
2350 "show profiling" command. This is essentially meant to be used when an
2351 abnormal memory usage is observed that cannot be explained by the pools and
2352 other info are required. The performance hit will typically be around 1%,
2353 maybe a bit more on highly threaded machines, so it is normally suitable for
2354 use in production. The same may be achieved at run time on the CLI using the
2355 "set profiling memory" command, please consult the management manual.
2356
Willy Tarreaud2d33482019-04-25 17:09:07 +02002357profiling.tasks { auto | on | off }
2358 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
2359 the profiling automatically turns on a thread when it starts to suffer from
2360 an average latency of 1000 microseconds or higher as reported in the
2361 "avg_loop_us" activity field, and automatically turns off when the latency
John Roeslerfb2fce12019-07-10 15:45:51 -05002362 returns below 990 microseconds (this value is an average over the last 1024
Willy Tarreaud2d33482019-04-25 17:09:07 +02002363 loops so it does not vary quickly and tends to significantly smooth short
2364 spikes). It may also spontaneously trigger from time to time on overloaded
2365 systems, containers, or virtual machines, or when the system swaps (which
2366 must absolutely never happen on a load balancer).
2367
2368 CPU profiling per task can be very convenient to report where the time is
2369 spent and which requests have what effect on which other request. Enabling
2370 it will typically affect the overall's performance by less than 1%, thus it
2371 is recommended to leave it to the default 'auto' value so that it only
2372 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01002373 supporting the clock_gettime(2) syscall with clock identifiers
2374 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
2375 be zero. This option may be changed at run time using "set profiling" on the
2376 CLI.
2377
Willy Tarreaufe255b72007-10-14 23:09:26 +02002378spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09002379 Sometimes it is desirable to avoid sending agent and health checks to
2380 servers at exact intervals, for instance when many logical servers are
2381 located on the same physical server. With the help of this parameter, it
2382 becomes possible to add some randomness in the check interval between 0
2383 and +/- 50%. A value between 2 and 5 seems to show good results. The
2384 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02002385
Davor Ocelice9ed2812017-12-25 17:49:28 +01002386ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00002387 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01002388 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00002389 multiple times, it will simply enable multiple crypto engines. Referencing an
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002390 unsupported engine will prevent HAProxy from starting. Note that many engines
Grant Zhang872f9c22017-01-21 01:10:18 +00002391 will lead to lower HTTPS performance than pure software with recent
2392 processors. The optional command "algo" sets the default algorithms an ENGINE
2393 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01002394 of "ALL" uses the engine for all cryptographic operations. If no list of
2395 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00002396 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
2397 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
2398 openssl configuration file uses:
2399 https://www.openssl.org/docs/man1.0.2/apps/config.html
2400
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00002401ssl-mode-async
2402 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02002403 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00002404 implementation supports a maximum of 32 engines. The Openssl ASYNC API
2405 doesn't support moving read/write buffers and is not compliant with
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002406 HAProxy's buffer management. So the asynchronous mode is disabled on
John Roeslerfb2fce12019-07-10 15:45:51 -05002407 read/write operations (it is only enabled during initial and renegotiation
Emeric Brunb5e42a82017-06-06 12:35:14 +00002408 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00002409
Willy Tarreau33cb0652014-12-23 22:52:37 +01002410tune.buffers.limit <number>
2411 Sets a hard limit on the number of buffers which may be allocated per process.
2412 The default value is zero which means unlimited. The minimum non-zero value
2413 will always be greater than "tune.buffers.reserve" and should ideally always
2414 be about twice as large. Forcing this value can be particularly useful to
2415 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01002416 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01002417 another one to be released by another session. Since buffers are dynamically
2418 allocated and released, the waiting time is very short and not perceptible
2419 provided that limits remain reasonable. In fact sometimes reducing the limit
2420 may even increase performance by increasing the CPU cache's efficiency. Tests
2421 have shown good results on average HTTP traffic with a limit to 1/10 of the
2422 expected global maxconn setting, which also significantly reduces memory
2423 usage. The memory savings come from the fact that a number of connections
2424 will not allocate 2*tune.bufsize. It is best not to touch this value unless
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002425 advised to do so by an HAProxy core developer.
Willy Tarreau33cb0652014-12-23 22:52:37 +01002426
Willy Tarreau1058ae72014-12-23 22:40:40 +01002427tune.buffers.reserve <number>
2428 Sets the number of buffers which are pre-allocated and reserved for use only
2429 during memory shortage conditions resulting in failed memory allocations. The
2430 minimum value is 2 and is also the default. There is no reason a user would
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002431 want to change this value, it's mostly aimed at HAProxy core developers.
Willy Tarreau1058ae72014-12-23 22:40:40 +01002432
Willy Tarreau27a674e2009-08-17 07:23:33 +02002433tune.bufsize <number>
2434 Sets the buffer size to this size (in bytes). Lower values allow more
2435 sessions to coexist in the same amount of RAM, and higher values allow some
2436 applications with very large cookies to work. The default value is 16384 and
2437 can be changed at build time. It is strongly recommended not to change this
2438 from the default value, as very low values will break some services such as
2439 statistics, and values larger than default size will increase memory usage,
2440 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01002441 parameter should be decreased by the same factor as this one is increased. In
2442 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002443 HTTP request is larger than (tune.bufsize - tune.maxrewrite), HAProxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04002444 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002445 than this size, HAProxy will return HTTP 502 (Bad Gateway). Note that the
Willy Tarreauc77d3642018-12-12 06:19:42 +01002446 value set using this parameter will automatically be rounded up to the next
2447 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02002448
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +01002449tune.chksize <number> (deprecated)
2450 This option is deprecated and ignored.
Willy Tarreau43961d52010-10-04 20:39:20 +02002451
William Lallemandf3747832012-11-09 12:33:10 +01002452tune.comp.maxlevel <number>
2453 Sets the maximum compression level. The compression level affects CPU
2454 usage during compression. This value affects CPU usage during compression.
2455 Each session using compression initializes the compression algorithm with
2456 this value. The default value is 1.
2457
Willy Tarreauc299e1e2019-02-27 11:35:12 +01002458tune.fail-alloc
2459 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
2460 allocation attempt fails. Must be between 0 (no failure) and 100 (no
2461 success). This is useful to debug and make sure memory failures are handled
2462 gracefully.
2463
Willy Tarreaubc52bec2020-06-18 08:58:47 +02002464tune.fd.edge-triggered { on | off } [ EXPERIMENTAL ]
2465 Enables ('on') or disables ('off') the edge-triggered polling mode for FDs
2466 that support it. This is currently only support with epoll. It may noticeably
2467 reduce the number of epoll_ctl() calls and slightly improve performance in
2468 certain scenarios. This is still experimental, it may result in frozen
2469 connections if bugs are still present, and is disabled by default.
2470
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02002471tune.h2.header-table-size <number>
2472 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
2473 cannot be larger than 65536 bytes. A larger value may help certain clients
2474 send more compact requests, depending on their capabilities. This amount of
2475 memory is consumed for each HTTP/2 connection. It is recommended not to
2476 change it.
2477
Willy Tarreaue6baec02017-07-27 11:45:11 +02002478tune.h2.initial-window-size <number>
2479 Sets the HTTP/2 initial window size, which is the number of bytes the client
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002480 can upload before waiting for an acknowledgment from HAProxy. This setting
Davor Ocelice9ed2812017-12-25 17:49:28 +01002481 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02002482 The default value is 65535, which roughly allows up to 5 Mbps of upload
2483 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
2484 over a 1-ms local network. It can make sense to increase this value to allow
2485 faster uploads, or to reduce it to increase fairness when dealing with many
2486 clients. It doesn't affect resource usage.
2487
Willy Tarreau5242ef82017-07-27 11:47:28 +02002488tune.h2.max-concurrent-streams <number>
2489 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
2490 number of outstanding requests on a single connection). The default value is
2491 100. A larger one may slightly improve page load time for complex sites when
2492 visited over high latency networks, but increases the amount of resources a
2493 single client may allocate. A value of zero disables the limit so a single
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002494 client may create as many streams as allocatable by HAProxy. It is highly
Willy Tarreau5242ef82017-07-27 11:47:28 +02002495 recommended not to change this value.
2496
Willy Tarreaua24b35c2019-02-21 13:24:36 +01002497tune.h2.max-frame-size <number>
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002498 Sets the HTTP/2 maximum frame size that HAProxy announces it is willing to
Willy Tarreaua24b35c2019-02-21 13:24:36 +01002499 receive to its peers. The default value is the largest between 16384 and the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002500 buffer size (tune.bufsize). In any case, HAProxy will not announce support
Willy Tarreaua24b35c2019-02-21 13:24:36 +01002501 for frame sizes larger than buffers. The main purpose of this setting is to
2502 allow to limit the maximum frame size setting when using large buffers. Too
2503 large frame sizes might have performance impact or cause some peers to
2504 misbehave. It is highly recommended not to change this value.
2505
Willy Tarreau193b8c62012-11-22 00:17:38 +01002506tune.http.cookielen <number>
2507 Sets the maximum length of captured cookies. This is the maximum value that
2508 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
2509 will automatically be truncated to this one. It is important not to set too
2510 high a value because all cookie captures still allocate this size whatever
2511 their configured value (they share a same pool). This value is per request
2512 per response, so the memory allocated is twice this value per connection.
2513 When not specified, the limit is set to 63 characters. It is recommended not
2514 to change this value.
2515
Stéphane Cottin23e9e932017-05-18 08:58:41 +02002516tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01002517 Sets the maximum length of request URI in logs. This prevents truncating long
2518 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02002519 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01002520 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02002521 configuration directives too.
2522 The default value is 1024.
2523
Willy Tarreauac1932d2011-10-24 19:14:41 +02002524tune.http.maxhdr <number>
2525 Sets the maximum number of headers in a request. When a request comes with a
2526 number of headers greater than this value (including the first line), it is
2527 rejected with a "400 Bad Request" status code. Similarly, too large responses
2528 are blocked with "502 Bad Gateway". The default value is 101, which is enough
2529 for all usages, considering that the widely deployed Apache server uses the
2530 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02002531 a buggy application to work by the time it gets fixed. The accepted range is
2532 1..32767. Keep in mind that each new header consumes 32bits of memory for
2533 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02002534
Willy Tarreau76cc6992020-07-01 18:49:24 +02002535tune.idle-pool.shared { on | off }
2536 Enables ('on') or disables ('off') sharing of idle connection pools between
2537 threads for a same server. The default is to share them between threads in
2538 order to minimize the number of persistent connections to a server, and to
2539 optimize the connection reuse rate. But to help with debugging or when
2540 suspecting a bug in HAProxy around connection reuse, it can be convenient to
2541 forcefully disable this idle pool sharing between multiple threads, and force
Willy Tarreau0784db82021-02-19 11:45:22 +01002542 this option to "off". The default is on. It is strongly recommended against
2543 disabling this option without setting a conservative value on "pool-low-conn"
2544 for all servers relying on connection reuse to achieve a high performance
2545 level, otherwise connections might be closed very often as the thread count
2546 increases.
Willy Tarreau76cc6992020-07-01 18:49:24 +02002547
Willy Tarreau7e312732014-02-12 16:35:14 +01002548tune.idletimer <timeout>
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002549 Sets the duration after which HAProxy will consider that an empty buffer is
Willy Tarreau7e312732014-02-12 16:35:14 +01002550 probably associated with an idle stream. This is used to optimally adjust
2551 some packet sizes while forwarding large and small data alternatively. The
2552 decision to use splice() or to send large buffers in SSL is modulated by this
2553 parameter. The value is in milliseconds between 0 and 65535. A value of zero
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002554 means that HAProxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01002555 which seems to correctly detect end user pauses (e.g. read a page before
John Roeslerfb2fce12019-07-10 15:45:51 -05002556 clicking). There should be no reason for changing this value. Please check
Willy Tarreau7e312732014-02-12 16:35:14 +01002557 tune.ssl.maxrecord below.
2558
Willy Tarreau7ac908b2019-02-27 12:02:18 +01002559tune.listener.multi-queue { on | off }
2560 Enables ('on') or disables ('off') the listener's multi-queue accept which
2561 spreads the incoming traffic to all threads a "bind" line is allowed to run
2562 on instead of taking them for itself. This provides a smoother traffic
2563 distribution and scales much better, especially in environments where threads
2564 may be unevenly loaded due to external activity (network interrupts colliding
2565 with one thread for example). This option is enabled by default, but it may
2566 be forcefully disabled for troubleshooting or for situations where it is
2567 estimated that the operating system already provides a good enough
2568 distribution and connections are extremely short-lived.
2569
Thierry FOURNIER90da1912015-03-05 11:17:06 +01002570tune.lua.forced-yield <number>
2571 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01002572 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01002573 HAProxy scheduler to process other tasks like accepting connections or
2574 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01002575 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01002576 lowered. If the Lua code is quite long and its result is absolutely required
2577 to process the data, the <number> can be increased.
2578
Willy Tarreau32f61e22015-03-18 17:54:59 +01002579tune.lua.maxmem
2580 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
2581 default it is zero which means unlimited. It is important to set a limit to
2582 ensure that a bug in a script will not result in the system running out of
2583 memory.
2584
Thierry FOURNIER90da1912015-03-05 11:17:06 +01002585tune.lua.session-timeout <timeout>
2586 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02002587 preventing infinite loops or spending too much time in Lua. This timeout
2588 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002589 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01002590
2591tune.lua.task-timeout <timeout>
2592 Purpose is the same as "tune.lua.session-timeout", but this timeout is
2593 dedicated to the tasks. By default, this timeout isn't set because a task may
2594 remain alive during of the lifetime of HAProxy. For example, a task used to
2595 check servers.
2596
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02002597tune.lua.service-timeout <timeout>
2598 This is the execution timeout for the Lua services. This is useful for
2599 preventing infinite loops or spending too much time in Lua. This timeout
2600 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002601 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02002602
Willy Tarreaua0250ba2008-01-06 11:22:57 +01002603tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01002604 Sets the maximum number of consecutive connections a process may accept in a
2605 row before switching to other work. In single process mode, higher numbers
Willy Tarreau66161322021-02-19 15:50:27 +01002606 used to give better performance at high connection rates, though this is not
2607 the case anymore with the multi-queue. This value applies individually to
2608 each listener, so that the number of processes a listener is bound to is
2609 taken into account. This value defaults to 4 which showed best results. If a
2610 significantly higher value was inherited from an ancient config, it might be
2611 worth removing it as it will both increase performance and lower response
2612 time. In multi-process mode, it is divided by twice the number of processes
2613 the listener is bound to. Setting this value to -1 completely disables the
2614 limitation. It should normally not be needed to tweak this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01002615
2616tune.maxpollevents <number>
2617 Sets the maximum amount of events that can be processed at once in a call to
2618 the polling system. The default value is adapted to the operating system. It
2619 has been noticed that reducing it below 200 tends to slightly decrease
2620 latency at the expense of network bandwidth, and increasing it above 200
2621 tends to trade latency for slightly increased bandwidth.
2622
Willy Tarreau27a674e2009-08-17 07:23:33 +02002623tune.maxrewrite <number>
2624 Sets the reserved buffer space to this size in bytes. The reserved space is
2625 used for header rewriting or appending. The first reads on sockets will never
2626 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
2627 bufsize, though that does not make much sense since there are rarely large
2628 numbers of headers to add. Setting it too high prevents processing of large
2629 requests or responses. Setting it too low prevents addition of new headers
2630 to already large requests or to POST requests. It is generally wise to set it
2631 to about 1024. It is automatically readjusted to half of bufsize if it is
2632 larger than that. This means you don't have to worry about it when changing
2633 bufsize.
2634
Willy Tarreauf3045d22015-04-29 16:24:50 +02002635tune.pattern.cache-size <number>
2636 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
2637 cache which reminds previous lookups and their results. It is used by ACLs
2638 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
2639 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
2640 strings. It applies to pattern expressions which means that it will be able
2641 to memorize the result of a lookup among all the patterns specified on a
2642 configuration line (including all those loaded from files). It automatically
2643 invalidates entries which are updated using HTTP actions or on the CLI. The
2644 default cache size is set to 10000 entries, which limits its footprint to
Willy Tarreau403bfbb2019-10-23 06:59:31 +02002645 about 5 MB per process/thread on 32-bit systems and 8 MB per process/thread
2646 on 64-bit systems, as caches are thread/process local. There is a very low
Willy Tarreauf3045d22015-04-29 16:24:50 +02002647 risk of collision in this cache, which is in the order of the size of the
2648 cache divided by 2^64. Typically, at 10000 requests per second with the
2649 default cache size of 10000 entries, there's 1% chance that a brute force
2650 attack could cause a single collision after 60 years, or 0.1% after 6 years.
2651 This is considered much lower than the risk of a memory corruption caused by
2652 aging components. If this is not acceptable, the cache can be disabled by
2653 setting this parameter to 0.
2654
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02002655tune.pipesize <number>
2656 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
2657 are the default size for the system. But sometimes when using TCP splicing,
2658 it can improve performance to increase pipe sizes, especially if it is
2659 suspected that pipes are not filled and that many calls to splice() are
2660 performed. This has an impact on the kernel's memory footprint, so this must
2661 not be changed if impacts are not understood.
2662
Olivier Houchard88698d92019-04-16 19:07:22 +02002663tune.pool-high-fd-ratio <number>
2664 This setting sets the max number of file descriptors (in percentage) used by
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002665 HAProxy globally against the maximum number of file descriptors HAProxy can
Olivier Houchard88698d92019-04-16 19:07:22 +02002666 use before we start killing idle connections when we can't reuse a connection
2667 and we have to create a new one. The default is 25 (one quarter of the file
2668 descriptor will mean that roughly half of the maximum front connections can
2669 keep an idle connection behind, anything beyond this probably doesn't make
John Roeslerfb2fce12019-07-10 15:45:51 -05002670 much sense in the general case when targeting connection reuse).
Olivier Houchard88698d92019-04-16 19:07:22 +02002671
Willy Tarreau83ca3052020-07-01 18:30:16 +02002672tune.pool-low-fd-ratio <number>
2673 This setting sets the max number of file descriptors (in percentage) used by
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002674 HAProxy globally against the maximum number of file descriptors HAProxy can
Willy Tarreau83ca3052020-07-01 18:30:16 +02002675 use before we stop putting connection into the idle pool for reuse. The
2676 default is 20.
2677
Willy Tarreaue803de22010-01-21 17:43:04 +01002678tune.rcvbuf.client <number>
2679tune.rcvbuf.server <number>
2680 Forces the kernel socket receive buffer size on the client or the server side
2681 to the specified value in bytes. This value applies to all TCP/HTTP frontends
2682 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05002683 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01002684 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01002685 order to save kernel memory by preventing it from buffering too large amounts
2686 of received data. Lower values will significantly increase CPU usage though.
2687
Willy Tarreaub22fc302015-12-14 12:04:35 +01002688tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01002689 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01002690 socket buffers. One of them is that a read returns more than <recv_enough>
2691 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
2692 may be changed by this setting to better deal with workloads involving lots
2693 of short messages such as telnet or SSH sessions.
2694
Olivier Houchard1599b802018-05-24 18:59:04 +02002695tune.runqueue-depth <number>
John Roeslerfb2fce12019-07-10 15:45:51 -05002696 Sets the maximum amount of task that can be processed at once when running
Willy Tarreau060a7612021-03-10 11:06:26 +01002697 tasks. The default value depends on the number of threads but sits between 35
2698 and 280, which tend to show the highest request rates and lowest latencies.
2699 Increasing it may incur latency when dealing with I/Os, making it too small
2700 can incur extra overhead. Higher thread counts benefit from lower values.
2701 When experimenting with much larger values, it may be useful to also enable
2702 tune.sched.low-latency and possibly tune.fd.edge-triggered to limit the
2703 maximum latency to the lowest possible.
Willy Tarreaue7723bd2020-06-24 11:11:02 +02002704
2705tune.sched.low-latency { on | off }
2706 Enables ('on') or disables ('off') the low-latency task scheduler. By default
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002707 HAProxy processes tasks from several classes one class at a time as this is
Willy Tarreaue7723bd2020-06-24 11:11:02 +02002708 the most efficient. But when running with large values of tune.runqueue-depth
2709 this can have a measurable effect on request or connection latency. When this
2710 low-latency setting is enabled, tasks of lower priority classes will always
2711 be executed before other ones if they exist. This will permit to lower the
2712 maximum latency experienced by new requests or connections in the middle of
2713 massive traffic, at the expense of a higher impact on this large traffic.
2714 For regular usage it is better to leave this off. The default value is off.
Olivier Houchard1599b802018-05-24 18:59:04 +02002715
Willy Tarreaue803de22010-01-21 17:43:04 +01002716tune.sndbuf.client <number>
2717tune.sndbuf.server <number>
2718 Forces the kernel socket send buffer size on the client or the server side to
2719 the specified value in bytes. This value applies to all TCP/HTTP frontends
2720 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05002721 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01002722 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01002723 order to save kernel memory by preventing it from buffering too large amounts
2724 of received data. Lower values will significantly increase CPU usage though.
2725 Another use case is to prevent write timeouts with extremely slow clients due
2726 to the kernel waiting for a large part of the buffer to be read before
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002727 notifying HAProxy again.
Willy Tarreaue803de22010-01-21 17:43:04 +01002728
Willy Tarreau6ec58db2012-11-16 16:32:15 +01002729tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01002730 Sets the size of the global SSL session cache, in a number of blocks. A block
William Dauchy9a4bbfe2021-02-12 15:58:46 +01002731 is large enough to contain an encoded session without peer certificate. An
2732 encoded session with peer certificate is stored in multiple blocks depending
2733 on the size of the peer certificate. A block uses approximately 200 bytes of
2734 memory (based on `sizeof(struct sh_ssl_sess_hdr) + SHSESS_BLOCK_MIN_SIZE`
2735 calculation used for `shctx_init` function). The default value may be forced
2736 at build time, otherwise defaults to 20000. When the cache is full, the most
2737 idle entries are purged and reassigned. Higher values reduce the occurrence
2738 of such a purge, hence the number of CPU-intensive SSL handshakes by ensuring
2739 that all users keep their session as long as possible. All entries are
2740 pre-allocated upon startup and are shared between all processes if "nbproc"
2741 is greater than 1. Setting this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01002742
Emeric Brun8dc60392014-05-09 13:52:00 +02002743tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02002744 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02002745 should normally not be used since it will force many renegotiations due to
2746 clients hitting a random process. But it may be required on some operating
2747 systems where none of the SSL cache synchronization method may be used. In
2748 this case, adding a first layer of hash-based load balancing before the SSL
2749 layer might limit the impact of the lack of session sharing.
2750
William Lallemand7d42ef52020-07-06 11:41:30 +02002751tune.ssl.keylog { on | off }
2752 This option activates the logging of the TLS keys. It should be used with
2753 care as it will consume more memory per SSL session and could decrease
2754 performances. This is disabled by default.
2755
2756 These sample fetches should be used to generate the SSLKEYLOGFILE that is
2757 required to decipher traffic with wireshark.
2758
2759 https://developer.mozilla.org/en-US/docs/Mozilla/Projects/NSS/Key_Log_Format
2760
2761 The SSLKEYLOG is a series of lines which are formatted this way:
2762
2763 <Label> <space> <ClientRandom> <space> <Secret>
2764
2765 The ClientRandom is provided by the %[ssl_fc_client_random,hex] sample
2766 fetch, the secret and the Label could be find in the array below. You need
2767 to generate a SSLKEYLOGFILE with all the labels in this array.
2768
2769 The following sample fetches are hexadecimal strings and does not need to be
2770 converted.
2771
2772 SSLKEYLOGFILE Label | Sample fetches for the Secrets
2773 --------------------------------|-----------------------------------------
2774 CLIENT_EARLY_TRAFFIC_SECRET | %[ssl_fc_client_early_traffic_secret]
2775 CLIENT_HANDSHAKE_TRAFFIC_SECRET | %[ssl_fc_client_handshake_traffic_secret]
2776 SERVER_HANDSHAKE_TRAFFIC_SECRET | %[ssl_fc_server_handshake_traffic_secret]
2777 CLIENT_TRAFFIC_SECRET_0 | %[ssl_fc_client_traffic_secret_0]
2778 SERVER_TRAFFIC_SECRET_0 | %[ssl_fc_server_traffic_secret_0]
William Lallemandd742b6c2020-07-07 10:14:56 +02002779 EXPORTER_SECRET | %[ssl_fc_exporter_secret]
2780 EARLY_EXPORTER_SECRET | %[ssl_fc_early_exporter_secret]
William Lallemand7d42ef52020-07-06 11:41:30 +02002781
2782 This is only available with OpenSSL 1.1.1, and useful with TLS1.3 session.
2783
2784 If you want to generate the content of a SSLKEYLOGFILE with TLS < 1.3, you
2785 only need this line:
2786
2787 "CLIENT_RANDOM %[ssl_fc_client_random,hex] %[ssl_fc_session_key,hex]"
2788
Emeric Brun4f65bff2012-11-16 15:11:00 +01002789tune.ssl.lifetime <timeout>
2790 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002791 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01002792 does not guarantee that sessions will last that long, because if the cache is
2793 full, the longest idle sessions will be purged despite their configured
2794 lifetime. The real usefulness of this setting is to prevent sessions from
2795 being used for too long.
2796
Willy Tarreaubfd59462013-02-21 07:46:09 +01002797tune.ssl.maxrecord <number>
2798 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
2799 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
2800 data only once it has received a full record. With large records, it means
2801 that clients might have to download up to 16kB of data before starting to
2802 process them. Limiting the value can improve page load times on browsers
2803 located over high latency or low bandwidth networks. It is suggested to find
2804 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
2805 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
2806 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
2807 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01002808 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01002809 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01002810
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002811tune.ssl.default-dh-param <number>
2812 Sets the maximum size of the Diffie-Hellman parameters used for generating
2813 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
2814 final size will try to match the size of the server's RSA (or DSA) key (e.g,
2815 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
Willy Tarreau3ba77d22020-05-08 09:31:18 +02002816 this maximum value. Default value if 2048. Only 1024 or higher values are
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002817 allowed. Higher values will increase the CPU load, and values greater than
2818 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02002819 used if static Diffie-Hellman parameters are supplied either directly
2820 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002821
Christopher Faulet31af49d2015-06-09 17:29:50 +02002822tune.ssl.ssl-ctx-cache-size <number>
2823 Sets the size of the cache used to store generated certificates to <number>
2824 entries. This is a LRU cache. Because generating a SSL certificate
2825 dynamically is expensive, they are cached. The default cache size is set to
2826 1000 entries.
2827
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01002828tune.ssl.capture-cipherlist-size <number>
2829 Sets the maximum size of the buffer used for capturing client-hello cipher
2830 list. If the value is 0 (default value) the capture is disabled, otherwise
2831 a buffer is allocated for each SSL/TLS connection.
2832
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002833tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002834tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002835tune.vars.reqres-max-size <size>
2836tune.vars.sess-max-size <size>
2837tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002838 These five tunes help to manage the maximum amount of memory used by the
2839 variables system. "global" limits the overall amount of memory available for
2840 all scopes. "proc" limits the memory for the process scope, "sess" limits the
2841 memory for the session scope, "txn" for the transaction scope, and "reqres"
2842 limits the memory for each request or response processing.
2843 Memory accounting is hierarchical, meaning more coarse grained limits include
2844 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
2845 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002846
Daniel Schneller0b547052016-03-21 20:46:57 +01002847 For example, when "tune.vars.sess-max-size" is limited to 100,
2848 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
2849 100 either. If we create a variable "txn.var" that contains 100 bytes,
2850 all available space is consumed.
2851 Notice that exceeding the limits at runtime will not result in an error
2852 message, but values might be cut off or corrupted. So make sure to accurately
2853 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002854
William Lallemanda509e4c2012-11-07 16:54:34 +01002855tune.zlib.memlevel <number>
2856 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002857 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01002858 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01002859 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01002860 between 1 and 9. The default value is 8.
2861
2862tune.zlib.windowsize <number>
2863 Sets the window size (the size of the history buffer) as a parameter of the
2864 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01002865 in better compression at the expense of memory usage. Can be a value between
2866 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002867
Willy Tarreauc57f0e22009-05-10 13:12:33 +020028683.3. Debugging
2869--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02002870
Willy Tarreau6a06a402007-07-15 20:15:28 +02002871quiet
2872 Do not display any message during startup. It is equivalent to the command-
2873 line argument "-q".
2874
Willy Tarreau3eb10b82020-04-15 16:42:39 +02002875zero-warning
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002876 When this option is set, HAProxy will refuse to start if any warning was
Willy Tarreau3eb10b82020-04-15 16:42:39 +02002877 emitted while processing the configuration. It is highly recommended to set
2878 this option on configurations that are not changed often, as it helps detect
2879 subtle mistakes and keep the configuration clean and forward-compatible. Note
2880 that "haproxy -c" will also report errors in such a case. This option is
2881 equivalent to command line argument "-dW".
2882
Emeric Brunf099e792010-09-27 12:05:28 +02002883
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010028843.4. Userlists
2885--------------
2886It is possible to control access to frontend/backend/listen sections or to
2887http stats by allowing only authenticated and authorized users. To do this,
2888it is required to create at least one userlist and to define users.
2889
2890userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01002891 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002892 used to store authentication & authorization data for independent customers.
2893
2894group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01002895 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002896 attach users to this group by using a comma separated list of names
2897 proceeded by "users" keyword.
2898
Cyril Bontéf0c60612010-02-06 14:44:47 +01002899user <username> [password|insecure-password <password>]
2900 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002901 Adds user <username> to the current userlist. Both secure (encrypted) and
2902 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002903 evaluated using the crypt(3) function, so depending on the system's
2904 capabilities, different algorithms are supported. For example, modern Glibc
2905 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
2906 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002907
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002908 Attention: Be aware that using encrypted passwords might cause significantly
2909 increased CPU usage, depending on the number of requests, and the algorithm
2910 used. For any of the hashed variants, the password for each request must
2911 be processed through the chosen algorithm, before it can be compared to the
2912 value specified in the config file. Most current algorithms are deliberately
2913 designed to be expensive to compute to achieve resistance against brute
2914 force attacks. They do not simply salt/hash the clear text password once,
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002915 but thousands of times. This can quickly become a major factor in HAProxy's
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002916 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002917
2918 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01002919 userlist L1
2920 group G1 users tiger,scott
2921 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002922
Cyril Bontéf0c60612010-02-06 14:44:47 +01002923 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
2924 user scott insecure-password elgato
2925 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002926
Cyril Bontéf0c60612010-02-06 14:44:47 +01002927 userlist L2
2928 group G1
2929 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002930
Cyril Bontéf0c60612010-02-06 14:44:47 +01002931 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
2932 user scott insecure-password elgato groups G1,G2
2933 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002934
2935 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002936
Emeric Brunf099e792010-09-27 12:05:28 +02002937
29383.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002939----------
Emeric Brun94900952015-06-11 18:25:54 +02002940It is possible to propagate entries of any data-types in stick-tables between
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002941several HAProxy instances over TCP connections in a multi-master fashion. Each
Emeric Brun94900952015-06-11 18:25:54 +02002942instance pushes its local updates and insertions to remote peers. The pushed
2943values overwrite remote ones without aggregation. Interrupted exchanges are
2944automatically detected and recovered from the last known point.
2945In addition, during a soft restart, the old process connects to the new one
2946using such a TCP connection to push all its entries before the new process
2947tries to connect to other peers. That ensures very fast replication during a
2948reload, it typically takes a fraction of a second even for large tables.
2949Note that Server IDs are used to identify servers remotely, so it is important
2950that configurations look similar or at least that the same IDs are forced on
2951each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002952
2953peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002954 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002955 which is referenced by one or more stick-tables.
2956
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002957bind [<address>]:<port_range> [, ...] [param*]
2958 Defines the binding parameters of the local peer of this "peers" section.
2959 Such lines are not supported with "peer" line in the same "peers" section.
2960
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002961disabled
2962 Disables a peers section. It disables both listening and any synchronization
2963 related to this section. This is provided to disable synchronization of stick
2964 tables without having to comment out all "peers" references.
2965
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002966default-bind [param*]
2967 Defines the binding parameters for the local peer, excepted its address.
2968
2969default-server [param*]
2970 Change default options for a server in a "peers" section.
2971
2972 Arguments:
2973 <param*> is a list of parameters for this server. The "default-server"
2974 keyword accepts an important number of options and has a complete
2975 section dedicated to it. Please refer to section 5 for more
2976 details.
2977
2978
2979 See also: "server" and section 5 about server options
2980
Emeric Brun7214dcf2021-09-29 10:29:52 +02002981enabled
2982 This re-enables a peers section which was previously disabled via the
2983 "disabled" keyword.
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002984
Jan Wagner3e678602020-12-17 22:22:32 +01002985log <address> [len <length>] [format <format>] [sample <ranges>:<sample_size>]
Frédéric Lécailleb6f759b2019-11-05 09:57:45 +01002986 <facility> [<level> [<minlevel>]]
2987 "peers" sections support the same "log" keyword as for the proxies to
2988 log information about the "peers" listener. See "log" option for proxies for
2989 more details.
2990
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002991peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002992 Defines a peer inside a peers section.
2993 If <peername> is set to the local peer name (by default hostname, or forced
Dragan Dosen13cd54c2020-06-18 18:24:05 +02002994 using "-L" command line option or "localpeer" global configuration setting),
Daniel Corbett9f0843f2021-05-08 10:50:37 -04002995 HAProxy will listen for incoming remote peer connection on <ip>:<port>.
Dragan Dosen13cd54c2020-06-18 18:24:05 +02002996 Otherwise, <ip>:<port> defines where to connect to in order to join the
2997 remote peer, and <peername> is used at the protocol level to identify and
2998 validate the remote peer on the server side.
Emeric Brunf099e792010-09-27 12:05:28 +02002999
3000 During a soft restart, local peer <ip>:<port> is used by the old instance to
3001 connect the new one and initiate a complete replication (teaching process).
3002
3003 It is strongly recommended to have the exact same peers declaration on all
Dragan Dosen13cd54c2020-06-18 18:24:05 +02003004 peers and to only rely on the "-L" command line argument or the "localpeer"
3005 global configuration setting to change the local peer name. This makes it
3006 easier to maintain coherent configuration files across all peers.
Emeric Brunf099e792010-09-27 12:05:28 +02003007
William Lallemandb2f07452015-05-12 14:27:13 +02003008 You may want to reference some environment variables in the address
3009 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01003010
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01003011 Note: "peer" keyword may transparently be replaced by "server" keyword (see
3012 "server" keyword explanation below).
3013
3014server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02003015 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01003016 with a support for all "server" parameters found in 5.2 paragraph.
3017 If the underlying peer is local, <ip>:<port> parameters must not be present.
3018 These parameters must be provided on a "bind" line (see "bind" keyword
3019 of this "peers" section).
3020 Some of these parameters are irrelevant for "peers" sections.
3021
3022
Cyril Bontédc4d9032012-04-08 21:57:39 +02003023 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01003024 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02003025 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01003026 peer haproxy1 192.168.0.1:1024
3027 peer haproxy2 192.168.0.2:1024
3028 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02003029
3030 backend mybackend
3031 mode tcp
3032 balance roundrobin
3033 stick-table type ip size 20k peers mypeers
3034 stick on src
3035
Willy Tarreauf7b30a92010-12-06 22:59:17 +01003036 server srv1 192.168.0.30:80
3037 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02003038
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01003039 Example:
3040 peers mypeers
3041 bind 127.0.0.11:10001 ssl crt mycerts/pem
3042 default-server ssl verify none
3043 server hostA 127.0.0.10:10000
3044 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02003045
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01003046
3047table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
3048 size <size> [expire <expire>] [nopurge] [store <data_type>]*
3049
3050 Configure a stickiness table for the current section. This line is parsed
3051 exactly the same way as the "stick-table" keyword in others section, except
John Roeslerfb2fce12019-07-10 15:45:51 -05003052 for the "peers" argument which is not required here and with an additional
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01003053 mandatory first parameter to designate the stick-table. Contrary to others
3054 sections, there may be several "table" lines in "peers" sections (see also
3055 "stick-table" keyword).
3056
3057 Also be aware of the fact that "peers" sections have their own stick-table
3058 namespaces to avoid collisions between stick-table names identical in
3059 different "peers" section. This is internally handled prepending the "peers"
3060 sections names to the name of the stick-tables followed by a '/' character.
3061 If somewhere else in the configuration file you have to refer to such
3062 stick-tables declared in "peers" sections you must use the prefixed version
3063 of the stick-table name as follows:
3064
3065 peers mypeers
3066 peer A ...
3067 peer B ...
3068 table t1 ...
3069
3070 frontend fe1
3071 tcp-request content track-sc0 src table mypeers/t1
3072
3073 This is also this prefixed version of the stick-table names which must be
3074 used to refer to stick-tables through the CLI.
3075
3076 About "peers" protocol, as only "peers" belonging to the same section may
3077 communicate with each others, there is no need to do such a distinction.
3078 Several "peers" sections may declare stick-tables with the same name.
3079 This is shorter version of the stick-table name which is sent over the network.
3080 There is only a '/' character as prefix to avoid stick-table name collisions between
3081 stick-tables declared as backends and stick-table declared in "peers" sections
3082 as follows in this weird but supported configuration:
3083
3084 peers mypeers
3085 peer A ...
3086 peer B ...
3087 table t1 type string size 10m store gpc0
3088
3089 backend t1
3090 stick-table type string size 10m store gpc0 peers mypeers
3091
Daniel Corbett67a82712020-07-06 23:01:19 -04003092 Here "t1" table declared in "mypeers" section has "mypeers/t1" as global name.
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01003093 "t1" table declared as a backend as "t1" as global name. But at peer protocol
3094 level the former table is named "/t1", the latter is again named "t1".
3095
Simon Horman51a1cf62015-02-03 13:00:44 +090030963.6. Mailers
3097------------
3098It is possible to send email alerts when the state of servers changes.
3099If configured email alerts are sent to each mailer that is configured
3100in a mailers section. Email is sent to mailers using SMTP.
3101
Pieter Baauw386a1272015-08-16 15:26:24 +02003102mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09003103 Creates a new mailer list with the name <mailersect>. It is an
3104 independent section which is referenced by one or more proxies.
3105
3106mailer <mailername> <ip>:<port>
3107 Defines a mailer inside a mailers section.
3108
3109 Example:
3110 mailers mymailers
3111 mailer smtp1 192.168.0.1:587
3112 mailer smtp2 192.168.0.2:587
3113
3114 backend mybackend
3115 mode tcp
3116 balance roundrobin
3117
3118 email-alert mailers mymailers
3119 email-alert from test1@horms.org
3120 email-alert to test2@horms.org
3121
3122 server srv1 192.168.0.30:80
3123 server srv2 192.168.0.31:80
3124
Pieter Baauw235fcfc2016-02-13 15:33:40 +01003125timeout mail <time>
3126 Defines the time available for a mail/connection to be made and send to
3127 the mail-server. If not defined the default value is 10 seconds. To allow
3128 for at least two SYN-ACK packets to be send during initial TCP handshake it
3129 is advised to keep this value above 4 seconds.
3130
3131 Example:
3132 mailers mymailers
3133 timeout mail 20s
3134 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09003135
William Lallemandc9515522019-06-12 16:32:11 +020031363.7. Programs
3137-------------
3138In master-worker mode, it is possible to launch external binaries with the
3139master, these processes are called programs. These programs are launched and
3140managed the same way as the workers.
3141
3142During a reload of HAProxy, those processes are dealing with the same
3143sequence as a worker:
3144
3145 - the master is re-executed
3146 - the master sends a SIGUSR1 signal to the program
3147 - if "option start-on-reload" is not disabled, the master launches a new
3148 instance of the program
3149
3150During a stop, or restart, a SIGTERM is sent to the programs.
3151
3152program <name>
3153 This is a new program section, this section will create an instance <name>
3154 which is visible in "show proc" on the master CLI. (See "9.4. Master CLI" in
3155 the management guide).
3156
3157command <command> [arguments*]
3158 Define the command to start with optional arguments. The command is looked
3159 up in the current PATH if it does not include an absolute path. This is a
3160 mandatory option of the program section. Arguments containing spaces must
3161 be enclosed in quotes or double quotes or be prefixed by a backslash.
3162
Andrew Heberle97236962019-07-12 11:50:26 +08003163user <user name>
3164 Changes the executed command user ID to the <user name> from /etc/passwd.
3165 See also "group".
3166
3167group <group name>
3168 Changes the executed command group ID to the <group name> from /etc/group.
3169 See also "user".
3170
William Lallemandc9515522019-06-12 16:32:11 +02003171option start-on-reload
3172no option start-on-reload
3173 Start (or not) a new instance of the program upon a reload of the master.
3174 The default is to start a new instance. This option may only be used in a
3175 program section.
3176
3177
Christopher Faulet76edc0f2020-01-13 15:52:01 +010031783.8. HTTP-errors
3179----------------
3180
3181It is possible to globally declare several groups of HTTP errors, to be
3182imported afterwards in any proxy section. Same group may be referenced at
3183several places and can be fully or partially imported.
3184
3185http-errors <name>
3186 Create a new http-errors group with the name <name>. It is an independent
3187 section that may be referenced by one or more proxies using its name.
3188
3189errorfile <code> <file>
3190 Associate a file contents to an HTTP error code
3191
3192 Arguments :
3193 <code> is the HTTP status code. Currently, HAProxy is capable of
Christopher Faulet612f2ea2020-05-27 09:57:28 +02003194 generating codes 200, 400, 401, 403, 404, 405, 407, 408, 410,
Christopher Faulete095f312020-12-07 11:22:24 +01003195 425, 429, 500, 501, 502, 503, and 504.
Christopher Faulet76edc0f2020-01-13 15:52:01 +01003196
3197 <file> designates a file containing the full HTTP response. It is
3198 recommended to follow the common practice of appending ".http" to
3199 the filename so that people do not confuse the response with HTML
3200 error pages, and to use absolute paths, since files are read
3201 before any chroot is performed.
3202
3203 Please referrers to "errorfile" keyword in section 4 for details.
3204
3205 Example:
3206 http-errors website-1
3207 errorfile 400 /etc/haproxy/errorfiles/site1/400.http
3208 errorfile 404 /etc/haproxy/errorfiles/site1/404.http
3209 errorfile 408 /dev/null # work around Chrome pre-connect bug
3210
3211 http-errors website-2
3212 errorfile 400 /etc/haproxy/errorfiles/site2/400.http
3213 errorfile 404 /etc/haproxy/errorfiles/site2/404.http
3214 errorfile 408 /dev/null # work around Chrome pre-connect bug
3215
Emeric Brun99c453d2020-05-25 15:01:04 +020032163.9. Rings
3217----------
3218
3219It is possible to globally declare ring-buffers, to be used as target for log
3220servers or traces.
3221
3222ring <ringname>
3223 Creates a new ring-buffer with name <ringname>.
3224
3225description <text>
Daniel Corbett67a82712020-07-06 23:01:19 -04003226 The description is an optional description string of the ring. It will
Emeric Brun99c453d2020-05-25 15:01:04 +02003227 appear on CLI. By default, <name> is reused to fill this field.
3228
3229format <format>
3230 Format used to store events into the ring buffer.
3231
3232 Arguments:
3233 <format> is the log format used when generating syslog messages. It may be
3234 one of the following :
3235
3236 iso A message containing only the ISO date, followed by the text.
3237 The PID, process name and system name are omitted. This is
3238 designed to be used with a local log server.
3239
Emeric Brun0237c4e2020-11-27 16:24:34 +01003240 local Analog to rfc3164 syslog message format except that hostname
3241 field is stripped. This is the default.
3242 Note: option "log-send-hostname" switches the default to
3243 rfc3164.
3244
Emeric Brun99c453d2020-05-25 15:01:04 +02003245 raw A message containing only the text. The level, PID, date, time,
3246 process name and system name are omitted. This is designed to be
3247 used in containers or during development, where the severity
3248 only depends on the file descriptor used (stdout/stderr). This
3249 is the default.
3250
Emeric Brun0237c4e2020-11-27 16:24:34 +01003251 rfc3164 The RFC3164 syslog message format.
Emeric Brun99c453d2020-05-25 15:01:04 +02003252 (https://tools.ietf.org/html/rfc3164)
3253
3254 rfc5424 The RFC5424 syslog message format.
3255 (https://tools.ietf.org/html/rfc5424)
3256
3257 short A message containing only a level between angle brackets such as
3258 '<3>', followed by the text. The PID, date, time, process name
3259 and system name are omitted. This is designed to be used with a
3260 local log server. This format is compatible with what the systemd
3261 logger consumes.
3262
Emeric Brun54648852020-07-06 15:54:06 +02003263 priority A message containing only a level plus syslog facility between angle
3264 brackets such as '<63>', followed by the text. The PID, date, time,
3265 process name and system name are omitted. This is designed to be used
3266 with a local log server.
3267
Emeric Brun99c453d2020-05-25 15:01:04 +02003268 timed A message containing only a level between angle brackets such as
3269 '<3>', followed by ISO date and by the text. The PID, process
3270 name and system name are omitted. This is designed to be
3271 used with a local log server.
3272
3273maxlen <length>
3274 The maximum length of an event message stored into the ring,
3275 including formatted header. If an event message is longer than
3276 <length>, it will be truncated to this length.
3277
Emeric Brun494c5052020-05-28 11:13:15 +02003278server <name> <address> [param*]
3279 Used to configure a syslog tcp server to forward messages from ring buffer.
3280 This supports for all "server" parameters found in 5.2 paragraph. Some of
3281 these parameters are irrelevant for "ring" sections. Important point: there
3282 is little reason to add more than one server to a ring, because all servers
3283 will receive the exact same copy of the ring contents, and as such the ring
3284 will progress at the speed of the slowest server. If one server does not
3285 respond, it will prevent old messages from being purged and may block new
3286 messages from being inserted into the ring. The proper way to send messages
3287 to multiple servers is to use one distinct ring per log server, not to
Emeric Brun97556472020-05-30 01:42:45 +02003288 attach multiple servers to the same ring. Note that specific server directive
3289 "log-proto" is used to set the protocol used to send messages.
Emeric Brun494c5052020-05-28 11:13:15 +02003290
Emeric Brun99c453d2020-05-25 15:01:04 +02003291size <size>
3292 This is the optional size in bytes for the ring-buffer. Default value is
3293 set to BUFSIZE.
3294
Emeric Brun494c5052020-05-28 11:13:15 +02003295timeout connect <timeout>
3296 Set the maximum time to wait for a connection attempt to a server to succeed.
3297
3298 Arguments :
3299 <timeout> is the timeout value specified in milliseconds by default, but
3300 can be in any other unit if the number is suffixed by the unit,
3301 as explained at the top of this document.
3302
3303timeout server <timeout>
3304 Set the maximum time for pending data staying into output buffer.
3305
3306 Arguments :
3307 <timeout> is the timeout value specified in milliseconds by default, but
3308 can be in any other unit if the number is suffixed by the unit,
3309 as explained at the top of this document.
3310
Emeric Brun99c453d2020-05-25 15:01:04 +02003311 Example:
3312 global
3313 log ring@myring local7
3314
3315 ring myring
3316 description "My local buffer"
3317 format rfc3164
3318 maxlen 1200
3319 size 32764
Emeric Brun494c5052020-05-28 11:13:15 +02003320 timeout connect 5s
3321 timeout server 10s
Emeric Brun97556472020-05-30 01:42:45 +02003322 server mysyslogsrv 127.0.0.1:6514 log-proto octet-count
Emeric Brun99c453d2020-05-25 15:01:04 +02003323
Emeric Brun12941c82020-07-07 14:19:42 +020033243.10. Log forwarding
3325-------------------
3326
3327It is possible to declare one or multiple log forwarding section,
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003328HAProxy will forward all received log messages to a log servers list.
Emeric Brun12941c82020-07-07 14:19:42 +02003329
3330log-forward <name>
3331 Creates a new log forwarder proxy identified as <name>.
3332
Emeric Bruncbb7bf72020-10-05 14:39:35 +02003333backlog <conns>
3334 Give hints to the system about the approximate listen backlog desired size
3335 on connections accept.
3336
3337bind <addr> [param*]
3338 Used to configure a stream log listener to receive messages to forward.
Emeric Brunda46c1c2020-10-08 08:39:02 +02003339 This supports the "bind" parameters found in 5.1 paragraph including
3340 those about ssl but some statements such as "alpn" may be irrelevant for
3341 syslog protocol over TCP.
3342 Those listeners support both "Octet Counting" and "Non-Transparent-Framing"
Emeric Bruncbb7bf72020-10-05 14:39:35 +02003343 modes as defined in rfc-6587.
3344
Willy Tarreau76aaa7f2020-09-16 15:07:22 +02003345dgram-bind <addr> [param*]
Emeric Bruncbb7bf72020-10-05 14:39:35 +02003346 Used to configure a datagram log listener to receive messages to forward.
3347 Addresses must be in IPv4 or IPv6 form,followed by a port. This supports
3348 for some of the "bind" parameters found in 5.1 paragraph among which
3349 "interface", "namespace" or "transparent", the other ones being
Willy Tarreau26ff5da2020-09-16 15:22:19 +02003350 silently ignored as irrelevant for UDP/syslog case.
Emeric Brun12941c82020-07-07 14:19:42 +02003351
3352log global
Jan Wagner3e678602020-12-17 22:22:32 +01003353log <address> [len <length>] [format <format>] [sample <ranges>:<sample_size>]
Emeric Brun12941c82020-07-07 14:19:42 +02003354 <facility> [<level> [<minlevel>]]
3355 Used to configure target log servers. See more details on proxies
3356 documentation.
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003357 If no format specified, HAProxy tries to keep the incoming log format.
Emeric Brun12941c82020-07-07 14:19:42 +02003358 Configured facility is ignored, except if incoming message does not
3359 present a facility but one is mandatory on the outgoing format.
3360 If there is no timestamp available in the input format, but the field
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003361 exists in output format, HAProxy will use the local date.
Emeric Brun12941c82020-07-07 14:19:42 +02003362
3363 Example:
3364 global
3365 log stderr format iso local7
3366
3367 ring myring
3368 description "My local buffer"
3369 format rfc5424
3370 maxlen 1200
3371 size 32764
3372 timeout connect 5s
3373 timeout server 10s
3374 # syslog tcp server
3375 server mysyslogsrv 127.0.0.1:514 log-proto octet-count
3376
3377 log-forward sylog-loadb
Emeric Bruncbb7bf72020-10-05 14:39:35 +02003378 dgram-bind 127.0.0.1:1514
3379 bind 127.0.0.1:1514
Emeric Brun12941c82020-07-07 14:19:42 +02003380 # all messages on stderr
3381 log global
3382 # all messages on local tcp syslog server
3383 log ring@myring local0
3384 # load balance messages on 4 udp syslog servers
3385 log 127.0.0.1:10001 sample 1:4 local0
3386 log 127.0.0.1:10002 sample 2:4 local0
3387 log 127.0.0.1:10003 sample 3:4 local0
3388 log 127.0.0.1:10004 sample 4:4 local0
Christopher Faulet76edc0f2020-01-13 15:52:01 +01003389
Emeric Bruncbb7bf72020-10-05 14:39:35 +02003390maxconn <conns>
3391 Fix the maximum number of concurrent connections on a log forwarder.
3392 10 is the default.
3393
3394timeout client <timeout>
3395 Set the maximum inactivity time on the client side.
3396
Willy Tarreauc57f0e22009-05-10 13:12:33 +020033974. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02003398----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01003399
Willy Tarreau6a06a402007-07-15 20:15:28 +02003400Proxy configuration can be located in a set of sections :
Willy Tarreau7c0b4d82021-02-12 14:58:08 +01003401 - defaults [<name>] [ from <defaults_name> ]
3402 - frontend <name> [ from <defaults_name> ]
3403 - backend <name> [ from <defaults_name> ]
3404 - listen <name> [ from <defaults_name> ]
Willy Tarreau6a06a402007-07-15 20:15:28 +02003405
3406A "frontend" section describes a set of listening sockets accepting client
3407connections.
3408
3409A "backend" section describes a set of servers to which the proxy will connect
3410to forward incoming connections.
3411
3412A "listen" section defines a complete proxy with its frontend and backend
3413parts combined in one section. It is generally useful for TCP-only traffic.
3414
Willy Tarreau7c0b4d82021-02-12 14:58:08 +01003415A "defaults" section resets all settings to the documented ones and presets new
3416ones for use by subsequent sections. All of "frontend", "backend" and "listen"
3417sections always take their initial settings from a defaults section, by default
3418the latest one that appears before the newly created section. It is possible to
3419explicitly designate a specific "defaults" section to load the initial settings
3420from by indicating its name on the section line after the optional keyword
3421"from". While "defaults" section do not impose a name, this use is encouraged
3422for better readability. It is also the only way to designate a specific section
3423to use instead of the default previous one. Since "defaults" section names are
3424optional, by default a very permissive check is applied on their name and these
3425are even permitted to overlap. However if a "defaults" section is referenced by
3426any other section, its name must comply with the syntax imposed on all proxy
3427names, and this name must be unique among the defaults sections. Please note
3428that regardless of what is currently permitted, it is recommended to avoid
3429duplicate section names in general and to respect the same syntax as for proxy
3430names. This rule might be enforced in a future version.
3431
3432Note that it is even possible for a defaults section to take its initial
3433settings from another one, and as such, inherit settings across multiple levels
3434of defaults sections. This can be convenient to establish certain configuration
3435profiles to carry groups of default settings (e.g. TCP vs HTTP or short vs long
3436timeouts) but can quickly become confusing to follow.
3437
Willy Tarreau0ba27502007-12-24 16:55:16 +01003438All proxy names must be formed from upper and lower case letters, digits,
3439'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
3440case-sensitive, which means that "www" and "WWW" are two different proxies.
3441
3442Historically, all proxy names could overlap, it just caused troubles in the
3443logs. Since the introduction of content switching, it is mandatory that two
3444proxies with overlapping capabilities (frontend/backend) have different names.
3445However, it is still permitted that a frontend and a backend share the same
3446name, as this configuration seems to be commonly encountered.
3447
3448Right now, two major proxy modes are supported : "tcp", also known as layer 4,
3449and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003450bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003451protocol, and can interact with it by allowing, blocking, switching, adding,
3452modifying, or removing arbitrary contents in requests or responses, based on
3453arbitrary criteria.
3454
Willy Tarreau70dffda2014-01-30 03:07:23 +01003455In HTTP mode, the processing applied to requests and responses flowing over
3456a connection depends in the combination of the frontend's HTTP options and
Julien Pivotto21ad3152019-12-10 13:11:17 +01003457the backend's. HAProxy supports 3 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01003458
3459 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
3460 requests and responses are processed, and connections remain open but idle
3461 between responses and new requests.
3462
Willy Tarreau70dffda2014-01-30 03:07:23 +01003463 - SCL: server close ("option http-server-close") : the server-facing
3464 connection is closed after the end of the response is received, but the
3465 client-facing connection remains open.
3466
Christopher Faulet315b39c2018-09-21 16:26:19 +02003467 - CLO: close ("option httpclose"): the connection is closed after the end of
3468 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01003469
3470The effective mode that will be applied to a connection passing through a
3471frontend and a backend can be determined by both proxy modes according to the
3472following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02003473weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01003474
Christopher Faulet315b39c2018-09-21 16:26:19 +02003475 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01003476
Christopher Faulet315b39c2018-09-21 16:26:19 +02003477 | KAL | SCL | CLO
3478 ----+-----+-----+----
3479 KAL | KAL | SCL | CLO
3480 ----+-----+-----+----
Christopher Faulet315b39c2018-09-21 16:26:19 +02003481 mode SCL | SCL | SCL | CLO
3482 ----+-----+-----+----
3483 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01003484
Christopher Faulet4d37e532021-03-26 14:44:00 +01003485It is possible to chain a TCP frontend to an HTTP backend. It is pointless if
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003486only HTTP traffic is handled. But it may be used to handle several protocols
3487within the same frontend. In this case, the client's connection is first handled
Christopher Faulet4d37e532021-03-26 14:44:00 +01003488as a raw tcp connection before being upgraded to HTTP. Before the upgrade, the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003489content processings are performend on raw data. Once upgraded, data is parsed
Christopher Faulet4d37e532021-03-26 14:44:00 +01003490and stored using an internal representation called HTX and it is no longer
3491possible to rely on raw representation. There is no way to go back.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003492
Christopher Faulet4d37e532021-03-26 14:44:00 +01003493There are two kind of upgrades, in-place upgrades and destructive upgrades. The
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003494first ones involves a TCP to HTTP/1 upgrade. In HTTP/1, the request
Christopher Faulet4d37e532021-03-26 14:44:00 +01003495processings are serialized, thus the applicative stream can be preserved. The
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003496second one involves a TCP to HTTP/2 upgrade. Because it is a multiplexed
Christopher Faulet4d37e532021-03-26 14:44:00 +01003497protocol, the applicative stream cannot be associated to any HTTP/2 stream and
3498is destroyed. New applicative streams are then created when HAProxy receives
3499new HTTP/2 streams at the lower level, in the H2 multiplexer. It is important
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003500to understand this difference because that drastically changes the way to
Christopher Faulet4d37e532021-03-26 14:44:00 +01003501process data. When an HTTP/1 upgrade is performed, the content processings
3502already performed on raw data are neither lost nor reexecuted while for an
3503HTTP/2 upgrade, applicative streams are distinct and all frontend rules are
3504evaluated systematically on each one. And as said, the first stream, the TCP
3505one, is destroyed, but only after the frontend rules were evaluated.
3506
3507There is another importnat point to understand when HTTP processings are
3508performed from a TCP proxy. While HAProxy is able to parse HTTP/1 in-fly from
3509tcp-request content rules, it is not possible for HTTP/2. Only the HTTP/2
3510preface can be parsed. This is a huge limitation regarding the HTTP content
3511analysis in TCP. Concretely it is only possible to know if received data are
3512HTTP. For instance, it is not possible to choose a backend based on the Host
3513header value while it is trivial in HTTP/1. Hopefully, there is a solution to
3514mitigate this drawback.
3515
Daniel Corbett9f0843f2021-05-08 10:50:37 -04003516There are two ways to perform an HTTP upgrade. The first one, the historical
Christopher Faulet4d37e532021-03-26 14:44:00 +01003517method, is to select an HTTP backend. The upgrade happens when the backend is
3518set. Thus, for in-place upgrades, only the backend configuration is considered
3519in the HTTP data processing. For destructive upgrades, the applicative stream
3520is destroyed, thus its processing is stopped. With this method, possibilities
3521to choose a backend with an HTTP/2 connection are really limited, as mentioned
3522above, and a bit useless because the stream is destroyed. The second method is
3523to upgrade during the tcp-request content rules evaluation, thanks to the
3524"switch-mode http" action. In this case, the upgrade is performed in the
3525frontend context and it is possible to define HTTP directives in this
3526frontend. For in-place upgrades, it offers all the power of the HTTP analysis
3527as soon as possible. It is not that far from an HTTP frontend. For destructive
3528upgrades, it does not change anything except it is useless to choose a backend
3529on limited information. It is of course the recommended method. Thus, testing
3530the request protocol from the tcp-request content rules to perform an HTTP
3531upgrade is enough. All the remaining HTTP manipulation may be moved to the
3532frontend http-request ruleset. But keep in mind that tcp-request content rules
3533remains evaluated on each streams, that can't be changed.
Willy Tarreau70dffda2014-01-30 03:07:23 +01003534
Willy Tarreauc57f0e22009-05-10 13:12:33 +020035354.1. Proxy keywords matrix
3536--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01003537
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003538The following list of keywords is supported. Most of them may only be used in a
3539limited set of section types. Some of them are marked as "deprecated" because
3540they are inherited from an old syntax which may be confusing or functionally
3541limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01003542marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003543option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02003544and must be disabled for a specific instance. Such options may also be prefixed
3545with "default" in order to restore default settings regardless of what has been
3546specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003547
Willy Tarreau6a06a402007-07-15 20:15:28 +02003548
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003549 keyword defaults frontend listen backend
3550------------------------------------+----------+----------+---------+---------
3551acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003552backlog X X X -
3553balance X - X X
3554bind - X X -
3555bind-process X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003556capture cookie - X X -
3557capture request header - X X -
3558capture response header - X X -
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09003559clitcpka-cnt X X X -
3560clitcpka-idle X X X -
3561clitcpka-intvl X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02003562compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003563cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003564declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003565default-server X - X X
3566default_backend X X X -
3567description - X X X
3568disabled X X X X
3569dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09003570email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09003571email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09003572email-alert mailers X X X X
3573email-alert myhostname X X X X
3574email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003575enabled X X X X
3576errorfile X X X X
Christopher Faulet76edc0f2020-01-13 15:52:01 +01003577errorfiles X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003578errorloc X X X X
3579errorloc302 X X X X
3580-- keyword -------------------------- defaults - frontend - listen -- backend -
3581errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003582force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003583filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003584fullconn X - X X
3585grace X X X X
3586hash-type X - X X
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01003587http-after-response - X X X
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02003588http-check comment X - X X
Christopher Faulete5870d82020-04-15 11:32:03 +02003589http-check connect X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003590http-check disable-on-404 X - X X
Christopher Faulete5870d82020-04-15 11:32:03 +02003591http-check expect X - X X
Peter Gervai8912ae62020-06-11 18:26:36 +02003592http-check send X - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02003593http-check send-state X - X X
Christopher Faulete5870d82020-04-15 11:32:03 +02003594http-check set-var X - X X
3595http-check unset-var X - X X
Christopher Faulet3b967c12020-05-15 15:47:44 +02003596http-error X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003597http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003598http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02003599http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02003600http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003601id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003602ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02003603load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02003604log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01003605log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02003606log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01003607log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02003608max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003609maxconn X X X -
3610mode X X X X
3611monitor fail - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003612monitor-uri X X X -
3613option abortonclose (*) X - X X
3614option accept-invalid-http-request (*) X X X -
3615option accept-invalid-http-response (*) X - X X
3616option allbackups (*) X - X X
3617option checkcache (*) X - X X
3618option clitcpka (*) X X X -
3619option contstats (*) X X X -
Christopher Faulet89aed322020-06-02 17:33:56 +02003620option disable-h2-upgrade (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003621option dontlog-normal (*) X X X -
3622option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003623-- keyword -------------------------- defaults - frontend - listen -- backend -
3624option forwardfor X X X X
Christopher Faulet98fbe952019-07-22 16:18:24 +02003625option h1-case-adjust-bogus-client (*) X X X -
3626option h1-case-adjust-bogus-server (*) X - X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02003627option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02003628option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01003629option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02003630option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02003631option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003632option http-server-close (*) X X X X
3633option http-use-proxy-header (*) X X X -
3634option httpchk X - X X
3635option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01003636option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003637option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04003638option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02003639option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09003640option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003641option log-health-checks (*) X - X X
3642option log-separate-errors (*) X X X -
3643option logasap (*) X X X -
3644option mysql-check X - X X
3645option nolinger (*) X X X X
3646option originalto X X X X
3647option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02003648option pgsql-check X - X X
3649option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003650option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02003651option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003652option smtpchk X - X X
3653option socket-stats (*) X X X -
3654option splice-auto (*) X X X X
3655option splice-request (*) X X X X
3656option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01003657option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003658option srvtcpka (*) X - X X
3659option ssl-hello-chk X - X X
3660-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01003661option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003662option tcp-smart-accept (*) X X X -
3663option tcp-smart-connect (*) X - X X
3664option tcpka X X X X
3665option tcplog X X X X
3666option transparent (*) X - X X
William Dauchy4711a892022-01-05 22:53:24 +01003667option idle-close-on-response (*) X X X -
Simon Horman98637e52014-06-20 12:30:16 +09003668external-check command X - X X
3669external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003670persist rdp-cookie X - X X
3671rate-limit sessions X X X -
3672redirect - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003673-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003674retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02003675retry-on X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003676server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02003677server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02003678server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003679source X - X X
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09003680srvtcpka-cnt X - X X
3681srvtcpka-idle X - X X
3682srvtcpka-intvl X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02003683stats admin - X X X
3684stats auth X X X X
3685stats enable X X X X
3686stats hide-version X X X X
3687stats http-request - X X X
3688stats realm X X X X
3689stats refresh X X X X
3690stats scope X X X X
3691stats show-desc X X X X
3692stats show-legends X X X X
3693stats show-node X X X X
3694stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003695-- keyword -------------------------- defaults - frontend - listen -- backend -
3696stick match - - X X
3697stick on - - X X
3698stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02003699stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01003700stick-table - X X X
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02003701tcp-check comment X - X X
Christopher Faulet404f9192020-04-09 23:13:54 +02003702tcp-check connect X - X X
3703tcp-check expect X - X X
3704tcp-check send X - X X
Christopher Fauletb50b3e62020-05-05 18:43:43 +02003705tcp-check send-lf X - X X
Christopher Faulet404f9192020-04-09 23:13:54 +02003706tcp-check send-binary X - X X
Christopher Fauletb50b3e62020-05-05 18:43:43 +02003707tcp-check send-binary-lf X - X X
Christopher Faulet404f9192020-04-09 23:13:54 +02003708tcp-check set-var X - X X
3709tcp-check unset-var X - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02003710tcp-request connection - X X -
3711tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02003712tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02003713tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02003714tcp-response content - - X X
3715tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003716timeout check X - X X
3717timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02003718timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003719timeout connect X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003720timeout http-keep-alive X X X X
3721timeout http-request X X X X
3722timeout queue X - X X
3723timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02003724timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003725timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02003726timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003727transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01003728unique-id-format X X X -
3729unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003730use_backend - X X -
Christopher Fauletb30b3102019-09-12 23:03:09 +02003731use-fcgi-app - - X X
Willy Tarreau4a5cade2012-04-05 21:09:48 +02003732use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01003733------------------------------------+----------+----------+---------+---------
3734 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02003735
Willy Tarreau0ba27502007-12-24 16:55:16 +01003736
Willy Tarreauc57f0e22009-05-10 13:12:33 +020037374.2. Alphabetically sorted keywords reference
3738---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01003739
3740This section provides a description of each keyword and its usage.
3741
3742
3743acl <aclname> <criterion> [flags] [operator] <value> ...
3744 Declare or complete an access list.
3745 May be used in sections : defaults | frontend | listen | backend
3746 no | yes | yes | yes
3747 Example:
3748 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
3749 acl invalid_src src_port 0:1023
3750 acl local_dst hdr(host) -i localhost
3751
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003752 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003753
3754
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01003755backlog <conns>
3756 Give hints to the system about the approximate listen backlog desired size
3757 May be used in sections : defaults | frontend | listen | backend
3758 yes | yes | yes | no
3759 Arguments :
3760 <conns> is the number of pending connections. Depending on the operating
3761 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02003762 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01003763
3764 In order to protect against SYN flood attacks, one solution is to increase
3765 the system's SYN backlog size. Depending on the system, sometimes it is just
3766 tunable via a system parameter, sometimes it is not adjustable at all, and
3767 sometimes the system relies on hints given by the application at the time of
3768 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
3769 to the listen() syscall. On systems which can make use of this value, it can
3770 sometimes be useful to be able to specify a different value, hence this
3771 backlog parameter.
3772
3773 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
3774 used as a hint and the system accepts up to the smallest greater power of
3775 two, and never more than some limits (usually 32768).
3776
3777 See also : "maxconn" and the target operating system's tuning guide.
3778
3779
Willy Tarreau0ba27502007-12-24 16:55:16 +01003780balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02003781balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003782 Define the load balancing algorithm to be used in a backend.
3783 May be used in sections : defaults | frontend | listen | backend
3784 yes | no | yes | yes
3785 Arguments :
3786 <algorithm> is the algorithm used to select a server when doing load
3787 balancing. This only applies when no persistence information
3788 is available, or when a connection is redispatched to another
3789 server. <algorithm> may be one of the following :
3790
3791 roundrobin Each server is used in turns, according to their weights.
3792 This is the smoothest and fairest algorithm when the server's
3793 processing time remains equally distributed. This algorithm
3794 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02003795 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08003796 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02003797 large farms, when a server becomes up after having been down
3798 for a very short time, it may sometimes take a few hundreds
3799 requests for it to be re-integrated into the farm and start
3800 receiving traffic. This is normal, though very rare. It is
3801 indicated here in case you would have the chance to observe
3802 it, so that you don't worry.
3803
3804 static-rr Each server is used in turns, according to their weights.
3805 This algorithm is as similar to roundrobin except that it is
3806 static, which means that changing a server's weight on the
3807 fly will have no effect. On the other hand, it has no design
3808 limitation on the number of servers, and when a server goes
3809 up, it is always immediately reintroduced into the farm, once
3810 the full map is recomputed. It also uses slightly less CPU to
3811 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003812
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01003813 leastconn The server with the lowest number of connections receives the
3814 connection. Round-robin is performed within groups of servers
3815 of the same load to ensure that all servers will be used. Use
3816 of this algorithm is recommended where very long sessions are
3817 expected, such as LDAP, SQL, TSE, etc... but is not very well
3818 suited for protocols using short sessions such as HTTP. This
3819 algorithm is dynamic, which means that server weights may be
Willy Tarreau8c855f62020-10-22 17:41:45 +02003820 adjusted on the fly for slow starts for instance. It will
3821 also consider the number of queued connections in addition to
3822 the established ones in order to minimize queuing.
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01003823
Willy Tarreauf09c6602012-02-13 17:12:08 +01003824 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003825 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01003826 identifier to the highest (see server parameter "id"), which
3827 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02003828 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01003829 not make sense to use this algorithm without setting maxconn.
3830 The purpose of this algorithm is to always use the smallest
3831 number of servers so that extra servers can be powered off
3832 during non-intensive hours. This algorithm ignores the server
3833 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02003834 or IMAP than HTTP, though it can be useful there too. In
3835 order to use this algorithm efficiently, it is recommended
3836 that a cloud controller regularly checks server usage to turn
3837 them off when unused, and regularly checks backend queue to
3838 turn new servers on when the queue inflates. Alternatively,
3839 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01003840
Willy Tarreau0ba27502007-12-24 16:55:16 +01003841 source The source IP address is hashed and divided by the total
3842 weight of the running servers to designate which server will
3843 receive the request. This ensures that the same client IP
3844 address will always reach the same server as long as no
3845 server goes down or up. If the hash result changes due to the
3846 number of running servers changing, many clients will be
3847 directed to a different server. This algorithm is generally
3848 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003849 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01003850 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003851 static by default, which means that changing a server's
3852 weight on the fly will have no effect, but this can be
3853 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003854
Oskar Stolc8dc41842012-05-19 10:19:54 +01003855 uri This algorithm hashes either the left part of the URI (before
3856 the question mark) or the whole URI (if the "whole" parameter
3857 is present) and divides the hash value by the total weight of
3858 the running servers. The result designates which server will
3859 receive the request. This ensures that the same URI will
3860 always be directed to the same server as long as no server
3861 goes up or down. This is used with proxy caches and
3862 anti-virus proxies in order to maximize the cache hit rate.
3863 Note that this algorithm may only be used in an HTTP backend.
3864 This algorithm is static by default, which means that
3865 changing a server's weight on the fly will have no effect,
3866 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003867
Oskar Stolc8dc41842012-05-19 10:19:54 +01003868 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02003869 "depth", both followed by a positive integer number. These
3870 options may be helpful when it is needed to balance servers
3871 based on the beginning of the URI only. The "len" parameter
3872 indicates that the algorithm should only consider that many
3873 characters at the beginning of the URI to compute the hash.
3874 Note that having "len" set to 1 rarely makes sense since most
3875 URIs start with a leading "/".
3876
3877 The "depth" parameter indicates the maximum directory depth
3878 to be used to compute the hash. One level is counted for each
3879 slash in the request. If both parameters are specified, the
3880 evaluation stops when either is reached.
3881
Willy Tarreau57a37412020-09-23 08:56:29 +02003882 A "path-only" parameter indicates that the hashing key starts
3883 at the first '/' of the path. This can be used to ignore the
3884 authority part of absolute URIs, and to make sure that HTTP/1
3885 and HTTP/2 URIs will provide the same hash.
3886
Willy Tarreau0ba27502007-12-24 16:55:16 +01003887 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003888 the query string of each HTTP GET request.
3889
3890 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02003891 request entity will be searched for the parameter argument,
3892 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02003893 ('?') in the URL. The message body will only start to be
3894 analyzed once either the advertised amount of data has been
3895 received or the request buffer is full. In the unlikely event
3896 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02003897 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02003898 be randomly balanced if at all. This keyword used to support
3899 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003900
3901 If the parameter is found followed by an equal sign ('=') and
3902 a value, then the value is hashed and divided by the total
3903 weight of the running servers. The result designates which
3904 server will receive the request.
3905
3906 This is used to track user identifiers in requests and ensure
3907 that a same user ID will always be sent to the same server as
3908 long as no server goes up or down. If no value is found or if
3909 the parameter is not found, then a round robin algorithm is
3910 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003911 backend. This algorithm is static by default, which means
3912 that changing a server's weight on the fly will have no
3913 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003914
Cyril Bontédc4d9032012-04-08 21:57:39 +02003915 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
3916 request. Just as with the equivalent ACL 'hdr()' function,
3917 the header name in parenthesis is not case sensitive. If the
3918 header is absent or if it does not contain any value, the
3919 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01003920
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003921 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01003922 reducing the hash algorithm to the main domain part with some
3923 specific headers such as 'Host'. For instance, in the Host
3924 value "haproxy.1wt.eu", only "1wt" will be considered.
3925
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003926 This algorithm is static by default, which means that
3927 changing a server's weight on the fly will have no effect,
3928 but this can be changed using "hash-type".
3929
Willy Tarreau21c741a2019-01-14 18:14:27 +01003930 random
3931 random(<draws>)
3932 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02003933 hashing function. This means that the servers' weights are
3934 respected, dynamic weight changes immediately take effect, as
3935 well as new server additions. Random load balancing can be
3936 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01003937 or removed as it may avoid the hammering effect that could
3938 result from roundrobin or leastconn in this situation. The
3939 hash-balance-factor directive can be used to further improve
3940 fairness of the load balancing, especially in situations
3941 where servers show highly variable response times. When an
3942 argument <draws> is present, it must be an integer value one
3943 or greater, indicating the number of draws before selecting
3944 the least loaded of these servers. It was indeed demonstrated
3945 that picking the least loaded of two servers is enough to
3946 significantly improve the fairness of the algorithm, by
3947 always avoiding to pick the most loaded server within a farm
3948 and getting rid of any bias that could be induced by the
3949 unfair distribution of the consistent list. Higher values N
3950 will take away N-1 of the highest loaded servers at the
3951 expense of performance. With very high values, the algorithm
3952 will converge towards the leastconn's result but much slower.
3953 The default value is 2, which generally shows very good
3954 distribution and performance. This algorithm is also known as
3955 the Power of Two Random Choices and is described here :
3956 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02003957
Emeric Brun736aa232009-06-30 17:56:00 +02003958 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02003959 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02003960 The RDP cookie <name> (or "mstshash" if omitted) will be
3961 looked up and hashed for each incoming TCP request. Just as
3962 with the equivalent ACL 'req_rdp_cookie()' function, the name
3963 is not case-sensitive. This mechanism is useful as a degraded
3964 persistence mode, as it makes it possible to always send the
3965 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003966 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02003967 used instead.
3968
3969 Note that for this to work, the frontend must ensure that an
3970 RDP cookie is already present in the request buffer. For this
3971 you must use 'tcp-request content accept' rule combined with
3972 a 'req_rdp_cookie_cnt' ACL.
3973
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003974 This algorithm is static by default, which means that
3975 changing a server's weight on the fly will have no effect,
3976 but this can be changed using "hash-type".
3977
Cyril Bontédc4d9032012-04-08 21:57:39 +02003978 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09003979
Willy Tarreau0ba27502007-12-24 16:55:16 +01003980 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02003981 algorithms. Right now, only "url_param" and "uri" support an
3982 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003983
Willy Tarreau3cd9af22009-03-15 14:06:41 +01003984 The load balancing algorithm of a backend is set to roundrobin when no other
3985 algorithm, mode nor option have been set. The algorithm may only be set once
3986 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003987
Lukas Tribus80512b12018-10-27 20:07:40 +02003988 With authentication schemes that require the same connection like NTLM, URI
John Roeslerfb2fce12019-07-10 15:45:51 -05003989 based algorithms must not be used, as they would cause subsequent requests
Lukas Tribus80512b12018-10-27 20:07:40 +02003990 to be routed to different backend servers, breaking the invalid assumptions
3991 NTLM relies on.
3992
Willy Tarreau0ba27502007-12-24 16:55:16 +01003993 Examples :
3994 balance roundrobin
3995 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003996 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01003997 balance hdr(User-Agent)
3998 balance hdr(host)
3999 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02004000
4001 Note: the following caveats and limitations on using the "check_post"
4002 extension with "url_param" must be considered :
4003
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004004 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02004005 to determine if the parameters will be found in the body or entity which
4006 may contain binary data. Therefore another method may be required to
4007 restrict consideration of POST requests that have no URL parameters in
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02004008 the body. (see acl http_end)
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02004009
4010 - using a <max_wait> value larger than the request buffer size does not
4011 make sense and is useless. The buffer size is set at build time, and
4012 defaults to 16 kB.
4013
4014 - Content-Encoding is not supported, the parameter search will probably
4015 fail; and load balancing will fall back to Round Robin.
4016
4017 - Expect: 100-continue is not supported, load balancing will fall back to
4018 Round Robin.
4019
Lukas Tribus23953682017-04-28 13:24:30 +00004020 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02004021 If the entire parameter value is not present in the first chunk, the
4022 selection of server is undefined (actually, defined by how little
4023 actually appeared in the first chunk).
4024
4025 - This feature does not support generation of a 100, 411 or 501 response.
4026
4027 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004028 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02004029 white space or control characters are found, indicating the end of what
4030 might be a URL parameter list. This is probably not a concern with SGML
4031 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004032
Willy Tarreau294d0f02015-08-10 19:40:12 +02004033 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01004034
4035
Willy Tarreaub6205fd2012-09-24 12:27:33 +02004036bind [<address>]:<port_range> [, ...] [param*]
4037bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01004038 Define one or several listening addresses and/or ports in a frontend.
4039 May be used in sections : defaults | frontend | listen | backend
4040 no | yes | yes | no
4041 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01004042 <address> is optional and can be a host name, an IPv4 address, an IPv6
4043 address, or '*'. It designates the address the frontend will
4044 listen on. If unset, all IPv4 addresses of the system will be
4045 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01004046 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01004047 Optionally, an address family prefix may be used before the
4048 address to force the family regardless of the address format,
4049 which can be useful to specify a path to a unix socket with
4050 no slash ('/'). Currently supported prefixes are :
4051 - 'ipv4@' -> address is always IPv4
4052 - 'ipv6@' -> address is always IPv6
Emeric Brun3835c0d2020-07-07 09:46:09 +02004053 - 'udp@' -> address is resolved as IPv4 or IPv6 and
Emeric Brun12941c82020-07-07 14:19:42 +02004054 protocol UDP is used. Currently those listeners are
4055 supported only in log-forward sections.
Emeric Brun3835c0d2020-07-07 09:46:09 +02004056 - 'udp4@' -> address is always IPv4 and protocol UDP
Emeric Brun12941c82020-07-07 14:19:42 +02004057 is used. Currently those listeners are supported
4058 only in log-forward sections.
Emeric Brun3835c0d2020-07-07 09:46:09 +02004059 - 'udp6@' -> address is always IPv6 and protocol UDP
Emeric Brun12941c82020-07-07 14:19:42 +02004060 is used. Currently those listeners are supported
4061 only in log-forward sections.
Willy Tarreau24709282013-03-10 21:32:12 +01004062 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02004063 - 'abns@' -> address is in abstract namespace (Linux only).
4064 Note: since abstract sockets are not "rebindable", they
4065 do not cope well with multi-process mode during
4066 soft-restart, so it is better to avoid them if
4067 nbproc is greater than 1. The effect is that if the
4068 new process fails to start, only one of the old ones
4069 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01004070 - 'fd@<n>' -> use file descriptor <n> inherited from the
4071 parent. The fd must be bound and may or may not already
4072 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02004073 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
4074 connected unix socket or of a socketpair. The bind waits
4075 to receive a FD over the unix socket and uses it as if it
4076 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02004077 You may want to reference some environment variables in the
4078 address parameter, see section 2.3 about environment
4079 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01004080
Willy Tarreauc5011ca2010-03-22 11:53:56 +01004081 <port_range> is either a unique TCP port, or a port range for which the
4082 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01004083 above. The port is mandatory for TCP listeners. Note that in
4084 the case of an IPv6 address, the port is always the number
4085 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01004086 - a numerical port (ex: '80')
4087 - a dash-delimited ports range explicitly stating the lower
4088 and upper bounds (ex: '2000-2100') which are included in
4089 the range.
4090
4091 Particular care must be taken against port ranges, because
4092 every <address:port> couple consumes one socket (= a file
4093 descriptor), so it's easy to consume lots of descriptors
4094 with a simple range, and to run out of sockets. Also, each
4095 <address:port> couple must be used only once among all
4096 instances running on a same system. Please note that binding
4097 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004098 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01004099 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004100
Willy Tarreauceb24bc2010-11-09 12:46:41 +01004101 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01004102 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01004103 receive UNIX connections on the socket located at this place.
4104 The path must begin with a slash and by default is absolute.
4105 It can be relative to the prefix defined by "unix-bind" in
4106 the global section. Note that the total length of the prefix
4107 followed by the socket path cannot exceed some system limits
4108 for UNIX sockets, which commonly are set to 107 characters.
4109
Willy Tarreaub6205fd2012-09-24 12:27:33 +02004110 <param*> is a list of parameters common to all sockets declared on the
4111 same line. These numerous parameters depend on OS and build
4112 options and have a complete section dedicated to them. Please
4113 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02004114
Willy Tarreau0ba27502007-12-24 16:55:16 +01004115 It is possible to specify a list of address:port combinations delimited by
4116 commas. The frontend will then listen on all of these addresses. There is no
4117 fixed limit to the number of addresses and ports which can be listened on in
4118 a frontend, as well as there is no limit to the number of "bind" statements
4119 in a frontend.
4120
4121 Example :
4122 listen http_proxy
4123 bind :80,:443
4124 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01004125 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01004126
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02004127 listen http_https_proxy
4128 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02004129 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02004130
Willy Tarreau24709282013-03-10 21:32:12 +01004131 listen http_https_proxy_explicit
4132 bind ipv6@:80
4133 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
4134 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
4135
Willy Tarreaudad36a32013-03-11 01:20:04 +01004136 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02004137 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01004138
Willy Tarreau55dcaf62015-09-27 15:03:15 +02004139 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
4140 sun_path length is used for the address length. Some other programs
4141 such as socat use the string length only by default. Pass the option
4142 ",unix-tightsocklen=0" to any abstract socket definition in socat to
4143 make it compatible with HAProxy's.
4144
Willy Tarreauceb24bc2010-11-09 12:46:41 +01004145 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02004146 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004147
4148
Christopher Fauletff4121f2017-11-22 16:38:49 +01004149bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004150 Limit visibility of an instance to a certain set of processes numbers.
4151 May be used in sections : defaults | frontend | listen | backend
4152 yes | yes | yes | yes
4153 Arguments :
4154 all All process will see this instance. This is the default. It
4155 may be used to override a default value.
4156
Willy Tarreaua9db57e2013-01-18 11:29:29 +01004157 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004158 option may be combined with other numbers.
4159
Willy Tarreaua9db57e2013-01-18 11:29:29 +01004160 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004161 option may be combined with other numbers. Do not use it
4162 with less than 2 processes otherwise some instances might be
4163 missing from all processes.
4164
Christopher Fauletff4121f2017-11-22 16:38:49 +01004165 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01004166 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01004167 the machine's word size. Ranges can be partially defined. The
4168 higher bound can be omitted. In such case, it is replaced by
4169 the corresponding maximum value. If a proxy is bound to
4170 process numbers greater than the configured global.nbproc, it
4171 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02004172 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004173
4174 This keyword limits binding of certain instances to certain processes. This
4175 is useful in order not to have too many processes listening to the same
4176 ports. For instance, on a dual-core machine, it might make sense to set
4177 'nbproc 2' in the global section, then distributes the listeners among 'odd'
4178 and 'even' instances.
4179
Willy Tarreaua9db57e2013-01-18 11:29:29 +01004180 At the moment, it is not possible to reference more than 32 or 64 processes
4181 using this keyword, but this should be more than enough for most setups.
4182 Please note that 'all' really means all processes regardless of the machine's
4183 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004184
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02004185 Each "bind" line may further be limited to a subset of the proxy's processes,
4186 please consult the "process" bind keyword in section 5.1.
4187
Willy Tarreaub369a042014-09-16 13:21:03 +02004188 When a frontend has no explicit "bind-process" line, it tries to bind to all
4189 the processes referenced by its "bind" lines. That means that frontends can
4190 easily adapt to their listeners' processes.
4191
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004192 If some backends are referenced by frontends bound to other processes, the
4193 backend automatically inherits the frontend's processes.
4194
4195 Example :
4196 listen app_ip1
4197 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02004198 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004199
4200 listen app_ip2
4201 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02004202 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004203
4204 listen management
4205 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02004206 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004207
Willy Tarreau110ecc12012-11-15 17:50:01 +01004208 listen management
4209 bind 10.0.0.4:80
4210 bind-process 1-4
4211
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02004212 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01004213
4214
Willy Tarreau0ba27502007-12-24 16:55:16 +01004215capture cookie <name> len <length>
4216 Capture and log a cookie in the request and in the response.
4217 May be used in sections : defaults | frontend | listen | backend
4218 no | yes | yes | no
4219 Arguments :
4220 <name> is the beginning of the name of the cookie to capture. In order
4221 to match the exact name, simply suffix the name with an equal
4222 sign ('='). The full name will appear in the logs, which is
4223 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01004224 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01004225
4226 <length> is the maximum number of characters to report in the logs, which
4227 include the cookie name, the equal sign and the value, all in the
4228 standard "name=value" form. The string will be truncated on the
4229 right if it exceeds <length>.
4230
4231 Only the first cookie is captured. Both the "cookie" request headers and the
4232 "set-cookie" response headers are monitored. This is particularly useful to
4233 check for application bugs causing session crossing or stealing between
4234 users, because generally the user's cookies can only change on a login page.
4235
4236 When the cookie was not presented by the client, the associated log column
4237 will report "-". When a request does not cause a cookie to be assigned by the
4238 server, a "-" is reported in the response column.
4239
4240 The capture is performed in the frontend only because it is necessary that
4241 the log format does not change for a given frontend depending on the
4242 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01004243 "capture cookie" statement in a frontend. The maximum capture length is set
4244 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
4245 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004246
4247 Example:
4248 capture cookie ASPSESSION len 32
4249
4250 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004251 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004252
4253
4254capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01004255 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004256 May be used in sections : defaults | frontend | listen | backend
4257 no | yes | yes | no
4258 Arguments :
4259 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004260 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01004261 appear in the requests, with the first letter of each word in
4262 upper case. The header name will not appear in the logs, only the
4263 value is reported, but the position in the logs is respected.
4264
4265 <length> is the maximum number of characters to extract from the value and
4266 report in the logs. The string will be truncated on the right if
4267 it exceeds <length>.
4268
Willy Tarreau4460d032012-11-21 23:37:37 +01004269 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01004270 value will be added to the logs between braces ('{}'). If multiple headers
4271 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004272 in the same order they were declared in the configuration. Non-existent
4273 headers will be logged just as an empty string. Common uses for request
4274 header captures include the "Host" field in virtual hosting environments, the
4275 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004276 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004277 environments to find where the request came from.
4278
4279 Note that when capturing headers such as "User-agent", some spaces may be
4280 logged, making the log analysis more difficult. Thus be careful about what
4281 you log if you know your log parser is not smart enough to rely on the
4282 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004283
Willy Tarreau0900abb2012-11-22 00:21:46 +01004284 There is no limit to the number of captured request headers nor to their
4285 length, though it is wise to keep them low to limit memory usage per session.
4286 In order to keep log format consistent for a same frontend, header captures
4287 can only be declared in a frontend. It is not possible to specify a capture
4288 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004289
4290 Example:
4291 capture request header Host len 15
4292 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01004293 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01004294
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004295 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01004296 about logging.
4297
4298
4299capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01004300 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004301 May be used in sections : defaults | frontend | listen | backend
4302 no | yes | yes | no
4303 Arguments :
4304 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004305 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01004306 appear in the response, with the first letter of each word in
4307 upper case. The header name will not appear in the logs, only the
4308 value is reported, but the position in the logs is respected.
4309
4310 <length> is the maximum number of characters to extract from the value and
4311 report in the logs. The string will be truncated on the right if
4312 it exceeds <length>.
4313
Willy Tarreau4460d032012-11-21 23:37:37 +01004314 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01004315 result will be added to the logs between braces ('{}') after the captured
4316 request headers. If multiple headers are captured, they will be delimited by
4317 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004318 the configuration. Non-existent headers will be logged just as an empty
4319 string. Common uses for response header captures include the "Content-length"
4320 header which indicates how many bytes are expected to be returned, the
4321 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004322
Willy Tarreau0900abb2012-11-22 00:21:46 +01004323 There is no limit to the number of captured response headers nor to their
4324 length, though it is wise to keep them low to limit memory usage per session.
4325 In order to keep log format consistent for a same frontend, header captures
4326 can only be declared in a frontend. It is not possible to specify a capture
4327 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004328
4329 Example:
4330 capture response header Content-length len 9
4331 capture response header Location len 15
4332
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004333 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01004334 about logging.
4335
4336
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09004337clitcpka-cnt <count>
4338 Sets the maximum number of keepalive probes TCP should send before dropping
4339 the connection on the client side.
4340 May be used in sections : defaults | frontend | listen | backend
4341 yes | yes | yes | no
4342 Arguments :
4343 <count> is the maximum number of keepalive probes.
4344
4345 This keyword corresponds to the socket option TCP_KEEPCNT. If this keyword
4346 is not specified, system-wide TCP parameter (tcp_keepalive_probes) is used.
Willy Tarreau52543212020-07-09 05:58:51 +02004347 The availability of this setting depends on the operating system. It is
4348 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09004349
4350 See also : "option clitcpka", "clitcpka-idle", "clitcpka-intvl".
4351
4352
4353clitcpka-idle <timeout>
4354 Sets the time the connection needs to remain idle before TCP starts sending
4355 keepalive probes, if enabled the sending of TCP keepalive packets on the
4356 client side.
4357 May be used in sections : defaults | frontend | listen | backend
4358 yes | yes | yes | no
4359 Arguments :
4360 <timeout> is the time the connection needs to remain idle before TCP starts
4361 sending keepalive probes. It is specified in seconds by default,
4362 but can be in any other unit if the number is suffixed by the
4363 unit, as explained at the top of this document.
4364
4365 This keyword corresponds to the socket option TCP_KEEPIDLE. If this keyword
4366 is not specified, system-wide TCP parameter (tcp_keepalive_time) is used.
Willy Tarreau52543212020-07-09 05:58:51 +02004367 The availability of this setting depends on the operating system. It is
4368 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09004369
4370 See also : "option clitcpka", "clitcpka-cnt", "clitcpka-intvl".
4371
4372
4373clitcpka-intvl <timeout>
4374 Sets the time between individual keepalive probes on the client side.
4375 May be used in sections : defaults | frontend | listen | backend
4376 yes | yes | yes | no
4377 Arguments :
4378 <timeout> is the time between individual keepalive probes. It is specified
4379 in seconds by default, but can be in any other unit if the number
4380 is suffixed by the unit, as explained at the top of this
4381 document.
4382
4383 This keyword corresponds to the socket option TCP_KEEPINTVL. If this keyword
4384 is not specified, system-wide TCP parameter (tcp_keepalive_intvl) is used.
Willy Tarreau52543212020-07-09 05:58:51 +02004385 The availability of this setting depends on the operating system. It is
4386 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +09004387
4388 See also : "option clitcpka", "clitcpka-cnt", "clitcpka-idle".
4389
4390
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004391compression algo <algorithm> ...
4392compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02004393compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02004394 Enable HTTP compression.
4395 May be used in sections : defaults | frontend | listen | backend
4396 yes | yes | yes | yes
4397 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004398 algo is followed by the list of supported compression algorithms.
4399 type is followed by the list of MIME types that will be compressed.
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004400 offload makes HAProxy work as a compression offloader only (see notes).
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004401
4402 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01004403 identity this is mostly for debugging, and it was useful for developing
4404 the compression feature. Identity does not apply any change on
4405 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004406
Willy Tarreauc91840a2015-03-28 17:00:39 +01004407 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01004408 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01004409
4410 deflate same as "gzip", but with deflate algorithm and zlib format.
4411 Note that this algorithm has ambiguous support on many
4412 browsers and no support at all from recent ones. It is
4413 strongly recommended not to use it for anything else than
4414 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01004415 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004416
Willy Tarreauc91840a2015-03-28 17:00:39 +01004417 raw-deflate same as "deflate" without the zlib wrapper, and used as an
4418 alternative when the browser wants "deflate". All major
4419 browsers understand it and despite violating the standards,
4420 it is known to work better than "deflate", at least on MSIE
4421 and some versions of Safari. Do not use it in conjunction
4422 with "deflate", use either one or the other since both react
4423 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01004424 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004425
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04004426 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01004427 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04004428 If backend servers support HTTP compression, these directives
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004429 will be no-op: HAProxy will see the compressed response and will not
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04004430 compress again. If backend servers do not support HTTP compression and
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004431 there is Accept-Encoding header in request, HAProxy will compress the
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04004432 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02004433
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004434 The "offload" setting makes HAProxy remove the Accept-Encoding header to
Willy Tarreau70737d12012-10-27 00:34:28 +02004435 prevent backend servers from compressing responses. It is strongly
4436 recommended not to do this because this means that all the compression work
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004437 will be done on the single point where HAProxy is located. However in some
4438 deployment scenarios, HAProxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04004439 with broken HTTP compression implementation which can't be turned off.
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004440 In that case HAProxy can be used to prevent that gateway from emitting
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04004441 invalid payloads. In this case, simply removing the header in the
4442 configuration does not work because it applies before the header is parsed,
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004443 so that prevents HAProxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02004444 then be used for such scenarios. Note: for now, the "offload" setting is
4445 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02004446
William Lallemand05097442012-11-20 12:14:28 +01004447 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01004448 * the request does not advertise a supported compression algorithm in the
4449 "Accept-Encoding" header
Julien Pivottoff80c822021-03-29 12:41:40 +02004450 * the response message is not HTTP/1.1 or above
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01004451 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01004452 * response contain neither a "Content-Length" header nor a
4453 "Transfer-Encoding" whose last value is "chunked"
4454 * response contains a "Content-Type" header whose first value starts with
4455 "multipart"
4456 * the response contains the "no-transform" value in the "Cache-control"
4457 header
4458 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
4459 and later
4460 * The response contains a "Content-Encoding" header, indicating that the
4461 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01004462 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01004463
Tim Duesterhusb229f012019-01-29 16:38:56 +01004464 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01004465
William Lallemand82fe75c2012-10-23 10:25:10 +02004466 Examples :
4467 compression algo gzip
4468 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01004469
Christopher Fauletc3fe5332016-04-07 15:30:10 +02004470
Willy Tarreau55165fe2009-05-10 12:02:55 +02004471cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02004472 [ postonly ] [ preserve ] [ httponly ] [ secure ]
4473 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Christopher Faulet2f533902020-01-21 11:06:48 +01004474 [ dynamic ] [ attr <value> ]*
Willy Tarreau0ba27502007-12-24 16:55:16 +01004475 Enable cookie-based persistence in a backend.
4476 May be used in sections : defaults | frontend | listen | backend
4477 yes | no | yes | yes
4478 Arguments :
4479 <name> is the name of the cookie which will be monitored, modified or
4480 inserted in order to bring persistence. This cookie is sent to
4481 the client via a "Set-Cookie" header in the response, and is
4482 brought back by the client in a "Cookie" header in all requests.
4483 Special care should be taken to choose a name which does not
4484 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01004485 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004486 HTTP/HTTPS), care should be taken to use different cookie names
4487 between all backends if persistence between them is not desired.
4488
4489 rewrite This keyword indicates that the cookie will be provided by the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004490 server and that HAProxy will have to modify its value to set the
Willy Tarreau0ba27502007-12-24 16:55:16 +01004491 server's identifier in it. This mode is handy when the management
4492 of complex combinations of "Set-cookie" and "Cache-control"
4493 headers is left to the application. The application can then
4494 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01004495 cookie. Since all responses should be monitored, this mode
4496 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01004497 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01004498 start with this mode for new deployments. This keyword is
4499 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01004500
4501 insert This keyword indicates that the persistence cookie will have to
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004502 be inserted by HAProxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004503
Willy Tarreaua79094d2010-08-31 22:54:15 +02004504 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004505 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02004506 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01004507 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004508 existing configurations running in the "rewrite" mode. The cookie
4509 will only be a session cookie and will not be stored on the
4510 client's disk. By default, unless the "indirect" option is added,
4511 the server will see the cookies emitted by the client. Due to
4512 caching effects, it is generally wise to add the "nocache" or
4513 "postonly" keywords (see below). The "insert" keyword is not
4514 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01004515
4516 prefix This keyword indicates that instead of relying on a dedicated
4517 cookie for the persistence, an existing one will be completed.
4518 This may be needed in some specific environments where the client
4519 does not support more than one single cookie and the application
4520 already needs it. In this case, whenever the server sets a cookie
4521 named <name>, it will be prefixed with the server's identifier
4522 and a delimiter. The prefix will be removed from all client
4523 requests so that the server still finds the cookie it emitted.
4524 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01004525 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02004526 not compatible with "rewrite" and "insert". Note: it is highly
4527 recommended not to use "indirect" with "prefix", otherwise server
4528 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004529
Willy Tarreaua79094d2010-08-31 22:54:15 +02004530 indirect When this option is specified, no cookie will be emitted to a
4531 client which already has a valid one for the server which has
4532 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004533 it will be removed, unless the "preserve" option is also set. In
4534 "insert" mode, this will additionally remove cookies from the
4535 requests transmitted to the server, making the persistence
4536 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02004537 Note: it is highly recommended not to use "indirect" with
4538 "prefix", otherwise server cookie updates would not be sent to
4539 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004540
4541 nocache This option is recommended in conjunction with the insert mode
4542 when there is a cache between the client and HAProxy, as it
4543 ensures that a cacheable response will be tagged non-cacheable if
4544 a cookie needs to be inserted. This is important because if all
4545 persistence cookies are added on a cacheable home page for
4546 instance, then all customers will then fetch the page from an
4547 outer cache and will all share the same persistence cookie,
4548 leading to one server receiving much more traffic than others.
4549 See also the "insert" and "postonly" options.
4550
4551 postonly This option ensures that cookie insertion will only be performed
4552 on responses to POST requests. It is an alternative to the
4553 "nocache" option, because POST responses are not cacheable, so
4554 this ensures that the persistence cookie will never get cached.
4555 Since most sites do not need any sort of persistence before the
4556 first POST which generally is a login request, this is a very
4557 efficient method to optimize caching without risking to find a
4558 persistence cookie in the cache.
4559 See also the "insert" and "nocache" options.
4560
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004561 preserve This option may only be used with "insert" and/or "indirect". It
4562 allows the server to emit the persistence cookie itself. In this
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004563 case, if a cookie is found in the response, HAProxy will leave it
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004564 untouched. This is useful in order to end persistence after a
4565 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01004566 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02004567 the past. By combining this mechanism with the "disable-on-404"
4568 check option, it is possible to perform a completely graceful
4569 shutdown because users will definitely leave the server after
4570 they logout.
4571
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004572 httponly This option tells HAProxy to add an "HttpOnly" cookie attribute
Willy Tarreau4992dd22012-05-31 21:02:17 +02004573 when a cookie is inserted. This attribute is used so that a
4574 user agent doesn't share the cookie with non-HTTP components.
4575 Please check RFC6265 for more information on this attribute.
4576
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004577 secure This option tells HAProxy to add a "Secure" cookie attribute when
Willy Tarreau4992dd22012-05-31 21:02:17 +02004578 a cookie is inserted. This attribute is used so that a user agent
4579 never emits this cookie over non-secure channels, which means
4580 that a cookie learned with this flag will be presented only over
4581 SSL/TLS connections. Please check RFC6265 for more information on
4582 this attribute.
4583
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02004584 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004585 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01004586 name. If the domain begins with a dot, the browser is allowed to
4587 use it for any host ending with that name. It is also possible to
4588 specify several domain names by invoking this option multiple
4589 times. Some browsers might have small limits on the number of
4590 domains, so be careful when doing that. For the record, sending
4591 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02004592
Willy Tarreau996a92c2010-10-13 19:30:47 +02004593 maxidle This option allows inserted cookies to be ignored after some idle
4594 time. It only works with insert-mode cookies. When a cookie is
4595 sent to the client, the date this cookie was emitted is sent too.
4596 Upon further presentations of this cookie, if the date is older
4597 than the delay indicated by the parameter (in seconds), it will
4598 be ignored. Otherwise, it will be refreshed if needed when the
4599 response is sent to the client. This is particularly useful to
4600 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01004601 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02004602 this option is set and a cookie has no date, it is always
4603 accepted, but gets refreshed in the response. This maintains the
4604 ability for admins to access their sites. Cookies that have a
4605 date in the future further than 24 hours are ignored. Doing so
4606 lets admins fix timezone issues without risking kicking users off
4607 the site.
4608
4609 maxlife This option allows inserted cookies to be ignored after some life
4610 time, whether they're in use or not. It only works with insert
4611 mode cookies. When a cookie is first sent to the client, the date
4612 this cookie was emitted is sent too. Upon further presentations
4613 of this cookie, if the date is older than the delay indicated by
4614 the parameter (in seconds), it will be ignored. If the cookie in
4615 the request has no date, it is accepted and a date will be set.
4616 Cookies that have a date in the future further than 24 hours are
4617 ignored. Doing so lets admins fix timezone issues without risking
4618 kicking users off the site. Contrary to maxidle, this value is
4619 not refreshed, only the first visit date counts. Both maxidle and
4620 maxlife may be used at the time. This is particularly useful to
4621 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01004622 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02004623 is stronger than the maxidle method in that it forces a
4624 redispatch after some absolute delay.
4625
Olivier Houchard4e694042017-03-14 20:01:29 +01004626 dynamic Activate dynamic cookies. When used, a session cookie is
4627 dynamically created for each server, based on the IP and port
4628 of the server, and a secret key, specified in the
4629 "dynamic-cookie-key" backend directive.
4630 The cookie will be regenerated each time the IP address change,
4631 and is only generated for IPv4/IPv6.
4632
Daniel Corbett9f0843f2021-05-08 10:50:37 -04004633 attr This option tells HAProxy to add an extra attribute when a
Christopher Faulet2f533902020-01-21 11:06:48 +01004634 cookie is inserted. The attribute value can contain any
4635 characters except control ones or ";". This option may be
4636 repeated.
4637
Willy Tarreau0ba27502007-12-24 16:55:16 +01004638 There can be only one persistence cookie per HTTP backend, and it can be
4639 declared in a defaults section. The value of the cookie will be the value
4640 indicated after the "cookie" keyword in a "server" statement. If no cookie
4641 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02004642
Willy Tarreau0ba27502007-12-24 16:55:16 +01004643 Examples :
4644 cookie JSESSIONID prefix
4645 cookie SRV insert indirect nocache
4646 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02004647 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01004648
Willy Tarreau294d0f02015-08-10 19:40:12 +02004649 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01004650
Willy Tarreau983e01e2010-01-11 18:42:06 +01004651
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02004652declare capture [ request | response ] len <length>
4653 Declares a capture slot.
4654 May be used in sections : defaults | frontend | listen | backend
4655 no | yes | yes | no
4656 Arguments:
4657 <length> is the length allowed for the capture.
4658
4659 This declaration is only available in the frontend or listen section, but the
4660 reserved slot can be used in the backends. The "request" keyword allocates a
4661 capture slot for use in the request, and "response" allocates a capture slot
4662 for use in the response.
4663
4664 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02004665 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02004666 "http-request capture" and "http-response capture".
4667
4668
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01004669default-server [param*]
4670 Change default options for a server in a backend
4671 May be used in sections : defaults | frontend | listen | backend
4672 yes | no | yes | yes
4673 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01004674 <param*> is a list of parameters for this server. The "default-server"
4675 keyword accepts an important number of options and has a complete
4676 section dedicated to it. Please refer to section 5 for more
4677 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01004678
Willy Tarreau983e01e2010-01-11 18:42:06 +01004679 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01004680 default-server inter 1000 weight 13
4681
4682 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01004683
Willy Tarreau983e01e2010-01-11 18:42:06 +01004684
Willy Tarreau0ba27502007-12-24 16:55:16 +01004685default_backend <backend>
4686 Specify the backend to use when no "use_backend" rule has been matched.
4687 May be used in sections : defaults | frontend | listen | backend
4688 yes | yes | yes | no
4689 Arguments :
4690 <backend> is the name of the backend to use.
4691
4692 When doing content-switching between frontend and backends using the
4693 "use_backend" keyword, it is often useful to indicate which backend will be
4694 used when no rule has matched. It generally is the dynamic backend which
4695 will catch all undetermined requests.
4696
Willy Tarreau0ba27502007-12-24 16:55:16 +01004697 Example :
4698
4699 use_backend dynamic if url_dyn
4700 use_backend static if url_css url_img extension_img
4701 default_backend dynamic
4702
Willy Tarreau98d04852015-05-26 12:18:29 +02004703 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004704
Willy Tarreau0ba27502007-12-24 16:55:16 +01004705
Baptiste Assmann27f51342013-10-09 06:51:49 +02004706description <string>
4707 Describe a listen, frontend or backend.
4708 May be used in sections : defaults | frontend | listen | backend
4709 no | yes | yes | yes
4710 Arguments : string
4711
4712 Allows to add a sentence to describe the related object in the HAProxy HTML
4713 stats page. The description will be printed on the right of the object name
4714 it describes.
4715 No need to backslash spaces in the <string> arguments.
4716
4717
Willy Tarreau0ba27502007-12-24 16:55:16 +01004718disabled
4719 Disable a proxy, frontend or backend.
4720 May be used in sections : defaults | frontend | listen | backend
4721 yes | yes | yes | yes
4722 Arguments : none
4723
4724 The "disabled" keyword is used to disable an instance, mainly in order to
4725 liberate a listening port or to temporarily disable a service. The instance
4726 will still be created and its configuration will be checked, but it will be
4727 created in the "stopped" state and will appear as such in the statistics. It
4728 will not receive any traffic nor will it send any health-checks or logs. It
4729 is possible to disable many instances at once by adding the "disabled"
4730 keyword in a "defaults" section.
4731
4732 See also : "enabled"
4733
4734
Willy Tarreau5ce94572010-06-07 14:35:41 +02004735dispatch <address>:<port>
4736 Set a default server address
4737 May be used in sections : defaults | frontend | listen | backend
4738 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004739 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02004740
4741 <address> is the IPv4 address of the default server. Alternatively, a
4742 resolvable hostname is supported, but this name will be resolved
4743 during start-up.
4744
4745 <ports> is a mandatory port specification. All connections will be sent
4746 to this port, and it is not permitted to use port offsets as is
4747 possible with normal servers.
4748
Willy Tarreau787aed52011-04-15 06:45:37 +02004749 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02004750 server can take the connection. In the past it was used to forward non
4751 persistent connections to an auxiliary load balancer. Due to its simple
4752 syntax, it has also been used for simple TCP relays. It is recommended not to
4753 use it for more clarity, and to use the "server" directive instead.
4754
4755 See also : "server"
4756
Olivier Houchard4e694042017-03-14 20:01:29 +01004757
4758dynamic-cookie-key <string>
4759 Set the dynamic cookie secret key for a backend.
4760 May be used in sections : defaults | frontend | listen | backend
4761 yes | no | yes | yes
4762 Arguments : The secret key to be used.
4763
4764 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01004765 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01004766 specified on the "server" line), using a hash of the IP address of the
4767 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01004768 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01004769 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02004770
Willy Tarreau0ba27502007-12-24 16:55:16 +01004771enabled
4772 Enable a proxy, frontend or backend.
4773 May be used in sections : defaults | frontend | listen | backend
4774 yes | yes | yes | yes
4775 Arguments : none
4776
4777 The "enabled" keyword is used to explicitly enable an instance, when the
4778 defaults has been set to "disabled". This is very rarely used.
4779
4780 See also : "disabled"
4781
4782
4783errorfile <code> <file>
4784 Return a file contents instead of errors generated by HAProxy
4785 May be used in sections : defaults | frontend | listen | backend
4786 yes | yes | yes | yes
4787 Arguments :
4788 <code> is the HTTP status code. Currently, HAProxy is capable of
Christopher Faulet612f2ea2020-05-27 09:57:28 +02004789 generating codes 200, 400, 401, 403, 404, 405, 407, 408, 410,
Christopher Faulete095f312020-12-07 11:22:24 +01004790 413, 425, 429, 500, 501, 502, 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004791
4792 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004793 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01004794 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01004795 error pages, and to use absolute paths, since files are read
4796 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01004797
4798 It is important to understand that this keyword is not meant to rewrite
4799 errors returned by the server, but errors detected and returned by HAProxy.
4800 This is why the list of supported errors is limited to a small set.
4801
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004802 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
4803
Christopher Faulet70170672020-05-18 17:42:48 +02004804 The files are parsed when HAProxy starts and must be valid according to the
4805 HTTP specification. They should not exceed the configured buffer size
4806 (BUFSIZE), which generally is 16 kB, otherwise an internal error will be
4807 returned. It is also wise not to put any reference to local contents
4808 (e.g. images) in order to avoid loops between the client and HAProxy when all
4809 servers are down, causing an error to be returned instead of an
4810 image. Finally, The response cannot exceed (tune.bufsize - tune.maxrewrite)
4811 so that "http-after-response" rules still have room to operate (see
4812 "tune.maxrewrite").
Willy Tarreau59140a22009-02-22 12:02:30 +01004813
Willy Tarreau0ba27502007-12-24 16:55:16 +01004814 The files are read at the same time as the configuration and kept in memory.
4815 For this reason, the errors continue to be returned even when the process is
4816 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01004817 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01004818 403 status code and interrogating a blocked URL.
4819
Christopher Faulet3b967c12020-05-15 15:47:44 +02004820 See also : "http-error", "errorloc", "errorloc302", "errorloc303"
Willy Tarreau0ba27502007-12-24 16:55:16 +01004821
Willy Tarreau59140a22009-02-22 12:02:30 +01004822 Example :
4823 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01004824 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01004825 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
4826 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
4827
Willy Tarreau2769aa02007-12-27 18:26:09 +01004828
Christopher Faulet76edc0f2020-01-13 15:52:01 +01004829errorfiles <name> [<code> ...]
4830 Import, fully or partially, the error files defined in the <name> http-errors
4831 section.
4832 May be used in sections : defaults | frontend | listen | backend
4833 yes | yes | yes | yes
4834 Arguments :
4835 <name> is the name of an existing http-errors section.
4836
4837 <code> is a HTTP status code. Several status code may be listed.
Christopher Faulet612f2ea2020-05-27 09:57:28 +02004838 Currently, HAProxy is capable of generating codes 200, 400, 401,
Christopher Faulete095f312020-12-07 11:22:24 +01004839 403, 404, 405, 407, 408, 410, 413, 425, 429, 500, 501, 502, 503,
4840 and 504.
Christopher Faulet76edc0f2020-01-13 15:52:01 +01004841
4842 Errors defined in the http-errors section with the name <name> are imported
4843 in the current proxy. If no status code is specified, all error files of the
4844 http-errors section are imported. Otherwise, only error files associated to
4845 the listed status code are imported. Those error files override the already
4846 defined custom errors for the proxy. And they may be overridden by following
Daniel Corbett67a82712020-07-06 23:01:19 -04004847 ones. Functionally, it is exactly the same as declaring all error files by
Christopher Faulet76edc0f2020-01-13 15:52:01 +01004848 hand using "errorfile" directives.
4849
Christopher Faulet3b967c12020-05-15 15:47:44 +02004850 See also : "http-error", "errorfile", "errorloc", "errorloc302" ,
4851 "errorloc303" and section 3.8 about http-errors.
Christopher Faulet76edc0f2020-01-13 15:52:01 +01004852
4853 Example :
4854 errorfiles generic
4855 errorfiles site-1 403 404
4856
4857
Willy Tarreau2769aa02007-12-27 18:26:09 +01004858errorloc <code> <url>
4859errorloc302 <code> <url>
4860 Return an HTTP redirection to a URL instead of errors generated by HAProxy
4861 May be used in sections : defaults | frontend | listen | backend
4862 yes | yes | yes | yes
4863 Arguments :
4864 <code> is the HTTP status code. Currently, HAProxy is capable of
Christopher Faulet612f2ea2020-05-27 09:57:28 +02004865 generating codes 200, 400, 401, 403, 404, 405, 407, 408, 410,
Christopher Faulete095f312020-12-07 11:22:24 +01004866 413, 425, 429, 500, 501, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004867
4868 <url> it is the exact contents of the "Location" header. It may contain
4869 either a relative URI to an error page hosted on the same site,
4870 or an absolute URI designating an error page on another site.
4871 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01004872 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01004873
4874 It is important to understand that this keyword is not meant to rewrite
4875 errors returned by the server, but errors detected and returned by HAProxy.
4876 This is why the list of supported errors is limited to a small set.
4877
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004878 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
4879
Willy Tarreau2769aa02007-12-27 18:26:09 +01004880 Note that both keyword return the HTTP 302 status code, which tells the
4881 client to fetch the designated URL using the same HTTP method. This can be
4882 quite problematic in case of non-GET methods such as POST, because the URL
4883 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01004884 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01004885 status code, indicating to the client that the URL must be fetched with a GET
4886 request.
4887
Christopher Faulet3b967c12020-05-15 15:47:44 +02004888 See also : "http-error", "errorfile", "errorloc303"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004889
4890
4891errorloc303 <code> <url>
4892 Return an HTTP redirection to a URL instead of errors generated by HAProxy
4893 May be used in sections : defaults | frontend | listen | backend
4894 yes | yes | yes | yes
4895 Arguments :
4896 <code> is the HTTP status code. Currently, HAProxy is capable of
Christopher Faulet612f2ea2020-05-27 09:57:28 +02004897 generating codes 200, 400, 401, 403, 404, 405, 407, 408, 410,
Christopher Faulete095f312020-12-07 11:22:24 +01004898 413, 425, 429, 500, 501, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004899
4900 <url> it is the exact contents of the "Location" header. It may contain
4901 either a relative URI to an error page hosted on the same site,
4902 or an absolute URI designating an error page on another site.
4903 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01004904 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01004905
4906 It is important to understand that this keyword is not meant to rewrite
4907 errors returned by the server, but errors detected and returned by HAProxy.
4908 This is why the list of supported errors is limited to a small set.
4909
Willy Tarreauae94d4d2011-05-11 16:28:49 +02004910 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
4911
Willy Tarreau2769aa02007-12-27 18:26:09 +01004912 Note that both keyword return the HTTP 303 status code, which tells the
4913 client to fetch the designated URL using the same HTTP GET method. This
4914 solves the usual problems associated with "errorloc" and the 302 code. It is
4915 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004916 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004917
Christopher Faulet3b967c12020-05-15 15:47:44 +02004918 See also : "http-error", "errorfile", "errorloc", "errorloc302"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004919
4920
Simon Horman51a1cf62015-02-03 13:00:44 +09004921email-alert from <emailaddr>
4922 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01004923 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09004924 May be used in sections: defaults | frontend | listen | backend
4925 yes | yes | yes | yes
4926
4927 Arguments :
4928
4929 <emailaddr> is the from email address to use when sending email alerts
4930
4931 Also requires "email-alert mailers" and "email-alert to" to be set
4932 and if so sending email alerts is enabled for the proxy.
4933
Simon Horman64e34162015-02-06 11:11:57 +09004934 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02004935 "email-alert myhostname", "email-alert to", section 3.6 about
4936 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09004937
4938
4939email-alert level <level>
4940 Declare the maximum log level of messages for which email alerts will be
4941 sent. This acts as a filter on the sending of email alerts.
4942 May be used in sections: defaults | frontend | listen | backend
4943 yes | yes | yes | yes
4944
4945 Arguments :
4946
4947 <level> One of the 8 syslog levels:
4948 emerg alert crit err warning notice info debug
4949 The above syslog levels are ordered from lowest to highest.
4950
4951 By default level is alert
4952
4953 Also requires "email-alert from", "email-alert mailers" and
4954 "email-alert to" to be set and if so sending email alerts is enabled
4955 for the proxy.
4956
Simon Horman1421e212015-04-30 13:10:35 +09004957 Alerts are sent when :
4958
4959 * An un-paused server is marked as down and <level> is alert or lower
4960 * A paused server is marked as down and <level> is notice or lower
4961 * A server is marked as up or enters the drain state and <level>
4962 is notice or lower
4963 * "option log-health-checks" is enabled, <level> is info or lower,
4964 and a health check status update occurs
4965
Simon Horman64e34162015-02-06 11:11:57 +09004966 See also : "email-alert from", "email-alert mailers",
4967 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09004968 section 3.6 about mailers.
4969
4970
4971email-alert mailers <mailersect>
4972 Declare the mailers to be used when sending email alerts
4973 May be used in sections: defaults | frontend | listen | backend
4974 yes | yes | yes | yes
4975
4976 Arguments :
4977
4978 <mailersect> is the name of the mailers section to send email alerts.
4979
4980 Also requires "email-alert from" and "email-alert to" to be set
4981 and if so sending email alerts is enabled for the proxy.
4982
Simon Horman64e34162015-02-06 11:11:57 +09004983 See also : "email-alert from", "email-alert level", "email-alert myhostname",
4984 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09004985
4986
4987email-alert myhostname <hostname>
4988 Declare the to hostname address to be used when communicating with
4989 mailers.
4990 May be used in sections: defaults | frontend | listen | backend
4991 yes | yes | yes | yes
4992
4993 Arguments :
4994
Baptiste Assmann738bad92015-12-21 15:27:53 +01004995 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09004996
4997 By default the systems hostname is used.
4998
4999 Also requires "email-alert from", "email-alert mailers" and
5000 "email-alert to" to be set and if so sending email alerts is enabled
5001 for the proxy.
5002
Simon Horman64e34162015-02-06 11:11:57 +09005003 See also : "email-alert from", "email-alert level", "email-alert mailers",
5004 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09005005
5006
5007email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01005008 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09005009 header of email alerts. This is the address that email alerts are sent to.
5010 May be used in sections: defaults | frontend | listen | backend
5011 yes | yes | yes | yes
5012
5013 Arguments :
5014
5015 <emailaddr> is the to email address to use when sending email alerts
5016
5017 Also requires "email-alert mailers" and "email-alert to" to be set
5018 and if so sending email alerts is enabled for the proxy.
5019
Simon Horman64e34162015-02-06 11:11:57 +09005020 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09005021 "email-alert myhostname", section 3.6 about mailers.
5022
5023
Willy Tarreau4de91492010-01-22 19:10:05 +01005024force-persist { if | unless } <condition>
5025 Declare a condition to force persistence on down servers
5026 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005027 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01005028
5029 By default, requests are not dispatched to down servers. It is possible to
5030 force this using "option persist", but it is unconditional and redispatches
5031 to a valid server if "option redispatch" is set. That leaves with very little
5032 possibilities to force some requests to reach a server which is artificially
5033 marked down for maintenance operations.
5034
5035 The "force-persist" statement allows one to declare various ACL-based
5036 conditions which, when met, will cause a request to ignore the down status of
5037 a server and still try to connect to it. That makes it possible to start a
5038 server, still replying an error to the health checks, and run a specially
5039 configured browser to test the service. Among the handy methods, one could
5040 use a specific source IP address, or a specific cookie. The cookie also has
5041 the advantage that it can easily be added/removed on the browser from a test
5042 page. Once the service is validated, it is then possible to open the service
5043 to the world by returning a valid response to health checks.
5044
5045 The forced persistence is enabled when an "if" condition is met, or unless an
5046 "unless" condition is met. The final redispatch is always disabled when this
5047 is used.
5048
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005049 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02005050 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01005051
Christopher Fauletc3fe5332016-04-07 15:30:10 +02005052
5053filter <name> [param*]
5054 Add the filter <name> in the filter list attached to the proxy.
5055 May be used in sections : defaults | frontend | listen | backend
5056 no | yes | yes | yes
5057 Arguments :
5058 <name> is the name of the filter. Officially supported filters are
5059 referenced in section 9.
5060
Tim Düsterhus4896c442016-11-29 02:15:19 +01005061 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02005062 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01005063 filter. Please refer to the documentation of the corresponding
5064 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02005065
5066 Multiple occurrences of the filter line can be used for the same proxy. The
5067 same filter can be referenced many times if needed.
5068
5069 Example:
5070 listen
5071 bind *:80
5072
5073 filter trace name BEFORE-HTTP-COMP
5074 filter compression
5075 filter trace name AFTER-HTTP-COMP
5076
5077 compression algo gzip
5078 compression offload
5079
5080 server srv1 192.168.0.1:80
5081
5082 See also : section 9.
5083
Willy Tarreau4de91492010-01-22 19:10:05 +01005084
Willy Tarreau2769aa02007-12-27 18:26:09 +01005085fullconn <conns>
5086 Specify at what backend load the servers will reach their maxconn
5087 May be used in sections : defaults | frontend | listen | backend
5088 yes | no | yes | yes
5089 Arguments :
5090 <conns> is the number of connections on the backend which will make the
5091 servers use the maximal number of connections.
5092
Willy Tarreau198a7442008-01-17 12:05:32 +01005093 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01005094 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01005095 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01005096 load. The server will then always accept at least <minconn> connections,
5097 never more than <maxconn>, and the limit will be on the ramp between both
5098 values when the backend has less than <conns> concurrent connections. This
5099 makes it possible to limit the load on the servers during normal loads, but
5100 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005101 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005102
Daniel Corbett9f0843f2021-05-08 10:50:37 -04005103 Since it's hard to get this value right, HAProxy automatically sets it to
Willy Tarreaufbb78422011-06-05 15:38:35 +02005104 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01005105 backend (based on "use_backend" and "default_backend" rules). That way it's
5106 safe to leave it unset. However, "use_backend" involving dynamic names are
5107 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02005108
Willy Tarreau2769aa02007-12-27 18:26:09 +01005109 Example :
5110 # The servers will accept between 100 and 1000 concurrent connections each
5111 # and the maximum of 1000 will be reached when the backend reaches 10000
5112 # connections.
5113 backend dynamic
5114 fullconn 10000
5115 server srv1 dyn1:80 minconn 100 maxconn 1000
5116 server srv2 dyn2:80 minconn 100 maxconn 1000
5117
5118 See also : "maxconn", "server"
5119
5120
Willy Tarreauab0a5192020-10-09 19:07:01 +02005121grace <time> (deprecated)
Willy Tarreau2769aa02007-12-27 18:26:09 +01005122 Maintain a proxy operational for some time after a soft stop
5123 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01005124 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01005125 Arguments :
5126 <time> is the time (by default in milliseconds) for which the instance
5127 will remain operational with the frontend sockets still listening
5128 when a soft-stop is received via the SIGUSR1 signal.
5129
5130 This may be used to ensure that the services disappear in a certain order.
5131 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005132 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01005133 needed by the equipment to detect the failure.
5134
5135 Note that currently, there is very little benefit in using this parameter,
5136 and it may in fact complicate the soft-reconfiguration process more than
5137 simplify it.
5138
Willy Tarreau0ba27502007-12-24 16:55:16 +01005139
Andrew Rodland17be45e2016-10-25 17:04:12 -04005140hash-balance-factor <factor>
5141 Specify the balancing factor for bounded-load consistent hashing
5142 May be used in sections : defaults | frontend | listen | backend
5143 yes | no | no | yes
5144 Arguments :
5145 <factor> is the control for the maximum number of concurrent requests to
5146 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01005147 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04005148
5149 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
5150 enables an algorithm that prevents any one server from getting too many
5151 requests at once, even if some hash buckets receive many more requests than
5152 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
5153 <factor> is a percentage greater than 100. For example, if <factor> is 150,
5154 then no server will be allowed to have a load more than 1.5 times the average.
5155 If server weights are used, they will be respected.
5156
5157 If the first-choice server is disqualified, the algorithm will choose another
5158 server based on the request hash, until a server with additional capacity is
5159 found. A higher <factor> allows more imbalance between the servers, while a
5160 lower <factor> means that more servers will be checked on average, affecting
5161 performance. Reasonable values are from 125 to 200.
5162
Willy Tarreau760e81d2018-05-03 07:20:40 +02005163 This setting is also used by "balance random" which internally relies on the
5164 consistent hashing mechanism.
5165
Andrew Rodland17be45e2016-10-25 17:04:12 -04005166 See also : "balance" and "hash-type".
5167
5168
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05005169hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02005170 Specify a method to use for mapping hashes to servers
5171 May be used in sections : defaults | frontend | listen | backend
5172 yes | no | yes | yes
5173 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04005174 <method> is the method used to select a server from the hash computed by
5175 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02005176
Bhaskar98634f02013-10-29 23:30:51 -04005177 map-based the hash table is a static array containing all alive servers.
5178 The hashes will be very smooth, will consider weights, but
5179 will be static in that weight changes while a server is up
5180 will be ignored. This means that there will be no slow start.
5181 Also, since a server is selected by its position in the array,
5182 most mappings are changed when the server count changes. This
5183 means that when a server goes up or down, or when a server is
5184 added to a farm, most connections will be redistributed to
5185 different servers. This can be inconvenient with caches for
5186 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01005187
Bhaskar98634f02013-10-29 23:30:51 -04005188 consistent the hash table is a tree filled with many occurrences of each
5189 server. The hash key is looked up in the tree and the closest
5190 server is chosen. This hash is dynamic, it supports changing
5191 weights while the servers are up, so it is compatible with the
5192 slow start feature. It has the advantage that when a server
5193 goes up or down, only its associations are moved. When a
5194 server is added to the farm, only a few part of the mappings
5195 are redistributed, making it an ideal method for caches.
5196 However, due to its principle, the distribution will never be
5197 very smooth and it may sometimes be necessary to adjust a
5198 server's weight or its ID to get a more balanced distribution.
5199 In order to get the same distribution on multiple load
5200 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05005201 same IDs. Note: consistent hash uses sdbm and avalanche if no
5202 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04005203
5204 <function> is the hash function to be used :
5205
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005206 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04005207 reimplementation of ndbm) database library. It was found to do
5208 well in scrambling bits, causing better distribution of the keys
5209 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05005210 function with good distribution, unless the total server weight
5211 is a multiple of 64, in which case applying the avalanche
5212 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04005213
5214 djb2 this function was first proposed by Dan Bernstein many years ago
5215 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05005216 function provides a better distribution than sdbm. It generally
5217 works well with text-based inputs though it can perform extremely
5218 poorly with numeric-only input or when the total server weight is
5219 a multiple of 33, unless the avalanche modifier is also used.
5220
Daniel Corbett9f0843f2021-05-08 10:50:37 -04005221 wt6 this function was designed for HAProxy while testing other
Willy Tarreaua0f42712013-11-14 14:30:35 +01005222 functions in the past. It is not as smooth as the other ones, but
5223 is much less sensible to the input data set or to the number of
5224 servers. It can make sense as an alternative to sdbm+avalanche or
5225 djb2+avalanche for consistent hashing or when hashing on numeric
5226 data such as a source IP address or a visitor identifier in a URL
5227 parameter.
5228
Willy Tarreau324f07f2015-01-20 19:44:50 +01005229 crc32 this is the most common CRC32 implementation as used in Ethernet,
5230 gzip, PNG, etc. It is slower than the other ones but may provide
5231 a better distribution or less predictable results especially when
5232 used on strings.
5233
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05005234 <modifier> indicates an optional method applied after hashing the key :
5235
5236 avalanche This directive indicates that the result from the hash
5237 function above should not be used in its raw form but that
5238 a 4-byte full avalanche hash must be applied first. The
5239 purpose of this step is to mix the resulting bits from the
5240 previous hash in order to avoid any undesired effect when
5241 the input contains some limited values or when the number of
5242 servers is a multiple of one of the hash's components (64
5243 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
5244 result less predictable, but it's also not as smooth as when
5245 using the original function. Some testing might be needed
5246 with some workloads. This hash is one of the many proposed
5247 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02005248
Bhaskar98634f02013-10-29 23:30:51 -04005249 The default hash type is "map-based" and is recommended for most usages. The
5250 default function is "sdbm", the selection of a function should be based on
5251 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02005252
Andrew Rodland17be45e2016-10-25 17:04:12 -04005253 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02005254
5255
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005256http-after-response <action> <options...> [ { if | unless } <condition> ]
5257 Access control for all Layer 7 responses (server, applet/service and internal
5258 ones).
5259
5260 May be used in sections: defaults | frontend | listen | backend
5261 no | yes | yes | yes
5262
5263 The http-after-response statement defines a set of rules which apply to layer
5264 7 processing. The rules are evaluated in their declaration order when they
5265 are met in a frontend, listen or backend section. Any rule may optionally be
5266 followed by an ACL-based condition, in which case it will only be evaluated
5267 if the condition is true. Since these rules apply on responses, the backend
5268 rules are applied first, followed by the frontend's rules.
5269
5270 Unlike http-response rules, these ones are applied on all responses, the
5271 server ones but also to all responses generated by HAProxy. These rules are
5272 evaluated at the end of the responses analysis, before the data forwarding.
5273
5274 The first keyword is the rule's action. The supported actions are described
5275 below.
5276
5277 There is no limit to the number of http-after-response statements per
5278 instance.
5279
Christopher Fauletd5ac6de2020-12-02 08:40:14 +01005280 Note: Errors emitted in early stage of the request parsing are handled by the
5281 multiplexer at a lower level, before any http analysis. Thus no
5282 http-after-response ruleset is evaluated on these errors.
5283
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005284 Example:
5285 http-after-response set-header Strict-Transport-Security "max-age=31536000"
5286 http-after-response set-header Cache-Control "no-store,no-cache,private"
5287 http-after-response set-header Pragma "no-cache"
5288
5289http-after-response add-header <name> <fmt> [ { if | unless } <condition> ]
5290
5291 This appends an HTTP header field whose name is specified in <name> and whose
5292 value is defined by <fmt> which follows the log-format rules (see Custom Log
5293 Format in section 8.2.4). This may be used to send a cookie to a client for
5294 example, or to pass some internal information.
5295 This rule is not final, so it is possible to add other similar rules.
5296 Note that header addition is performed immediately, so one rule might reuse
5297 the resulting header from a previous rule.
5298
5299http-after-response allow [ { if | unless } <condition> ]
5300
5301 This stops the evaluation of the rules and lets the response pass the check.
5302 No further "http-after-response" rules are evaluated.
5303
Maciej Zdebebdd4c52020-11-20 13:58:48 +00005304http-after-response del-header <name> [ -m <meth> ] [ { if | unless } <condition> ]
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005305
Maciej Zdebebdd4c52020-11-20 13:58:48 +00005306 This removes all HTTP header fields whose name is specified in <name>. <meth>
5307 is the matching method, applied on the header name. Supported matching methods
5308 are "str" (exact match), "beg" (prefix match), "end" (suffix match), "sub"
5309 (substring match) and "reg" (regex match). If not specified, exact matching
5310 method is used.
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005311
5312http-after-response replace-header <name> <regex-match> <replace-fmt>
5313 [ { if | unless } <condition> ]
5314
5315 This works like "http-response replace-header".
5316
5317 Example:
5318 http-after-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
5319
5320 # applied to:
5321 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
5322
5323 # outputs:
5324 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
5325
5326 # assuming the backend IP is 192.168.1.20.
5327
5328http-after-response replace-value <name> <regex-match> <replace-fmt>
5329 [ { if | unless } <condition> ]
5330
5331 This works like "http-response replace-value".
5332
5333 Example:
5334 http-after-response replace-value Cache-control ^public$ private
5335
5336 # applied to:
5337 Cache-Control: max-age=3600, public
5338
5339 # outputs:
5340 Cache-Control: max-age=3600, private
5341
5342http-after-response set-header <name> <fmt> [ { if | unless } <condition> ]
5343
5344 This does the same as "add-header" except that the header name is first
5345 removed if it existed. This is useful when passing security information to
5346 the server, where the header must not be manipulated by external users.
5347
5348http-after-response set-status <status> [reason <str>]
5349 [ { if | unless } <condition> ]
5350
5351 This replaces the response status code with <status> which must be an integer
5352 between 100 and 999. Optionally, a custom reason text can be provided defined
5353 by <str>, or the default reason for the specified code will be used as a
5354 fallback.
5355
5356 Example:
5357 # return "431 Request Header Fields Too Large"
5358 http-response set-status 431
5359 # return "503 Slow Down", custom reason
5360 http-response set-status 503 reason "Slow Down"
5361
5362http-after-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5363
5364 This is used to set the contents of a variable. The variable is declared
5365 inline.
5366
5367 Arguments:
5368 <var-name> The name of the variable starts with an indication about its
5369 scope. The scopes allowed are:
5370 "proc" : the variable is shared with the whole process
5371 "sess" : the variable is shared with the whole session
5372 "txn" : the variable is shared with the transaction
5373 (request and response)
5374 "req" : the variable is shared only during request
5375 processing
5376 "res" : the variable is shared only during response
5377 processing
5378 This prefix is followed by a name. The separator is a '.'.
5379 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5380 and '_'.
5381
5382 <expr> Is a standard HAProxy expression formed by a sample-fetch
5383 followed by some converters.
5384
5385 Example:
5386 http-after-response set-var(sess.last_redir) res.hdr(location)
5387
5388http-after-response strict-mode { on | off }
5389
5390 This enables or disables the strict rewriting mode for following rules. It
5391 does not affect rules declared before it and it is only applicable on rules
5392 performing a rewrite on the responses. When the strict mode is enabled, any
5393 rewrite failure triggers an internal error. Otherwise, such errors are
5394 silently ignored. The purpose of the strict rewriting mode is to make some
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05005395 rewrites optional while others must be performed to continue the response
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005396 processing.
5397
5398 By default, the strict rewriting mode is enabled. Its value is also reset
5399 when a ruleset evaluation ends. So, for instance, if you change the mode on
Daniel Corbett67a82712020-07-06 23:01:19 -04005400 the backend, the default mode is restored when HAProxy starts the frontend
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005401 rules evaluation.
5402
5403http-after-response unset-var(<var-name>) [ { if | unless } <condition> ]
5404
5405 This is used to unset a variable. See "http-after-response set-var" for
5406 details about <var-name>.
5407
5408 Example:
5409 http-after-response unset-var(sess.last_redir)
5410
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005411
5412http-check comment <string>
5413 Defines a comment for the following the http-check rule, reported in logs if
5414 it fails.
5415 May be used in sections : defaults | frontend | listen | backend
5416 yes | no | yes | yes
5417
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005418 Arguments :
5419 <string> is the comment message to add in logs if the following http-check
5420 rule fails.
5421
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005422 It only works for connect, send and expect rules. It is useful to make
5423 user-friendly error reporting.
5424
Daniel Corbett67a82712020-07-06 23:01:19 -04005425 See also : "option httpchk", "http-check connect", "http-check send" and
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005426 "http-check expect".
5427
5428
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005429http-check connect [default] [port <expr>] [addr <ip>] [send-proxy]
5430 [via-socks4] [ssl] [sni <sni>] [alpn <alpn>] [linger]
Christopher Fauletedc6ed92020-04-23 16:27:59 +02005431 [proto <name>] [comment <msg>]
Christopher Faulete5870d82020-04-15 11:32:03 +02005432 Opens a new connection to perform an HTTP health check
5433 May be used in sections : defaults | frontend | listen | backend
5434 yes | no | yes | yes
5435
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005436 Arguments :
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005437 comment <msg> defines a message to report if the rule evaluation fails.
5438
Christopher Faulete5870d82020-04-15 11:32:03 +02005439 default Use default options of the server line to do the health
Daniel Corbett67a82712020-07-06 23:01:19 -04005440 checks. The server options are used only if not redefined.
Christopher Faulete5870d82020-04-15 11:32:03 +02005441
5442 port <expr> if not set, check port or server port is used.
5443 It tells HAProxy where to open the connection to.
5444 <port> must be a valid TCP port source integer, from 1 to
5445 65535 or an sample-fetch expression.
5446
5447 addr <ip> defines the IP address to do the health check.
5448
5449 send-proxy send a PROXY protocol string
5450
5451 via-socks4 enables outgoing health checks using upstream socks4 proxy.
5452
5453 ssl opens a ciphered connection
5454
5455 sni <sni> specifies the SNI to use to do health checks over SSL.
5456
5457 alpn <alpn> defines which protocols to advertise with ALPN. The protocol
5458 list consists in a comma-delimited list of protocol names,
5459 for instance: "h2,http/1.1". If it is not set, the server ALPN
5460 is used.
5461
Christopher Fauletedc6ed92020-04-23 16:27:59 +02005462 proto <name> forces the multiplexer's protocol to use for this connection.
5463 It must be an HTTP mux protocol and it must be usable on the
5464 backend side. The list of available protocols is reported in
5465 haproxy -vv.
5466
Christopher Faulete5870d82020-04-15 11:32:03 +02005467 linger cleanly close the connection instead of using a single RST.
5468
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005469 Just like tcp-check health checks, it is possible to configure the connection
5470 to use to perform HTTP health check. This directive should also be used to
5471 describe a scenario involving several request/response exchanges, possibly on
5472 different ports or with different servers.
5473
5474 When there are no TCP port configured on the server line neither server port
5475 directive, then the first step of the http-check sequence must be to specify
5476 the port with a "http-check connect".
5477
5478 In an http-check ruleset a 'connect' is required, it is also mandatory to start
5479 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
5480 do.
5481
5482 When a connect must start the ruleset, if may still be preceded by set-var,
5483 unset-var or comment rules.
5484
5485 Examples :
Christopher Faulete5870d82020-04-15 11:32:03 +02005486 # check HTTP and HTTPs services on a server.
5487 # first open port 80 thanks to server line port directive, then
5488 # tcp-check opens port 443, ciphered and run a request on it:
5489 option httpchk
5490
5491 http-check connect
Christopher Fauleta5c14ef2020-04-29 14:19:13 +02005492 http-check send meth GET uri / ver HTTP/1.1 hdr host haproxy.1wt.eu
Christopher Faulet8021a5f2020-04-24 13:53:12 +02005493 http-check expect status 200-399
Christopher Faulete5870d82020-04-15 11:32:03 +02005494 http-check connect port 443 ssl sni haproxy.1wt.eu
Christopher Fauleta5c14ef2020-04-29 14:19:13 +02005495 http-check send meth GET uri / ver HTTP/1.1 hdr host haproxy.1wt.eu
Christopher Faulet8021a5f2020-04-24 13:53:12 +02005496 http-check expect status 200-399
Christopher Faulete5870d82020-04-15 11:32:03 +02005497
5498 server www 10.0.0.1 check port 80
5499
5500 See also : "option httpchk", "http-check send", "http-check expect"
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01005501
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005502
Willy Tarreau0ba27502007-12-24 16:55:16 +01005503http-check disable-on-404
5504 Enable a maintenance mode upon HTTP/404 response to health-checks
5505 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01005506 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01005507 Arguments : none
5508
5509 When this option is set, a server which returns an HTTP code 404 will be
5510 excluded from further load-balancing, but will still receive persistent
5511 connections. This provides a very convenient method for Web administrators
5512 to perform a graceful shutdown of their servers. It is also important to note
5513 that a server which is detected as failed while it was in this mode will not
5514 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
5515 will immediately be reinserted into the farm. The status on the stats page
5516 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01005517 option only works in conjunction with the "httpchk" option. If this option
5518 is used with "http-check expect", then it has precedence over it so that 404
Christopher Fauletfa8b89a2020-11-20 18:54:13 +01005519 responses will still be considered as soft-stop. Note also that a stopped
5520 server will stay stopped even if it replies 404s. This option is only
5521 evaluated for running servers.
Willy Tarreaubd741542010-03-16 18:46:54 +01005522
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005523 See also : "option httpchk" and "http-check expect".
Willy Tarreaubd741542010-03-16 18:46:54 +01005524
5525
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005526http-check expect [min-recv <int>] [comment <msg>]
Christopher Faulete5870d82020-04-15 11:32:03 +02005527 [ok-status <st>] [error-status <st>] [tout-status <st>]
5528 [on-success <fmt>] [on-error <fmt>] [status-code <expr>]
5529 [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005530 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01005531 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02005532 yes | no | yes | yes
Christopher Faulete5870d82020-04-15 11:32:03 +02005533
Willy Tarreaubd741542010-03-16 18:46:54 +01005534 Arguments :
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005535 comment <msg> defines a message to report if the rule evaluation fails.
5536
Christopher Faulete5870d82020-04-15 11:32:03 +02005537 min-recv is optional and can define the minimum amount of data required to
5538 evaluate the current expect rule. If the number of received bytes
5539 is under this limit, the check will wait for more data. This
5540 option can be used to resolve some ambiguous matching rules or to
5541 avoid executing costly regex matches on content known to be still
5542 incomplete. If an exact string is used, the minimum between the
5543 string length and this parameter is used. This parameter is
5544 ignored if it is set to -1. If the expect rule does not match,
5545 the check will wait for more data. If set to 0, the evaluation
5546 result is always conclusive.
5547
5548 ok-status <st> is optional and can be used to set the check status if
5549 the expect rule is successfully evaluated and if it is
5550 the last rule in the tcp-check ruleset. "L7OK", "L7OKC",
Christopher Fauletd888f0f2020-05-07 07:40:17 +02005551 "L6OK" and "L4OK" are supported :
5552 - L7OK : check passed on layer 7
Christopher Faulet83662b52020-11-20 17:47:47 +01005553 - L7OKC : check conditionally passed on layer 7, set
5554 server to NOLB state.
Christopher Fauletd888f0f2020-05-07 07:40:17 +02005555 - L6OK : check passed on layer 6
5556 - L4OK : check passed on layer 4
5557 By default "L7OK" is used.
Christopher Faulete5870d82020-04-15 11:32:03 +02005558
5559 error-status <st> is optional and can be used to set the check status if
5560 an error occurred during the expect rule evaluation.
Christopher Faulet83662b52020-11-20 17:47:47 +01005561 "L7OKC", "L7RSP", "L7STS", "L6RSP" and "L4CON" are
5562 supported :
5563 - L7OKC : check conditionally passed on layer 7, set
5564 server to NOLB state.
Christopher Fauletd888f0f2020-05-07 07:40:17 +02005565 - L7RSP : layer 7 invalid response - protocol error
5566 - L7STS : layer 7 response error, for example HTTP 5xx
5567 - L6RSP : layer 6 invalid response - protocol error
5568 - L4CON : layer 1-4 connection problem
5569 By default "L7RSP" is used.
Christopher Faulete5870d82020-04-15 11:32:03 +02005570
5571 tout-status <st> is optional and can be used to set the check status if
5572 a timeout occurred during the expect rule evaluation.
Christopher Fauletd888f0f2020-05-07 07:40:17 +02005573 "L7TOUT", "L6TOUT", and "L4TOUT" are supported :
5574 - L7TOUT : layer 7 (HTTP/SMTP) timeout
5575 - L6TOUT : layer 6 (SSL) timeout
5576 - L4TOUT : layer 1-4 timeout
Christopher Faulete5870d82020-04-15 11:32:03 +02005577 By default "L7TOUT" is used.
5578
5579 on-success <fmt> is optional and can be used to customize the
5580 informational message reported in logs if the expect
5581 rule is successfully evaluated and if it is the last rule
5582 in the tcp-check ruleset. <fmt> is a log-format string.
5583
5584 on-error <fmt> is optional and can be used to customize the
5585 informational message reported in logs if an error
5586 occurred during the expect rule evaluation. <fmt> is a
5587 log-format string.
5588
Willy Tarreaubd741542010-03-16 18:46:54 +01005589 <match> is a keyword indicating how to look for a specific pattern in the
Christopher Fauletb5594262020-05-05 20:23:13 +02005590 response. The keyword may be one of "status", "rstatus", "hdr",
5591 "fhdr", "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01005592 exclamation mark ("!") to negate the match. Spaces are allowed
5593 between the exclamation mark and the keyword. See below for more
5594 details on the supported keywords.
5595
Christopher Faulet39708192020-05-05 10:47:36 +02005596 <pattern> is the pattern to look for. It may be a string, a regular
5597 expression or a more complex pattern with several arguments. If
5598 the string pattern contains spaces, they must be escaped with the
5599 usual backslash ('\').
Willy Tarreaubd741542010-03-16 18:46:54 +01005600
5601 By default, "option httpchk" considers that response statuses 2xx and 3xx
5602 are valid, and that others are invalid. When "http-check expect" is used,
5603 it defines what is considered valid or invalid. Only one "http-check"
5604 statement is supported in a backend. If a server fails to respond or times
5605 out, the check obviously fails. The available matches are :
5606
Christopher Faulet8021a5f2020-04-24 13:53:12 +02005607 status <codes> : test the status codes found parsing <codes> string. it
5608 must be a comma-separated list of status codes or range
5609 codes. A health check response will be considered as
5610 valid if the response's status code matches any status
5611 code or is inside any range of the list. If the "status"
5612 keyword is prefixed with "!", then the response will be
5613 considered invalid if the status code matches.
Willy Tarreaubd741542010-03-16 18:46:54 +01005614
5615 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005616 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01005617 response's status code matches the expression. If the
5618 "rstatus" keyword is prefixed with "!", then the response
5619 will be considered invalid if the status code matches.
5620 This is mostly used to check for multiple codes.
5621
Christopher Fauletb5594262020-05-05 20:23:13 +02005622 hdr { name | name-lf } [ -m <meth> ] <name>
5623 [ { value | value-lf } [ -m <meth> ] <value> :
Christopher Faulet39708192020-05-05 10:47:36 +02005624 test the specified header pattern on the HTTP response
5625 headers. The name pattern is mandatory but the value
5626 pattern is optional. If not specified, only the header
5627 presence is verified. <meth> is the matching method,
5628 applied on the header name or the header value. Supported
5629 matching methods are "str" (exact match), "beg" (prefix
5630 match), "end" (suffix match), "sub" (substring match) or
5631 "reg" (regex match). If not specified, exact matching
Christopher Fauletb5594262020-05-05 20:23:13 +02005632 method is used. If the "name-lf" parameter is used,
5633 <name> is evaluated as a log-format string. If "value-lf"
5634 parameter is used, <value> is evaluated as a log-format
5635 string. These parameters cannot be used with the regex
5636 matching method. Finally, the header value is considered
5637 as comma-separated list. Note that matchings are case
5638 insensitive on the header names.
5639
5640 fhdr { name | name-lf } [ -m <meth> ] <name>
5641 [ { value | value-lf } [ -m <meth> ] <value> :
5642 test the specified full header pattern on the HTTP
5643 response headers. It does exactly the same than "hdr"
5644 keyword, except the full header value is tested, commas
5645 are not considered as delimiters.
Christopher Faulet39708192020-05-05 10:47:36 +02005646
Willy Tarreaubd741542010-03-16 18:46:54 +01005647 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005648 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01005649 response's body contains this exact string. If the
5650 "string" keyword is prefixed with "!", then the response
5651 will be considered invalid if the body contains this
5652 string. This can be used to look for a mandatory word at
5653 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005654 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01005655 trace).
5656
5657 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005658 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01005659 response's body matches this expression. If the "rstring"
5660 keyword is prefixed with "!", then the response will be
5661 considered invalid if the body matches the expression.
5662 This can be used to look for a mandatory word at the end
5663 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01005664 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01005665
Christopher Fauletaaab0832020-05-05 15:54:22 +02005666 string-lf <fmt> : test a log-format string match in the HTTP response body.
5667 A health check response will be considered valid if the
5668 response's body contains the string resulting of the
5669 evaluation of <fmt>, which follows the log-format rules.
5670 If prefixed with "!", then the response will be
5671 considered invalid if the body contains the string.
5672
Willy Tarreaubd741542010-03-16 18:46:54 +01005673 It is important to note that the responses will be limited to a certain size
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +01005674 defined by the global "tune.bufsize" option, which defaults to 16384 bytes.
Willy Tarreaubd741542010-03-16 18:46:54 +01005675 Thus, too large responses may not contain the mandatory pattern when using
5676 "string" or "rstring". If a large response is absolutely required, it is
5677 possible to change the default max size by setting the global variable.
5678 However, it is worth keeping in mind that parsing very large responses can
5679 waste some CPU cycles, especially when regular expressions are used, and that
5680 it is always better to focus the checks on smaller resources.
5681
Christopher Faulete5870d82020-04-15 11:32:03 +02005682 In an http-check ruleset, the last expect rule may be implicit. If no expect
5683 rule is specified after the last "http-check send", an implicit expect rule
5684 is defined to match on 2xx or 3xx status codes. It means this rule is also
5685 defined if there is no "http-check" rule at all, when only "option httpchk"
5686 is set.
Cyril Bonté32602d22015-01-30 00:07:07 +01005687
Willy Tarreaubd741542010-03-16 18:46:54 +01005688 Last, if "http-check expect" is combined with "http-check disable-on-404",
5689 then this last one has precedence when the server responds with 404.
5690
5691 Examples :
5692 # only accept status 200 as valid
Christopher Faulet8021a5f2020-04-24 13:53:12 +02005693 http-check expect status 200,201,300-310
Willy Tarreaubd741542010-03-16 18:46:54 +01005694
Christopher Faulet39708192020-05-05 10:47:36 +02005695 # be sure a sessid coookie is set
5696 http-check expect header name "set-cookie" value -m beg "sessid="
5697
Willy Tarreaubd741542010-03-16 18:46:54 +01005698 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01005699 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01005700
5701 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01005702 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01005703
5704 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005705 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01005706
Christopher Faulete5870d82020-04-15 11:32:03 +02005707 See also : "option httpchk", "http-check connect", "http-check disable-on-404"
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005708 and "http-check send".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005709
5710
Christopher Faulet7c95f5f2020-05-06 15:06:34 +02005711http-check send [meth <method>] [{ uri <uri> | uri-lf <fmt> }>] [ver <version>]
Christopher Faulet574e7bd2020-05-06 15:38:58 +02005712 [hdr <name> <fmt>]* [{ body <string> | body-lf <fmt> }]
5713 [comment <msg>]
Christopher Faulet8acb1282020-04-09 08:44:06 +02005714 Add a possible list of headers and/or a body to the request sent during HTTP
5715 health checks.
5716 May be used in sections : defaults | frontend | listen | backend
5717 yes | no | yes | yes
5718 Arguments :
Christopher Faulet4f5c2e22020-04-23 15:22:33 +02005719 comment <msg> defines a message to report if the rule evaluation fails.
5720
Christopher Faulete5870d82020-04-15 11:32:03 +02005721 meth <method> is the optional HTTP method used with the requests. When not
5722 set, the "OPTIONS" method is used, as it generally requires
5723 low server processing and is easy to filter out from the
5724 logs. Any method may be used, though it is not recommended
5725 to invent non-standard ones.
5726
Christopher Faulet7c95f5f2020-05-06 15:06:34 +02005727 uri <uri> is optional and set the URI referenced in the HTTP requests
5728 to the string <uri>. It defaults to "/" which is accessible
5729 by default on almost any server, but may be changed to any
5730 other URI. Query strings are permitted.
5731
5732 uri-lf <fmt> is optional and set the URI referenced in the HTTP requests
5733 using the log-format string <fmt>. It defaults to "/" which
5734 is accessible by default on almost any server, but may be
5735 changed to any other URI. Query strings are permitted.
Christopher Faulet8acb1282020-04-09 08:44:06 +02005736
Christopher Faulet907701b2020-04-28 09:37:00 +02005737 ver <version> is the optional HTTP version string. It defaults to
Christopher Faulete5870d82020-04-15 11:32:03 +02005738 "HTTP/1.0" but some servers might behave incorrectly in HTTP
Daniel Corbett67a82712020-07-06 23:01:19 -04005739 1.0, so turning it to HTTP/1.1 may sometimes help. Note that
Christopher Faulete5870d82020-04-15 11:32:03 +02005740 the Host field is mandatory in HTTP/1.1, use "hdr" argument
5741 to add it.
5742
5743 hdr <name> <fmt> adds the HTTP header field whose name is specified in
5744 <name> and whose value is defined by <fmt>, which follows
5745 to the log-format rules.
5746
5747 body <string> add the body defined by <string> to the request sent during
5748 HTTP health checks. If defined, the "Content-Length" header
5749 is thus automatically added to the request.
Christopher Faulet8acb1282020-04-09 08:44:06 +02005750
Christopher Faulet574e7bd2020-05-06 15:38:58 +02005751 body-lf <fmt> add the body defined by the log-format string <fmt> to the
5752 request sent during HTTP health checks. If defined, the
5753 "Content-Length" header is thus automatically added to the
5754 request.
5755
Christopher Faulet8acb1282020-04-09 08:44:06 +02005756 In addition to the request line defined by the "option httpchk" directive,
5757 this one is the valid way to add some headers and optionally a body to the
5758 request sent during HTTP health checks. If a body is defined, the associate
Christopher Faulet9df910c2020-04-29 14:20:47 +02005759 "Content-Length" header is automatically added. Thus, this header or
5760 "Transfer-encoding" header should not be present in the request provided by
5761 "http-check send". If so, it will be ignored. The old trick consisting to add
5762 headers after the version string on the "option httpchk" line is now
Amaury Denoyelle6d975f02020-12-22 14:08:52 +01005763 deprecated.
Christopher Faulet8acb1282020-04-09 08:44:06 +02005764
Christopher Faulete5870d82020-04-15 11:32:03 +02005765 Also "http-check send" doesn't support HTTP keep-alive. Keep in mind that it
Amaury Denoyelle6d975f02020-12-22 14:08:52 +01005766 will automatically append a "Connection: close" header, unless a Connection
5767 header has already already been configured via a hdr entry.
Christopher Faulet9df910c2020-04-29 14:20:47 +02005768
5769 Note that the Host header and the request authority, when both defined, are
5770 automatically synchronized. It means when the HTTP request is sent, when a
5771 Host is inserted in the request, the request authority is accordingly
5772 updated. Thus, don't be surprised if the Host header value overwrites the
5773 configured request authority.
5774
5775 Note also for now, no Host header is automatically added in HTTP/1.1 or above
5776 requests. You should add it explicitly.
Christopher Faulete5870d82020-04-15 11:32:03 +02005777
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005778 See also : "option httpchk", "http-check send-state" and "http-check expect".
Christopher Faulet8acb1282020-04-09 08:44:06 +02005779
5780
Willy Tarreauef781042010-01-27 11:53:01 +01005781http-check send-state
5782 Enable emission of a state header with HTTP health checks
5783 May be used in sections : defaults | frontend | listen | backend
5784 yes | no | yes | yes
5785 Arguments : none
5786
Daniel Corbett9f0843f2021-05-08 10:50:37 -04005787 When this option is set, HAProxy will systematically send a special header
Willy Tarreauef781042010-01-27 11:53:01 +01005788 "X-Haproxy-Server-State" with a list of parameters indicating to each server
Daniel Corbett9f0843f2021-05-08 10:50:37 -04005789 how they are seen by HAProxy. This can be used for instance when a server is
5790 manipulated without access to HAProxy and the operator needs to know whether
5791 HAProxy still sees it up or not, or if the server is the last one in a farm.
Willy Tarreauef781042010-01-27 11:53:01 +01005792
5793 The header is composed of fields delimited by semi-colons, the first of which
5794 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
5795 checks on the total number before transition, just as appears in the stats
5796 interface. Next headers are in the form "<variable>=<value>", indicating in
5797 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08005798 - a variable "address", containing the address of the backend server.
5799 This corresponds to the <address> field in the server declaration. For
5800 unix domain sockets, it will read "unix".
5801
5802 - a variable "port", containing the port of the backend server. This
5803 corresponds to the <port> field in the server declaration. For unix
5804 domain sockets, it will read "unix".
5805
Willy Tarreauef781042010-01-27 11:53:01 +01005806 - a variable "name", containing the name of the backend followed by a slash
5807 ("/") then the name of the server. This can be used when a server is
5808 checked in multiple backends.
5809
Daniel Corbett9f0843f2021-05-08 10:50:37 -04005810 - a variable "node" containing the name of the HAProxy node, as set in the
Willy Tarreauef781042010-01-27 11:53:01 +01005811 global "node" variable, otherwise the system's hostname if unspecified.
5812
5813 - a variable "weight" indicating the weight of the server, a slash ("/")
5814 and the total weight of the farm (just counting usable servers). This
5815 helps to know if other servers are available to handle the load when this
5816 one fails.
5817
5818 - a variable "scur" indicating the current number of concurrent connections
5819 on the server, followed by a slash ("/") then the total number of
5820 connections on all servers of the same backend.
5821
5822 - a variable "qcur" indicating the current number of requests in the
5823 server's queue.
5824
5825 Example of a header received by the application server :
5826 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
5827 scur=13/22; qcur=0
5828
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005829 See also : "option httpchk", "http-check disable-on-404" and
5830 "http-check send".
Willy Tarreauef781042010-01-27 11:53:01 +01005831
Christopher Faulete5870d82020-04-15 11:32:03 +02005832
5833http-check set-var(<var-name>) <expr>
Christopher Faulete5870d82020-04-15 11:32:03 +02005834 This operation sets the content of a variable. The variable is declared inline.
Christopher Faulete5870d82020-04-15 11:32:03 +02005835 May be used in sections: defaults | frontend | listen | backend
5836 yes | no | yes | yes
5837
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005838 Arguments :
Christopher Faulete5870d82020-04-15 11:32:03 +02005839 <var-name> The name of the variable starts with an indication about its
5840 scope. The scopes allowed for http-check are:
5841 "proc" : the variable is shared with the whole process.
5842 "sess" : the variable is shared with the tcp-check session.
5843 "check": the variable is declared for the lifetime of the tcp-check.
5844 This prefix is followed by a name. The separator is a '.'.
5845 The name may only contain characters 'a-z', 'A-Z', '0-9', '.',
5846 and '-'.
5847
5848 <expr> Is a sample-fetch expression potentially followed by converters.
5849
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005850 Examples :
5851 http-check set-var(check.port) int(1234)
Christopher Faulete5870d82020-04-15 11:32:03 +02005852
5853
5854http-check unset-var(<var-name>)
Christopher Faulete5870d82020-04-15 11:32:03 +02005855 Free a reference to a variable within its scope.
Christopher Faulete5870d82020-04-15 11:32:03 +02005856 May be used in sections: defaults | frontend | listen | backend
5857 yes | no | yes | yes
5858
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005859 Arguments :
Christopher Faulete5870d82020-04-15 11:32:03 +02005860 <var-name> The name of the variable starts with an indication about its
5861 scope. The scopes allowed for http-check are:
5862 "proc" : the variable is shared with the whole process.
5863 "sess" : the variable is shared with the tcp-check session.
5864 "check": the variable is declared for the lifetime of the tcp-check.
5865 This prefix is followed by a name. The separator is a '.'.
5866 The name may only contain characters 'a-z', 'A-Z', '0-9', '.',
5867 and '-'.
5868
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02005869 Examples :
5870 http-check unset-var(check.port)
Christopher Faulete5870d82020-04-15 11:32:03 +02005871
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005872
Christopher Faulet3b967c12020-05-15 15:47:44 +02005873http-error status <code> [content-type <type>]
5874 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
5875 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
5876 [ hdr <name> <fmt> ]*
5877 Defines a custom error message to use instead of errors generated by HAProxy.
5878 May be used in sections : defaults | frontend | listen | backend
5879 yes | yes | yes | yes
5880 Arguments :
Ilya Shipitsin11057a32020-06-21 21:18:27 +05005881 status <code> is the HTTP status code. It must be specified.
Christopher Faulet3b967c12020-05-15 15:47:44 +02005882 Currently, HAProxy is capable of generating codes
Anthonin Bonnefoy85048f82020-06-22 09:17:01 +02005883 200, 400, 401, 403, 404, 405, 407, 408, 410, 413, 425,
Christopher Faulete095f312020-12-07 11:22:24 +01005884 429, 500, 501, 502, 503, and 504.
Christopher Faulet3b967c12020-05-15 15:47:44 +02005885
5886 content-type <type> is the response content type, for instance
5887 "text/plain". This parameter is ignored and should be
5888 omitted when an errorfile is configured or when the
5889 payload is empty. Otherwise, it must be defined.
5890
5891 default-errorfiles Reset the previously defined error message for current
5892 proxy for the status <code>. If used on a backend, the
5893 frontend error message is used, if defined. If used on
5894 a frontend, the default error message is used.
5895
5896 errorfile <file> designates a file containing the full HTTP response.
5897 It is recommended to follow the common practice of
5898 appending ".http" to the filename so that people do
5899 not confuse the response with HTML error pages, and to
5900 use absolute paths, since files are read before any
5901 chroot is performed.
5902
5903 errorfiles <name> designates the http-errors section to use to import
5904 the error message with the status code <code>. If no
5905 such message is found, the proxy's error messages are
5906 considered.
5907
5908 file <file> specifies the file to use as response payload. If the
5909 file is not empty, its content-type must be set as
5910 argument to "content-type", otherwise, any
5911 "content-type" argument is ignored. <file> is
5912 considered as a raw string.
5913
5914 string <str> specifies the raw string to use as response payload.
5915 The content-type must always be set as argument to
5916 "content-type".
5917
5918 lf-file <file> specifies the file to use as response payload. If the
5919 file is not empty, its content-type must be set as
5920 argument to "content-type", otherwise, any
5921 "content-type" argument is ignored. <file> is
5922 evaluated as a log-format string.
5923
5924 lf-string <str> specifies the log-format string to use as response
5925 payload. The content-type must always be set as
5926 argument to "content-type".
5927
5928 hdr <name> <fmt> adds to the response the HTTP header field whose name
5929 is specified in <name> and whose value is defined by
5930 <fmt>, which follows to the log-format rules.
5931 This parameter is ignored if an errorfile is used.
5932
5933 This directive may be used instead of "errorfile", to define a custom error
5934 message. As "errorfile" directive, it is used for errors detected and
5935 returned by HAProxy. If an errorfile is defined, it is parsed when HAProxy
5936 starts and must be valid according to the HTTP standards. The generated
5937 response must not exceed the configured buffer size (BUFFSIZE), otherwise an
5938 internal error will be returned. Finally, if you consider to use some
5939 http-after-response rules to rewrite these errors, the reserved buffer space
5940 should be available (see "tune.maxrewrite").
5941
5942 The files are read at the same time as the configuration and kept in memory.
5943 For this reason, the errors continue to be returned even when the process is
5944 chrooted, and no file change is considered while the process is running.
5945
Christopher Fauletd5ac6de2020-12-02 08:40:14 +01005946 Note: 400/408/500 errors emitted in early stage of the request parsing are
5947 handled by the multiplexer at a lower level. No custom formatting is
5948 supported at this level. Thus only static error messages, defined with
5949 "errorfile" directive, are supported. However, this limitation only
5950 exists during the request headers parsing or between two transactions.
5951
Christopher Faulet3b967c12020-05-15 15:47:44 +02005952 See also : "errorfile", "errorfiles", "errorloc", "errorloc302",
5953 "errorloc303" and section 3.8 about http-errors.
5954
5955
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005956http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005957 Access control for Layer 7 requests
5958
5959 May be used in sections: defaults | frontend | listen | backend
5960 no | yes | yes | yes
5961
Willy Tarreau20b0de52012-12-24 15:45:22 +01005962 The http-request statement defines a set of rules which apply to layer 7
5963 processing. The rules are evaluated in their declaration order when they are
5964 met in a frontend, listen or backend section. Any rule may optionally be
5965 followed by an ACL-based condition, in which case it will only be evaluated
5966 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005967
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005968 The first keyword is the rule's action. The supported actions are described
5969 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01005970
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005971 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01005972
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005973 Example:
5974 acl nagios src 192.168.129.3
5975 acl local_net src 192.168.0.0/16
5976 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01005977
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005978 http-request allow if nagios
5979 http-request allow if local_net auth_ok
5980 http-request auth realm Gimme if local_net auth_ok
5981 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01005982
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005983 Example:
5984 acl key req.hdr(X-Add-Acl-Key) -m found
5985 acl add path /addacl
5986 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01005987
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005988 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01005989
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005990 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
5991 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02005992
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005993 Example:
5994 acl value req.hdr(X-Value) -m found
5995 acl setmap path /setmap
5996 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06005997
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005998 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06005999
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006000 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
6001 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06006002
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006003 See also : "stats http-request", section 3.4 about userlists and section 7
6004 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06006005
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006006http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06006007
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006008 This is used to add a new entry into an ACL. The ACL must be loaded from a
6009 file (even a dummy empty file). The file name of the ACL to be updated is
6010 passed between parentheses. It takes one argument: <key fmt>, which follows
6011 log-format rules, to collect content of the new entry. It performs a lookup
6012 in the ACL before insertion, to avoid duplicated (or more) values. This
6013 lookup is done by a linear search and can be expensive with large lists!
6014 It is the equivalent of the "add acl" command from the stats socket, but can
6015 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06006016
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006017http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06006018
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006019 This appends an HTTP header field whose name is specified in <name> and
6020 whose value is defined by <fmt> which follows the log-format rules (see
6021 Custom Log Format in section 8.2.4). This is particularly useful to pass
6022 connection-specific information to the server (e.g. the client's SSL
6023 certificate), or to combine several headers into one. This rule is not
6024 final, so it is possible to add other similar rules. Note that header
6025 addition is performed immediately, so one rule might reuse the resulting
6026 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06006027
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006028http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06006029
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006030 This stops the evaluation of the rules and lets the request pass the check.
6031 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06006032
Sasha Pachev218f0642014-06-16 12:05:59 -06006033
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006034http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06006035
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006036 This stops the evaluation of the rules and immediately responds with an
6037 HTTP 401 or 407 error code to invite the user to present a valid user name
6038 and password. No further "http-request" rules are evaluated. An optional
6039 "realm" parameter is supported, it sets the authentication realm that is
6040 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06006041
Christopher Faulet612f2ea2020-05-27 09:57:28 +02006042 The corresponding proxy's error message is used. It may be customized using
6043 an "errorfile" or an "http-error" directive. For 401 responses, all
6044 occurrences of the WWW-Authenticate header are removed and replaced by a new
6045 one with a basic authentication challenge for realm "<realm>". For 407
6046 responses, the same is done on the Proxy-Authenticate header. If the error
6047 message must not be altered, consider to use "http-request return" rule
6048 instead.
6049
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006050 Example:
6051 acl auth_ok http_auth_group(L1) G1
6052 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06006053
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02006054http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06006055
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02006056 See section 6.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006057
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006058http-request capture <sample> [ len <length> | id <id> ]
6059 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006060
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006061 This captures sample expression <sample> from the request buffer, and
6062 converts it to a string of at most <len> characters. The resulting string is
6063 stored into the next request "capture" slot, so it will possibly appear next
6064 to some captured HTTP headers. It will then automatically appear in the logs,
6065 and it will be possible to extract it using sample fetch rules to feed it
6066 into headers or anything. The length should be limited given that this size
6067 will be allocated for each capture during the whole session life.
6068 Please check section 7.3 (Fetching samples) and "capture request header" for
6069 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006070
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006071 If the keyword "id" is used instead of "len", the action tries to store the
6072 captured string in a previously declared capture slot. This is useful to run
6073 captures in backends. The slot id can be declared by a previous directive
Baptiste Assmann19a69b32020-01-16 14:34:22 +01006074 "http-request capture" or with the "declare capture" keyword.
6075
6076 When using this action in a backend, double check that the relevant
6077 frontend(s) have the required capture slots otherwise, this rule will be
6078 ignored at run time. This can't be detected at configuration parsing time
6079 due to HAProxy's ability to dynamically resolve backend name at runtime.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006080
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006081http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006082
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006083 This is used to delete an entry from an ACL. The ACL must be loaded from a
6084 file (even a dummy empty file). The file name of the ACL to be updated is
6085 passed between parentheses. It takes one argument: <key fmt>, which follows
6086 log-format rules, to collect content of the entry to delete.
6087 It is the equivalent of the "del acl" command from the stats socket, but can
6088 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01006089
Maciej Zdebebdd4c52020-11-20 13:58:48 +00006090http-request del-header <name> [ -m <meth> ] [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02006091
Maciej Zdebebdd4c52020-11-20 13:58:48 +00006092 This removes all HTTP header fields whose name is specified in <name>. <meth>
6093 is the matching method, applied on the header name. Supported matching methods
6094 are "str" (exact match), "beg" (prefix match), "end" (suffix match), "sub"
6095 (substring match) and "reg" (regex match). If not specified, exact matching
6096 method is used.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02006097
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006098http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02006099
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006100 This is used to delete an entry from a MAP. The MAP must be loaded from a
6101 file (even a dummy empty file). The file name of the MAP to be updated is
6102 passed between parentheses. It takes one argument: <key fmt>, which follows
6103 log-format rules, to collect content of the entry to delete.
6104 It takes one argument: "file name" It is the equivalent of the "del map"
6105 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02006106
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006107http-request deny [deny_status <status>] [ { if | unless } <condition> ]
6108http-request deny [ { status | deny_status } <code>] [content-type <type>]
6109 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
6110 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
6111 [ hdr <name> <fmt> ]*
6112 [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04006113
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006114 This stops the evaluation of the rules and immediately rejects the request.
6115 By default an HTTP 403 error is returned. But the response may be customized
6116 using same syntax than "http-request return" rules. Thus, see "http-request
Ilya Shipitsin11057a32020-06-21 21:18:27 +05006117 return" for details. For compatibility purpose, when no argument is defined,
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006118 or only "deny_status", the argument "default-errorfiles" is implied. It means
6119 "http-request deny [deny_status <status>]" is an alias of
6120 "http-request deny [status <status>] default-errorfiles".
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006121 No further "http-request" rules are evaluated.
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006122 See also "http-request return".
Patrick Hemmer268a7072018-05-11 12:52:31 -04006123
Olivier Houchard602bf7d2019-05-10 13:59:15 +02006124http-request disable-l7-retry [ { if | unless } <condition> ]
6125 This disables any attempt to retry the request if it fails for any other
6126 reason than a connection failure. This can be useful for example to make
6127 sure POST requests aren't retried on failure.
6128
Baptiste Assmann333939c2019-01-21 08:34:50 +01006129http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
6130
6131 This action performs a DNS resolution of the output of <expr> and stores
6132 the result in the variable <var>. It uses the DNS resolvers section
6133 pointed by <resolvers>.
6134 It is possible to choose a resolution preference using the optional
6135 arguments 'ipv4' or 'ipv6'.
6136 When performing the DNS resolution, the client side connection is on
6137 pause waiting till the end of the resolution.
6138 If an IP address can be found, it is stored into <var>. If any kind of
6139 error occurs, then <var> is not set.
6140 One can use this action to discover a server IP address at run time and
6141 based on information found in the request (IE a Host header).
6142 If this action is used to find the server's IP address (using the
6143 "set-dst" action), then the server IP address in the backend must be set
6144 to 0.0.0.0.
6145
6146 Example:
6147 resolvers mydns
6148 nameserver local 127.0.0.53:53
6149 nameserver google 8.8.8.8:53
6150 timeout retry 1s
6151 hold valid 10s
6152 hold nx 3s
6153 hold other 3s
6154 hold obsolete 0s
6155 accepted_payload_size 8192
6156
6157 frontend fe
6158 bind 10.42.0.1:80
6159 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
6160 http-request capture var(txn.myip) len 40
6161
6162 # return 503 when the variable is not set,
6163 # which mean DNS resolution error
6164 use_backend b_503 unless { var(txn.myip) -m found }
6165
6166 default_backend be
6167
6168 backend b_503
6169 # dummy backend used to return 503.
6170 # one can use the errorfile directive to send a nice
6171 # 503 error page to end users
6172
6173 backend be
6174 # rule to prevent HAProxy from reconnecting to services
6175 # on the local network (forged DNS name used to scan the network)
6176 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
6177 http-request set-dst var(txn.myip)
6178 server clear 0.0.0.0:0
6179
6180 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
6181 be used to scan the network or worst won't loop over itself...
6182
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01006183http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
6184
6185 This is used to build an HTTP 103 Early Hints response prior to any other one.
6186 This appends an HTTP header field to this response whose name is specified in
6187 <name> and whose value is defined by <fmt> which follows the log-format rules
6188 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01006189 to the client some Link headers to preload resources required to render the
6190 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01006191
6192 See RFC 8297 for more information.
6193
Tim Duesterhusd371e992021-04-15 21:45:58 +02006194http-request normalize-uri <normalizer> [ { if | unless } <condition> ]
Tim Duesterhusdec1c362021-05-10 17:28:26 +02006195http-request normalize-uri fragment-encode [ { if | unless } <condition> ]
Tim Duesterhusc9e05ab2021-05-10 17:28:25 +02006196http-request normalize-uri fragment-strip [ { if | unless } <condition> ]
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006197http-request normalize-uri path-merge-slashes [ { if | unless } <condition> ]
Maximilian Maderff3bb8b2021-04-21 00:22:50 +02006198http-request normalize-uri path-strip-dot [ { if | unless } <condition> ]
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006199http-request normalize-uri path-strip-dotdot [ full ] [ { if | unless } <condition> ]
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006200http-request normalize-uri percent-decode-unreserved [ strict ] [ { if | unless } <condition> ]
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006201http-request normalize-uri percent-to-uppercase [ strict ] [ { if | unless } <condition> ]
6202http-request normalize-uri query-sort-by-name [ { if | unless } <condition> ]
Tim Duesterhusd371e992021-04-15 21:45:58 +02006203
Tim Duesterhusb918a4a2021-04-16 23:52:29 +02006204 Performs normalization of the request's URI.
6205
Tim Duesterhus2963fd32021-04-17 00:24:56 +02006206 URI normalization in HAProxy 2.4 is currently available as an experimental
Amaury Denoyellea9e639a2021-05-06 15:50:12 +02006207 technical preview. As such, it requires the global directive
6208 'expose-experimental-directives' first to be able to invoke it. You should be
6209 prepared that the behavior of normalizers might change to fix possible
6210 issues, possibly breaking proper request processing in your infrastructure.
Tim Duesterhus2963fd32021-04-17 00:24:56 +02006211
Tim Duesterhusb918a4a2021-04-16 23:52:29 +02006212 Each normalizer handles a single type of normalization to allow for a
6213 fine-grained selection of the level of normalization that is appropriate for
6214 the supported backend.
6215
6216 As an example the "path-strip-dotdot" normalizer might be useful for a static
6217 fileserver that directly maps the requested URI to the path within the local
6218 filesystem. However it might break routing of an API that expects a specific
6219 number of segments in the path.
6220
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006221 It is important to note that some normalizers might result in unsafe
6222 transformations for broken URIs. It might also be possible that a combination
6223 of normalizers that are safe by themselves results in unsafe transformations
6224 when improperly combined.
6225
6226 As an example the "percent-decode-unreserved" normalizer might result in
6227 unexpected results when a broken URI includes bare percent characters. One
6228 such a broken URI is "/%%36%36" which would be decoded to "/%66" which in
6229 turn is equivalent to "/f". By specifying the "strict" option requests to
6230 such a broken URI would safely be rejected.
6231
Tim Duesterhusb918a4a2021-04-16 23:52:29 +02006232 The following normalizers are available:
Tim Duesterhusd371e992021-04-15 21:45:58 +02006233
Tim Duesterhusdec1c362021-05-10 17:28:26 +02006234 - fragment-encode: Encodes "#" as "%23".
6235
6236 The "fragment-strip" normalizer should be preferred, unless it is known
6237 that broken clients do not correctly encode '#' within the path component.
6238
6239 Example:
6240 - /#foo -> /%23foo
6241
Tim Duesterhusc9e05ab2021-05-10 17:28:25 +02006242 - fragment-strip: Removes the URI's "fragment" component.
6243
6244 According to RFC 3986#3.5 the "fragment" component of an URI should not
6245 be sent, but handled by the User Agent after retrieving a resource.
6246
6247 This normalizer should be applied first to ensure that the fragment is
6248 not interpreted as part of the request's path component.
6249
6250 Example:
6251 - /#foo -> /
6252
Tim Duesterhusd6d33de2021-04-21 21:20:35 +02006253 - path-strip-dot: Removes "/./" segments within the "path" component
6254 (RFC 3986#6.2.2.3).
Maximilian Maderff3bb8b2021-04-21 00:22:50 +02006255
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006256 Segments including percent encoded dots ("%2E") will not be detected. Use
6257 the "percent-decode-unreserved" normalizer first if this is undesired.
6258
Tim Duesterhus7a95f412021-04-21 21:20:33 +02006259 Example:
6260 - /. -> /
6261 - /./bar/ -> /bar/
6262 - /a/./a -> /a/a
6263 - /.well-known/ -> /.well-known/ (no change)
Maximilian Maderff3bb8b2021-04-21 00:22:50 +02006264
Tim Duesterhusd6d33de2021-04-21 21:20:35 +02006265 - path-strip-dotdot: Normalizes "/../" segments within the "path" component
6266 (RFC 3986#6.2.2.3).
6267
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006268 This merges segments that attempt to access the parent directory with
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006269 their preceding segment.
6270
6271 Empty segments do not receive special treatment. Use the "merge-slashes"
6272 normalizer first if this is undesired.
6273
6274 Segments including percent encoded dots ("%2E") will not be detected. Use
6275 the "percent-decode-unreserved" normalizer first if this is undesired.
Tim Duesterhus9982fc22021-04-15 21:45:59 +02006276
6277 Example:
6278 - /foo/../ -> /
6279 - /foo/../bar/ -> /bar/
6280 - /foo/bar/../ -> /foo/
6281 - /../bar/ -> /../bar/
Tim Duesterhus560e1a62021-04-15 21:46:00 +02006282 - /bar/../../ -> /../
Tim Duesterhus9982fc22021-04-15 21:45:59 +02006283 - /foo//../ -> /foo/
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006284 - /foo/%2E%2E/ -> /foo/%2E%2E/
Tim Duesterhus9982fc22021-04-15 21:45:59 +02006285
Tim Duesterhus560e1a62021-04-15 21:46:00 +02006286 If the "full" option is specified then "../" at the beginning will be
6287 removed as well:
6288
6289 Example:
6290 - /../bar/ -> /bar/
6291 - /bar/../../ -> /
6292
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006293 - path-merge-slashes: Merges adjacent slashes within the "path" component
6294 into a single slash.
Tim Duesterhusd371e992021-04-15 21:45:58 +02006295
6296 Example:
6297 - // -> /
6298 - /foo//bar -> /foo/bar
6299
Tim Duesterhus2e4a18e2021-04-21 21:20:36 +02006300 - percent-decode-unreserved: Decodes unreserved percent encoded characters to
6301 their representation as a regular character (RFC 3986#6.2.2.2).
6302
6303 The set of unreserved characters includes all letters, all digits, "-",
6304 ".", "_", and "~".
6305
6306 Example:
6307 - /%61dmin -> /admin
6308 - /foo%3Fbar=baz -> /foo%3Fbar=baz (no change)
6309 - /%%36%36 -> /%66 (unsafe)
6310 - /%ZZ -> /%ZZ
6311
6312 If the "strict" option is specified then invalid sequences will result
6313 in a HTTP 400 Bad Request being returned.
6314
6315 Example:
6316 - /%%36%36 -> HTTP 400
6317 - /%ZZ -> HTTP 400
6318
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006319 - percent-to-uppercase: Uppercases letters within percent-encoded sequences
Tim Duesterhusc315efd2021-04-21 21:20:34 +02006320 (RFC 3986#6.2.2.1).
Tim Duesterhusa4071932021-04-15 21:46:02 +02006321
6322 Example:
6323 - /%6f -> /%6F
6324 - /%zz -> /%zz
6325
6326 If the "strict" option is specified then invalid sequences will result
6327 in a HTTP 400 Bad Request being returned.
6328
6329 Example:
6330 - /%zz -> HTTP 400
6331
Tim Duesterhus5be6ab22021-04-17 11:21:10 +02006332 - query-sort-by-name: Sorts the query string parameters by parameter name.
Tim Duesterhusd7b89be2021-04-15 21:46:01 +02006333 Parameters are assumed to be delimited by '&'. Shorter names sort before
6334 longer names and identical parameter names maintain their relative order.
6335
6336 Example:
6337 - /?c=3&a=1&b=2 -> /?a=1&b=2&c=3
6338 - /?aaa=3&a=1&aa=2 -> /?a=1&aa=2&aaa=3
6339 - /?a=3&b=4&a=1&b=5&a=2 -> /?a=3&a=1&a=2&b=4&b=5
6340
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006341http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006342
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006343 This performs an HTTP redirection based on a redirect rule. This is exactly
6344 the same as the "redirect" statement except that it inserts a redirect rule
6345 which can be processed in the middle of other "http-request" rules and that
6346 these rules use the "log-format" strings. See the "redirect" keyword for the
6347 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006348
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006349http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006350
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006351 This stops the evaluation of the rules and immediately closes the connection
6352 without sending any response. It acts similarly to the
6353 "tcp-request content reject" rules. It can be useful to force an immediate
6354 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006355
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006356http-request replace-header <name> <match-regex> <replace-fmt>
6357 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02006358
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05006359 This matches the value of all occurrences of header field <name> against
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006360 <match-regex>. Matching is performed case-sensitively. Matching values are
6361 completely replaced by <replace-fmt>. Format characters are allowed in
6362 <replace-fmt> and work like <fmt> arguments in "http-request add-header".
6363 Standard back-references using the backslash ('\') followed by a number are
6364 supported.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02006365
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006366 This action acts on whole header lines, regardless of the number of values
6367 they may contain. Thus it is well-suited to process headers naturally
6368 containing commas in their value, such as If-Modified-Since. Headers that
6369 contain a comma-separated list of values, such as Accept, should be processed
6370 using "http-request replace-value".
William Lallemand86d0df02017-11-24 21:36:45 +01006371
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006372 Example:
6373 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
6374
6375 # applied to:
6376 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
6377
6378 # outputs:
6379 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
6380
6381 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02006382
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006383 http-request replace-header User-Agent curl foo
6384
6385 # applied to:
6386 User-Agent: curl/7.47.0
Willy Tarreau09448f72014-06-25 18:12:15 +02006387
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006388 # outputs:
6389 User-Agent: foo
Willy Tarreau09448f72014-06-25 18:12:15 +02006390
Willy Tarreau262c3f12019-12-17 06:52:51 +01006391http-request replace-path <match-regex> <replace-fmt>
6392 [ { if | unless } <condition> ]
6393
6394 This works like "replace-header" except that it works on the request's path
6395 component instead of a header. The path component starts at the first '/'
Christopher Faulet82c83322020-09-02 14:16:59 +02006396 after an optional scheme+authority and ends before the question mark. Thus,
6397 the replacement does not modify the scheme, the authority and the
6398 query-string.
Willy Tarreau262c3f12019-12-17 06:52:51 +01006399
6400 It is worth noting that regular expressions may be more expensive to evaluate
6401 than certain ACLs, so rare replacements may benefit from a condition to avoid
6402 performing the evaluation at all if it does not match.
6403
6404 Example:
6405 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
6406 http-request replace-path (.*) /foo\1
6407
Willy Tarreau262c3f12019-12-17 06:52:51 +01006408 # strip /foo : turn /foo/bar?q=1 into /bar?q=1
6409 http-request replace-path /foo/(.*) /\1
6410 # or more efficient if only some requests match :
6411 http-request replace-path /foo/(.*) /\1 if { url_beg /foo/ }
6412
Christopher Faulet312294f2020-09-02 17:17:44 +02006413http-request replace-pathq <match-regex> <replace-fmt>
6414 [ { if | unless } <condition> ]
6415
6416 This does the same as "http-request replace-path" except that the path
6417 contains the query-string if any is present. Thus, the path and the
6418 query-string are replaced.
6419
6420 Example:
6421 # suffix /foo : turn /bar?q=1 into /bar/foo?q=1 :
6422 http-request replace-pathq ([^?]*)(\?(.*))? \1/foo\2
6423
Willy Tarreau33810222019-06-12 17:44:02 +02006424http-request replace-uri <match-regex> <replace-fmt>
6425 [ { if | unless } <condition> ]
6426
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006427 This works like "replace-header" except that it works on the request's URI part
6428 instead of a header. The URI part may contain an optional scheme, authority or
6429 query string. These are considered to be part of the value that is matched
6430 against.
6431
6432 It is worth noting that regular expressions may be more expensive to evaluate
6433 than certain ACLs, so rare replacements may benefit from a condition to avoid
6434 performing the evaluation at all if it does not match.
Willy Tarreau33810222019-06-12 17:44:02 +02006435
Willy Tarreau62b59132019-12-17 06:51:20 +01006436 IMPORTANT NOTE: historically in HTTP/1.x, the vast majority of requests sent
6437 by browsers use the "origin form", which differs from the "absolute form" in
6438 that they do not contain a scheme nor authority in the URI portion. Mostly
6439 only requests sent to proxies, those forged by hand and some emitted by
6440 certain applications use the absolute form. As such, "replace-uri" usually
6441 works fine most of the time in HTTP/1.x with rules starting with a "/". But
6442 with HTTP/2, clients are encouraged to send absolute URIs only, which look
6443 like the ones HTTP/1 clients use to talk to proxies. Such partial replace-uri
6444 rules may then fail in HTTP/2 when they work in HTTP/1. Either the rules need
Willy Tarreau262c3f12019-12-17 06:52:51 +01006445 to be adapted to optionally match a scheme and authority, or replace-path
6446 should be used.
Willy Tarreau33810222019-06-12 17:44:02 +02006447
Willy Tarreau62b59132019-12-17 06:51:20 +01006448 Example:
6449 # rewrite all "http" absolute requests to "https":
6450 http-request replace-uri ^http://(.*) https://\1
Willy Tarreau33810222019-06-12 17:44:02 +02006451
Willy Tarreau62b59132019-12-17 06:51:20 +01006452 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
6453 http-request replace-uri ([^/:]*://[^/]*)?(.*) \1/foo\2
Willy Tarreau33810222019-06-12 17:44:02 +02006454
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006455http-request replace-value <name> <match-regex> <replace-fmt>
6456 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02006457
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006458 This works like "replace-header" except that it matches the regex against
6459 every comma-delimited value of the header field <name> instead of the
6460 entire header. This is suited for all headers which are allowed to carry
6461 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02006462
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006463 Example:
6464 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02006465
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006466 # applied to:
6467 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02006468
Tim Duesterhus2252beb2019-10-29 00:05:13 +01006469 # outputs:
6470 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01006471
Christopher Faulet24231ab2020-01-24 17:44:23 +01006472http-request return [status <code>] [content-type <type>]
6473 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
6474 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
Christopher Faulet4a2c1422020-01-31 17:36:01 +01006475 [ hdr <name> <fmt> ]*
Christopher Faulet24231ab2020-01-24 17:44:23 +01006476 [ { if | unless } <condition> ]
6477
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05006478 This stops the evaluation of the rules and immediately returns a response. The
Christopher Faulet24231ab2020-01-24 17:44:23 +01006479 default status code used for the response is 200. It can be optionally
6480 specified as an arguments to "status". The response content-type may also be
Daniel Corbett67a82712020-07-06 23:01:19 -04006481 specified as an argument to "content-type". Finally the response itself may
Sébastien Grossab877122020-10-08 10:06:03 +02006482 be defined. It can be a full HTTP response specifying the errorfile to use,
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05006483 or the response payload specifying the file or the string to use. These rules
Christopher Faulet24231ab2020-01-24 17:44:23 +01006484 are followed to create the response :
6485
6486 * If neither the errorfile nor the payload to use is defined, a dummy
6487 response is returned. Only the "status" argument is considered. It can be
6488 any code in the range [200, 599]. The "content-type" argument, if any, is
6489 ignored.
6490
6491 * If "default-errorfiles" argument is set, the proxy's errorfiles are
6492 considered. If the "status" argument is defined, it must be one of the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04006493 status code handled by HAProxy (200, 400, 403, 404, 405, 408, 410, 413,
Christopher Faulete095f312020-12-07 11:22:24 +01006494 425, 429, 500, 501, 502, 503, and 504). The "content-type" argument, if
6495 any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01006496
6497 * If a specific errorfile is defined, with an "errorfile" argument, the
6498 corresponding file, containing a full HTTP response, is returned. Only the
6499 "status" argument is considered. It must be one of the status code handled
Daniel Corbett9f0843f2021-05-08 10:50:37 -04006500 by HAProxy (200, 400, 403, 404, 405, 408, 410, 413, 425, 429, 500, 501,
Christopher Faulete095f312020-12-07 11:22:24 +01006501 502, 503, and 504). The "content-type" argument, if any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01006502
6503 * If an http-errors section is defined, with an "errorfiles" argument, the
6504 corresponding file in the specified http-errors section, containing a full
6505 HTTP response, is returned. Only the "status" argument is considered. It
Daniel Corbett9f0843f2021-05-08 10:50:37 -04006506 must be one of the status code handled by HAProxy (200, 400, 403, 404, 405,
Christopher Faulete095f312020-12-07 11:22:24 +01006507 408, 410, 413, 425, 429, 500, 501, 502, 503, and 504). The "content-type"
Anthonin Bonnefoy85048f82020-06-22 09:17:01 +02006508 argument, if any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01006509
6510 * If a "file" or a "lf-file" argument is specified, the file's content is
6511 used as the response payload. If the file is not empty, its content-type
6512 must be set as argument to "content-type". Otherwise, any "content-type"
6513 argument is ignored. With a "lf-file" argument, the file's content is
6514 evaluated as a log-format string. With a "file" argument, it is considered
6515 as a raw content.
6516
6517 * If a "string" or "lf-string" argument is specified, the defined string is
6518 used as the response payload. The content-type must always be set as
6519 argument to "content-type". With a "lf-string" argument, the string is
6520 evaluated as a log-format string. With a "string" argument, it is
6521 considered as a raw string.
6522
Sébastien Grossab877122020-10-08 10:06:03 +02006523 When the response is not based on an errorfile, it is possible to append HTTP
Christopher Faulet4a2c1422020-01-31 17:36:01 +01006524 header fields to the response using "hdr" arguments. Otherwise, all "hdr"
6525 arguments are ignored. For each one, the header name is specified in <name>
6526 and its value is defined by <fmt> which follows the log-format rules.
6527
Christopher Faulet24231ab2020-01-24 17:44:23 +01006528 Note that the generated response must be smaller than a buffer. And to avoid
6529 any warning, when an errorfile or a raw file is loaded, the buffer space
Sébastien Grossab877122020-10-08 10:06:03 +02006530 reserved for the headers rewriting should also be free.
Christopher Faulet24231ab2020-01-24 17:44:23 +01006531
6532 No further "http-request" rules are evaluated.
6533
6534 Example:
Daniel Corbett67a82712020-07-06 23:01:19 -04006535 http-request return errorfile /etc/haproxy/errorfiles/200.http \
Christopher Faulet24231ab2020-01-24 17:44:23 +01006536 if { path /ping }
6537
6538 http-request return content-type image/x-icon file /var/www/favicon.ico \
6539 if { path /favicon.ico }
6540
6541 http-request return status 403 content-type text/plain \
6542 lf-string "Access denied. IP %[src] is blacklisted." \
6543 if { src -f /etc/haproxy/blacklist.lst }
6544
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006545http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
6546http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02006547
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006548 This actions increments the GPC0 or GPC1 counter according with the sticky
6549 counter designated by <sc-id>. If an error occurs, this action silently fails
6550 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02006551
Cédric Dufour0d7712d2019-11-06 18:38:53 +01006552http-request sc-set-gpt0(<sc-id>) { <int> | <expr> }
6553 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02006554
Cédric Dufour0d7712d2019-11-06 18:38:53 +01006555 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
6556 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
6557 boolean. If an error occurs, this action silently fails and the actions
6558 evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02006559
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006560http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02006561
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006562 This is used to set the destination IP address to the value of specified
6563 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
6564 but provides the correct IP in a HTTP header; or you want to mask the IP for
6565 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
6566 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01006567
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006568 Arguments:
6569 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
6570 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01006571
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006572 Example:
6573 http-request set-dst hdr(x-dst)
6574 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01006575
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006576 When possible, set-dst preserves the original destination port as long as the
6577 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02006578
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006579http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02006580
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006581 This is used to set the destination port address to the value of specified
6582 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
6583 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02006584
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006585 Arguments:
6586 <expr> Is a standard HAProxy expression formed by a sample-fetch
6587 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02006588
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006589 Example:
6590 http-request set-dst-port hdr(x-port)
6591 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02006592
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006593 When possible, set-dst-port preserves the original destination address as
6594 long as the address family supports a port, otherwise it forces the
6595 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02006596
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006597http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02006598
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006599 This does the same as "http-request add-header" except that the header name
6600 is first removed if it existed. This is useful when passing security
6601 information to the server, where the header must not be manipulated by
6602 external users. Note that the new value is computed before the removal so it
6603 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02006604
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006605 Example:
6606 http-request set-header X-Haproxy-Current-Date %T
6607 http-request set-header X-SSL %[ssl_fc]
6608 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
6609 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
6610 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
6611 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
6612 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
6613 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
6614 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02006615
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006616http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02006617
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006618 This is used to change the log level of the current request when a certain
6619 condition is met. Valid levels are the 8 syslog levels (see the "log"
6620 keyword) plus the special level "silent" which disables logging for this
6621 request. This rule is not final so the last matching rule wins. This rule
6622 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02006623
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006624http-request set-map(<file-name>) <key fmt> <value fmt>
6625 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02006626
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006627 This is used to add a new entry into a MAP. The MAP must be loaded from a
6628 file (even a dummy empty file). The file name of the MAP to be updated is
6629 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
6630 log-format rules, used to collect MAP key, and <value fmt>, which follows
6631 log-format rules, used to collect content for the new entry.
6632 It performs a lookup in the MAP before insertion, to avoid duplicated (or
6633 more) values. This lookup is done by a linear search and can be expensive
6634 with large lists! It is the equivalent of the "set map" command from the
6635 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02006636
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006637http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02006638
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006639 This is used to set the Netfilter MARK on all packets sent to the client to
6640 the value passed in <mark> on platforms which support it. This value is an
6641 unsigned 32 bit value which can be matched by netfilter and by the routing
6642 table. It can be expressed both in decimal or hexadecimal format (prefixed by
6643 "0x"). This can be useful to force certain packets to take a different route
6644 (for example a cheaper network path for bulk downloads). This works on Linux
6645 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02006646
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006647http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02006648
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006649 This rewrites the request method with the result of the evaluation of format
6650 string <fmt>. There should be very few valid reasons for having to do so as
6651 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02006652
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006653http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02006654
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006655 This sets the "nice" factor of the current request being processed. It only
6656 has effect against the other requests being processed at the same time.
6657 The default value is 0, unless altered by the "nice" setting on the "bind"
6658 line. The accepted range is -1024..1024. The higher the value, the nicest
6659 the request will be. Lower values will make the request more important than
6660 other ones. This can be useful to improve the speed of some requests, or
6661 lower the priority of non-important requests. Using this setting without
6662 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02006663
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006664http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02006665
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006666 This rewrites the request path with the result of the evaluation of format
6667 string <fmt>. The query string, if any, is left intact. If a scheme and
6668 authority is found before the path, they are left intact as well. If the
6669 request doesn't have a path ("*"), this one is replaced with the format.
6670 This can be used to prepend a directory component in front of a path for
6671 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02006672
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006673 Example :
6674 # prepend the host name before the path
6675 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02006676
Christopher Faulet312294f2020-09-02 17:17:44 +02006677http-request set-pathq <fmt> [ { if | unless } <condition> ]
6678
6679 This does the same as "http-request set-path" except that the query-string is
6680 also rewritten. It may be used to remove the query-string, including the
6681 question mark (it is not possible using "http-request set-query").
6682
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006683http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02006684
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006685 This is used to set the queue priority class of the current request.
6686 The value must be a sample expression which converts to an integer in the
6687 range -2047..2047. Results outside this range will be truncated.
6688 The priority class determines the order in which queued requests are
6689 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02006690
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006691http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02006692
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006693 This is used to set the queue priority timestamp offset of the current
6694 request. The value must be a sample expression which converts to an integer
6695 in the range -524287..524287. Results outside this range will be truncated.
6696 When a request is queued, it is ordered first by the priority class, then by
6697 the current timestamp adjusted by the given offset in milliseconds. Lower
6698 values have higher priority.
6699 Note that the resulting timestamp is is only tracked with enough precision
6700 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
6701 adjusted timestamp exceeds this value, it will be misidentified as highest
6702 priority. Thus it is important to set "timeout queue" to a value, where when
6703 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02006704
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006705http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01006706
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006707 This rewrites the request's query string which appears after the first
6708 question mark ("?") with the result of the evaluation of format string <fmt>.
6709 The part prior to the question mark is left intact. If the request doesn't
6710 contain a question mark and the new value is not empty, then one is added at
6711 the end of the URI, followed by the new value. If a question mark was
6712 present, it will never be removed even if the value is empty. This can be
6713 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08006714
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006715 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01006716
6717 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006718 # replace "%3D" with "=" in the query string
6719 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01006720
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006721http-request set-src <expr> [ { if | unless } <condition> ]
6722 This is used to set the source IP address to the value of specified
6723 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
6724 provides the correct IP in a HTTP header; or you want to mask source IP for
Olivier Doucet56e31202020-04-21 09:32:56 +02006725 privacy. All subsequent calls to "src" fetch will return this value
6726 (see example).
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006727
6728 Arguments :
6729 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
6730 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01006731
Olivier Doucet56e31202020-04-21 09:32:56 +02006732 See also "option forwardfor".
6733
Cyril Bonté78caf842010-03-10 22:41:43 +01006734 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006735 http-request set-src hdr(x-forwarded-for)
6736 http-request set-src src,ipmask(24)
6737
Olivier Doucet56e31202020-04-21 09:32:56 +02006738 # After the masking this will track connections
6739 # based on the IP address with the last byte zeroed out.
6740 http-request track-sc0 src
6741
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006742 When possible, set-src preserves the original source port as long as the
6743 address family allows it, otherwise the source port is set to 0.
6744
6745http-request set-src-port <expr> [ { if | unless } <condition> ]
6746
6747 This is used to set the source port address to the value of specified
6748 expression.
6749
6750 Arguments:
6751 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
6752 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01006753
Willy Tarreau20b0de52012-12-24 15:45:22 +01006754 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006755 http-request set-src-port hdr(x-port)
6756 http-request set-src-port int(4000)
6757
6758 When possible, set-src-port preserves the original source address as long as
6759 the address family supports a port, otherwise it forces the source address to
6760 IPv4 "0.0.0.0" before rewriting the port.
6761
Alex59c53352021-04-27 12:57:07 +02006762http-request set-timeout { server | tunnel } { <timeout> | <expr> }
Amaury Denoyelle8d228232020-12-10 13:43:54 +01006763 [ { if | unless } <condition> ]
6764
6765 This action overrides the specified "server" or "tunnel" timeout for the
6766 current stream only. The timeout can be specified in millisecond or with any
6767 other unit if the number is suffixed by the unit as explained at the top of
6768 this document. It is also possible to write an expression which must returns
6769 a number interpreted as a timeout in millisecond.
6770
6771 Note that the server/tunnel timeouts are only relevant on the backend side
6772 and thus this rule is only available for the proxies with backend
6773 capabilities. Also the timeout value must be non-null to obtain the expected
6774 results.
6775
6776 Example:
Alex59c53352021-04-27 12:57:07 +02006777 http-request set-timeout tunnel 5s
6778 http-request set-timeout server req.hdr(host),map_int(host.lst)
Amaury Denoyelle8d228232020-12-10 13:43:54 +01006779
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006780http-request set-tos <tos> [ { if | unless } <condition> ]
6781
6782 This is used to set the TOS or DSCP field value of packets sent to the client
6783 to the value passed in <tos> on platforms which support this. This value
6784 represents the whole 8 bits of the IP TOS field, and can be expressed both in
6785 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
6786 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
6787 be used to adjust some routing behavior on border routers based on some
6788 information from the request.
6789
6790 See RFC 2474, 2597, 3260 and 4594 for more information.
6791
6792http-request set-uri <fmt> [ { if | unless } <condition> ]
6793
6794 This rewrites the request URI with the result of the evaluation of format
6795 string <fmt>. The scheme, authority, path and query string are all replaced
6796 at once. This can be used to rewrite hosts in front of proxies, or to
6797 perform complex modifications to the URI such as moving parts between the
6798 path and the query string.
6799 See also "http-request set-path" and "http-request set-query".
6800
6801http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
6802
6803 This is used to set the contents of a variable. The variable is declared
6804 inline.
6805
6806 Arguments:
6807 <var-name> The name of the variable starts with an indication about its
6808 scope. The scopes allowed are:
6809 "proc" : the variable is shared with the whole process
6810 "sess" : the variable is shared with the whole session
6811 "txn" : the variable is shared with the transaction
6812 (request and response)
6813 "req" : the variable is shared only during request
6814 processing
6815 "res" : the variable is shared only during response
6816 processing
6817 This prefix is followed by a name. The separator is a '.'.
6818 The name may only contain characters 'a-z', 'A-Z', '0-9'
6819 and '_'.
6820
6821 <expr> Is a standard HAProxy expression formed by a sample-fetch
6822 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01006823
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006824 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006825 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006826
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006827http-request send-spoe-group <engine-name> <group-name>
6828 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006829
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006830 This action is used to trigger sending of a group of SPOE messages. To do so,
6831 the SPOE engine used to send messages must be defined, as well as the SPOE
6832 group to send. Of course, the SPOE engine must refer to an existing SPOE
6833 filter. If not engine name is provided on the SPOE filter line, the SPOE
6834 agent name must be used.
6835
6836 Arguments:
6837 <engine-name> The SPOE engine name.
6838
6839 <group-name> The SPOE group name as specified in the engine
6840 configuration.
6841
6842http-request silent-drop [ { if | unless } <condition> ]
6843
6844 This stops the evaluation of the rules and makes the client-facing connection
6845 suddenly disappear using a system-dependent way that tries to prevent the
6846 client from being notified. The effect it then that the client still sees an
6847 established connection while there's none on HAProxy. The purpose is to
6848 achieve a comparable effect to "tarpit" except that it doesn't use any local
6849 resource at all on the machine running HAProxy. It can resist much higher
6850 loads than "tarpit", and slow down stronger attackers. It is important to
6851 understand the impact of using this mechanism. All stateful equipment placed
6852 between the client and HAProxy (firewalls, proxies, load balancers) will also
6853 keep the established connection for a long time and may suffer from this
6854 action.
6855 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
6856 option is used to block the emission of a TCP reset. On other systems, the
6857 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
6858 router, though it's still delivered to local networks. Do not use it unless
6859 you fully understand how it works.
6860
Christopher Faulet46f95542019-12-20 10:07:22 +01006861http-request strict-mode { on | off }
6862
6863 This enables or disables the strict rewriting mode for following rules. It
6864 does not affect rules declared before it and it is only applicable on rules
6865 performing a rewrite on the requests. When the strict mode is enabled, any
6866 rewrite failure triggers an internal error. Otherwise, such errors are
6867 silently ignored. The purpose of the strict rewriting mode is to make some
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05006868 rewrites optional while others must be performed to continue the request
Christopher Faulet46f95542019-12-20 10:07:22 +01006869 processing.
6870
Christopher Faulet1aea50e2020-01-17 16:03:53 +01006871 By default, the strict rewriting mode is enabled. Its value is also reset
Christopher Faulet46f95542019-12-20 10:07:22 +01006872 when a ruleset evaluation ends. So, for instance, if you change the mode on
6873 the frontend, the default mode is restored when HAProxy starts the backend
6874 rules evaluation.
6875
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006876http-request tarpit [deny_status <status>] [ { if | unless } <condition> ]
6877http-request tarpit [ { status | deny_status } <code>] [content-type <type>]
6878 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
6879 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
6880 [ hdr <name> <fmt> ]*
6881 [ { if | unless } <condition> ]
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006882
6883 This stops the evaluation of the rules and immediately blocks the request
6884 without responding for a delay specified by "timeout tarpit" or
6885 "timeout connect" if the former is not set. After that delay, if the client
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006886 is still connected, a response is returned so that the client does not
6887 suspect it has been tarpitted. Logs will report the flags "PT". The goal of
6888 the tarpit rule is to slow down robots during an attack when they're limited
6889 on the number of concurrent requests. It can be very efficient against very
6890 dumb robots, and will significantly reduce the load on firewalls compared to
6891 a "deny" rule. But when facing "correctly" developed robots, it can make
Daniel Corbett9f0843f2021-05-08 10:50:37 -04006892 things worse by forcing HAProxy and the front firewall to support insane
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006893 number of concurrent connections. By default an HTTP error 500 is returned.
6894 But the response may be customized using same syntax than
6895 "http-request return" rules. Thus, see "http-request return" for details.
Ilya Shipitsin11057a32020-06-21 21:18:27 +05006896 For compatibility purpose, when no argument is defined, or only "deny_status",
Christopher Faulet5cb513a2020-05-13 17:56:56 +02006897 the argument "default-errorfiles" is implied. It means
6898 "http-request tarpit [deny_status <status>]" is an alias of
6899 "http-request tarpit [status <status>] default-errorfiles".
6900 No further "http-request" rules are evaluated.
6901 See also "http-request return" and "http-request silent-drop".
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006902
6903http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
6904http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
6905http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
6906
6907 This enables tracking of sticky counters from current request. These rules do
6908 not stop evaluation and do not change default action. The number of counters
6909 that may be simultaneously tracked by the same connection is set in
6910 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
Matteo Contrini1857b8c2020-10-16 17:35:54 +02006911 so the track-sc number is between 0 and (MAX_SESS_STKCTR-1). The first
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006912 "track-sc0" rule executed enables tracking of the counters of the specified
6913 table as the first set. The first "track-sc1" rule executed enables tracking
6914 of the counters of the specified table as the second set. The first
6915 "track-sc2" rule executed enables tracking of the counters of the specified
6916 table as the third set. It is a recommended practice to use the first set of
6917 counters for the per-frontend counters and the second set for the per-backend
6918 ones. But this is just a guideline, all may be used everywhere.
6919
6920 Arguments :
6921 <key> is mandatory, and is a sample expression rule as described in
6922 section 7.3. It describes what elements of the incoming request or
6923 connection will be analyzed, extracted, combined, and used to
6924 select which table entry to update the counters.
6925
6926 <table> is an optional table to be used instead of the default one, which
6927 is the stick-table declared in the current proxy. All the counters
6928 for the matches and updates for the key will then be performed in
6929 that table until the session ends.
6930
6931 Once a "track-sc*" rule is executed, the key is looked up in the table and if
6932 it is not found, an entry is allocated for it. Then a pointer to that entry
6933 is kept during all the session's life, and this entry's counters are updated
6934 as often as possible, every time the session's counters are updated, and also
6935 systematically when the session ends. Counters are only updated for events
6936 that happen after the tracking has been started. As an exception, connection
6937 counters and request counters are systematically updated so that they reflect
6938 useful information.
6939
6940 If the entry tracks concurrent connection counters, one connection is counted
6941 for as long as the entry is tracked, and the entry will not expire during
6942 that time. Tracking counters also provides a performance advantage over just
6943 checking the keys, because only one table lookup is performed for all ACL
6944 checks that make use of it.
6945
6946http-request unset-var(<var-name>) [ { if | unless } <condition> ]
6947
6948 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006949
6950 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006951 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006952
Christopher Faulet579d83b2019-11-22 15:34:17 +01006953http-request use-service <service-name> [ { if | unless } <condition> ]
6954
6955 This directive executes the configured HTTP service to reply to the request
6956 and stops the evaluation of the rules. An HTTP service may choose to reply by
6957 sending any valid HTTP response or it may immediately close the connection
6958 without sending any response. Outside natives services, for instance the
6959 Prometheus exporter, it is possible to write your own services in Lua. No
6960 further "http-request" rules are evaluated.
6961
6962 Arguments :
6963 <service-name> is mandatory. It is the service to call
6964
6965 Example:
6966 http-request use-service prometheus-exporter if { path /metrics }
6967
Christopher Faulet021a8e42021-03-29 10:46:38 +02006968http-request wait-for-body time <time> [ at-least <bytes> ]
6969 [ { if | unless } <condition> ]
6970
6971 This will delay the processing of the request waiting for the payload for at
6972 most <time> milliseconds. if "at-least" argument is specified, HAProxy stops
6973 to wait the payload when the first <bytes> bytes are received. 0 means no
6974 limit, it is the default value. Regardless the "at-least" argument value,
6975 HAProxy stops to wait if the whole payload is received or if the request
6976 buffer is full. This action may be used as a replacement to "option
6977 http-buffer-request".
6978
6979 Arguments :
6980
6981 <time> is mandatory. It is the maximum time to wait for the body. It
6982 follows the HAProxy time format and is expressed in milliseconds.
6983
6984 <bytes> is optional. It is the minimum payload size to receive to stop to
Ilya Shipitsinb2be9a12021-04-24 13:25:42 +05006985 wait. It follows the HAProxy size format and is expressed in
Christopher Faulet021a8e42021-03-29 10:46:38 +02006986 bytes.
6987
6988 Example:
6989 http-request wait-for-body time 1s at-least 1k if METH_POST
6990
6991 See also : "option http-buffer-request"
6992
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006993http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006994
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02006995 This will delay the processing of the request until the SSL handshake
6996 happened. This is mostly useful to delay processing early data until we're
6997 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02006998
Willy Tarreauef781042010-01-27 11:53:01 +01006999
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007000http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007001 Access control for Layer 7 responses
7002
7003 May be used in sections: defaults | frontend | listen | backend
7004 no | yes | yes | yes
7005
7006 The http-response statement defines a set of rules which apply to layer 7
7007 processing. The rules are evaluated in their declaration order when they are
7008 met in a frontend, listen or backend section. Any rule may optionally be
7009 followed by an ACL-based condition, in which case it will only be evaluated
7010 if the condition is true. Since these rules apply on responses, the backend
7011 rules are applied first, followed by the frontend's rules.
7012
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007013 The first keyword is the rule's action. The supported actions are described
7014 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007015
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007016 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007017
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007018 Example:
7019 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02007020
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007021 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06007022
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007023 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
7024 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06007025
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007026 Example:
7027 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06007028
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007029 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06007030
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007031 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
7032 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06007033
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007034 See also : "http-request", section 3.4 about userlists and section 7 about
7035 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06007036
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007037http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06007038
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007039 This is used to add a new entry into an ACL. The ACL must be loaded from a
7040 file (even a dummy empty file). The file name of the ACL to be updated is
7041 passed between parentheses. It takes one argument: <key fmt>, which follows
7042 log-format rules, to collect content of the new entry. It performs a lookup
7043 in the ACL before insertion, to avoid duplicated (or more) values.
7044 This lookup is done by a linear search and can be expensive with large lists!
7045 It is the equivalent of the "add acl" command from the stats socket, but can
7046 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06007047
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007048http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06007049
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007050 This appends an HTTP header field whose name is specified in <name> and whose
7051 value is defined by <fmt> which follows the log-format rules (see Custom Log
7052 Format in section 8.2.4). This may be used to send a cookie to a client for
7053 example, or to pass some internal information.
7054 This rule is not final, so it is possible to add other similar rules.
7055 Note that header addition is performed immediately, so one rule might reuse
7056 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06007057
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007058http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06007059
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007060 This stops the evaluation of the rules and lets the response pass the check.
7061 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06007062
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02007063http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06007064
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02007065 See section 6.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06007066
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007067http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06007068
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007069 This captures sample expression <sample> from the response buffer, and
7070 converts it to a string. The resulting string is stored into the next request
7071 "capture" slot, so it will possibly appear next to some captured HTTP
7072 headers. It will then automatically appear in the logs, and it will be
7073 possible to extract it using sample fetch rules to feed it into headers or
7074 anything. Please check section 7.3 (Fetching samples) and
7075 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02007076
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007077 The keyword "id" is the id of the capture slot which is used for storing the
7078 string. The capture slot must be defined in an associated frontend.
7079 This is useful to run captures in backends. The slot id can be declared by a
7080 previous directive "http-response capture" or with the "declare capture"
7081 keyword.
Baptiste Assmann19a69b32020-01-16 14:34:22 +01007082
7083 When using this action in a backend, double check that the relevant
7084 frontend(s) have the required capture slots otherwise, this rule will be
7085 ignored at run time. This can't be detected at configuration parsing time
7086 due to HAProxy's ability to dynamically resolve backend name at runtime.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02007087
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007088http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02007089
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007090 This is used to delete an entry from an ACL. The ACL must be loaded from a
7091 file (even a dummy empty file). The file name of the ACL to be updated is
7092 passed between parentheses. It takes one argument: <key fmt>, which follows
7093 log-format rules, to collect content of the entry to delete.
7094 It is the equivalent of the "del acl" command from the stats socket, but can
7095 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02007096
Maciej Zdebebdd4c52020-11-20 13:58:48 +00007097http-response del-header <name> [ -m <meth> ] [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02007098
Maciej Zdebebdd4c52020-11-20 13:58:48 +00007099 This removes all HTTP header fields whose name is specified in <name>. <meth>
7100 is the matching method, applied on the header name. Supported matching methods
7101 are "str" (exact match), "beg" (prefix match), "end" (suffix match), "sub"
7102 (substring match) and "reg" (regex match). If not specified, exact matching
7103 method is used.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02007104
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007105http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02007106
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007107 This is used to delete an entry from a MAP. The MAP must be loaded from a
7108 file (even a dummy empty file). The file name of the MAP to be updated is
7109 passed between parentheses. It takes one argument: <key fmt>, which follows
7110 log-format rules, to collect content of the entry to delete.
7111 It takes one argument: "file name" It is the equivalent of the "del map"
7112 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007113
Christopher Faulet5cb513a2020-05-13 17:56:56 +02007114http-response deny [deny_status <status>] [ { if | unless } <condition> ]
7115http-response deny [ { status | deny_status } <code>] [content-type <type>]
7116 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
7117 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
7118 [ hdr <name> <fmt> ]*
7119 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007120
Christopher Faulet5cb513a2020-05-13 17:56:56 +02007121 This stops the evaluation of the rules and immediately rejects the response.
7122 By default an HTTP 502 error is returned. But the response may be customized
7123 using same syntax than "http-response return" rules. Thus, see
Ilya Shipitsin11057a32020-06-21 21:18:27 +05007124 "http-response return" for details. For compatibility purpose, when no
Christopher Faulet5cb513a2020-05-13 17:56:56 +02007125 argument is defined, or only "deny_status", the argument "default-errorfiles"
7126 is implied. It means "http-response deny [deny_status <status>]" is an alias
7127 of "http-response deny [status <status>] default-errorfiles".
Christopher Faulet040c8cd2020-01-13 16:43:45 +01007128 No further "http-response" rules are evaluated.
Christopher Faulet5cb513a2020-05-13 17:56:56 +02007129 See also "http-response return".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007130
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007131http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007132
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007133 This performs an HTTP redirection based on a redirect rule.
7134 This supports a format string similarly to "http-request redirect" rules,
7135 with the exception that only the "location" type of redirect is possible on
7136 the response. See the "redirect" keyword for the rule's syntax. When a
7137 redirect rule is applied during a response, connections to the server are
7138 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02007139
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007140http-response replace-header <name> <regex-match> <replace-fmt>
7141 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02007142
Tim Duesterhus2252beb2019-10-29 00:05:13 +01007143 This works like "http-request replace-header" except that it works on the
7144 server's response instead of the client's request.
William Lallemand86d0df02017-11-24 21:36:45 +01007145
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007146 Example:
7147 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02007148
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007149 # applied to:
7150 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02007151
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007152 # outputs:
7153 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02007154
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007155 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02007156
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007157http-response replace-value <name> <regex-match> <replace-fmt>
7158 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02007159
Tim Duesterhus6bd909b2020-01-17 15:53:18 +01007160 This works like "http-request replace-value" except that it works on the
Tim Duesterhus2252beb2019-10-29 00:05:13 +01007161 server's response instead of the client's request.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02007162
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007163 Example:
7164 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01007165
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007166 # applied to:
7167 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01007168
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007169 # outputs:
7170 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01007171
Christopher Faulet24231ab2020-01-24 17:44:23 +01007172http-response return [status <code>] [content-type <type>]
7173 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
7174 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
Christopher Faulet4a2c1422020-01-31 17:36:01 +01007175 [ hdr <name> <value> ]*
Christopher Faulet24231ab2020-01-24 17:44:23 +01007176 [ { if | unless } <condition> ]
7177
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05007178 This stops the evaluation of the rules and immediately returns a response. The
Christopher Faulet24231ab2020-01-24 17:44:23 +01007179 default status code used for the response is 200. It can be optionally
7180 specified as an arguments to "status". The response content-type may also be
Daniel Corbett67a82712020-07-06 23:01:19 -04007181 specified as an argument to "content-type". Finally the response itself may
Christopher Faulet24231ab2020-01-24 17:44:23 +01007182 be defined. If can be a full HTTP response specifying the errorfile to use,
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05007183 or the response payload specifying the file or the string to use. These rules
Christopher Faulet24231ab2020-01-24 17:44:23 +01007184 are followed to create the response :
7185
7186 * If neither the errorfile nor the payload to use is defined, a dummy
7187 response is returned. Only the "status" argument is considered. It can be
7188 any code in the range [200, 599]. The "content-type" argument, if any, is
7189 ignored.
7190
7191 * If "default-errorfiles" argument is set, the proxy's errorfiles are
7192 considered. If the "status" argument is defined, it must be one of the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007193 status code handled by HAProxy (200, 400, 403, 404, 405, 408, 410, 413,
Christopher Faulete095f312020-12-07 11:22:24 +01007194 425, 429, 500, 501, 502, 503, and 504). The "content-type" argument, if
7195 any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01007196
7197 * If a specific errorfile is defined, with an "errorfile" argument, the
7198 corresponding file, containing a full HTTP response, is returned. Only the
7199 "status" argument is considered. It must be one of the status code handled
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007200 by HAProxy (200, 400, 403, 404, 405, 408, 410, 413, 425, 429, 500, 501,
Christopher Faulete095f312020-12-07 11:22:24 +01007201 502, 503, and 504). The "content-type" argument, if any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01007202
7203 * If an http-errors section is defined, with an "errorfiles" argument, the
7204 corresponding file in the specified http-errors section, containing a full
7205 HTTP response, is returned. Only the "status" argument is considered. It
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007206 must be one of the status code handled by HAProxy (200, 400, 403, 404, 405,
Christopher Faulete095f312020-12-07 11:22:24 +01007207 408, 410, 413, 425, 429, 500, 501, 502, 503, and 504). The "content-type"
Anthonin Bonnefoy85048f82020-06-22 09:17:01 +02007208 argument, if any, is ignored.
Christopher Faulet24231ab2020-01-24 17:44:23 +01007209
7210 * If a "file" or a "lf-file" argument is specified, the file's content is
7211 used as the response payload. If the file is not empty, its content-type
7212 must be set as argument to "content-type". Otherwise, any "content-type"
7213 argument is ignored. With a "lf-file" argument, the file's content is
7214 evaluated as a log-format string. With a "file" argument, it is considered
7215 as a raw content.
7216
7217 * If a "string" or "lf-string" argument is specified, the defined string is
7218 used as the response payload. The content-type must always be set as
7219 argument to "content-type". With a "lf-string" argument, the string is
7220 evaluated as a log-format string. With a "string" argument, it is
7221 considered as a raw string.
7222
Christopher Faulet4a2c1422020-01-31 17:36:01 +01007223 When the response is not based an errorfile, it is possible to appends HTTP
7224 header fields to the response using "hdr" arguments. Otherwise, all "hdr"
7225 arguments are ignored. For each one, the header name is specified in <name>
7226 and its value is defined by <fmt> which follows the log-format rules.
7227
Christopher Faulet24231ab2020-01-24 17:44:23 +01007228 Note that the generated response must be smaller than a buffer. And to avoid
7229 any warning, when an errorfile or a raw file is loaded, the buffer space
Ilya Shipitsin11057a32020-06-21 21:18:27 +05007230 reserved to the headers rewriting should also be free.
Christopher Faulet24231ab2020-01-24 17:44:23 +01007231
7232 No further "http-response" rules are evaluated.
7233
7234 Example:
Daniel Corbett67a82712020-07-06 23:01:19 -04007235 http-response return errorfile /etc/haproxy/errorfiles/200.http \
Christopher Faulet24231ab2020-01-24 17:44:23 +01007236 if { status eq 404 }
7237
7238 http-response return content-type text/plain \
7239 string "This is the end !" \
7240 if { status eq 500 }
7241
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007242http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
7243http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08007244
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007245 This action increments the GPC0 or GPC1 counter according with the sticky
7246 counter designated by <sc-id>. If an error occurs, this action silently fails
7247 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02007248
Cédric Dufour0d7712d2019-11-06 18:38:53 +01007249http-response sc-set-gpt0(<sc-id>) { <int> | <expr> }
7250 [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02007251
Cédric Dufour0d7712d2019-11-06 18:38:53 +01007252 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
7253 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
7254 boolean. If an error occurs, this action silently fails and the actions
7255 evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01007256
Christopher Faulet68fc3a12021-08-12 09:32:07 +02007257http-response send-spoe-group <engine-name> <group-name>
7258 [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02007259
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007260 This action is used to trigger sending of a group of SPOE messages. To do so,
7261 the SPOE engine used to send messages must be defined, as well as the SPOE
7262 group to send. Of course, the SPOE engine must refer to an existing SPOE
7263 filter. If not engine name is provided on the SPOE filter line, the SPOE
7264 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02007265
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007266 Arguments:
7267 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02007268
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007269 <group-name> The SPOE group name as specified in the engine
7270 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02007271
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007272http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007273
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007274 This does the same as "add-header" except that the header name is first
7275 removed if it existed. This is useful when passing security information to
7276 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007277
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007278http-response set-log-level <level> [ { if | unless } <condition> ]
7279
7280 This is used to change the log level of the current request when a certain
7281 condition is met. Valid levels are the 8 syslog levels (see the "log"
7282 keyword) plus the special level "silent" which disables logging for this
7283 request. This rule is not final so the last matching rule wins. This rule can
7284 be useful to disable health checks coming from another equipment.
7285
7286http-response set-map(<file-name>) <key fmt> <value fmt>
7287
7288 This is used to add a new entry into a MAP. The MAP must be loaded from a
7289 file (even a dummy empty file). The file name of the MAP to be updated is
7290 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
7291 log-format rules, used to collect MAP key, and <value fmt>, which follows
7292 log-format rules, used to collect content for the new entry. It performs a
7293 lookup in the MAP before insertion, to avoid duplicated (or more) values.
7294 This lookup is done by a linear search and can be expensive with large lists!
7295 It is the equivalent of the "set map" command from the stats socket, but can
7296 be triggered by an HTTP response.
7297
7298http-response set-mark <mark> [ { if | unless } <condition> ]
7299
7300 This is used to set the Netfilter MARK on all packets sent to the client to
7301 the value passed in <mark> on platforms which support it. This value is an
7302 unsigned 32 bit value which can be matched by netfilter and by the routing
7303 table. It can be expressed both in decimal or hexadecimal format (prefixed
7304 by "0x"). This can be useful to force certain packets to take a different
7305 route (for example a cheaper network path for bulk downloads). This works on
7306 Linux kernels 2.6.32 and above and requires admin privileges.
7307
7308http-response set-nice <nice> [ { if | unless } <condition> ]
7309
7310 This sets the "nice" factor of the current request being processed.
7311 It only has effect against the other requests being processed at the same
7312 time. The default value is 0, unless altered by the "nice" setting on the
7313 "bind" line. The accepted range is -1024..1024. The higher the value, the
7314 nicest the request will be. Lower values will make the request more important
7315 than other ones. This can be useful to improve the speed of some requests, or
7316 lower the priority of non-important requests. Using this setting without
7317 prior experimentation can cause some major slowdown.
7318
7319http-response set-status <status> [reason <str>]
7320 [ { if | unless } <condition> ]
7321
7322 This replaces the response status code with <status> which must be an integer
7323 between 100 and 999. Optionally, a custom reason text can be provided defined
7324 by <str>, or the default reason for the specified code will be used as a
7325 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007326
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007327 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007328 # return "431 Request Header Fields Too Large"
7329 http-response set-status 431
7330 # return "503 Slow Down", custom reason
7331 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007332
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007333http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007334
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007335 This is used to set the TOS or DSCP field value of packets sent to the client
7336 to the value passed in <tos> on platforms which support this.
7337 This value represents the whole 8 bits of the IP TOS field, and can be
7338 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
7339 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
7340 always 0. This can be used to adjust some routing behavior on border routers
7341 based on some information from the request.
7342
7343 See RFC 2474, 2597, 3260 and 4594 for more information.
7344
7345http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
7346
7347 This is used to set the contents of a variable. The variable is declared
7348 inline.
7349
7350 Arguments:
7351 <var-name> The name of the variable starts with an indication about its
7352 scope. The scopes allowed are:
7353 "proc" : the variable is shared with the whole process
7354 "sess" : the variable is shared with the whole session
7355 "txn" : the variable is shared with the transaction
7356 (request and response)
7357 "req" : the variable is shared only during request
7358 processing
7359 "res" : the variable is shared only during response
7360 processing
7361 This prefix is followed by a name. The separator is a '.'.
7362 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
7363 and '_'.
7364
7365 <expr> Is a standard HAProxy expression formed by a sample-fetch
7366 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007367
7368 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007369 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007370
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007371http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007372
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007373 This stops the evaluation of the rules and makes the client-facing connection
7374 suddenly disappear using a system-dependent way that tries to prevent the
7375 client from being notified. The effect it then that the client still sees an
7376 established connection while there's none on HAProxy. The purpose is to
7377 achieve a comparable effect to "tarpit" except that it doesn't use any local
7378 resource at all on the machine running HAProxy. It can resist much higher
7379 loads than "tarpit", and slow down stronger attackers. It is important to
7380 understand the impact of using this mechanism. All stateful equipment placed
7381 between the client and HAProxy (firewalls, proxies, load balancers) will also
7382 keep the established connection for a long time and may suffer from this
7383 action.
7384 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
7385 option is used to block the emission of a TCP reset. On other systems, the
7386 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
7387 router, though it's still delivered to local networks. Do not use it unless
7388 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02007389
Christopher Faulet46f95542019-12-20 10:07:22 +01007390http-response strict-mode { on | off }
7391
7392 This enables or disables the strict rewriting mode for following rules. It
7393 does not affect rules declared before it and it is only applicable on rules
7394 performing a rewrite on the responses. When the strict mode is enabled, any
7395 rewrite failure triggers an internal error. Otherwise, such errors are
7396 silently ignored. The purpose of the strict rewriting mode is to make some
Ilya Shipitsin8525fd92020-02-29 12:34:59 +05007397 rewrites optional while others must be performed to continue the response
Christopher Faulet46f95542019-12-20 10:07:22 +01007398 processing.
7399
Christopher Faulet1aea50e2020-01-17 16:03:53 +01007400 By default, the strict rewriting mode is enabled. Its value is also reset
Christopher Faulet46f95542019-12-20 10:07:22 +01007401 when a ruleset evaluation ends. So, for instance, if you change the mode on
Daniel Corbett67a82712020-07-06 23:01:19 -04007402 the backend, the default mode is restored when HAProxy starts the frontend
Christopher Faulet46f95542019-12-20 10:07:22 +01007403 rules evaluation.
7404
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007405http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
7406http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
7407http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02007408
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007409 This enables tracking of sticky counters from current response. Please refer
7410 to "http-request track-sc" for a complete description. The only difference
7411 from "http-request track-sc" is the <key> sample expression can only make use
7412 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
7413 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007414 supported, HAProxy will fail and warn while parsing the config.
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02007415
7416http-response unset-var(<var-name>) [ { if | unless } <condition> ]
7417
7418 This is used to unset a variable. See "http-response set-var" for details
7419 about <var-name>.
7420
7421 Example:
7422 http-response unset-var(sess.last_redir)
7423
Christopher Faulet021a8e42021-03-29 10:46:38 +02007424http-response wait-for-body time <time> [ at-least <bytes> ]
7425 [ { if | unless } <condition> ]
7426
7427 This will delay the processing of the response waiting for the payload for at
7428 most <time> milliseconds. if "at-least" argument is specified, HAProxy stops
7429 to wait the payload when the first <bytes> bytes are received. 0 means no
7430 limit, it is the default value. Regardless the "at-least" argument value,
7431 HAProxy stops to wait if the whole payload is received or if the response
7432 buffer is full.
7433
7434 Arguments :
7435
7436 <time> is mandatory. It is the maximum time to wait for the body. It
7437 follows the HAProxy time format and is expressed in milliseconds.
7438
7439 <bytes> is optional. It is the minimum payload size to receive to stop to
Ilya Shipitsinb2be9a12021-04-24 13:25:42 +05007440 wait. It follows the HAProxy size format and is expressed in
Christopher Faulet021a8e42021-03-29 10:46:38 +02007441 bytes.
7442
7443 Example:
7444 http-response wait-for-body time 1s at-least 10k
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02007445
Willy Tarreau30631952015-08-06 15:05:24 +02007446http-reuse { never | safe | aggressive | always }
7447 Declare how idle HTTP connections may be shared between requests
7448
7449 May be used in sections: defaults | frontend | listen | backend
7450 yes | no | yes | yes
7451
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007452 By default, a connection established between HAProxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01007453 which is considered safe for reuse is moved back to the server's idle
7454 connections pool so that any other request can make use of it. This is the
7455 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02007456
7457 The argument indicates the desired connection reuse strategy :
7458
Olivier Houchard86006a52018-12-14 19:37:49 +01007459 - "never" : idle connections are never shared between sessions. This mode
7460 may be enforced to cancel a different strategy inherited from
7461 a defaults section or for troubleshooting. For example, if an
7462 old bogus application considers that multiple requests over
7463 the same connection come from the same client and it is not
7464 possible to fix the application, it may be desirable to
7465 disable connection sharing in a single backend. An example of
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007466 such an application could be an old HAProxy using cookie
Olivier Houchard86006a52018-12-14 19:37:49 +01007467 insertion in tunnel mode and not checking any request past the
7468 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02007469
Olivier Houchard86006a52018-12-14 19:37:49 +01007470 - "safe" : this is the default and the recommended strategy. The first
7471 request of a session is always sent over its own connection,
7472 and only subsequent requests may be dispatched over other
7473 existing connections. This ensures that in case the server
7474 closes the connection when the request is being sent, the
7475 browser can decide to silently retry it. Since it is exactly
7476 equivalent to regular keep-alive, there should be no side
Amaury Denoyelle27179652020-10-14 18:17:12 +02007477 effects. There is also a special handling for the connections
7478 using protocols subject to Head-of-line blocking (backend with
7479 h2 or fcgi). In this case, when at least one stream is
7480 processed, the used connection is reserved to handle streams
7481 of the same session. When no more streams are processed, the
7482 connection is released and can be reused.
Willy Tarreau30631952015-08-06 15:05:24 +02007483
7484 - "aggressive" : this mode may be useful in webservices environments where
7485 all servers are not necessarily known and where it would be
7486 appreciable to deliver most first requests over existing
7487 connections. In this case, first requests are only delivered
7488 over existing connections that have been reused at least once,
7489 proving that the server correctly supports connection reuse.
7490 It should only be used when it's sure that the client can
7491 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02007492 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02007493 downsides of rare connection failures.
7494
7495 - "always" : this mode is only recommended when the path to the server is
7496 known for never breaking existing connections quickly after
7497 releasing them. It allows the first request of a session to be
7498 sent to an existing connection. This can provide a significant
7499 performance increase over the "safe" strategy when the backend
7500 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01007501 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02007502 sharing. It is recommended that the "http-keep-alive" timeout
7503 remains low in this mode so that no dead connections remain
7504 usable. In most cases, this will lead to the same performance
7505 gains as "aggressive" but with more risks. It should only be
7506 used when it improves the situation over "aggressive".
7507
7508 When http connection sharing is enabled, a great care is taken to respect the
Amaury Denoyelled773a4e2021-01-29 15:18:49 +01007509 connection properties and compatibility. Indeed, some properties are specific
7510 and it is not possibly to reuse it blindly. Those are the SSL SNI, source
7511 and destination address and proxy protocol block. A connection is reused only
7512 if it shares the same set of properties with the request.
Willy Tarreau30631952015-08-06 15:05:24 +02007513
Amaury Denoyelled773a4e2021-01-29 15:18:49 +01007514 Also note that connections with certain bogus authentication schemes (relying
7515 on the connection) like NTLM are marked private and never shared.
Willy Tarreau30631952015-08-06 15:05:24 +02007516
Lukas Tribuse8adfeb2019-11-06 11:50:25 +01007517 A connection pool is involved and configurable with "pool-max-conn".
Willy Tarreau30631952015-08-06 15:05:24 +02007518
7519 Note: connection reuse improves the accuracy of the "server maxconn" setting,
7520 because almost no new connection will be established while idle connections
7521 remain available. This is particularly true with the "always" strategy.
7522
7523 See also : "option http-keep-alive", "server maxconn"
7524
7525
Mark Lamourinec2247f02012-01-04 13:02:01 -05007526http-send-name-header [<header>]
7527 Add the server name to a request. Use the header string given by <header>
Mark Lamourinec2247f02012-01-04 13:02:01 -05007528 May be used in sections: defaults | frontend | listen | backend
7529 yes | no | yes | yes
Mark Lamourinec2247f02012-01-04 13:02:01 -05007530 Arguments :
Mark Lamourinec2247f02012-01-04 13:02:01 -05007531 <header> The header string to use to send the server name
7532
Willy Tarreau81bef7e2019-10-07 14:58:02 +02007533 The "http-send-name-header" statement causes the header field named <header>
7534 to be set to the name of the target server at the moment the request is about
7535 to be sent on the wire. Any existing occurrences of this header are removed.
7536 Upon retries and redispatches, the header field is updated to always reflect
7537 the server being attempted to connect to. Given that this header is modified
7538 very late in the connection setup, it may have unexpected effects on already
7539 modified headers. For example using it with transport-level header such as
7540 connection, content-length, transfer-encoding and so on will likely result in
7541 invalid requests being sent to the server. Additionally it has been reported
7542 that this directive is currently being used as a way to overwrite the Host
7543 header field in outgoing requests; while this trick has been known to work
7544 as a side effect of the feature for some time, it is not officially supported
7545 and might possibly not work anymore in a future version depending on the
7546 technical difficulties this feature induces. A long-term solution instead
7547 consists in fixing the application which required this trick so that it binds
7548 to the correct host name.
Mark Lamourinec2247f02012-01-04 13:02:01 -05007549
7550 See also : "server"
7551
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01007552id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02007553 Set a persistent ID to a proxy.
7554 May be used in sections : defaults | frontend | listen | backend
7555 no | yes | yes | yes
7556 Arguments : none
7557
7558 Set a persistent ID for the proxy. This ID must be unique and positive.
7559 An unused ID will automatically be assigned if unset. The first assigned
7560 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01007561
7562
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007563ignore-persist { if | unless } <condition>
7564 Declare a condition to ignore persistence
7565 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01007566 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007567
7568 By default, when cookie persistence is enabled, every requests containing
7569 the cookie are unconditionally persistent (assuming the target server is up
7570 and running).
7571
7572 The "ignore-persist" statement allows one to declare various ACL-based
7573 conditions which, when met, will cause a request to ignore persistence.
7574 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007575 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007576 persistence for a specific User-Agent (for example, some web crawler bots).
7577
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007578 The persistence is ignored when an "if" condition is met, or unless an
7579 "unless" condition is met.
7580
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007581 Example:
7582 acl url_static path_beg /static /images /img /css
7583 acl url_static path_end .gif .png .jpg .css .js
7584 ignore-persist if url_static
7585
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007586 See also : "force-persist", "cookie", and section 7 about ACL usage.
7587
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007588load-server-state-from-file { global | local | none }
7589 Allow seamless reload of HAProxy
7590 May be used in sections: defaults | frontend | listen | backend
7591 yes | no | yes | yes
7592
7593 This directive points HAProxy to a file where server state from previous
7594 running process has been saved. That way, when starting up, before handling
7595 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01007596 reload occurred. The purpose of the "load-server-state-from-file" directive is
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007597 to tell HAProxy which file to use. For now, only 2 arguments to either prevent
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007598 loading state or load states from a file containing all backends and servers.
7599 The state file can be generated by running the command "show servers state"
7600 over the stats socket and redirect output.
7601
Davor Ocelice9ed2812017-12-25 17:49:28 +01007602 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007603 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02007604 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007605
7606 Arguments:
7607 global load the content of the file pointed by the global directive
7608 named "server-state-file".
7609
7610 local load the content of the file pointed by the directive
7611 "server-state-file-name" if set. If not set, then the backend
7612 name is used as a file name.
7613
7614 none don't load any stat for this backend
7615
7616 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01007617 - server's IP address is preserved across reloads by default, but the
7618 order can be changed thanks to the server's "init-addr" setting. This
7619 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01007620 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01007621 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007622
7623 - server's weight is applied from previous running process unless it has
7624 has changed between previous and new configuration files.
7625
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007626 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007627
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007628 global
7629 stats socket /tmp/socket
7630 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007631
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007632 defaults
7633 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007634
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007635 backend bk
7636 server s1 127.0.0.1:22 check weight 11
7637 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007638
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007639
7640 Then one can run :
7641
7642 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
7643
7644 Content of the file /tmp/server_state would be like this:
7645
7646 1
7647 # <field names skipped for the doc example>
7648 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
7649 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
7650
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007651 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007652
7653 global
7654 stats socket /tmp/socket
7655 server-state-base /etc/haproxy/states
7656
7657 defaults
7658 load-server-state-from-file local
7659
7660 backend bk
7661 server s1 127.0.0.1:22 check weight 11
7662 server s2 127.0.0.1:22 check weight 12
7663
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02007664
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007665 Then one can run :
7666
7667 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
7668
7669 Content of the file /etc/haproxy/states/bk would be like this:
7670
7671 1
7672 # <field names skipped for the doc example>
7673 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
7674 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
7675
7676 See also: "server-state-file", "server-state-file-name", and
7677 "show servers state"
7678
Cyril Bonté0d4bf012010-04-25 23:21:46 +02007679
Willy Tarreau2769aa02007-12-27 18:26:09 +01007680log global
Jan Wagner3e678602020-12-17 22:22:32 +01007681log <address> [len <length>] [format <format>] [sample <ranges>:<sample_size>]
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02007682 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02007683no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01007684 Enable per-instance logging of events and traffic.
7685 May be used in sections : defaults | frontend | listen | backend
7686 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02007687
7688 Prefix :
7689 no should be used when the logger list must be flushed. For example,
7690 if you don't want to inherit from the default logger list. This
7691 prefix does not allow arguments.
7692
Willy Tarreau2769aa02007-12-27 18:26:09 +01007693 Arguments :
7694 global should be used when the instance's logging parameters are the
7695 same as the global ones. This is the most common usage. "global"
7696 replaces <address>, <facility> and <level> with those of the log
7697 entries found in the "global" section. Only one "log global"
7698 statement may be used per instance, and this form takes no other
7699 parameter.
7700
7701 <address> indicates where to send the logs. It takes the same format as
7702 for the "global" section's logs, and can be one of :
7703
7704 - An IPv4 address optionally followed by a colon (':') and a UDP
7705 port. If no port is specified, 514 is used by default (the
7706 standard syslog port).
7707
David du Colombier24bb5f52011-03-17 10:40:23 +01007708 - An IPv6 address followed by a colon (':') and optionally a UDP
7709 port. If no port is specified, 514 is used by default (the
7710 standard syslog port).
7711
Willy Tarreau2769aa02007-12-27 18:26:09 +01007712 - A filesystem path to a UNIX domain socket, keeping in mind
7713 considerations for chroot (be sure the path is accessible
7714 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007715 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01007716
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01007717 - A file descriptor number in the form "fd@<number>", which may
7718 point to a pipe, terminal, or socket. In this case unbuffered
7719 logs are used and one writev() call per log is performed. This
7720 is a bit expensive but acceptable for most workloads. Messages
7721 sent this way will not be truncated but may be dropped, in
7722 which case the DroppedLogs counter will be incremented. The
7723 writev() call is atomic even on pipes for messages up to
7724 PIPE_BUF size, which POSIX recommends to be at least 512 and
7725 which is 4096 bytes on most modern operating systems. Any
7726 larger message may be interleaved with messages from other
7727 processes. Exceptionally for debugging purposes the file
7728 descriptor may also be directed to a file, but doing so will
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007729 significantly slow HAProxy down as non-blocking calls will be
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01007730 ignored. Also there will be no way to purge nor rotate this
7731 file without restarting the process. Note that the configured
7732 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01007733 with a TCP syslog server. See also the "short" and "raw"
7734 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01007735
7736 - "stdout" / "stderr", which are respectively aliases for "fd@1"
7737 and "fd@2", see above.
7738
Willy Tarreauc046d162019-08-30 15:24:59 +02007739 - A ring buffer in the form "ring@<name>", which will correspond
7740 to an in-memory ring buffer accessible over the CLI using the
7741 "show events" command, which will also list existing rings and
7742 their sizes. Such buffers are lost on reload or restart but
7743 when used as a complement this can help troubleshooting by
7744 having the logs instantly available.
7745
Emeric Brun94aab062021-04-02 10:41:36 +02007746 - An explicit stream address prefix such as "tcp@","tcp6@",
7747 "tcp4@" or "uxst@" will allocate an implicit ring buffer with
7748 a stream forward server targeting the given address.
7749
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01007750 You may want to reference some environment variables in the
7751 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01007752
Willy Tarreau18324f52014-06-27 18:10:07 +02007753 <length> is an optional maximum line length. Log lines larger than this
7754 value will be truncated before being sent. The reason is that
7755 syslog servers act differently on log line length. All servers
7756 support the default value of 1024, but some servers simply drop
7757 larger lines while others do log them. If a server supports long
7758 lines, it may make sense to set this value here in order to avoid
7759 truncating long lines. Similarly, if a server drops long lines,
7760 it is preferable to truncate them before sending them. Accepted
7761 values are 80 to 65535 inclusive. The default value of 1024 is
7762 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01007763 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02007764
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02007765 <ranges> A list of comma-separated ranges to identify the logs to sample.
7766 This is used to balance the load of the logs to send to the log
7767 server. The limits of the ranges cannot be null. They are numbered
7768 from 1. The size or period (in number of logs) of the sample must
7769 be set with <sample_size> parameter.
7770
7771 <sample_size>
7772 The size of the sample in number of logs to consider when balancing
7773 their logging loads. It is used to balance the load of the logs to
7774 send to the syslog server. This size must be greater or equal to the
7775 maximum of the high limits of the ranges.
7776 (see also <ranges> parameter).
7777
Willy Tarreauadb345d2018-11-12 07:56:13 +01007778 <format> is the log format used when generating syslog messages. It may be
7779 one of the following :
7780
Emeric Brun0237c4e2020-11-27 16:24:34 +01007781 local Analog to rfc3164 syslog message format except that hostname
7782 field is stripped. This is the default.
7783 Note: option "log-send-hostname" switches the default to
7784 rfc3164.
7785
7786 rfc3164 The RFC3164 syslog message format.
Willy Tarreauadb345d2018-11-12 07:56:13 +01007787 (https://tools.ietf.org/html/rfc3164)
7788
7789 rfc5424 The RFC5424 syslog message format.
7790 (https://tools.ietf.org/html/rfc5424)
7791
Emeric Brun54648852020-07-06 15:54:06 +02007792 priority A message containing only a level plus syslog facility between
7793 angle brackets such as '<63>', followed by the text. The PID,
7794 date, time, process name and system name are omitted. This is
7795 designed to be used with a local log server.
7796
Willy Tarreaue8746a02018-11-12 08:45:00 +01007797 short A message containing only a level between angle brackets such as
7798 '<3>', followed by the text. The PID, date, time, process name
7799 and system name are omitted. This is designed to be used with a
7800 local log server. This format is compatible with what the
7801 systemd logger consumes.
7802
Emeric Brun54648852020-07-06 15:54:06 +02007803 timed A message containing only a level between angle brackets such as
7804 '<3>', followed by ISO date and by the text. The PID, process
7805 name and system name are omitted. This is designed to be
7806 used with a local log server.
7807
7808 iso A message containing only the ISO date, followed by the text.
7809 The PID, process name and system name are omitted. This is
7810 designed to be used with a local log server.
7811
Willy Tarreauc1b06452018-11-12 11:57:56 +01007812 raw A message containing only the text. The level, PID, date, time,
7813 process name and system name are omitted. This is designed to
7814 be used in containers or during development, where the severity
7815 only depends on the file descriptor used (stdout/stderr).
7816
Willy Tarreau2769aa02007-12-27 18:26:09 +01007817 <facility> must be one of the 24 standard syslog facilities :
7818
Willy Tarreaue8746a02018-11-12 08:45:00 +01007819 kern user mail daemon auth syslog lpr news
7820 uucp cron auth2 ftp ntp audit alert cron2
7821 local0 local1 local2 local3 local4 local5 local6 local7
7822
Willy Tarreauc1b06452018-11-12 11:57:56 +01007823 Note that the facility is ignored for the "short" and "raw"
7824 formats, but still required as a positional field. It is
7825 recommended to use "daemon" in this case to make it clear that
7826 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01007827
7828 <level> is optional and can be specified to filter outgoing messages. By
7829 default, all messages are sent. If a level is specified, only
7830 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02007831 will be sent. An optional minimum level can be specified. If it
7832 is set, logs emitted with a more severe level than this one will
7833 be capped to this level. This is used to avoid sending "emerg"
7834 messages on all terminals on some default syslog configurations.
7835 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01007836
7837 emerg alert crit err warning notice info debug
7838
William Lallemand0f99e342011-10-12 17:50:54 +02007839 It is important to keep in mind that it is the frontend which decides what to
7840 log from a connection, and that in case of content switching, the log entries
7841 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01007842
7843 However, backend log declaration define how and where servers status changes
7844 will be logged. Level "notice" will be used to indicate a server going up,
7845 "warning" will be used for termination signals and definitive service
7846 termination, and "alert" will be used for when a server goes down.
7847
7848 Note : According to RFC3164, messages are truncated to 1024 bytes before
7849 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01007850
7851 Example :
7852 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01007853 log stdout format short daemon # send log to systemd
7854 log stdout format raw daemon # send everything to stdout
7855 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02007856 log 127.0.0.1:514 local0 notice # only send important events
Emeric Brun94aab062021-04-02 10:41:36 +02007857 log tcp@127.0.0.1:514 local0 notice notice # same but limit output
7858 # level and send in tcp
William Lallemandb2f07452015-05-12 14:27:13 +02007859 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01007860
Willy Tarreau2769aa02007-12-27 18:26:09 +01007861
William Lallemand48940402012-01-30 16:47:22 +01007862log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01007863 Specifies the log format string to use for traffic logs
7864 May be used in sections: defaults | frontend | listen | backend
7865 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01007866
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01007867 This directive specifies the log format string that will be used for all logs
7868 resulting from traffic passing through the frontend using this line. If the
7869 directive is used in a defaults section, all subsequent frontends will use
7870 the same log format. Please see section 8.2.4 which covers the log format
7871 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01007872
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007873 "log-format" directive overrides previous "option tcplog", "log-format" and
7874 "option httplog" directives.
7875
Dragan Dosen7ad31542015-09-28 17:16:47 +02007876log-format-sd <string>
7877 Specifies the RFC5424 structured-data log format string
7878 May be used in sections: defaults | frontend | listen | backend
7879 yes | yes | yes | no
7880
7881 This directive specifies the RFC5424 structured-data log format string that
7882 will be used for all logs resulting from traffic passing through the frontend
7883 using this line. If the directive is used in a defaults section, all
7884 subsequent frontends will use the same log format. Please see section 8.2.4
7885 which covers the log format string in depth.
7886
7887 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
7888 about the RFC5424 structured-data part.
7889
7890 Note : This log format string will be used only for loggers that have set
7891 log format to "rfc5424".
7892
7893 Example :
7894 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
7895
7896
Willy Tarreau094af4e2015-01-07 15:03:42 +01007897log-tag <string>
7898 Specifies the log tag to use for all outgoing logs
7899 May be used in sections: defaults | frontend | listen | backend
7900 yes | yes | yes | yes
7901
7902 Sets the tag field in the syslog header to this string. It defaults to the
7903 log-tag set in the global section, otherwise the program name as launched
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007904 from the command line, which usually is "HAProxy". Sometimes it can be useful
Willy Tarreau094af4e2015-01-07 15:03:42 +01007905 to differentiate between multiple processes running on the same host, or to
7906 differentiate customer instances running in the same process. In the backend,
7907 logs about servers up/down will use this tag. As a hint, it can be convenient
7908 to set a log-tag related to a hosted customer in a defaults section then put
7909 all the frontends and backends for that customer, then start another customer
7910 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01007911
Willy Tarreauc35362a2014-04-25 13:58:37 +02007912max-keep-alive-queue <value>
7913 Set the maximum server queue size for maintaining keep-alive connections
7914 May be used in sections: defaults | frontend | listen | backend
7915 yes | no | yes | yes
7916
7917 HTTP keep-alive tries to reuse the same server connection whenever possible,
7918 but sometimes it can be counter-productive, for example if a server has a lot
7919 of connections while other ones are idle. This is especially true for static
7920 servers.
7921
7922 The purpose of this setting is to set a threshold on the number of queued
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007923 connections at which HAProxy stops trying to reuse the same server and prefers
Willy Tarreauc35362a2014-04-25 13:58:37 +02007924 to find another one. The default value, -1, means there is no limit. A value
7925 of zero means that keep-alive requests will never be queued. For very close
7926 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007927 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02007928 use a value of 10 or less). For remote servers suffering from a high latency,
7929 higher values might be needed to cover for the latency and/or the cost of
7930 picking a different server.
7931
7932 Note that this has no impact on responses which are maintained to the same
7933 server consecutively to a 401 response. They will still go to the same server
7934 even if they have to be queued.
7935
7936 See also : "option http-server-close", "option prefer-last-server", server
7937 "maxconn" and cookie persistence.
7938
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01007939max-session-srv-conns <nb>
7940 Set the maximum number of outgoing connections we can keep idling for a given
7941 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
7942 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02007943
Willy Tarreau2769aa02007-12-27 18:26:09 +01007944maxconn <conns>
7945 Fix the maximum number of concurrent connections on a frontend
7946 May be used in sections : defaults | frontend | listen | backend
7947 yes | yes | yes | no
7948 Arguments :
7949 <conns> is the maximum number of concurrent connections the frontend will
7950 accept to serve. Excess connections will be queued by the system
7951 in the socket's listen queue and will be served once a connection
7952 closes.
7953
7954 If the system supports it, it can be useful on big sites to raise this limit
Daniel Corbett9f0843f2021-05-08 10:50:37 -04007955 very high so that HAProxy manages connection queues, instead of leaving the
Willy Tarreau2769aa02007-12-27 18:26:09 +01007956 clients with unanswered connection attempts. This value should not exceed the
7957 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01007958 of tune.bufsize (16kB by default) each, as well as some other data resulting
7959 in about 33 kB of RAM being consumed per established connection. That means
7960 that a medium system equipped with 1GB of RAM can withstand around
7961 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01007962
7963 Also, when <conns> is set to large values, it is possible that the servers
7964 are not sized to accept such loads, and for this reason it is generally wise
7965 to assign them some reasonable connection limits.
7966
Willy Tarreauc8d5b952019-02-27 17:25:52 +01007967 When this value is set to zero, which is the default, the global "maxconn"
7968 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02007969
Willy Tarreau2769aa02007-12-27 18:26:09 +01007970 See also : "server", global section's "maxconn", "fullconn"
7971
7972
Willy Tarreau77e0dae2020-10-14 15:44:27 +02007973mode { tcp|http }
Willy Tarreau2769aa02007-12-27 18:26:09 +01007974 Set the running mode or protocol of the instance
7975 May be used in sections : defaults | frontend | listen | backend
7976 yes | yes | yes | yes
7977 Arguments :
7978 tcp The instance will work in pure TCP mode. A full-duplex connection
7979 will be established between clients and servers, and no layer 7
7980 examination will be performed. This is the default mode. It
7981 should be used for SSL, SSH, SMTP, ...
7982
7983 http The instance will work in HTTP mode. The client request will be
7984 analyzed in depth before connecting to any server. Any request
7985 which is not RFC-compliant will be rejected. Layer 7 filtering,
7986 processing and switching will be possible. This is the mode which
7987 brings HAProxy most of its value.
7988
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007989 When doing content switching, it is mandatory that the frontend and the
7990 backend are in the same mode (generally HTTP), otherwise the configuration
7991 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01007992
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007993 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01007994 defaults http_instances
7995 mode http
7996
Willy Tarreau0ba27502007-12-24 16:55:16 +01007997
Cyril Bontéf0c60612010-02-06 14:44:47 +01007998monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01007999 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008000 May be used in sections : defaults | frontend | listen | backend
8001 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01008002 Arguments :
8003 if <cond> the monitor request will fail if the condition is satisfied,
8004 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008005 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01008006 are met, for instance a low number of servers both in a
8007 backend and its backup.
8008
8009 unless <cond> the monitor request will succeed only if the condition is
8010 satisfied, and will fail otherwise. Such a condition may be
8011 based on a test on the presence of a minimum number of active
8012 servers in a list of backends.
8013
8014 This statement adds a condition which can force the response to a monitor
8015 request to report a failure. By default, when an external component queries
8016 the URI dedicated to monitoring, a 200 response is returned. When one of the
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008017 conditions above is met, HAProxy will return 503 instead of 200. This is
Willy Tarreau0ba27502007-12-24 16:55:16 +01008018 very useful to report a site failure to an external component which may base
8019 routing advertisements between multiple sites on the availability reported by
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008020 HAProxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02008021 criterion. Note that "monitor fail" only works in HTTP mode. Both status
8022 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008023
8024 Example:
8025 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01008026 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01008027 acl site_dead nbsrv(dynamic) lt 2
8028 acl site_dead nbsrv(static) lt 2
8029 monitor-uri /site_alive
8030 monitor fail if site_dead
8031
Willy Tarreau9e9919d2020-10-14 15:55:23 +02008032 See also : "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01008033
8034
Willy Tarreau2769aa02007-12-27 18:26:09 +01008035monitor-uri <uri>
8036 Intercept a URI used by external components' monitor requests
8037 May be used in sections : defaults | frontend | listen | backend
8038 yes | yes | yes | no
8039 Arguments :
8040 <uri> is the exact URI which we want to intercept to return HAProxy's
8041 health status instead of forwarding the request.
8042
8043 When an HTTP request referencing <uri> will be received on a frontend,
8044 HAProxy will not forward it nor log it, but instead will return either
8045 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
8046 conditions defined with "monitor fail". This is normally enough for any
8047 front-end HTTP probe to detect that the service is UP and running without
8048 forwarding the request to a backend server. Note that the HTTP method, the
8049 version and all headers are ignored, but the request must at least be valid
8050 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
8051
Willy Tarreau721d8e02017-12-01 18:25:08 +01008052 Monitor requests are processed very early, just after the request is parsed
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02008053 and even before any "http-request". The only rulesets applied before are the
8054 tcp-request ones. They cannot be logged either, and it is the intended
8055 purpose. They are only used to report HAProxy's health to an upper component,
8056 nothing more. However, it is possible to add any number of conditions using
8057 "monitor fail" and ACLs so that the result can be adjusted to whatever check
8058 can be imagined (most often the number of available servers in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01008059
Christopher Faulet6072beb2020-02-18 15:34:58 +01008060 Note: if <uri> starts by a slash ('/'), the matching is performed against the
8061 request's path instead of the request's uri. It is a workaround to let
8062 the HTTP/2 requests match the monitor-uri. Indeed, in HTTP/2, clients
8063 are encouraged to send absolute URIs only.
8064
Willy Tarreau2769aa02007-12-27 18:26:09 +01008065 Example :
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008066 # Use /haproxy_test to report HAProxy's status
Willy Tarreau2769aa02007-12-27 18:26:09 +01008067 frontend www
8068 mode http
8069 monitor-uri /haproxy_test
8070
Willy Tarreau9e9919d2020-10-14 15:55:23 +02008071 See also : "monitor fail"
Willy Tarreau2769aa02007-12-27 18:26:09 +01008072
Willy Tarreau0ba27502007-12-24 16:55:16 +01008073
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008074option abortonclose
8075no option abortonclose
8076 Enable or disable early dropping of aborted requests pending in queues.
8077 May be used in sections : defaults | frontend | listen | backend
8078 yes | no | yes | yes
8079 Arguments : none
8080
8081 In presence of very high loads, the servers will take some time to respond.
8082 The per-instance connection queue will inflate, and the response time will
8083 increase respective to the size of the queue times the average per-session
8084 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01008085 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008086 the queue, and slowing down other users, and the servers as well, because the
8087 request will eventually be served, then aborted at the first error
8088 encountered while delivering the response.
8089
8090 As there is no way to distinguish between a full STOP and a simple output
8091 close on the client side, HTTP agents should be conservative and consider
8092 that the client might only have closed its output channel while waiting for
8093 the response. However, this introduces risks of congestion when lots of users
8094 do the same, and is completely useless nowadays because probably no client at
8095 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01008096 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008097 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01008098 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008099 of being the single component to break rare but valid traffic is extremely
8100 low, which adds to the temptation to be able to abort a session early while
8101 still not served and not pollute the servers.
8102
Davor Ocelice9ed2812017-12-25 17:49:28 +01008103 In HAProxy, the user can choose the desired behavior using the option
8104 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008105 compliant and aborted requests will be served. But when the option is
8106 specified, a session with an incoming channel closed will be aborted while
8107 it is still possible, either pending in the queue for a connection slot, or
8108 during the connection establishment if the server has not yet acknowledged
8109 the connection request. This considerably reduces the queue size and the load
8110 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01008111 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008112
8113 If this option has been enabled in a "defaults" section, it can be disabled
8114 in a specific instance by prepending the "no" keyword before it.
8115
8116 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
8117
8118
Willy Tarreau4076a152009-04-02 15:18:36 +02008119option accept-invalid-http-request
8120no option accept-invalid-http-request
8121 Enable or disable relaxing of HTTP request parsing
8122 May be used in sections : defaults | frontend | listen | backend
8123 yes | yes | yes | no
8124 Arguments : none
8125
Willy Tarreau91852eb2015-05-01 13:26:00 +02008126 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02008127 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01008128 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02008129 forbidden characters are essentially used to build attacks exploiting server
8130 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
8131 server will emit invalid header names for whatever reason (configuration,
8132 implementation) and the issue will not be immediately fixed. In such a case,
8133 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01008134 even if that does not make sense, by specifying this option. Similarly, the
8135 list of characters allowed to appear in a URI is well defined by RFC3986, and
8136 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
8137 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01008138 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02008139 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02008140 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
8141 to pass through (no version specified) and multiple digits for both the major
8142 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02008143
8144 This option should never be enabled by default as it hides application bugs
8145 and open security breaches. It should only be deployed after a problem has
8146 been confirmed.
8147
8148 When this option is enabled, erroneous header names will still be accepted in
8149 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01008150 analysis using the "show errors" request on the UNIX stats socket. Similarly,
8151 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02008152 also helps confirming that the issue has been solved.
8153
8154 If this option has been enabled in a "defaults" section, it can be disabled
8155 in a specific instance by prepending the "no" keyword before it.
8156
8157 See also : "option accept-invalid-http-response" and "show errors" on the
8158 stats socket.
8159
8160
8161option accept-invalid-http-response
8162no option accept-invalid-http-response
8163 Enable or disable relaxing of HTTP response parsing
8164 May be used in sections : defaults | frontend | listen | backend
8165 yes | no | yes | yes
8166 Arguments : none
8167
Willy Tarreau91852eb2015-05-01 13:26:00 +02008168 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02008169 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01008170 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02008171 forbidden characters are essentially used to build attacks exploiting server
8172 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
8173 server will emit invalid header names for whatever reason (configuration,
8174 implementation) and the issue will not be immediately fixed. In such a case,
8175 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02008176 even if that does not make sense, by specifying this option. This option also
8177 relaxes the test on the HTTP version format, it allows multiple digits for
8178 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02008179
8180 This option should never be enabled by default as it hides application bugs
8181 and open security breaches. It should only be deployed after a problem has
8182 been confirmed.
8183
8184 When this option is enabled, erroneous header names will still be accepted in
8185 responses, but the complete response will be captured in order to permit
8186 later analysis using the "show errors" request on the UNIX stats socket.
8187 Doing this also helps confirming that the issue has been solved.
8188
8189 If this option has been enabled in a "defaults" section, it can be disabled
8190 in a specific instance by prepending the "no" keyword before it.
8191
8192 See also : "option accept-invalid-http-request" and "show errors" on the
8193 stats socket.
8194
8195
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008196option allbackups
8197no option allbackups
8198 Use either all backup servers at a time or only the first one
8199 May be used in sections : defaults | frontend | listen | backend
8200 yes | no | yes | yes
8201 Arguments : none
8202
8203 By default, the first operational backup server gets all traffic when normal
8204 servers are all down. Sometimes, it may be preferred to use multiple backups
8205 at once, because one will not be enough. When "option allbackups" is enabled,
8206 the load balancing will be performed among all backup servers when all normal
8207 ones are unavailable. The same load balancing algorithm will be used and the
8208 servers' weights will be respected. Thus, there will not be any priority
8209 order between the backup servers anymore.
8210
8211 This option is mostly used with static server farms dedicated to return a
8212 "sorry" page when an application is completely offline.
8213
8214 If this option has been enabled in a "defaults" section, it can be disabled
8215 in a specific instance by prepending the "no" keyword before it.
8216
8217
8218option checkcache
8219no option checkcache
Godbach7056a352013-12-11 20:01:07 +08008220 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008221 May be used in sections : defaults | frontend | listen | backend
8222 yes | no | yes | yes
8223 Arguments : none
8224
8225 Some high-level frameworks set application cookies everywhere and do not
8226 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008227 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008228 high risk of session crossing or stealing between users traversing the same
8229 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008230 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008231
8232 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008233 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01008234 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008235 response to check if there's a risk of caching a cookie on a client-side
8236 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01008237 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01008238 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01008239 - all those with a return code other than 200, 203, 204, 206, 300, 301,
8240 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01008241 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01008242 - all those that result from a request using a method other than GET, HEAD,
8243 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01008244 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008245 - those with a 'Pragma: no-cache' header
8246 - those with a 'Cache-control: private' header
8247 - those with a 'Cache-control: no-store' header
8248 - those with a 'Cache-control: max-age=0' header
8249 - those with a 'Cache-control: s-maxage=0' header
8250 - those with a 'Cache-control: no-cache' header
8251 - those with a 'Cache-control: no-cache="set-cookie"' header
8252 - those with a 'Cache-control: no-cache="set-cookie,' header
8253 (allowing other fields after set-cookie)
8254
8255 If a response doesn't respect these requirements, then it will be blocked
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02008256 just as if it was from an "http-response deny" rule, with an "HTTP 502 bad
8257 gateway". The session state shows "PH--" meaning that the proxy blocked the
8258 response during headers processing. Additionally, an alert will be sent in
8259 the logs so that admins are informed that there's something to be fixed.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008260
8261 Due to the high impact on the application, the application should be tested
8262 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008263 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01008264 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008265
8266 If this option has been enabled in a "defaults" section, it can be disabled
8267 in a specific instance by prepending the "no" keyword before it.
8268
8269
8270option clitcpka
8271no option clitcpka
8272 Enable or disable the sending of TCP keepalive packets on the client side
8273 May be used in sections : defaults | frontend | listen | backend
8274 yes | yes | yes | no
8275 Arguments : none
8276
8277 When there is a firewall or any session-aware component between a client and
8278 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01008279 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008280 components decides to expire a session which has remained idle for too long.
8281
8282 Enabling socket-level TCP keep-alives makes the system regularly send packets
8283 to the other end of the connection, leaving it active. The delay between
8284 keep-alive probes is controlled by the system only and depends both on the
8285 operating system and its tuning parameters.
8286
8287 It is important to understand that keep-alive packets are neither emitted nor
8288 received at the application level. It is only the network stacks which sees
8289 them. For this reason, even if one side of the proxy already uses keep-alives
8290 to maintain its connection alive, those keep-alive packets will not be
8291 forwarded to the other side of the proxy.
8292
8293 Please note that this has nothing to do with HTTP keep-alive.
8294
8295 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
8296 client side of a connection, which should help when session expirations are
8297 noticed between HAProxy and a client.
8298
8299 If this option has been enabled in a "defaults" section, it can be disabled
8300 in a specific instance by prepending the "no" keyword before it.
8301
8302 See also : "option srvtcpka", "option tcpka"
8303
8304
Willy Tarreau0ba27502007-12-24 16:55:16 +01008305option contstats
8306 Enable continuous traffic statistics updates
8307 May be used in sections : defaults | frontend | listen | backend
8308 yes | yes | yes | no
8309 Arguments : none
8310
8311 By default, counters used for statistics calculation are incremented
8312 only when a session finishes. It works quite well when serving small
8313 objects, but with big ones (for example large images or archives) or
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008314 with A/V streaming, a graph generated from HAProxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01008315 a hedgehog. With this option enabled counters get incremented frequently
8316 along the session, typically every 5 seconds, which is often enough to
8317 produce clean graphs. Recounting touches a hotpath directly so it is not
8318 not enabled by default, as it can cause a lot of wakeups for very large
8319 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008320
Christopher Faulet89aed322020-06-02 17:33:56 +02008321option disable-h2-upgrade
8322no option disable-h2-upgrade
8323 Enable or disable the implicit HTTP/2 upgrade from an HTTP/1.x client
8324 connection.
8325 May be used in sections : defaults | frontend | listen | backend
8326 yes | yes | yes | no
8327 Arguments : none
8328
8329 By default, HAProxy is able to implicitly upgrade an HTTP/1.x client
8330 connection to an HTTP/2 connection if the first request it receives from a
8331 given HTTP connection matches the HTTP/2 connection preface (i.e. the string
8332 "PRI * HTTP/2.0\r\n\r\nSM\r\n\r\n"). This way, it is possible to support
Christopher Faulet982e17d2021-03-26 14:44:18 +01008333 HTTP/1.x and HTTP/2 clients on a non-SSL connections. This option must be
8334 used to disable the implicit upgrade. Note this implicit upgrade is only
8335 supported for HTTP proxies, thus this option too. Note also it is possible to
8336 force the HTTP/2 on clear connections by specifying "proto h2" on the bind
8337 line. Finally, this option is applied on all bind lines. To disable implicit
8338 HTTP/2 upgrades for a specific bind line, it is possible to use "proto h1".
Christopher Faulet89aed322020-06-02 17:33:56 +02008339
8340 If this option has been enabled in a "defaults" section, it can be disabled
8341 in a specific instance by prepending the "no" keyword before it.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008342
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02008343option dontlog-normal
8344no option dontlog-normal
8345 Enable or disable logging of normal, successful connections
8346 May be used in sections : defaults | frontend | listen | backend
8347 yes | yes | yes | no
8348 Arguments : none
8349
8350 There are large sites dealing with several thousand connections per second
8351 and for which logging is a major pain. Some of them are even forced to turn
8352 logs off and cannot debug production issues. Setting this option ensures that
8353 normal connections, those which experience no error, no timeout, no retry nor
8354 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
8355 mode, the response status code is checked and return codes 5xx will still be
8356 logged.
8357
8358 It is strongly discouraged to use this option as most of the time, the key to
8359 complex issues is in the normal logs which will not be logged here. If you
8360 need to separate logs, see the "log-separate-errors" option instead.
8361
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008362 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02008363 logging.
8364
8365
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008366option dontlognull
8367no option dontlognull
8368 Enable or disable logging of null connections
8369 May be used in sections : defaults | frontend | listen | backend
8370 yes | yes | yes | no
8371 Arguments : none
8372
8373 In certain environments, there are components which will regularly connect to
8374 various systems to ensure that they are still alive. It can be the case from
8375 another load balancer as well as from monitoring systems. By default, even a
8376 simple port probe or scan will produce a log. If those connections pollute
8377 the logs too much, it is possible to enable option "dontlognull" to indicate
8378 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02008379 which typically corresponds to those probes. Note that errors will still be
8380 returned to the client and accounted for in the stats. If this is not what is
8381 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008382
8383 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01008384 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008385 would not be logged.
8386
8387 If this option has been enabled in a "defaults" section, it can be disabled
8388 in a specific instance by prepending the "no" keyword before it.
8389
Willy Tarreau9e9919d2020-10-14 15:55:23 +02008390 See also : "log", "http-ignore-probes", "monitor-uri", and
Willy Tarreau0f228a02015-05-01 15:37:53 +02008391 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008392
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008393
Willy Tarreau87cf5142011-08-19 22:57:24 +02008394option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01008395 Enable insertion of the X-Forwarded-For header to requests sent to servers
8396 May be used in sections : defaults | frontend | listen | backend
8397 yes | yes | yes | yes
8398 Arguments :
8399 <network> is an optional argument used to disable this option for sources
8400 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02008401 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01008402 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008403
8404 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
8405 their client address. This is sometimes annoying when the client's IP address
8406 is expected in server logs. To solve this problem, the well-known HTTP header
8407 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
8408 This header contains a value representing the client's IP address. Since this
8409 header is always appended at the end of the existing header list, the server
8410 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02008411 the server's manual to find how to enable use of this standard header. Note
8412 that only the last occurrence of the header must be used, since it is really
8413 possible that the client has already brought one.
8414
Willy Tarreaud72758d2010-01-12 10:42:19 +01008415 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02008416 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01008417 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01008418 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008419 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02008420 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01008421
8422 Sometimes, a same HAProxy instance may be shared between a direct client
8423 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
8424 used to decrypt HTTPS traffic). It is possible to disable the addition of the
8425 header for a known source address or network by adding the "except" keyword
8426 followed by the network address. In this case, any source IP matching the
8427 network will not cause an addition of this header. Most common uses are with
Christopher Faulet5d1def62021-02-26 09:19:15 +01008428 private networks or 127.0.0.1. IPv4 and IPv6 are both supported.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008429
Willy Tarreau87cf5142011-08-19 22:57:24 +02008430 Alternatively, the keyword "if-none" states that the header will only be
8431 added if it is not present. This should only be used in perfectly trusted
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008432 environment, as this might cause a security issue if headers reaching HAProxy
Willy Tarreau87cf5142011-08-19 22:57:24 +02008433 are under the control of the end-user.
8434
Willy Tarreauc27debf2008-01-06 08:57:02 +01008435 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02008436 least one of them uses it, the header will be added. Note that the backend's
8437 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02008438 both are defined. In the case of the "if-none" argument, if at least one of
8439 the frontend or the backend does not specify it, it wants the addition to be
8440 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008441
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02008442 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01008443 # Public HTTP address also used by stunnel on the same machine
8444 frontend www
8445 mode http
8446 option forwardfor except 127.0.0.1 # stunnel already adds the header
8447
Ross Westaf72a1d2008-08-03 10:51:45 +02008448 # Those servers want the IP Address in X-Client
8449 backend www
8450 mode http
8451 option forwardfor header X-Client
8452
Willy Tarreau87cf5142011-08-19 22:57:24 +02008453 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02008454 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01008455
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008456
Christopher Faulet98fbe952019-07-22 16:18:24 +02008457option h1-case-adjust-bogus-client
8458no option h1-case-adjust-bogus-client
8459 Enable or disable the case adjustment of HTTP/1 headers sent to bogus clients
8460 May be used in sections : defaults | frontend | listen | backend
8461 yes | yes | yes | no
8462 Arguments : none
8463
8464 There is no standard case for header names because, as stated in RFC7230,
8465 they are case-insensitive. So applications must handle them in a case-
8466 insensitive manner. But some bogus applications violate the standards and
8467 erroneously rely on the cases most commonly used by browsers. This problem
8468 becomes critical with HTTP/2 because all header names must be exchanged in
8469 lower case, and HAProxy follows the same convention. All header names are
8470 sent in lower case to clients and servers, regardless of the HTTP version.
8471
8472 When HAProxy receives an HTTP/1 response, its header names are converted to
8473 lower case and manipulated and sent this way to the clients. If a client is
8474 known to violate the HTTP standards and to fail to process a response coming
8475 from HAProxy, it is possible to transform the lower case header names to a
8476 different format when the response is formatted and sent to the client, by
8477 enabling this option and specifying the list of headers to be reformatted
8478 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
8479 must only be a temporary workaround for the time it takes the client to be
8480 fixed, because clients which require such workarounds might be vulnerable to
8481 content smuggling attacks and must absolutely be fixed.
8482
8483 Please note that this option will not affect standards-compliant clients.
8484
8485 If this option has been enabled in a "defaults" section, it can be disabled
8486 in a specific instance by prepending the "no" keyword before it.
8487
8488 See also: "option h1-case-adjust-bogus-server", "h1-case-adjust",
8489 "h1-case-adjust-file".
8490
8491
8492option h1-case-adjust-bogus-server
8493no option h1-case-adjust-bogus-server
8494 Enable or disable the case adjustment of HTTP/1 headers sent to bogus servers
8495 May be used in sections : defaults | frontend | listen | backend
8496 yes | no | yes | yes
8497 Arguments : none
8498
8499 There is no standard case for header names because, as stated in RFC7230,
8500 they are case-insensitive. So applications must handle them in a case-
8501 insensitive manner. But some bogus applications violate the standards and
8502 erroneously rely on the cases most commonly used by browsers. This problem
8503 becomes critical with HTTP/2 because all header names must be exchanged in
8504 lower case, and HAProxy follows the same convention. All header names are
8505 sent in lower case to clients and servers, regardless of the HTTP version.
8506
8507 When HAProxy receives an HTTP/1 request, its header names are converted to
8508 lower case and manipulated and sent this way to the servers. If a server is
8509 known to violate the HTTP standards and to fail to process a request coming
8510 from HAProxy, it is possible to transform the lower case header names to a
8511 different format when the request is formatted and sent to the server, by
8512 enabling this option and specifying the list of headers to be reformatted
8513 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
8514 must only be a temporary workaround for the time it takes the server to be
8515 fixed, because servers which require such workarounds might be vulnerable to
8516 content smuggling attacks and must absolutely be fixed.
8517
8518 Please note that this option will not affect standards-compliant servers.
8519
8520 If this option has been enabled in a "defaults" section, it can be disabled
8521 in a specific instance by prepending the "no" keyword before it.
8522
8523 See also: "option h1-case-adjust-bogus-client", "h1-case-adjust",
8524 "h1-case-adjust-file".
8525
8526
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02008527option http-buffer-request
8528no option http-buffer-request
8529 Enable or disable waiting for whole HTTP request body before proceeding
8530 May be used in sections : defaults | frontend | listen | backend
8531 yes | yes | yes | yes
8532 Arguments : none
8533
8534 It is sometimes desirable to wait for the body of an HTTP request before
8535 taking a decision. This is what is being done by "balance url_param" for
8536 example. The first use case is to buffer requests from slow clients before
8537 connecting to the server. Another use case consists in taking the routing
8538 decision based on the request body's contents. This option placed in a
8539 frontend or backend forces the HTTP processing to wait until either the whole
Christopher Faulet6db8a2e2019-11-19 16:27:25 +01008540 body is received or the request buffer is full. It can have undesired side
8541 effects with some applications abusing HTTP by expecting unbuffered
8542 transmissions between the frontend and the backend, so this should definitely
8543 not be used by default.
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02008544
Christopher Faulet021a8e42021-03-29 10:46:38 +02008545 See also : "option http-no-delay", "timeout http-request",
8546 "http-request wait-for-body"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02008547
8548
Willy Tarreau0f228a02015-05-01 15:37:53 +02008549option http-ignore-probes
8550no option http-ignore-probes
8551 Enable or disable logging of null connections and request timeouts
8552 May be used in sections : defaults | frontend | listen | backend
8553 yes | yes | yes | no
8554 Arguments : none
8555
8556 Recently some browsers started to implement a "pre-connect" feature
8557 consisting in speculatively connecting to some recently visited web sites
8558 just in case the user would like to visit them. This results in many
8559 connections being established to web sites, which end up in 408 Request
8560 Timeout if the timeout strikes first, or 400 Bad Request when the browser
8561 decides to close them first. These ones pollute the log and feed the error
8562 counters. There was already "option dontlognull" but it's insufficient in
8563 this case. Instead, this option does the following things :
8564 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01008565 was received over a connection before it was closed;
8566 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02008567 - prevent any error counter from being incremented
8568
8569 That way the empty connection is silently ignored. Note that it is better
8570 not to use this unless it is clear that it is needed, because it will hide
8571 real problems. The most common reason for not receiving a request and seeing
8572 a 408 is due to an MTU inconsistency between the client and an intermediary
8573 element such as a VPN, which blocks too large packets. These issues are
8574 generally seen with POST requests as well as GET with large cookies. The logs
8575 are often the only way to detect them.
8576
8577 If this option has been enabled in a "defaults" section, it can be disabled
8578 in a specific instance by prepending the "no" keyword before it.
8579
8580 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
8581
8582
Willy Tarreau16bfb022010-01-16 19:48:41 +01008583option http-keep-alive
8584no option http-keep-alive
8585 Enable or disable HTTP keep-alive from client to server
8586 May be used in sections : defaults | frontend | listen | backend
8587 yes | yes | yes | yes
8588 Arguments : none
8589
Willy Tarreau70dffda2014-01-30 03:07:23 +01008590 By default HAProxy operates in keep-alive mode with regards to persistent
8591 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02008592 leaves the connection idle on both sides between the end of a response and
8593 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02008594 as "option http-server-close" or "option httpclose". This option allows to
8595 set back the keep-alive mode, which can be useful when another mode was used
8596 in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01008597
8598 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
8599 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01008600 network) and the fastest session reuse on the server side at the expense
8601 of maintaining idle connections to the servers. In general, it is possible
8602 with this option to achieve approximately twice the request rate that the
8603 "http-server-close" option achieves on small objects. There are mainly two
8604 situations where this option may be useful :
8605
8606 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008607 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01008608
8609 - when the cost of establishing the connection to the server is significant
8610 compared to the cost of retrieving the associated object from the server.
8611
8612 This last case can happen when the server is a fast static server of cache.
8613 In this case, the server will need to be properly tuned to support high enough
8614 connection counts because connections will last until the client sends another
8615 request.
8616
8617 If the client request has to go to another backend or another server due to
8618 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01008619 immediately be closed and a new one re-opened. Option "prefer-last-server" is
8620 available to try optimize server selection so that if the server currently
8621 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01008622
Willy Tarreau16bfb022010-01-16 19:48:41 +01008623 At the moment, logs will not indicate whether requests came from the same
8624 session or not. The accept date reported in the logs corresponds to the end
8625 of the previous request, and the request time corresponds to the time spent
8626 waiting for a new request. The keep-alive request time is still bound to the
8627 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
8628 not set.
8629
Christopher Faulet159e6672019-07-16 15:09:52 +02008630 This option disables and replaces any previous "option httpclose" or "option
8631 http-server-close". When backend and frontend options differ, all of these 4
8632 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01008633
Christopher Faulet315b39c2018-09-21 16:26:19 +02008634 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01008635 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01008636 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01008637
8638
Willy Tarreau96e31212011-05-30 18:10:30 +02008639option http-no-delay
8640no option http-no-delay
8641 Instruct the system to favor low interactive delays over performance in HTTP
8642 May be used in sections : defaults | frontend | listen | backend
8643 yes | yes | yes | yes
8644 Arguments : none
8645
8646 In HTTP, each payload is unidirectional and has no notion of interactivity.
8647 Any agent is expected to queue data somewhat for a reasonably low delay.
8648 There are some very rare server-to-server applications that abuse the HTTP
8649 protocol and expect the payload phase to be highly interactive, with many
8650 interleaved data chunks in both directions within a single request. This is
8651 absolutely not supported by the HTTP specification and will not work across
8652 most proxies or servers. When such applications attempt to do this through
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008653 HAProxy, it works but they will experience high delays due to the network
Willy Tarreau96e31212011-05-30 18:10:30 +02008654 optimizations which favor performance by instructing the system to wait for
8655 enough data to be available in order to only send full packets. Typical
8656 delays are around 200 ms per round trip. Note that this only happens with
8657 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
8658 affected.
8659
8660 When "option http-no-delay" is present in either the frontend or the backend
8661 used by a connection, all such optimizations will be disabled in order to
8662 make the exchanges as fast as possible. Of course this offers no guarantee on
8663 the functionality, as it may break at any other place. But if it works via
8664 HAProxy, it will work as fast as possible. This option should never be used
8665 by default, and should never be used at all unless such a buggy application
8666 is discovered. The impact of using this option is an increase of bandwidth
8667 usage and CPU usage, which may significantly lower performance in high
8668 latency environments.
8669
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02008670 See also : "option http-buffer-request"
8671
Willy Tarreau96e31212011-05-30 18:10:30 +02008672
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008673option http-pretend-keepalive
8674no option http-pretend-keepalive
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008675 Define whether HAProxy will announce keepalive to the server or not
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008676 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02008677 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008678 Arguments : none
8679
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008680 When running with "option http-server-close" or "option httpclose", HAProxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008681 adds a "Connection: close" header to the request forwarded to the server.
8682 Unfortunately, when some servers see this header, they automatically refrain
8683 from using the chunked encoding for responses of unknown length, while this
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008684 is totally unrelated. The immediate effect is that this prevents HAProxy from
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008685 maintaining the client connection alive. A second effect is that a client or
8686 a cache could receive an incomplete response without being aware of it, and
8687 consider the response complete.
8688
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008689 By setting "option http-pretend-keepalive", HAProxy will make the server
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008690 believe it will keep the connection alive. The server will then not fall back
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008691 to the abnormal undesired above. When HAProxy gets the whole response, it
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008692 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02008693 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008694 connection is correctly closed on the server side.
8695
8696 It is recommended not to enable this option by default, because most servers
8697 will more efficiently close the connection themselves after the last packet,
8698 and release its buffers slightly earlier. Also, the added packet on the
8699 network could slightly reduce the overall peak performance. However it is
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008700 worth noting that when this option is enabled, HAProxy will have slightly
8701 less work to do. So if HAProxy is the bottleneck on the whole architecture,
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008702 enabling this option might save a few CPU cycles.
8703
Christopher Faulet98db9762018-09-21 10:25:19 +02008704 This option may be set in backend and listen sections. Using it in a frontend
8705 section will be ignored and a warning will be reported during startup. It is
8706 a backend related option, so there is no real reason to set it on a
8707 frontend. This option may be combined with "option httpclose", which will
8708 cause keepalive to be announced to the server and close to be announced to
8709 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008710
8711 If this option has been enabled in a "defaults" section, it can be disabled
8712 in a specific instance by prepending the "no" keyword before it.
8713
Christopher Faulet315b39c2018-09-21 16:26:19 +02008714 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01008715 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02008716
Willy Tarreauc27debf2008-01-06 08:57:02 +01008717
Willy Tarreaub608feb2010-01-02 22:47:18 +01008718option http-server-close
8719no option http-server-close
8720 Enable or disable HTTP connection closing on the server side
8721 May be used in sections : defaults | frontend | listen | backend
8722 yes | yes | yes | yes
8723 Arguments : none
8724
Willy Tarreau70dffda2014-01-30 03:07:23 +01008725 By default HAProxy operates in keep-alive mode with regards to persistent
8726 connections: for each connection it processes each request and response, and
8727 leaves the connection idle on both sides between the end of a response and
8728 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02008729 as "option http-server-close" or "option httpclose". Setting "option
8730 http-server-close" enables HTTP connection-close mode on the server side
8731 while keeping the ability to support HTTP keep-alive and pipelining on the
8732 client side. This provides the lowest latency on the client side (slow
8733 network) and the fastest session reuse on the server side to save server
8734 resources, similarly to "option httpclose". It also permits non-keepalive
8735 capable servers to be served in keep-alive mode to the clients if they
8736 conform to the requirements of RFC7230. Please note that some servers do not
8737 always conform to those requirements when they see "Connection: close" in the
8738 request. The effect will be that keep-alive will never be used. A workaround
8739 consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01008740
8741 At the moment, logs will not indicate whether requests came from the same
8742 session or not. The accept date reported in the logs corresponds to the end
8743 of the previous request, and the request time corresponds to the time spent
8744 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01008745 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
8746 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01008747
8748 This option may be set both in a frontend and in a backend. It is enabled if
8749 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02008750 It disables and replaces any previous "option httpclose" or "option
8751 http-keep-alive". Please check section 4 ("Proxies") to see how this option
8752 combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01008753
8754 If this option has been enabled in a "defaults" section, it can be disabled
8755 in a specific instance by prepending the "no" keyword before it.
8756
Christopher Faulet315b39c2018-09-21 16:26:19 +02008757 See also : "option httpclose", "option http-pretend-keepalive",
8758 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01008759
Willy Tarreau88d349d2010-01-25 12:15:43 +01008760option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01008761no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01008762 Make use of non-standard Proxy-Connection header instead of Connection
8763 May be used in sections : defaults | frontend | listen | backend
8764 yes | yes | yes | no
8765 Arguments : none
8766
Lukas Tribus23953682017-04-28 13:24:30 +00008767 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01008768 Connection header to indicate their wish of persistent or non-persistent
8769 connections, both browsers and proxies ignore this header for proxied
8770 connections and make use of the undocumented, non-standard Proxy-Connection
8771 header instead. The issue begins when trying to put a load balancer between
8772 browsers and such proxies, because there will be a difference between what
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008773 HAProxy understands and what the client and the proxy agree on.
Willy Tarreau88d349d2010-01-25 12:15:43 +01008774
Daniel Corbett9f0843f2021-05-08 10:50:37 -04008775 By setting this option in a frontend, HAProxy can automatically switch to use
Willy Tarreau88d349d2010-01-25 12:15:43 +01008776 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01008777 defined here as one where the URI begins with neither a '/' nor a '*'. This
8778 is incompatible with the HTTP tunnel mode. Note that this option can only be
8779 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01008780
Willy Tarreau844a7e72010-01-31 21:46:18 +01008781 Also, when this option is set, a request which requires authentication will
8782 automatically switch to use proxy authentication headers if it is itself a
8783 proxied request. That makes it possible to check or enforce authentication in
8784 front of an existing proxy.
8785
Willy Tarreau88d349d2010-01-25 12:15:43 +01008786 This option should normally never be used, except in front of a proxy.
8787
Christopher Faulet315b39c2018-09-21 16:26:19 +02008788 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01008789
Willy Tarreaud63335a2010-02-26 12:56:52 +01008790option httpchk
8791option httpchk <uri>
8792option httpchk <method> <uri>
8793option httpchk <method> <uri> <version>
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02008794 Enables HTTP protocol to check on the servers health
Willy Tarreaud63335a2010-02-26 12:56:52 +01008795 May be used in sections : defaults | frontend | listen | backend
8796 yes | no | yes | yes
8797 Arguments :
8798 <method> is the optional HTTP method used with the requests. When not set,
8799 the "OPTIONS" method is used, as it generally requires low server
8800 processing and is easy to filter out from the logs. Any method
8801 may be used, though it is not recommended to invent non-standard
8802 ones.
8803
8804 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
8805 which is accessible by default on almost any server, but may be
8806 changed to any other URI. Query strings are permitted.
8807
8808 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
8809 but some servers might behave incorrectly in HTTP 1.0, so turning
8810 it to HTTP/1.1 may sometimes help. Note that the Host field is
Christopher Faulet8acb1282020-04-09 08:44:06 +02008811 mandatory in HTTP/1.1, use "http-check send" directive to add it.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008812
8813 By default, server health checks only consist in trying to establish a TCP
8814 connection. When "option httpchk" is specified, a complete HTTP request is
8815 sent once the TCP connection is established, and responses 2xx and 3xx are
8816 considered valid, while all other ones indicate a server failure, including
8817 the lack of any response.
8818
Christopher Faulete5870d82020-04-15 11:32:03 +02008819 Combined with "http-check" directives, it is possible to customize the
8820 request sent during the HTTP health checks or the matching rules on the
8821 response. It is also possible to configure a send/expect sequence, just like
8822 with the directive "tcp-check" for TCP health checks.
8823
8824 The server configuration is used by default to open connections to perform
8825 HTTP health checks. By it is also possible to overwrite server parameters
8826 using "http-check connect" rules.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008827
Christopher Faulete5870d82020-04-15 11:32:03 +02008828 "httpchk" option does not necessarily require an HTTP backend, it also works
8829 with plain TCP backends. This is particularly useful to check simple scripts
Christopher Faulet14cd3162020-04-16 14:50:06 +02008830 bound to some dedicated ports using the inetd daemon. However, it will always
Daniel Corbett67a82712020-07-06 23:01:19 -04008831 internally relies on an HTX multiplexer. Thus, it means the request
Christopher Faulet14cd3162020-04-16 14:50:06 +02008832 formatting and the response parsing will be strict.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008833
Christopher Faulet8acb1282020-04-09 08:44:06 +02008834 Note : For a while, there was no way to add headers or body in the request
8835 used for HTTP health checks. So a workaround was to hide it at the end
8836 of the version string with a "\r\n" after the version. It is now
8837 deprecated. The directive "http-check send" must be used instead.
8838
Willy Tarreaud63335a2010-02-26 12:56:52 +01008839 Examples :
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02008840 # Relay HTTPS traffic to Apache instance and check service availability
8841 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
8842 backend https_relay
8843 mode tcp
8844 option httpchk OPTIONS * HTTP/1.1
8845 http-check send hdr Host www
8846 server apache1 192.168.1.1:443 check port 80
Willy Tarreaud63335a2010-02-26 12:56:52 +01008847
Simon Hormanafc47ee2013-11-25 10:46:35 +09008848 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
8849 "option pgsql-check", "http-check" and the "check", "port" and
8850 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008851
8852
Willy Tarreauc27debf2008-01-06 08:57:02 +01008853option httpclose
8854no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02008855 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01008856 May be used in sections : defaults | frontend | listen | backend
8857 yes | yes | yes | yes
8858 Arguments : none
8859
Willy Tarreau70dffda2014-01-30 03:07:23 +01008860 By default HAProxy operates in keep-alive mode with regards to persistent
8861 connections: for each connection it processes each request and response, and
8862 leaves the connection idle on both sides between the end of a response and
8863 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02008864 as "option http-server-close" or "option httpclose".
Willy Tarreau70dffda2014-01-30 03:07:23 +01008865
Christopher Faulet315b39c2018-09-21 16:26:19 +02008866 If "option httpclose" is set, HAProxy will close connections with the server
8867 and the client as soon as the request and the response are received. It will
John Roeslerfb2fce12019-07-10 15:45:51 -05008868 also check if a "Connection: close" header is already set in each direction,
Christopher Faulet315b39c2018-09-21 16:26:19 +02008869 and will add one if missing. Any "Connection" header different from "close"
8870 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008871
Christopher Faulet315b39c2018-09-21 16:26:19 +02008872 This option may also be combined with "option http-pretend-keepalive", which
8873 will disable sending of the "Connection: close" header, but will still cause
8874 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008875
8876 This option may be set both in a frontend and in a backend. It is enabled if
8877 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02008878 It disables and replaces any previous "option http-server-close" or "option
8879 http-keep-alive". Please check section 4 ("Proxies") to see how this option
8880 combines with others when frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008881
8882 If this option has been enabled in a "defaults" section, it can be disabled
8883 in a specific instance by prepending the "no" keyword before it.
8884
Christopher Faulet315b39c2018-09-21 16:26:19 +02008885 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01008886
8887
Emeric Brun3a058f32009-06-30 18:26:00 +02008888option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01008889 Enable logging of HTTP request, session state and timers
8890 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01008891 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02008892 Arguments :
8893 clf if the "clf" argument is added, then the output format will be
8894 the CLF format instead of HAProxy's default HTTP format. You can
8895 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01008896 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02008897 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008898
8899 By default, the log output format is very poor, as it only contains the
8900 source and destination addresses, and the instance name. By specifying
8901 "option httplog", each log line turns into a much richer format including,
8902 but not limited to, the HTTP request, the connection timers, the session
8903 status, the connections numbers, the captured headers and cookies, the
8904 frontend, backend and server name, and of course the source address and
8905 ports.
8906
PiBa-NLbd556bf2014-12-11 21:31:54 +01008907 Specifying only "option httplog" will automatically clear the 'clf' mode
8908 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02008909
Guillaume de Lafond29f45602017-03-31 19:52:15 +02008910 "option httplog" overrides any previous "log-format" directive.
8911
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008912 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01008913
Willy Tarreau55165fe2009-05-10 12:02:55 +02008914
8915option http_proxy
8916no option http_proxy
8917 Enable or disable plain HTTP proxy mode
8918 May be used in sections : defaults | frontend | listen | backend
8919 yes | yes | yes | yes
8920 Arguments : none
8921
8922 It sometimes happens that people need a pure HTTP proxy which understands
8923 basic proxy requests without caching nor any fancy feature. In this case,
8924 it may be worth setting up an HAProxy instance with the "option http_proxy"
8925 set. In this mode, no server is declared, and the connection is forwarded to
8926 the IP address and port found in the URL after the "http://" scheme.
8927
8928 No host address resolution is performed, so this only works when pure IP
8929 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01008930 it will probably be used only by experts who know they need exactly it. This
8931 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02008932
8933 If this option has been enabled in a "defaults" section, it can be disabled
8934 in a specific instance by prepending the "no" keyword before it.
8935
8936 Example :
8937 # this backend understands HTTP proxy requests and forwards them directly.
8938 backend direct_forward
8939 option httpclose
8940 option http_proxy
8941
8942 See also : "option httpclose"
8943
Willy Tarreau211ad242009-10-03 21:45:07 +02008944
Jamie Gloudon801a0a32012-08-25 00:18:33 -04008945option independent-streams
8946no option independent-streams
8947 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02008948 May be used in sections : defaults | frontend | listen | backend
8949 yes | yes | yes | yes
8950 Arguments : none
8951
8952 By default, when data is sent over a socket, both the write timeout and the
8953 read timeout for that socket are refreshed, because we consider that there is
8954 activity on that socket, and we have no other means of guessing if we should
8955 receive data or not.
8956
Davor Ocelice9ed2812017-12-25 17:49:28 +01008957 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02008958 exists a situation where it is desirable to disable it, and only refresh the
8959 read timeout if there are incoming data. This happens on sessions with large
8960 timeouts and low amounts of exchanged data such as telnet session. If the
8961 server suddenly disappears, the output data accumulates in the system's
8962 socket buffers, both timeouts are correctly refreshed, and there is no way
8963 to know the server does not receive them, so we don't timeout. However, when
8964 the underlying protocol always echoes sent data, it would be enough by itself
8965 to detect the issue using the read timeout. Note that this problem does not
8966 happen with more verbose protocols because data won't accumulate long in the
8967 socket buffers.
8968
8969 When this option is set on the frontend, it will disable read timeout updates
8970 on data sent to the client. There probably is little use of this case. When
8971 the option is set on the backend, it will disable read timeout updates on
8972 data sent to the server. Doing so will typically break large HTTP posts from
8973 slow lines, so use it with caution.
8974
Willy Tarreauce887fd2012-05-12 12:50:00 +02008975 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02008976
8977
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02008978option ldap-check
8979 Use LDAPv3 health checks for server testing
8980 May be used in sections : defaults | frontend | listen | backend
8981 yes | no | yes | yes
8982 Arguments : none
8983
8984 It is possible to test that the server correctly talks LDAPv3 instead of just
8985 testing that it accepts the TCP connection. When this option is set, an
8986 LDAPv3 anonymous simple bind message is sent to the server, and the response
8987 is analyzed to find an LDAPv3 bind response message.
8988
8989 The server is considered valid only when the LDAP response contains success
8990 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
8991
8992 Logging of bind requests is server dependent see your documentation how to
8993 configure it.
8994
8995 Example :
8996 option ldap-check
8997
8998 See also : "option httpchk"
8999
9000
Simon Horman98637e52014-06-20 12:30:16 +09009001option external-check
9002 Use external processes for server health checks
9003 May be used in sections : defaults | frontend | listen | backend
9004 yes | no | yes | yes
9005
9006 It is possible to test the health of a server using an external command.
9007 This is achieved by running the executable set using "external-check
9008 command".
9009
9010 Requires the "external-check" global to be set.
9011
9012 See also : "external-check", "external-check command", "external-check path"
9013
9014
William Dauchy4711a892022-01-05 22:53:24 +01009015option idle-close-on-response
9016no option idle-close-on-response
9017 Avoid closing idle frontend connections if a soft stop is in progress
9018 May be used in sections : defaults | frontend | listen | backend
9019 yes | yes | yes | no
9020 Arguments : none
9021
9022 By default, idle connections will be closed during a soft stop. In some
9023 environments, a client talking to the proxy may have prepared some idle
9024 connections in order to send requests later. If there is no proper retry on
9025 write errors, this can result in errors while haproxy is reloading. Even
9026 though a proper implementation should retry on connection/write errors, this
9027 option was introduced to support backwards compatibility with haproxy prior
9028 to version 2.4. Indeed before v2.4, haproxy used to wait for a last request
9029 and response to add a "connection: close" header before closing, thus
9030 notifying the client that the connection would not be reusable.
9031
9032 In a real life example, this behavior was seen in AWS using the ALB in front
9033 of a haproxy. The end result was ALB sending 502 during haproxy reloads.
9034
9035 Users are warned that using this option may increase the number of old
9036 processes if connections remain idle for too long. Adjusting the client
9037 timeouts and/or the "hard-stop-after" parameter accordingly might be
9038 needed in case of frequent reloads.
9039
9040 See also: "timeout client", "timeout client-fin", "timeout http-request",
9041 "hard-stop-after"
9042
9043
Willy Tarreau211ad242009-10-03 21:45:07 +02009044option log-health-checks
9045no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02009046 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02009047 May be used in sections : defaults | frontend | listen | backend
9048 yes | no | yes | yes
9049 Arguments : none
9050
Willy Tarreaubef1b322014-05-13 21:01:39 +02009051 By default, failed health check are logged if server is UP and successful
9052 health checks are logged if server is DOWN, so the amount of additional
9053 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02009054
Willy Tarreaubef1b322014-05-13 21:01:39 +02009055 When this option is enabled, any change of the health check status or to
9056 the server's health will be logged, so that it becomes possible to know
9057 that a server was failing occasional checks before crashing, or exactly when
9058 it failed to respond a valid HTTP status, then when the port started to
9059 reject connections, then when the server stopped responding at all.
9060
Davor Ocelice9ed2812017-12-25 17:49:28 +01009061 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02009062 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02009063
Willy Tarreaubef1b322014-05-13 21:01:39 +02009064 See also: "option httpchk", "option ldap-check", "option mysql-check",
9065 "option pgsql-check", "option redis-check", "option smtpchk",
9066 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02009067
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02009068
9069option log-separate-errors
9070no option log-separate-errors
9071 Change log level for non-completely successful connections
9072 May be used in sections : defaults | frontend | listen | backend
9073 yes | yes | yes | no
9074 Arguments : none
9075
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009076 Sometimes looking for errors in logs is not easy. This option makes HAProxy
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02009077 raise the level of logs containing potentially interesting information such
9078 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
9079 level changes from "info" to "err". This makes it possible to log them
9080 separately to a different file with most syslog daemons. Be careful not to
9081 remove them from the original file, otherwise you would lose ordering which
9082 provides very important information.
9083
9084 Using this option, large sites dealing with several thousand connections per
9085 second may log normal traffic to a rotating buffer and only archive smaller
9086 error logs.
9087
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009088 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02009089 logging.
9090
Willy Tarreauc27debf2008-01-06 08:57:02 +01009091
9092option logasap
9093no option logasap
Jerome Magnin95fb57b2020-04-23 19:01:17 +02009094 Enable or disable early logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01009095 May be used in sections : defaults | frontend | listen | backend
9096 yes | yes | yes | no
9097 Arguments : none
9098
Jerome Magnin95fb57b2020-04-23 19:01:17 +02009099 By default, logs are emitted when all the log format variables and sample
9100 fetches used in the definition of the log-format string return a value, or
9101 when the session is terminated. This allows the built in log-format strings
9102 to account for the transfer time, or the number of bytes in log messages.
9103
9104 When handling long lived connections such as large file transfers or RDP,
9105 it may take a while for the request or connection to appear in the logs.
9106 Using "option logasap", the log message is created as soon as the server
9107 connection is established in mode tcp, or as soon as the server sends the
9108 complete headers in mode http. Missing information in the logs will be the
Ilya Shipitsin4329a9a2020-05-05 21:17:10 +05009109 total number of bytes which will only indicate the amount of data transferred
Jerome Magnin95fb57b2020-04-23 19:01:17 +02009110 before the message was created and the total time which will not take the
9111 remainder of the connection life or transfer time into account. For the case
9112 of HTTP, it is good practice to capture the Content-Length response header
9113 so that the logs at least indicate how many bytes are expected to be
Ilya Shipitsin4329a9a2020-05-05 21:17:10 +05009114 transferred.
Willy Tarreauc27debf2008-01-06 08:57:02 +01009115
Willy Tarreaucc6c8912009-02-22 10:53:55 +01009116 Examples :
9117 listen http_proxy 0.0.0.0:80
9118 mode http
9119 option httplog
9120 option logasap
9121 log 192.168.2.200 local3
9122
9123 >>> Feb 6 12:14:14 localhost \
9124 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
9125 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
9126 "GET /image.iso HTTP/1.0"
9127
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009128 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01009129 logging.
9130
9131
Christopher Faulet62f79fe2020-05-18 18:13:03 +02009132option mysql-check [ user <username> [ { post-41 | pre-41 } ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009133 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01009134 May be used in sections : defaults | frontend | listen | backend
9135 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009136 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009137 <username> This is the username which will be used when connecting to MySQL
9138 server.
Christopher Faulet62f79fe2020-05-18 18:13:03 +02009139 post-41 Send post v4.1 client compatible checks (the default)
9140 pre-41 Send pre v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009141
9142 If you specify a username, the check consists of sending two MySQL packet,
9143 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01009144 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009145 Error packet. It is a basic but useful test which does not produce error nor
Daniel Black9dc310d2021-07-01 12:09:32 +10009146 aborted connect on the server. However, it requires an unlocked authorised
9147 user without a password. To create a basic limited user in MySQL with
9148 optional resource limits:
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009149
Daniel Black9dc310d2021-07-01 12:09:32 +10009150 CREATE USER '<username>'@'<ip_of_haproxy|network_of_haproxy/netmask>'
9151 /*!50701 WITH MAX_QUERIES_PER_HOUR 1 MAX_UPDATES_PER_HOUR 0 */
9152 /*M!100201 MAX_STATEMENT_TIME 0.0001 */;
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009153
9154 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009155 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02009156 Error packet, we don't send anything in this mode. It was reported that it
9157 can generate lockout if check is too frequent and/or if there is not enough
9158 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
9159 value as if a connection is established successfully within fewer than MySQL
9160 "max_connect_errors" attempts after a previous connection was interrupted,
9161 the error count for the host is cleared to zero. If HAProxy's server get
9162 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
9163
9164 Remember that this does not check database presence nor database consistency.
9165 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01009166
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02009167 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01009168
9169 Most often, an incoming MySQL server needs to see the client's IP address for
9170 various purposes, including IP privilege matching and connection logging.
9171 When possible, it is often wise to masquerade the client's IP address when
9172 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02009173 which requires the transparent proxy feature to be compiled in, and the MySQL
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009174 server to route the client via the machine hosting HAProxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01009175
9176 See also: "option httpchk"
9177
9178
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009179option nolinger
9180no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009181 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009182 May be used in sections: defaults | frontend | listen | backend
9183 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009184 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009185
Davor Ocelice9ed2812017-12-25 17:49:28 +01009186 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009187 physically disconnected), the session timeouts triggers and the session is
9188 closed. But it will remain in FIN_WAIT1 state for some time in the system,
9189 using some resources and possibly limiting the ability to establish newer
9190 connections.
9191
9192 When this happens, it is possible to activate "option nolinger" which forces
9193 the system to immediately remove any socket's pending data on close. Thus,
Willy Tarreau4a321032020-10-16 04:55:19 +02009194 a TCP RST is emitted, any pending data are truncated, and the session is
9195 instantly purged from the system's tables. The generally visible effect for
9196 a client is that responses are truncated if the close happens with a last
9197 block of data (e.g. on a redirect or error response). On the server side,
9198 it may help release the source ports immediately when forwarding a client
9199 aborts in tunnels. In both cases, TCP resets are emitted and given that
9200 the session is instantly destroyed, there will be no retransmit. On a lossy
9201 network this can increase problems, especially when there is a firewall on
9202 the lossy side, because the firewall might see and process the reset (hence
9203 purge its session) and block any further traffic for this session,, including
9204 retransmits from the other side. So if the other side doesn't receive it,
9205 it will never receive any RST again, and the firewall might log many blocked
9206 packets.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009207
Willy Tarreau4a321032020-10-16 04:55:19 +02009208 For all these reasons, it is strongly recommended NOT to use this option,
9209 unless absolutely needed as a last resort. In most situations, using the
9210 "client-fin" or "server-fin" timeouts achieves similar results with a more
9211 reliable behavior. On Linux it's also possible to use the "tcp-ut" bind or
9212 server setting.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009213
9214 This option may be used both on frontends and backends, depending on the side
9215 where it is required. Use it on the frontend for clients, and on the backend
Willy Tarreau4a321032020-10-16 04:55:19 +02009216 for servers. While this option is technically supported in "defaults"
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +05009217 sections, it must really not be used there as it risks to accidentally
Willy Tarreau4a321032020-10-16 04:55:19 +02009218 propagate to sections that must no use it and to cause problems there.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009219
9220 If this option has been enabled in a "defaults" section, it can be disabled
9221 in a specific instance by prepending the "no" keyword before it.
9222
Willy Tarreau4a321032020-10-16 04:55:19 +02009223 See also: "timeout client-fin", "timeout server-fin", "tcp-ut" bind or server
9224 keywords.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009225
Willy Tarreau55165fe2009-05-10 12:02:55 +02009226option originalto [ except <network> ] [ header <name> ]
9227 Enable insertion of the X-Original-To header to requests sent to servers
9228 May be used in sections : defaults | frontend | listen | backend
9229 yes | yes | yes | yes
9230 Arguments :
9231 <network> is an optional argument used to disable this option for sources
9232 matching <network>
9233 <name> an optional argument to specify a different "X-Original-To"
9234 header name.
9235
9236 Since HAProxy can work in transparent mode, every request from a client can
9237 be redirected to the proxy and HAProxy itself can proxy every request to a
9238 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
9239 be lost. This is annoying when you want access rules based on destination ip
9240 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
9241 added by HAProxy to all requests sent to the server. This header contains a
9242 value representing the original destination IP address. Since this must be
9243 configured to always use the last occurrence of this header only. Note that
9244 only the last occurrence of the header must be used, since it is really
9245 possible that the client has already brought one.
9246
9247 The keyword "header" may be used to supply a different header name to replace
9248 the default "X-Original-To". This can be useful where you might already
9249 have a "X-Original-To" header from a different application, and you need
9250 preserve it. Also if your backend server doesn't use the "X-Original-To"
9251 header and requires different one.
9252
9253 Sometimes, a same HAProxy instance may be shared between a direct client
9254 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
9255 used to decrypt HTTPS traffic). It is possible to disable the addition of the
Amaury Denoyellef8b42922021-03-04 18:41:14 +01009256 header for a known destination address or network by adding the "except"
9257 keyword followed by the network address. In this case, any destination IP
9258 matching the network will not cause an addition of this header. Most common
9259 uses are with private networks or 127.0.0.1. IPv4 and IPv6 are both
9260 supported.
Willy Tarreau55165fe2009-05-10 12:02:55 +02009261
9262 This option may be specified either in the frontend or in the backend. If at
9263 least one of them uses it, the header will be added. Note that the backend's
9264 setting of the header subargument takes precedence over the frontend's if
9265 both are defined.
9266
Willy Tarreau55165fe2009-05-10 12:02:55 +02009267 Examples :
9268 # Original Destination address
9269 frontend www
9270 mode http
9271 option originalto except 127.0.0.1
9272
9273 # Those servers want the IP Address in X-Client-Dst
9274 backend www
9275 mode http
9276 option originalto header X-Client-Dst
9277
Christopher Faulet315b39c2018-09-21 16:26:19 +02009278 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02009279
9280
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009281option persist
9282no option persist
9283 Enable or disable forced persistence on down servers
9284 May be used in sections: defaults | frontend | listen | backend
9285 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009286 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009287
9288 When an HTTP request reaches a backend with a cookie which references a dead
9289 server, by default it is redispatched to another server. It is possible to
9290 force the request to be sent to the dead server first using "option persist"
9291 if absolutely needed. A common use case is when servers are under extreme
9292 load and spend their time flapping. In this case, the users would still be
9293 directed to the server they opened the session on, in the hope they would be
9294 correctly served. It is recommended to use "option redispatch" in conjunction
9295 with this option so that in the event it would not be possible to connect to
9296 the server at all (server definitely dead), the client would finally be
9297 redirected to another valid server.
9298
9299 If this option has been enabled in a "defaults" section, it can be disabled
9300 in a specific instance by prepending the "no" keyword before it.
9301
Willy Tarreau4de91492010-01-22 19:10:05 +01009302 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009303
9304
Willy Tarreau0c122822013-12-15 18:49:01 +01009305option pgsql-check [ user <username> ]
9306 Use PostgreSQL health checks for server testing
9307 May be used in sections : defaults | frontend | listen | backend
9308 yes | no | yes | yes
9309 Arguments :
9310 <username> This is the username which will be used when connecting to
9311 PostgreSQL server.
9312
9313 The check sends a PostgreSQL StartupMessage and waits for either
9314 Authentication request or ErrorResponse message. It is a basic but useful
9315 test which does not produce error nor aborted connect on the server.
9316 This check is identical with the "mysql-check".
9317
9318 See also: "option httpchk"
9319
9320
Willy Tarreau9420b122013-12-15 18:58:25 +01009321option prefer-last-server
9322no option prefer-last-server
9323 Allow multiple load balanced requests to remain on the same server
9324 May be used in sections: defaults | frontend | listen | backend
9325 yes | no | yes | yes
9326 Arguments : none
9327
9328 When the load balancing algorithm in use is not deterministic, and a previous
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009329 request was sent to a server to which HAProxy still holds a connection, it is
Willy Tarreau9420b122013-12-15 18:58:25 +01009330 sometimes desirable that subsequent requests on a same session go to the same
9331 server as much as possible. Note that this is different from persistence, as
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009332 we only indicate a preference which HAProxy tries to apply without any form
Willy Tarreau9420b122013-12-15 18:58:25 +01009333 of warranty. The real use is for keep-alive connections sent to servers. When
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009334 this option is used, HAProxy will try to reuse the same connection that is
Willy Tarreau9420b122013-12-15 18:58:25 +01009335 attached to the server instead of rebalancing to another server, causing a
9336 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01009337 not make much sense to use this in combination with hashing algorithms. Note,
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009338 HAProxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02009339 to a proxy which sends a 407 (authentication required), when the load
9340 balancing algorithm is not deterministic. This is mandatory for use with the
9341 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01009342 troubleshooting some faulty applications. Option prefer-last-server might be
9343 desirable in these environments as well, to avoid redistributing the traffic
9344 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01009345
9346 If this option has been enabled in a "defaults" section, it can be disabled
9347 in a specific instance by prepending the "no" keyword before it.
9348
9349 See also: "option http-keep-alive"
9350
9351
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009352option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07009353option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009354no option redispatch
9355 Enable or disable session redistribution in case of connection failure
9356 May be used in sections: defaults | frontend | listen | backend
9357 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07009358 Arguments :
9359 <interval> The optional integer value that controls how often redispatches
9360 occur when retrying connections. Positive value P indicates a
9361 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01009362 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07009363 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009364 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07009365 positive value of 1 would indicate a redispatch on every retry,
9366 and a positive value of 3 would indicate a redispatch on every
9367 third retry. You can disable redispatches with a value of 0.
9368
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009369
9370 In HTTP mode, if a server designated by a cookie is down, clients may
9371 definitely stick to it because they cannot flush the cookie, so they will not
9372 be able to access the service anymore.
9373
Willy Tarreau59884a62019-01-02 14:48:31 +01009374 Specifying "option redispatch" will allow the proxy to break cookie or
9375 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009376
Olivier Carrère6e6f59b2020-04-15 11:30:18 +02009377 Active servers are selected from a subset of the list of available
9378 servers. Active servers that are not down or in maintenance (i.e., whose
9379 health is not checked or that have been checked as "up"), are selected in the
9380 following order:
9381
9382 1. Any active, non-backup server, if any, or,
9383
9384 2. If the "allbackups" option is not set, the first backup server in the
9385 list, or
9386
9387 3. If the "allbackups" option is set, any backup server.
9388
9389 When a retry occurs, HAProxy tries to select another server than the last
9390 one. The new server is selected from the current list of servers.
9391
9392 Sometimes, if the list is updated between retries (e.g., if numerous retries
9393 occur and last longer than the time needed to check that a server is down,
9394 remove it from the list and fall back on the list of backup servers),
9395 connections may be redirected to a backup server, though.
9396
Joseph Lynch726ab712015-05-11 23:25:34 -07009397 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009398 connection failures. Of course, it requires having "retries" set to a nonzero
9399 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01009400
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009401 If this option has been enabled in a "defaults" section, it can be disabled
9402 in a specific instance by prepending the "no" keyword before it.
9403
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02009404 See also : "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009405
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009406
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02009407option redis-check
9408 Use redis health checks for server testing
9409 May be used in sections : defaults | frontend | listen | backend
9410 yes | no | yes | yes
9411 Arguments : none
9412
9413 It is possible to test that the server correctly talks REDIS protocol instead
9414 of just testing that it accepts the TCP connection. When this option is set,
9415 a PING redis command is sent to the server, and the response is analyzed to
9416 find the "+PONG" response message.
9417
9418 Example :
9419 option redis-check
9420
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009421 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02009422
9423
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009424option smtpchk
9425option smtpchk <hello> <domain>
9426 Use SMTP health checks for server testing
9427 May be used in sections : defaults | frontend | listen | backend
9428 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01009429 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009430 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02009431 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009432 values will be turned into the default command ("HELO").
9433
9434 <domain> is the domain name to present to the server. It may only be
9435 specified (and is mandatory) if the hello command has been
9436 specified. By default, "localhost" is used.
9437
9438 When "option smtpchk" is set, the health checks will consist in TCP
9439 connections followed by an SMTP command. By default, this command is
9440 "HELO localhost". The server's return code is analyzed and only return codes
9441 starting with a "2" will be considered as valid. All other responses,
9442 including a lack of response will constitute an error and will indicate a
9443 dead server.
9444
9445 This test is meant to be used with SMTP servers or relays. Depending on the
9446 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01009447 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009448 25 is often easier than adjusting the configuration.
9449
9450 Most often, an incoming SMTP server needs to see the client's IP address for
9451 various purposes, including spam filtering, anti-spoofing and logging. When
9452 possible, it is often wise to masquerade the client's IP address when
9453 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02009454 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009455
9456 Example :
9457 option smtpchk HELO mydomain.org
9458
9459 See also : "option httpchk", "source"
9460
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01009461
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02009462option socket-stats
9463no option socket-stats
9464
9465 Enable or disable collecting & providing separate statistics for each socket.
9466 May be used in sections : defaults | frontend | listen | backend
9467 yes | yes | yes | no
9468
9469 Arguments : none
9470
9471
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009472option splice-auto
9473no option splice-auto
9474 Enable or disable automatic kernel acceleration on sockets in both directions
9475 May be used in sections : defaults | frontend | listen | backend
9476 yes | yes | yes | yes
9477 Arguments : none
9478
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009479 When this option is enabled either on a frontend or on a backend, HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009480 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01009481 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009482 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009483 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009484 are not much aggressive in order to limit excessive use of splicing. This
9485 option requires splicing to be enabled at compile time, and may be globally
9486 disabled with the global option "nosplice". Since splice uses pipes, using it
9487 requires that there are enough spare pipes.
9488
9489 Important note: kernel-based TCP splicing is a Linux-specific feature which
9490 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
9491 transfer data between sockets without copying these data to user-space, thus
9492 providing noticeable performance gains and CPU cycles savings. Since many
9493 early implementations are buggy, corrupt data and/or are inefficient, this
9494 feature is not enabled by default, and it should be used with extreme care.
9495 While it is not possible to detect the correctness of an implementation,
9496 2.6.29 is the first version offering a properly working implementation. In
9497 case of doubt, splicing may be globally disabled using the global "nosplice"
9498 keyword.
9499
9500 Example :
9501 option splice-auto
9502
9503 If this option has been enabled in a "defaults" section, it can be disabled
9504 in a specific instance by prepending the "no" keyword before it.
9505
9506 See also : "option splice-request", "option splice-response", and global
9507 options "nosplice" and "maxpipes"
9508
9509
9510option splice-request
9511no option splice-request
9512 Enable or disable automatic kernel acceleration on sockets for requests
9513 May be used in sections : defaults | frontend | listen | backend
9514 yes | yes | yes | yes
9515 Arguments : none
9516
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009517 When this option is enabled either on a frontend or on a backend, HAProxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009518 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009519 the client to the server. It might still use the recv/send scheme if there
9520 are no spare pipes left. This option requires splicing to be enabled at
9521 compile time, and may be globally disabled with the global option "nosplice".
9522 Since splice uses pipes, using it requires that there are enough spare pipes.
9523
9524 Important note: see "option splice-auto" for usage limitations.
9525
9526 Example :
9527 option splice-request
9528
9529 If this option has been enabled in a "defaults" section, it can be disabled
9530 in a specific instance by prepending the "no" keyword before it.
9531
9532 See also : "option splice-auto", "option splice-response", and global options
9533 "nosplice" and "maxpipes"
9534
9535
9536option splice-response
9537no option splice-response
9538 Enable or disable automatic kernel acceleration on sockets for responses
9539 May be used in sections : defaults | frontend | listen | backend
9540 yes | yes | yes | yes
9541 Arguments : none
9542
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009543 When this option is enabled either on a frontend or on a backend, HAProxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009544 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01009545 the server to the client. It might still use the recv/send scheme if there
9546 are no spare pipes left. This option requires splicing to be enabled at
9547 compile time, and may be globally disabled with the global option "nosplice".
9548 Since splice uses pipes, using it requires that there are enough spare pipes.
9549
9550 Important note: see "option splice-auto" for usage limitations.
9551
9552 Example :
9553 option splice-response
9554
9555 If this option has been enabled in a "defaults" section, it can be disabled
9556 in a specific instance by prepending the "no" keyword before it.
9557
9558 See also : "option splice-auto", "option splice-request", and global options
9559 "nosplice" and "maxpipes"
9560
9561
Christopher Fauletba7bc162016-11-07 21:07:38 +01009562option spop-check
9563 Use SPOP health checks for server testing
9564 May be used in sections : defaults | frontend | listen | backend
9565 no | no | no | yes
9566 Arguments : none
9567
9568 It is possible to test that the server correctly talks SPOP protocol instead
9569 of just testing that it accepts the TCP connection. When this option is set,
9570 a HELLO handshake is performed between HAProxy and the server, and the
9571 response is analyzed to check no error is reported.
9572
9573 Example :
9574 option spop-check
9575
9576 See also : "option httpchk"
9577
9578
Willy Tarreaubf1f8162007-12-28 17:42:56 +01009579option srvtcpka
9580no option srvtcpka
9581 Enable or disable the sending of TCP keepalive packets on the server side
9582 May be used in sections : defaults | frontend | listen | backend
9583 yes | no | yes | yes
9584 Arguments : none
9585
9586 When there is a firewall or any session-aware component between a client and
9587 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01009588 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01009589 components decides to expire a session which has remained idle for too long.
9590
9591 Enabling socket-level TCP keep-alives makes the system regularly send packets
9592 to the other end of the connection, leaving it active. The delay between
9593 keep-alive probes is controlled by the system only and depends both on the
9594 operating system and its tuning parameters.
9595
9596 It is important to understand that keep-alive packets are neither emitted nor
9597 received at the application level. It is only the network stacks which sees
9598 them. For this reason, even if one side of the proxy already uses keep-alives
9599 to maintain its connection alive, those keep-alive packets will not be
9600 forwarded to the other side of the proxy.
9601
9602 Please note that this has nothing to do with HTTP keep-alive.
9603
9604 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
9605 server side of a connection, which should help when session expirations are
9606 noticed between HAProxy and a server.
9607
9608 If this option has been enabled in a "defaults" section, it can be disabled
9609 in a specific instance by prepending the "no" keyword before it.
9610
9611 See also : "option clitcpka", "option tcpka"
9612
9613
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009614option ssl-hello-chk
9615 Use SSLv3 client hello health checks for server testing
9616 May be used in sections : defaults | frontend | listen | backend
9617 yes | no | yes | yes
9618 Arguments : none
9619
9620 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
9621 possible to test that the server correctly talks SSL instead of just testing
9622 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
9623 SSLv3 client hello messages are sent once the connection is established to
9624 the server, and the response is analyzed to find an SSL server hello message.
9625 The server is considered valid only when the response contains this server
9626 hello message.
9627
9628 All servers tested till there correctly reply to SSLv3 client hello messages,
9629 and most servers tested do not even log the requests containing only hello
9630 messages, which is appreciable.
9631
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009632 Note that this check works even when SSL support was not built into HAProxy
Willy Tarreau763a95b2012-10-04 23:15:39 +02009633 because it forges the SSL message. When SSL support is available, it is best
9634 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009635
Willy Tarreau763a95b2012-10-04 23:15:39 +02009636 See also: "option httpchk", "check-ssl"
9637
Willy Tarreaua453bdd2008-01-08 19:50:52 +01009638
Willy Tarreaued179852013-12-16 01:07:00 +01009639option tcp-check
9640 Perform health checks using tcp-check send/expect sequences
9641 May be used in sections: defaults | frontend | listen | backend
9642 yes | no | yes | yes
9643
9644 This health check method is intended to be combined with "tcp-check" command
9645 lists in order to support send/expect types of health check sequences.
9646
9647 TCP checks currently support 4 modes of operations :
9648 - no "tcp-check" directive : the health check only consists in a connection
9649 attempt, which remains the default mode.
9650
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009651 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01009652 used to send a string along with a connection opening. With some
9653 protocols, it helps sending a "QUIT" message for example that prevents
9654 the server from logging a connection error for each health check. The
9655 check result will still be based on the ability to open the connection
9656 only.
9657
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009658 - "tcp-check expect" only is mentioned : this is used to test a banner.
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009659 The connection is opened and HAProxy waits for the server to present some
Willy Tarreaued179852013-12-16 01:07:00 +01009660 contents which must validate some rules. The check result will be based
9661 on the matching between the contents and the rules. This is suited for
9662 POP, IMAP, SMTP, FTP, SSH, TELNET.
9663
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009664 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009665 used to test a hello-type protocol. HAProxy sends a message, the server
9666 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009667 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01009668 suited for protocols which require a binding or a request/response model.
9669 LDAP, MySQL, Redis and SSL are example of such protocols, though they
9670 already all have their dedicated checks with a deeper understanding of
9671 the respective protocols.
9672 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01009673 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01009674
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02009675 A fifth mode can be used to insert comments in different steps of the script.
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009676
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02009677 For each tcp-check rule you create, you can add a "comment" directive,
9678 followed by a string. This string will be reported in the log and stderr in
9679 debug mode. It is useful to make user-friendly error reporting. The
9680 "comment" is of course optional.
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009681
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02009682 During the execution of a health check, a variable scope is made available to
9683 store data samples, using the "tcp-check set-var" operation. Freeing those
9684 variable is possible using "tcp-check unset-var".
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +01009685
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009686
Willy Tarreaued179852013-12-16 01:07:00 +01009687 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01009688 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01009689 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009690 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01009691
Davor Ocelice9ed2812017-12-25 17:49:28 +01009692 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01009693 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009694 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01009695
9696 # look for the redis master server after ensuring it speaks well
9697 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009698 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01009699 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009700 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01009701 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009702 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009703 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01009704 tcp-check send info\ replication\r\n
9705 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009706 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01009707 tcp-check send QUIT\r\n
9708 tcp-check expect string +OK
9709
Davor Ocelice9ed2812017-12-25 17:49:28 +01009710 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01009711 (send many headers before analyzing)
9712 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009713 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01009714 tcp-check send HEAD\ /\ HTTP/1.1\r\n
9715 tcp-check send Host:\ www.mydomain.com\r\n
9716 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
9717 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02009718 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01009719
9720
Christopher Fauletc52ea4d2020-04-23 15:43:35 +02009721 See also : "tcp-check connect", "tcp-check expect" and "tcp-check send".
Willy Tarreaued179852013-12-16 01:07:00 +01009722
9723
Willy Tarreau9ea05a72009-06-14 12:07:01 +02009724option tcp-smart-accept
9725no option tcp-smart-accept
9726 Enable or disable the saving of one ACK packet during the accept sequence
9727 May be used in sections : defaults | frontend | listen | backend
9728 yes | yes | yes | no
9729 Arguments : none
9730
9731 When an HTTP connection request comes in, the system acknowledges it on
9732 behalf of HAProxy, then the client immediately sends its request, and the
9733 system acknowledges it too while it is notifying HAProxy about the new
9734 connection. HAProxy then reads the request and responds. This means that we
9735 have one TCP ACK sent by the system for nothing, because the request could
9736 very well be acknowledged by HAProxy when it sends its response.
9737
9738 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
9739 sending this useless ACK on platforms which support it (currently at least
9740 Linux). It must not cause any problem, because the system will send it anyway
9741 after 40 ms if the response takes more time than expected to come.
9742
9743 During complex network debugging sessions, it may be desirable to disable
9744 this optimization because delayed ACKs can make troubleshooting more complex
9745 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01009746 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02009747
9748 It is also possible to force it for non-HTTP proxies by simply specifying
9749 "option tcp-smart-accept". For instance, it can make sense with some services
9750 such as SMTP where the server speaks first.
9751
9752 It is recommended to avoid forcing this option in a defaults section. In case
9753 of doubt, consider setting it back to automatic values by prepending the
9754 "default" keyword before it, or disabling it using the "no" keyword.
9755
Willy Tarreaud88edf22009-06-14 15:48:17 +02009756 See also : "option tcp-smart-connect"
9757
9758
9759option tcp-smart-connect
9760no option tcp-smart-connect
9761 Enable or disable the saving of one ACK packet during the connect sequence
9762 May be used in sections : defaults | frontend | listen | backend
9763 yes | no | yes | yes
9764 Arguments : none
9765
9766 On certain systems (at least Linux), HAProxy can ask the kernel not to
9767 immediately send an empty ACK upon a connection request, but to directly
9768 send the buffer request instead. This saves one packet on the network and
9769 thus boosts performance. It can also be useful for some servers, because they
9770 immediately get the request along with the incoming connection.
9771
9772 This feature is enabled when "option tcp-smart-connect" is set in a backend.
9773 It is not enabled by default because it makes network troubleshooting more
9774 complex.
9775
9776 It only makes sense to enable it with protocols where the client speaks first
9777 such as HTTP. In other situations, if there is no data to send in place of
9778 the ACK, a normal ACK is sent.
9779
9780 If this option has been enabled in a "defaults" section, it can be disabled
9781 in a specific instance by prepending the "no" keyword before it.
9782
9783 See also : "option tcp-smart-accept"
9784
Willy Tarreau9ea05a72009-06-14 12:07:01 +02009785
Willy Tarreaubf1f8162007-12-28 17:42:56 +01009786option tcpka
9787 Enable or disable the sending of TCP keepalive packets on both sides
9788 May be used in sections : defaults | frontend | listen | backend
9789 yes | yes | yes | yes
9790 Arguments : none
9791
9792 When there is a firewall or any session-aware component between a client and
9793 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01009794 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01009795 components decides to expire a session which has remained idle for too long.
9796
9797 Enabling socket-level TCP keep-alives makes the system regularly send packets
9798 to the other end of the connection, leaving it active. The delay between
9799 keep-alive probes is controlled by the system only and depends both on the
9800 operating system and its tuning parameters.
9801
9802 It is important to understand that keep-alive packets are neither emitted nor
9803 received at the application level. It is only the network stacks which sees
9804 them. For this reason, even if one side of the proxy already uses keep-alives
9805 to maintain its connection alive, those keep-alive packets will not be
9806 forwarded to the other side of the proxy.
9807
9808 Please note that this has nothing to do with HTTP keep-alive.
9809
9810 Using option "tcpka" enables the emission of TCP keep-alive probes on both
9811 the client and server sides of a connection. Note that this is meaningful
9812 only in "defaults" or "listen" sections. If this option is used in a
9813 frontend, only the client side will get keep-alives, and if this option is
9814 used in a backend, only the server side will get keep-alives. For this
9815 reason, it is strongly recommended to explicitly use "option clitcpka" and
9816 "option srvtcpka" when the configuration is split between frontends and
9817 backends.
9818
9819 See also : "option clitcpka", "option srvtcpka"
9820
Willy Tarreau844e3c52008-01-11 16:28:18 +01009821
9822option tcplog
9823 Enable advanced logging of TCP connections with session state and timers
9824 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01009825 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01009826 Arguments : none
9827
9828 By default, the log output format is very poor, as it only contains the
9829 source and destination addresses, and the instance name. By specifying
9830 "option tcplog", each log line turns into a much richer format including, but
9831 not limited to, the connection timers, the session status, the connections
9832 numbers, the frontend, backend and server name, and of course the source
9833 address and ports. This option is useful for pure TCP proxies in order to
9834 find which of the client or server disconnects or times out. For normal HTTP
9835 proxies, it's better to use "option httplog" which is even more complete.
9836
Guillaume de Lafond29f45602017-03-31 19:52:15 +02009837 "option tcplog" overrides any previous "log-format" directive.
9838
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009839 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009840
9841
Willy Tarreau844e3c52008-01-11 16:28:18 +01009842option transparent
9843no option transparent
9844 Enable client-side transparent proxying
9845 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01009846 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01009847 Arguments : none
9848
9849 This option was introduced in order to provide layer 7 persistence to layer 3
9850 load balancers. The idea is to use the OS's ability to redirect an incoming
9851 connection for a remote address to a local process (here HAProxy), and let
9852 this process know what address was initially requested. When this option is
9853 used, sessions without cookies will be forwarded to the original destination
9854 IP address of the incoming request (which should match that of another
9855 equipment), while requests with cookies will still be forwarded to the
9856 appropriate server.
9857
9858 Note that contrary to a common belief, this option does NOT make HAProxy
9859 present the client's IP to the server when establishing the connection.
9860
Willy Tarreaua1146052011-03-01 09:51:54 +01009861 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009862 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01009863
Willy Tarreaubf1f8162007-12-28 17:42:56 +01009864
Simon Horman98637e52014-06-20 12:30:16 +09009865external-check command <command>
9866 Executable to run when performing an external-check
9867 May be used in sections : defaults | frontend | listen | backend
9868 yes | no | yes | yes
9869
9870 Arguments :
9871 <command> is the external command to run
9872
Simon Horman98637e52014-06-20 12:30:16 +09009873 The arguments passed to the to the command are:
9874
Cyril Bonté777be862014-12-02 21:21:35 +01009875 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09009876
Cyril Bonté777be862014-12-02 21:21:35 +01009877 The <proxy_address> and <proxy_port> are derived from the first listener
9878 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
9879 listener the proxy_address will be the path of the socket and the
9880 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
9881 possible to determine a listener, and both <proxy_address> and <proxy_port>
9882 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09009883
Cyril Bonté72cda2a2014-12-27 22:28:39 +01009884 Some values are also provided through environment variables.
9885
9886 Environment variables :
9887 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
9888 applicable, for example in a "backend" section).
9889
9890 HAPROXY_PROXY_ID The backend id.
9891
9892 HAPROXY_PROXY_NAME The backend name.
9893
9894 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
9895 applicable, for example in a "backend" section or
9896 for a UNIX socket).
9897
9898 HAPROXY_SERVER_ADDR The server address.
9899
9900 HAPROXY_SERVER_CURCONN The current number of connections on the server.
9901
9902 HAPROXY_SERVER_ID The server id.
9903
9904 HAPROXY_SERVER_MAXCONN The server max connections.
9905
9906 HAPROXY_SERVER_NAME The server name.
9907
9908 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
9909 socket).
9910
9911 PATH The PATH environment variable used when executing
9912 the command may be set using "external-check path".
9913
William Lallemand4d03e432019-06-14 15:35:37 +02009914 See also "2.3. Environment variables" for other variables.
9915
Simon Horman98637e52014-06-20 12:30:16 +09009916 If the command executed and exits with a zero status then the check is
9917 considered to have passed, otherwise the check is considered to have
9918 failed.
9919
9920 Example :
9921 external-check command /bin/true
9922
9923 See also : "external-check", "option external-check", "external-check path"
9924
9925
9926external-check path <path>
9927 The value of the PATH environment variable used when running an external-check
9928 May be used in sections : defaults | frontend | listen | backend
9929 yes | no | yes | yes
9930
9931 Arguments :
9932 <path> is the path used when executing external command to run
9933
9934 The default path is "".
9935
9936 Example :
9937 external-check path "/usr/bin:/bin"
9938
9939 See also : "external-check", "option external-check",
9940 "external-check command"
9941
9942
Emeric Brun647caf12009-06-30 17:57:00 +02009943persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02009944persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02009945 Enable RDP cookie-based persistence
9946 May be used in sections : defaults | frontend | listen | backend
9947 yes | no | yes | yes
9948 Arguments :
9949 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02009950 default cookie name "msts" will be used. There currently is no
9951 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02009952
9953 This statement enables persistence based on an RDP cookie. The RDP cookie
9954 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01009955 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02009956 and if an RDP cookie is found, it is decoded. If it matches a known server
9957 which is still UP (or if "option persist" is set), then the connection is
9958 forwarded to this server.
9959
9960 Note that this only makes sense in a TCP backend, but for this to work, the
9961 frontend must have waited long enough to ensure that an RDP cookie is present
9962 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009963 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02009964 a single "listen" section.
9965
Willy Tarreau61e28f22010-05-16 22:31:05 +02009966 Also, it is important to understand that the terminal server will emit this
9967 RDP cookie only if it is configured for "token redirection mode", which means
9968 that the "IP address redirection" option is disabled.
9969
Emeric Brun647caf12009-06-30 17:57:00 +02009970 Example :
9971 listen tse-farm
9972 bind :3389
9973 # wait up to 5s for an RDP cookie in the request
9974 tcp-request inspect-delay 5s
9975 tcp-request content accept if RDP_COOKIE
9976 # apply RDP cookie persistence
9977 persist rdp-cookie
9978 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02009979 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02009980 balance rdp-cookie
9981 server srv1 1.1.1.1:3389
9982 server srv2 1.1.1.2:3389
9983
Simon Hormanab814e02011-06-24 14:50:20 +09009984 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
9985 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02009986
9987
Willy Tarreau3a7d2072009-03-05 23:48:25 +01009988rate-limit sessions <rate>
9989 Set a limit on the number of new sessions accepted per second on a frontend
9990 May be used in sections : defaults | frontend | listen | backend
9991 yes | yes | yes | no
9992 Arguments :
9993 <rate> The <rate> parameter is an integer designating the maximum number
9994 of new sessions per second to accept on the frontend.
9995
9996 When the frontend reaches the specified number of new sessions per second, it
9997 stops accepting new connections until the rate drops below the limit again.
9998 During this time, the pending sessions will be kept in the socket's backlog
Daniel Corbett9f0843f2021-05-08 10:50:37 -04009999 (in system buffers) and HAProxy will not even be aware that sessions are
Willy Tarreau3a7d2072009-03-05 23:48:25 +010010000 pending. When applying very low limit on a highly loaded service, it may make
10001 sense to increase the socket's backlog using the "backlog" keyword.
10002
10003 This feature is particularly efficient at blocking connection-based attacks
10004 or service abuse on fragile servers. Since the session rate is measured every
10005 millisecond, it is extremely accurate. Also, the limit applies immediately,
10006 no delay is needed at all to detect the threshold.
10007
10008 Example : limit the connection rate on SMTP to 10 per second max
10009 listen smtp
10010 mode tcp
10011 bind :25
10012 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +020010013 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +010010014
Willy Tarreaua17c2d92011-07-25 08:16:20 +020010015 Note : when the maximum rate is reached, the frontend's status is not changed
10016 but its sockets appear as "WAITING" in the statistics if the
10017 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +010010018
10019 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
10020
10021
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010022redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
10023redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
10024redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +020010025 Return an HTTP redirection if/unless a condition is matched
10026 May be used in sections : defaults | frontend | listen | backend
10027 no | yes | yes | yes
10028
10029 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +010010030 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +020010031
Willy Tarreau0140f252008-11-19 21:07:09 +010010032 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010033 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +010010034 the HTTP "Location" header. When used in an "http-request" rule,
10035 <loc> value follows the log-format rules and can include some
10036 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010037
10038 <pfx> With "redirect prefix", the "Location" header is built from the
10039 concatenation of <pfx> and the complete URI path, including the
10040 query string, unless the "drop-query" option is specified (see
10041 below). As a special case, if <pfx> equals exactly "/", then
10042 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +010010043 redirect to the same URL (for instance, to insert a cookie). When
10044 used in an "http-request" rule, <pfx> value follows the log-format
10045 rules and can include some dynamic values (see Custom Log Format
10046 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010047
10048 <sch> With "redirect scheme", then the "Location" header is built by
10049 concatenating <sch> with "://" then the first occurrence of the
10050 "Host" header, and then the URI path, including the query string
10051 unless the "drop-query" option is specified (see below). If no
10052 path is found or if the path is "*", then "/" is used instead. If
10053 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010054 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010055 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +010010056 HTTPS. When used in an "http-request" rule, <sch> value follows
10057 the log-format rules and can include some dynamic values (see
10058 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +010010059
10060 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +010010061 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
10062 with 302 used by default if no code is specified. 301 means
10063 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +020010064 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +010010065 cache the redirection. 303 is equivalent to 302 except that the
10066 browser will fetch the location with a GET method. 307 is just
10067 like 302 but makes it clear that the same method must be reused.
10068 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +010010069
10070 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +010010071 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +010010072
10073 - "drop-query"
10074 When this keyword is used in a prefix-based redirection, then the
10075 location will be set without any possible query-string, which is useful
10076 for directing users to a non-secure page for instance. It has no effect
10077 with a location-type redirect.
10078
Willy Tarreau81e3b4f2010-01-10 00:42:19 +010010079 - "append-slash"
10080 This keyword may be used in conjunction with "drop-query" to redirect
10081 users who use a URL not ending with a '/' to the same one with the '/'.
10082 It can be useful to ensure that search engines will only see one URL.
10083 For this, a return code 301 is preferred.
10084
Willy Tarreau0140f252008-11-19 21:07:09 +010010085 - "set-cookie NAME[=value]"
10086 A "Set-Cookie" header will be added with NAME (and optionally "=value")
10087 to the response. This is sometimes used to indicate that a user has
10088 been seen, for instance to protect against some types of DoS. No other
10089 cookie option is added, so the cookie will be a session cookie. Note
10090 that for a browser, a sole cookie name without an equal sign is
10091 different from a cookie with an equal sign.
10092
10093 - "clear-cookie NAME[=]"
10094 A "Set-Cookie" header will be added with NAME (and optionally "="), but
10095 with the "Max-Age" attribute set to zero. This will tell the browser to
10096 delete this cookie. It is useful for instance on logout pages. It is
10097 important to note that clearing the cookie "NAME" will not remove a
10098 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
10099 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +020010100
10101 Example: move the login URL only to HTTPS.
10102 acl clear dst_port 80
10103 acl secure dst_port 8080
10104 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +010010105 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +010010106 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +010010107 acl cookie_set hdr_sub(cookie) SEEN=1
10108
10109 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +010010110 redirect prefix https://mysite.com if login_page !secure
10111 redirect prefix http://mysite.com drop-query if login_page !uid_given
10112 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +010010113 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +020010114
Willy Tarreau81e3b4f2010-01-10 00:42:19 +010010115 Example: send redirects for request for articles without a '/'.
10116 acl missing_slash path_reg ^/article/[^/]*$
10117 redirect code 301 prefix / drop-query append-slash if missing_slash
10118
Daniel Corbett9f0843f2021-05-08 10:50:37 -040010119 Example: redirect all HTTP traffic to HTTPS when SSL is handled by HAProxy.
David BERARDe7153042012-11-03 00:11:31 +010010120 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +020010121
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +010010122 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +020010123 http-request redirect code 301 location \
10124 http://www.%[hdr(host)]%[capture.req.uri] \
10125 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +010010126
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010127 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +020010128
Willy Tarreau303c0352008-01-17 19:01:39 +010010129
Willy Tarreaue5c5ce92008-06-20 17:27:19 +020010130retries <value>
10131 Set the number of retries to perform on a server after a connection failure
10132 May be used in sections: defaults | frontend | listen | backend
10133 yes | no | yes | yes
10134 Arguments :
10135 <value> is the number of times a connection attempt should be retried on
10136 a server when a connection either is refused or times out. The
10137 default value is 3.
10138
10139 It is important to understand that this value applies to the number of
10140 connection attempts, not full requests. When a connection has effectively
10141 been established to a server, there will be no more retry.
10142
10143 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -070010144 a turn-around timer of min("timeout connect", one second) is applied before
10145 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +020010146
10147 When "option redispatch" is set, the last retry may be performed on another
10148 server even if a cookie references a different server.
10149
10150 See also : "option redispatch"
10151
10152
Lukas Tribusbb474ff2021-12-08 11:33:01 +010010153retry-on [space-delimited list of keywords]
Jerome Magnin5ce3c142020-05-13 20:09:57 +020010154 Specify when to attempt to automatically retry a failed request.
10155 This setting is only valid when "mode" is set to http and is silently ignored
10156 otherwise.
Olivier Houcharda254a372019-04-05 15:30:12 +020010157 May be used in sections: defaults | frontend | listen | backend
10158 yes | no | yes | yes
10159 Arguments :
Lukas Tribusbb474ff2021-12-08 11:33:01 +010010160 <keywords> is a space-delimited list of keywords or HTTP status codes, each
10161 representing a type of failure event on which an attempt to
10162 retry the request is desired. Please read the notes at the
10163 bottom before changing this setting. The following keywords are
10164 supported :
Olivier Houcharda254a372019-04-05 15:30:12 +020010165
10166 none never retry
10167
10168 conn-failure retry when the connection or the SSL handshake failed
10169 and the request could not be sent. This is the default.
10170
10171 empty-response retry when the server connection was closed after part
10172 of the request was sent, and nothing was received from
10173 the server. This type of failure may be caused by the
10174 request timeout on the server side, poor network
10175 condition, or a server crash or restart while
10176 processing the request.
10177
Olivier Houcharde3249a92019-05-03 23:01:47 +020010178 junk-response retry when the server returned something not looking
10179 like a complete HTTP response. This includes partial
10180 responses headers as well as non-HTTP contents. It
10181 usually is a bad idea to retry on such events, which
10182 may be caused a configuration issue (wrong server port)
10183 or by the request being harmful to the server (buffer
10184 overflow attack for example).
10185
Olivier Houcharda254a372019-04-05 15:30:12 +020010186 response-timeout the server timeout stroke while waiting for the server
10187 to respond to the request. This may be caused by poor
10188 network condition, the reuse of an idle connection
10189 which has expired on the path, or by the request being
10190 extremely expensive to process. It generally is a bad
10191 idea to retry on such events on servers dealing with
10192 heavy database processing (full scans, etc) as it may
10193 amplify denial of service attacks.
10194
Olivier Houchard865d8392019-05-03 22:46:27 +020010195 0rtt-rejected retry requests which were sent over early data and were
10196 rejected by the server. These requests are generally
10197 considered to be safe to retry.
10198
Julien Pivotto2de240a2020-11-12 11:14:05 +010010199 <status> any HTTP status code among "401" (Unauthorized), "403"
10200 (Forbidden), "404" (Not Found), "408" (Request Timeout),
10201 "425" (Too Early), "500" (Server Error), "501" (Not
10202 Implemented), "502" (Bad Gateway), "503" (Service
10203 Unavailable), "504" (Gateway Timeout).
Olivier Houcharda254a372019-04-05 15:30:12 +020010204
Olivier Houchardddf0e032019-05-10 18:05:40 +020010205 all-retryable-errors
10206 retry request for any error that are considered
10207 retryable. This currently activates "conn-failure",
10208 "empty-response", "junk-response", "response-timeout",
10209 "0rtt-rejected", "500", "502", "503", and "504".
10210
Olivier Houcharda254a372019-04-05 15:30:12 +020010211 Using this directive replaces any previous settings with the new ones; it is
10212 not cumulative.
10213
10214 Please note that using anything other than "none" and "conn-failure" requires
10215 to allocate a buffer and copy the whole request into it, so it has memory and
10216 performance impacts. Requests not fitting in a single buffer will never be
10217 retried (see the global tune.bufsize setting).
10218
10219 You have to make sure the application has a replay protection mechanism built
10220 in such as a unique transaction IDs passed in requests, or that replaying the
10221 same request has no consequence, or it is very dangerous to use any retry-on
10222 value beside "conn-failure" and "none". Static file servers and caches are
10223 generally considered safe against any type of retry. Using a status code can
10224 be useful to quickly leave a server showing an abnormal behavior (out of
10225 memory, file system issues, etc), but in this case it may be a good idea to
10226 immediately redispatch the connection to another server (please see "option
10227 redispatch" for this). Last, it is important to understand that most causes
10228 of failures are the requests themselves and that retrying a request causing a
10229 server to misbehave will often make the situation even worse for this server,
10230 or for the whole service in case of redispatch.
10231
10232 Unless you know exactly how the application deals with replayed requests, you
10233 should not use this directive.
10234
10235 The default is "conn-failure".
10236
Lukas Tribusbb474ff2021-12-08 11:33:01 +010010237 Example:
10238 retry-on 503 504
10239
Olivier Houcharda254a372019-04-05 15:30:12 +020010240 See also: "retries", "option redispatch", "tune.bufsize"
10241
David du Colombier486df472011-03-17 10:40:26 +010010242server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010243 Declare a server in a backend
10244 May be used in sections : defaults | frontend | listen | backend
10245 no | no | yes | yes
10246 Arguments :
10247 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +010010248 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -050010249 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010250
David du Colombier486df472011-03-17 10:40:26 +010010251 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
10252 resolvable hostname is supported, but this name will be resolved
10253 during start-up. Address "0.0.0.0" or "*" has a special meaning.
10254 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +020010255 address as the one from the client connection. This is useful in
10256 transparent proxy architectures where the client's connection is
Daniel Corbett9f0843f2021-05-08 10:50:37 -040010257 intercepted and HAProxy must forward to the original destination
Willy Tarreaud669a4f2010-07-13 14:49:50 +020010258 address. This is more or less what the "transparent" keyword does
10259 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +010010260 to report statistics. Optionally, an address family prefix may be
10261 used before the address to force the family regardless of the
10262 address format, which can be useful to specify a path to a unix
10263 socket with no slash ('/'). Currently supported prefixes are :
10264 - 'ipv4@' -> address is always IPv4
10265 - 'ipv6@' -> address is always IPv6
10266 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +020010267 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +020010268 - 'sockpair@' -> address is the FD of a connected unix
10269 socket or of a socketpair. During a connection, the
10270 backend creates a pair of connected sockets, and passes
10271 one of them over the FD. The bind part will use the
10272 received socket as the client FD. Should be used
10273 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +020010274 You may want to reference some environment variables in the
10275 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +010010276 variables. The "init-addr" setting can be used to modify the way
10277 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010278
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010279 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010280 be sent to this port. If unset, the same port the client
10281 connected to will be used. The port may also be prefixed by a "+"
10282 or a "-". In this case, the server's port will be determined by
10283 adding this value to the client's port.
10284
10285 <param*> is a list of parameters for this server. The "server" keywords
10286 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010287 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010288
10289 Examples :
10290 server first 10.1.1.1:1080 cookie first check inter 1000
10291 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +010010292 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +020010293 server backup "${SRV_BACKUP}:1080" backup
10294 server www1_dc1 "${LAN_DC1}.101:80"
10295 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010296
Willy Tarreau55dcaf62015-09-27 15:03:15 +020010297 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
10298 sun_path length is used for the address length. Some other programs
10299 such as socat use the string length only by default. Pass the option
10300 ",unix-tightsocklen=0" to any abstract socket definition in socat to
10301 make it compatible with HAProxy's.
10302
Mark Lamourinec2247f02012-01-04 13:02:01 -050010303 See also: "default-server", "http-send-name-header" and section 5 about
10304 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010305
Christopher Faulet583b6de2021-02-12 09:27:10 +010010306server-state-file-name [ { use-backend-name | <file> } ]
Baptiste Assmann01c6cc32015-08-23 11:45:29 +020010307 Set the server state file to read, load and apply to servers available in
Christopher Faulet583b6de2021-02-12 09:27:10 +010010308 this backend.
10309 May be used in sections: defaults | frontend | listen | backend
10310 no | no | yes | yes
10311
10312 It only applies when the directive "load-server-state-from-file" is set to
10313 "local". When <file> is not provided, if "use-backend-name" is used or if
10314 this directive is not set, then backend name is used. If <file> starts with a
10315 slash '/', then it is considered as an absolute path. Otherwise, <file> is
10316 concatenated to the global directive "server-state-base".
Baptiste Assmann01c6cc32015-08-23 11:45:29 +020010317
10318 Example: the minimal configuration below would make HAProxy look for the
10319 state server file '/etc/haproxy/states/bk':
10320
10321 global
10322 server-state-file-base /etc/haproxy/states
10323
Willy Tarreauc9c6cdb2020-03-05 16:03:58 +010010324 backend bk
Baptiste Assmann01c6cc32015-08-23 11:45:29 +020010325 load-server-state-from-file
10326
Christopher Faulet583b6de2021-02-12 09:27:10 +010010327 See also: "server-state-base", "load-server-state-from-file", and
Baptiste Assmann01c6cc32015-08-23 11:45:29 +020010328 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010329
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +020010330server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
10331 Set a template to initialize servers with shared parameters.
10332 The names of these servers are built from <prefix> and <num | range> parameters.
10333 May be used in sections : defaults | frontend | listen | backend
10334 no | no | yes | yes
10335
10336 Arguments:
10337 <prefix> A prefix for the server names to be built.
10338
10339 <num | range>
10340 If <num> is provided, this template initializes <num> servers
10341 with 1 up to <num> as server name suffixes. A range of numbers
10342 <num_low>-<num_high> may also be used to use <num_low> up to
10343 <num_high> as server name suffixes.
10344
10345 <fqdn> A FQDN for all the servers this template initializes.
10346
10347 <port> Same meaning as "server" <port> argument (see "server" keyword).
10348
10349 <params*>
10350 Remaining server parameters among all those supported by "server"
10351 keyword.
10352
10353 Examples:
10354 # Initializes 3 servers with srv1, srv2 and srv3 as names,
10355 # google.com as FQDN, and health-check enabled.
10356 server-template srv 1-3 google.com:80 check
10357
10358 # or
10359 server-template srv 3 google.com:80 check
10360
10361 # would be equivalent to:
10362 server srv1 google.com:80 check
10363 server srv2 google.com:80 check
10364 server srv3 google.com:80 check
10365
10366
10367
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010368source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020010369source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +010010370source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010371 Set the source address for outgoing connections
10372 May be used in sections : defaults | frontend | listen | backend
10373 yes | no | yes | yes
10374 Arguments :
10375 <addr> is the IPv4 address HAProxy will bind to before connecting to a
10376 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +010010377
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010378 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +010010379 the most appropriate address to reach its destination. Optionally
10380 an address family prefix may be used before the address to force
10381 the family regardless of the address format, which can be useful
10382 to specify a path to a unix socket with no slash ('/'). Currently
10383 supported prefixes are :
10384 - 'ipv4@' -> address is always IPv4
10385 - 'ipv6@' -> address is always IPv6
10386 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +020010387 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +020010388 You may want to reference some environment variables in the
10389 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010390
10391 <port> is an optional port. It is normally not needed but may be useful
10392 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020010393 the system will select a free port. Note that port ranges are not
10394 supported in the backend. If you want to force port ranges, you
10395 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010396
10397 <addr2> is the IP address to present to the server when connections are
10398 forwarded in full transparent proxy mode. This is currently only
10399 supported on some patched Linux kernels. When this address is
10400 specified, clients connecting to the server will be presented
10401 with this address, while health checks will still use the address
10402 <addr>.
10403
10404 <port2> is the optional port to present to the server when connections
10405 are forwarded in full transparent proxy mode (see <addr2> above).
10406 The default value of zero means the system will select a free
10407 port.
10408
Willy Tarreaubce70882009-09-07 11:51:47 +020010409 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
10410 This is the name of a comma-separated header list which can
10411 contain multiple IP addresses. By default, the last occurrence is
10412 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +010010413 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +020010414 by previous proxy, typically Stunnel. In order to use another
10415 occurrence from the last one, please see the <occ> parameter
10416 below. When the header (or occurrence) is not found, no binding
10417 is performed so that the proxy's default IP address is used. Also
10418 keep in mind that the header name is case insensitive, as for any
10419 HTTP header.
10420
10421 <occ> is the occurrence number of a value to be used in a multi-value
10422 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010423 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +020010424 address. Positive values indicate a position from the first
10425 occurrence, 1 being the first one. Negative values indicate
10426 positions relative to the last one, -1 being the last one. This
10427 is helpful for situations where an X-Forwarded-For header is set
10428 at the entry point of an infrastructure and must be used several
10429 proxy layers away. When this value is not specified, -1 is
10430 assumed. Passing a zero here disables the feature.
10431
Willy Tarreaud53f96b2009-02-04 18:46:54 +010010432 <name> is an optional interface name to which to bind to for outgoing
10433 traffic. On systems supporting this features (currently, only
10434 Linux), this allows one to bind all traffic to the server to
10435 this interface even if it is not the one the system would select
10436 based on routing tables. This should be used with extreme care.
10437 Note that using this option requires root privileges.
10438
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010439 The "source" keyword is useful in complex environments where a specific
10440 address only is allowed to connect to the servers. It may be needed when a
10441 private address must be used through a public gateway for instance, and it is
10442 known that the system cannot determine the adequate source address by itself.
10443
10444 An extension which is available on certain patched Linux kernels may be used
10445 through the "usesrc" optional keyword. It makes it possible to connect to the
10446 servers with an IP address which does not belong to the system itself. This
10447 is called "full transparent proxy mode". For this to work, the destination
10448 servers have to route their traffic back to this address through the machine
10449 running HAProxy, and IP forwarding must generally be enabled on this machine.
10450
10451 In this "full transparent proxy" mode, it is possible to force a specific IP
10452 address to be presented to the servers. This is not much used in fact. A more
10453 common use is to tell HAProxy to present the client's IP address. For this,
10454 there are two methods :
10455
10456 - present the client's IP and port addresses. This is the most transparent
10457 mode, but it can cause problems when IP connection tracking is enabled on
10458 the machine, because a same connection may be seen twice with different
10459 states. However, this solution presents the huge advantage of not
10460 limiting the system to the 64k outgoing address+port couples, because all
10461 of the client ranges may be used.
10462
10463 - present only the client's IP address and select a spare port. This
10464 solution is still quite elegant but slightly less transparent (downstream
10465 firewalls logs will not match upstream's). It also presents the downside
10466 of limiting the number of concurrent connections to the usual 64k ports.
10467 However, since the upstream and downstream ports are different, local IP
10468 connection tracking on the machine will not be upset by the reuse of the
10469 same session.
10470
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010471 This option sets the default source for all servers in the backend. It may
10472 also be specified in a "defaults" section. Finer source address specification
10473 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010474 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010475
Baptiste Assmann91bd3372015-07-17 21:59:42 +020010476 In order to work, "usesrc" requires root privileges.
10477
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010478 Examples :
10479 backend private
10480 # Connect to the servers using our 192.168.1.200 source address
10481 source 192.168.1.200
10482
10483 backend transparent_ssl1
10484 # Connect to the SSL farm from the client's source address
10485 source 192.168.1.200 usesrc clientip
10486
10487 backend transparent_ssl2
10488 # Connect to the SSL farm from the client's source address and port
10489 # not recommended if IP conntrack is present on the local machine.
10490 source 192.168.1.200 usesrc client
10491
10492 backend transparent_ssl3
10493 # Connect to the SSL farm from the client's source address. It
10494 # is more conntrack-friendly.
10495 source 192.168.1.200 usesrc clientip
10496
10497 backend transparent_smtp
10498 # Connect to the SMTP farm from the client's source address/port
10499 # with Tproxy version 4.
10500 source 0.0.0.0 usesrc clientip
10501
Willy Tarreaubce70882009-09-07 11:51:47 +020010502 backend transparent_http
10503 # Connect to the servers using the client's IP as seen by previous
10504 # proxy.
10505 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
10506
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010507 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010508 the Linux kernel on www.balabit.com, the "bind" keyword.
10509
Willy Tarreau844e3c52008-01-11 16:28:18 +010010510
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +090010511srvtcpka-cnt <count>
10512 Sets the maximum number of keepalive probes TCP should send before dropping
10513 the connection on the server side.
10514 May be used in sections : defaults | frontend | listen | backend
10515 yes | no | yes | yes
10516 Arguments :
10517 <count> is the maximum number of keepalive probes.
10518
10519 This keyword corresponds to the socket option TCP_KEEPCNT. If this keyword
10520 is not specified, system-wide TCP parameter (tcp_keepalive_probes) is used.
Willy Tarreau52543212020-07-09 05:58:51 +020010521 The availability of this setting depends on the operating system. It is
10522 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +090010523
10524 See also : "option srvtcpka", "srvtcpka-idle", "srvtcpka-intvl".
10525
10526
10527srvtcpka-idle <timeout>
10528 Sets the time the connection needs to remain idle before TCP starts sending
10529 keepalive probes, if enabled the sending of TCP keepalive packets on the
10530 server side.
10531 May be used in sections : defaults | frontend | listen | backend
10532 yes | no | yes | yes
10533 Arguments :
10534 <timeout> is the time the connection needs to remain idle before TCP starts
10535 sending keepalive probes. It is specified in seconds by default,
10536 but can be in any other unit if the number is suffixed by the
10537 unit, as explained at the top of this document.
10538
10539 This keyword corresponds to the socket option TCP_KEEPIDLE. If this keyword
10540 is not specified, system-wide TCP parameter (tcp_keepalive_time) is used.
Willy Tarreau52543212020-07-09 05:58:51 +020010541 The availability of this setting depends on the operating system. It is
10542 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +090010543
10544 See also : "option srvtcpka", "srvtcpka-cnt", "srvtcpka-intvl".
10545
10546
10547srvtcpka-intvl <timeout>
10548 Sets the time between individual keepalive probes on the server side.
10549 May be used in sections : defaults | frontend | listen | backend
10550 yes | no | yes | yes
10551 Arguments :
10552 <timeout> is the time between individual keepalive probes. It is specified
10553 in seconds by default, but can be in any other unit if the number
10554 is suffixed by the unit, as explained at the top of this
10555 document.
10556
10557 This keyword corresponds to the socket option TCP_KEEPINTVL. If this keyword
10558 is not specified, system-wide TCP parameter (tcp_keepalive_intvl) is used.
Willy Tarreau52543212020-07-09 05:58:51 +020010559 The availability of this setting depends on the operating system. It is
10560 known to work on Linux.
MIZUTA Takeshib24bc0d2020-07-09 11:13:20 +090010561
10562 See also : "option srvtcpka", "srvtcpka-cnt", "srvtcpka-idle".
10563
10564
Cyril Bonté66c327d2010-10-12 00:14:37 +020010565stats admin { if | unless } <cond>
10566 Enable statistics admin level if/unless a condition is matched
10567 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010568 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +020010569
10570 This statement enables the statistics admin level if/unless a condition is
10571 matched.
10572
10573 The admin level allows to enable/disable servers from the web interface. By
10574 default, statistics page is read-only for security reasons.
10575
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010010576 Note : Consider not using this feature in multi-process mode (nbproc > 1)
10577 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +010010578 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010010579
Cyril Bonté23b39d92011-02-10 22:54:44 +010010580 Currently, the POST request is limited to the buffer size minus the reserved
10581 buffer space, which means that if the list of servers is too long, the
10582 request won't be processed. It is recommended to alter few servers at a
10583 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +020010584
10585 Example :
10586 # statistics admin level only for localhost
10587 backend stats_localhost
10588 stats enable
10589 stats admin if LOCALHOST
10590
10591 Example :
10592 # statistics admin level always enabled because of the authentication
10593 backend stats_auth
10594 stats enable
10595 stats auth admin:AdMiN123
10596 stats admin if TRUE
10597
10598 Example :
10599 # statistics admin level depends on the authenticated user
10600 userlist stats-auth
10601 group admin users admin
10602 user admin insecure-password AdMiN123
10603 group readonly users haproxy
10604 user haproxy insecure-password haproxy
10605
10606 backend stats_auth
10607 stats enable
10608 acl AUTH http_auth(stats-auth)
10609 acl AUTH_ADMIN http_auth_group(stats-auth) admin
10610 stats http-request auth unless AUTH
10611 stats admin if AUTH_ADMIN
10612
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010010613 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
10614 "bind-process", section 3.4 about userlists and section 7 about
10615 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +020010616
10617
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010618stats auth <user>:<passwd>
10619 Enable statistics with authentication and grant access to an account
10620 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010621 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010622 Arguments :
10623 <user> is a user name to grant access to
10624
10625 <passwd> is the cleartext password associated to this user
10626
10627 This statement enables statistics with default settings, and restricts access
10628 to declared users only. It may be repeated as many times as necessary to
10629 allow as many users as desired. When a user tries to access the statistics
10630 without a valid account, a "401 Forbidden" response will be returned so that
10631 the browser asks the user to provide a valid user and password. The real
10632 which will be returned to the browser is configurable using "stats realm".
10633
10634 Since the authentication method is HTTP Basic Authentication, the passwords
10635 circulate in cleartext on the network. Thus, it was decided that the
10636 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020010637 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010638
10639 It is also possible to reduce the scope of the proxies which appear in the
10640 report using "stats scope".
10641
10642 Though this statement alone is enough to enable statistics reporting, it is
10643 recommended to set all other settings in order to avoid relying on default
10644 unobvious parameters.
10645
10646 Example :
10647 # public access (limited to this backend only)
10648 backend public_www
10649 server srv1 192.168.0.1:80
10650 stats enable
10651 stats hide-version
10652 stats scope .
10653 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010654 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010655 stats auth admin1:AdMiN123
10656 stats auth admin2:AdMiN321
10657
10658 # internal monitoring access (unlimited)
10659 backend private_monitoring
10660 stats enable
10661 stats uri /admin?stats
10662 stats refresh 5s
10663
10664 See also : "stats enable", "stats realm", "stats scope", "stats uri"
10665
10666
10667stats enable
10668 Enable statistics reporting with default settings
10669 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010670 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010671 Arguments : none
10672
10673 This statement enables statistics reporting with default settings defined
10674 at build time. Unless stated otherwise, these settings are used :
10675 - stats uri : /haproxy?stats
10676 - stats realm : "HAProxy Statistics"
10677 - stats auth : no authentication
10678 - stats scope : no restriction
10679
10680 Though this statement alone is enough to enable statistics reporting, it is
10681 recommended to set all other settings in order to avoid relying on default
10682 unobvious parameters.
10683
10684 Example :
10685 # public access (limited to this backend only)
10686 backend public_www
10687 server srv1 192.168.0.1:80
10688 stats enable
10689 stats hide-version
10690 stats scope .
10691 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010692 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010693 stats auth admin1:AdMiN123
10694 stats auth admin2:AdMiN321
10695
10696 # internal monitoring access (unlimited)
10697 backend private_monitoring
10698 stats enable
10699 stats uri /admin?stats
10700 stats refresh 5s
10701
10702 See also : "stats auth", "stats realm", "stats uri"
10703
10704
Willy Tarreaud63335a2010-02-26 12:56:52 +010010705stats hide-version
10706 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010707 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010708 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +010010709 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010710
Willy Tarreaud63335a2010-02-26 12:56:52 +010010711 By default, the stats page reports some useful status information along with
10712 the statistics. Among them is HAProxy's version. However, it is generally
10713 considered dangerous to report precise version to anyone, as it can help them
10714 target known weaknesses with specific attacks. The "stats hide-version"
10715 statement removes the version from the statistics report. This is recommended
10716 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010717
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +020010718 Though this statement alone is enough to enable statistics reporting, it is
10719 recommended to set all other settings in order to avoid relying on default
10720 unobvious parameters.
10721
Willy Tarreaud63335a2010-02-26 12:56:52 +010010722 Example :
10723 # public access (limited to this backend only)
10724 backend public_www
10725 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +020010726 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +010010727 stats hide-version
10728 stats scope .
10729 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010730 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +010010731 stats auth admin1:AdMiN123
10732 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010733
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010734 # internal monitoring access (unlimited)
10735 backend private_monitoring
10736 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +010010737 stats uri /admin?stats
10738 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +010010739
Willy Tarreaud63335a2010-02-26 12:56:52 +010010740 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +020010741
Willy Tarreau983e01e2010-01-11 18:42:06 +010010742
Cyril Bonté2be1b3f2010-09-30 23:46:30 +020010743stats http-request { allow | deny | auth [realm <realm>] }
10744 [ { if | unless } <condition> ]
10745 Access control for statistics
10746
10747 May be used in sections: defaults | frontend | listen | backend
10748 no | no | yes | yes
10749
10750 As "http-request", these set of options allow to fine control access to
10751 statistics. Each option may be followed by if/unless and acl.
10752 First option with matched condition (or option without condition) is final.
10753 For "deny" a 403 error will be returned, for "allow" normal processing is
10754 performed, for "auth" a 401/407 error code is returned so the client
10755 should be asked to enter a username and password.
10756
10757 There is no fixed limit to the number of http-request statements per
10758 instance.
10759
10760 See also : "http-request", section 3.4 about userlists and section 7
10761 about ACL usage.
10762
10763
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010764stats realm <realm>
10765 Enable statistics and set authentication realm
10766 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010767 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010768 Arguments :
10769 <realm> is the name of the HTTP Basic Authentication realm reported to
10770 the browser. The browser uses it to display it in the pop-up
10771 inviting the user to enter a valid username and password.
10772
10773 The realm is read as a single word, so any spaces in it should be escaped
10774 using a backslash ('\').
10775
10776 This statement is useful only in conjunction with "stats auth" since it is
10777 only related to authentication.
10778
10779 Though this statement alone is enough to enable statistics reporting, it is
10780 recommended to set all other settings in order to avoid relying on default
10781 unobvious parameters.
10782
10783 Example :
10784 # public access (limited to this backend only)
10785 backend public_www
10786 server srv1 192.168.0.1:80
10787 stats enable
10788 stats hide-version
10789 stats scope .
10790 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010791 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010792 stats auth admin1:AdMiN123
10793 stats auth admin2:AdMiN321
10794
10795 # internal monitoring access (unlimited)
10796 backend private_monitoring
10797 stats enable
10798 stats uri /admin?stats
10799 stats refresh 5s
10800
10801 See also : "stats auth", "stats enable", "stats uri"
10802
10803
10804stats refresh <delay>
10805 Enable statistics with automatic refresh
10806 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010807 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010808 Arguments :
10809 <delay> is the suggested refresh delay, specified in seconds, which will
10810 be returned to the browser consulting the report page. While the
10811 browser is free to apply any delay, it will generally respect it
10812 and refresh the page this every seconds. The refresh interval may
10813 be specified in any other non-default time unit, by suffixing the
10814 unit after the value, as explained at the top of this document.
10815
10816 This statement is useful on monitoring displays with a permanent page
10817 reporting the load balancer's activity. When set, the HTML report page will
10818 include a link "refresh"/"stop refresh" so that the user can select whether
Jackie Tapia749f74c2020-07-22 18:59:40 -050010819 they want automatic refresh of the page or not.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010820
10821 Though this statement alone is enough to enable statistics reporting, it is
10822 recommended to set all other settings in order to avoid relying on default
10823 unobvious parameters.
10824
10825 Example :
10826 # public access (limited to this backend only)
10827 backend public_www
10828 server srv1 192.168.0.1:80
10829 stats enable
10830 stats hide-version
10831 stats scope .
10832 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010833 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010834 stats auth admin1:AdMiN123
10835 stats auth admin2:AdMiN321
10836
10837 # internal monitoring access (unlimited)
10838 backend private_monitoring
10839 stats enable
10840 stats uri /admin?stats
10841 stats refresh 5s
10842
10843 See also : "stats auth", "stats enable", "stats realm", "stats uri"
10844
10845
10846stats scope { <name> | "." }
10847 Enable statistics and limit access scope
10848 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010849 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010850 Arguments :
10851 <name> is the name of a listen, frontend or backend section to be
10852 reported. The special name "." (a single dot) designates the
10853 section in which the statement appears.
10854
10855 When this statement is specified, only the sections enumerated with this
10856 statement will appear in the report. All other ones will be hidden. This
10857 statement may appear as many times as needed if multiple sections need to be
10858 reported. Please note that the name checking is performed as simple string
10859 comparisons, and that it is never checked that a give section name really
10860 exists.
10861
10862 Though this statement alone is enough to enable statistics reporting, it is
10863 recommended to set all other settings in order to avoid relying on default
10864 unobvious parameters.
10865
10866 Example :
10867 # public access (limited to this backend only)
10868 backend public_www
10869 server srv1 192.168.0.1:80
10870 stats enable
10871 stats hide-version
10872 stats scope .
10873 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010010874 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010875 stats auth admin1:AdMiN123
10876 stats auth admin2:AdMiN321
10877
10878 # internal monitoring access (unlimited)
10879 backend private_monitoring
10880 stats enable
10881 stats uri /admin?stats
10882 stats refresh 5s
10883
10884 See also : "stats auth", "stats enable", "stats realm", "stats uri"
10885
Willy Tarreaud63335a2010-02-26 12:56:52 +010010886
Willy Tarreauc9705a12010-07-27 20:05:50 +020010887stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +010010888 Enable reporting of a description on the statistics page.
10889 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010890 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +010010891
Willy Tarreauc9705a12010-07-27 20:05:50 +020010892 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +010010893 description from global section is automatically used instead.
10894
10895 This statement is useful for users that offer shared services to their
10896 customers, where node or description should be different for each customer.
10897
10898 Though this statement alone is enough to enable statistics reporting, it is
10899 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +010010900 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010901
10902 Example :
10903 # internal monitoring access (unlimited)
10904 backend private_monitoring
10905 stats enable
10906 stats show-desc Master node for Europe, Asia, Africa
10907 stats uri /admin?stats
10908 stats refresh 5s
10909
10910 See also: "show-node", "stats enable", "stats uri" and "description" in
10911 global section.
10912
10913
10914stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +020010915 Enable reporting additional information on the statistics page
10916 May be used in sections : defaults | frontend | listen | backend
10917 yes | yes | yes | yes
10918 Arguments : none
10919
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010920 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +010010921 - cap: capabilities (proxy)
10922 - mode: one of tcp, http or health (proxy)
10923 - id: SNMP ID (proxy, socket, server)
10924 - IP (socket, server)
10925 - cookie (backend, server)
10926
10927 Though this statement alone is enough to enable statistics reporting, it is
10928 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +010010929 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010930
10931 See also: "stats enable", "stats uri".
10932
10933
Amaury Denoyelle0b70a8a2020-10-05 11:49:45 +020010934stats show-modules
10935 Enable display of extra statistics module on the statistics page
10936 May be used in sections : defaults | frontend | listen | backend
10937 yes | yes | yes | yes
10938 Arguments : none
10939
10940 New columns are added at the end of the line containing the extra statistics
10941 values as a tooltip.
10942
10943 Though this statement alone is enough to enable statistics reporting, it is
10944 recommended to set all other settings in order to avoid relying on default
10945 unobvious parameters. Default behavior is not to show this information.
10946
10947 See also: "stats enable", "stats uri".
10948
10949
Willy Tarreaud63335a2010-02-26 12:56:52 +010010950stats show-node [ <name> ]
10951 Enable reporting of a host name on the statistics page.
10952 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010953 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +010010954 Arguments:
10955 <name> is an optional name to be reported. If unspecified, the
10956 node name from global section is automatically used instead.
10957
10958 This statement is useful for users that offer shared services to their
10959 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +010010960 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010961
10962 Though this statement alone is enough to enable statistics reporting, it is
10963 recommended to set all other settings in order to avoid relying on default
10964 unobvious parameters.
10965
10966 Example:
10967 # internal monitoring access (unlimited)
10968 backend private_monitoring
10969 stats enable
10970 stats show-node Europe-1
10971 stats uri /admin?stats
10972 stats refresh 5s
10973
10974 See also: "show-desc", "stats enable", "stats uri", and "node" in global
10975 section.
10976
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010977
10978stats uri <prefix>
10979 Enable statistics and define the URI prefix to access them
10980 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +020010981 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010982 Arguments :
10983 <prefix> is the prefix of any URI which will be redirected to stats. This
10984 prefix may contain a question mark ('?') to indicate part of a
10985 query string.
10986
10987 The statistics URI is intercepted on the relayed traffic, so it appears as a
10988 page within the normal application. It is strongly advised to ensure that the
10989 selected URI will never appear in the application, otherwise it will never be
10990 possible to reach it in the application.
10991
Daniel Corbett9f0843f2021-05-08 10:50:37 -040010992 The default URI compiled in HAProxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010993 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +010010994 It is generally a good idea to include a question mark in the URI so that
10995 intermediate proxies refrain from caching the results. Also, since any string
10996 beginning with the prefix will be accepted as a stats request, the question
10997 mark helps ensuring that no valid URI will begin with the same words.
10998
10999 It is sometimes very convenient to use "/" as the URI prefix, and put that
11000 statement in a "listen" instance of its own. That makes it easy to dedicate
11001 an address or a port to statistics only.
11002
11003 Though this statement alone is enough to enable statistics reporting, it is
11004 recommended to set all other settings in order to avoid relying on default
11005 unobvious parameters.
11006
11007 Example :
11008 # public access (limited to this backend only)
11009 backend public_www
11010 server srv1 192.168.0.1:80
11011 stats enable
11012 stats hide-version
11013 stats scope .
11014 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +010011015 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +010011016 stats auth admin1:AdMiN123
11017 stats auth admin2:AdMiN321
11018
11019 # internal monitoring access (unlimited)
11020 backend private_monitoring
11021 stats enable
11022 stats uri /admin?stats
11023 stats refresh 5s
11024
11025 See also : "stats auth", "stats enable", "stats realm"
11026
11027
Willy Tarreaud63335a2010-02-26 12:56:52 +010011028stick match <pattern> [table <table>] [{if | unless} <cond>]
11029 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +010011030 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +010011031 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011032
11033 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +020011034 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011035 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +010011036 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011037 stickiness table. This rule is mandatory.
11038
11039 <table> is an optional stickiness table name. If unspecified, the same
11040 backend's table is used. A stickiness table is declared using
11041 the "stick-table" statement.
11042
11043 <cond> is an optional matching condition. It makes it possible to match
11044 on a certain criterion only when other conditions are met (or
11045 not met). For instance, it could be used to match on a source IP
11046 address except when a request passes through a known proxy, in
11047 which case we'd match on a header containing that IP address.
11048
11049 Some protocols or applications require complex stickiness rules and cannot
11050 always simply rely on cookies nor hashing. The "stick match" statement
11051 describes a rule to extract the stickiness criterion from an incoming request
11052 or connection. See section 7 for a complete list of possible patterns and
11053 transformation rules.
11054
11055 The table has to be declared using the "stick-table" statement. It must be of
11056 a type compatible with the pattern. By default it is the one which is present
11057 in the same backend. It is possible to share a table with other backends by
11058 referencing it using the "table" keyword. If another table is referenced,
11059 the server's ID inside the backends are used. By default, all server IDs
11060 start at 1 in each backend, so the server ordering is enough. But in case of
11061 doubt, it is highly recommended to force server IDs using their "id" setting.
11062
11063 It is possible to restrict the conditions where a "stick match" statement
11064 will apply, using "if" or "unless" followed by a condition. See section 7 for
11065 ACL based conditions.
11066
11067 There is no limit on the number of "stick match" statements. The first that
11068 applies and matches will cause the request to be directed to the same server
11069 as was used for the request which created the entry. That way, multiple
11070 matches can be used as fallbacks.
11071
11072 The stick rules are checked after the persistence cookies, so they will not
11073 affect stickiness if a cookie has already been used to select a server. That
11074 way, it becomes very easy to insert cookies and match on IP addresses in
11075 order to maintain stickiness between HTTP and HTTPS.
11076
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011077 Note : Consider not using this feature in multi-process mode (nbproc > 1)
11078 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011079 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011080
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011081 Example :
11082 # forward SMTP users to the same server they just used for POP in the
11083 # last 30 minutes
11084 backend pop
11085 mode tcp
11086 balance roundrobin
11087 stick store-request src
11088 stick-table type ip size 200k expire 30m
11089 server s1 192.168.1.1:110
11090 server s2 192.168.1.1:110
11091
11092 backend smtp
11093 mode tcp
11094 balance roundrobin
11095 stick match src table pop
11096 server s1 192.168.1.1:25
11097 server s2 192.168.1.1:25
11098
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011099 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +020011100 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011101
11102
11103stick on <pattern> [table <table>] [{if | unless} <condition>]
11104 Define a request pattern to associate a user to a server
11105 May be used in sections : defaults | frontend | listen | backend
11106 no | no | yes | yes
11107
11108 Note : This form is exactly equivalent to "stick match" followed by
11109 "stick store-request", all with the same arguments. Please refer
11110 to both keywords for details. It is only provided as a convenience
11111 for writing more maintainable configurations.
11112
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011113 Note : Consider not using this feature in multi-process mode (nbproc > 1)
11114 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011115 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011116
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011117 Examples :
11118 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +010011119 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011120
11121 # ...is strictly equivalent to this one :
11122 stick match src table pop if !localhost
11123 stick store-request src table pop if !localhost
11124
11125
11126 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
11127 # well as HTTP without cookie. Share the same table between both accesses.
11128 backend http
11129 mode http
11130 balance roundrobin
11131 stick on src table https
11132 cookie SRV insert indirect nocache
11133 server s1 192.168.1.1:80 cookie s1
11134 server s2 192.168.1.1:80 cookie s2
11135
11136 backend https
11137 mode tcp
11138 balance roundrobin
11139 stick-table type ip size 200k expire 30m
11140 stick on src
11141 server s1 192.168.1.1:443
11142 server s2 192.168.1.1:443
11143
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011144 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011145
11146
11147stick store-request <pattern> [table <table>] [{if | unless} <condition>]
11148 Define a request pattern used to create an entry in a stickiness table
11149 May be used in sections : defaults | frontend | listen | backend
11150 no | no | yes | yes
11151
11152 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +020011153 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011154 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +010011155 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011156 server is selected.
11157
11158 <table> is an optional stickiness table name. If unspecified, the same
11159 backend's table is used. A stickiness table is declared using
11160 the "stick-table" statement.
11161
11162 <cond> is an optional storage condition. It makes it possible to store
11163 certain criteria only when some conditions are met (or not met).
11164 For instance, it could be used to store the source IP address
11165 except when the request passes through a known proxy, in which
11166 case we'd store a converted form of a header containing that IP
11167 address.
11168
11169 Some protocols or applications require complex stickiness rules and cannot
11170 always simply rely on cookies nor hashing. The "stick store-request" statement
11171 describes a rule to decide what to extract from the request and when to do
11172 it, in order to store it into a stickiness table for further requests to
11173 match it using the "stick match" statement. Obviously the extracted part must
11174 make sense and have a chance to be matched in a further request. Storing a
11175 client's IP address for instance often makes sense. Storing an ID found in a
11176 URL parameter also makes sense. Storing a source port will almost never make
11177 any sense because it will be randomly matched. See section 7 for a complete
11178 list of possible patterns and transformation rules.
11179
11180 The table has to be declared using the "stick-table" statement. It must be of
11181 a type compatible with the pattern. By default it is the one which is present
11182 in the same backend. It is possible to share a table with other backends by
11183 referencing it using the "table" keyword. If another table is referenced,
11184 the server's ID inside the backends are used. By default, all server IDs
11185 start at 1 in each backend, so the server ordering is enough. But in case of
11186 doubt, it is highly recommended to force server IDs using their "id" setting.
11187
11188 It is possible to restrict the conditions where a "stick store-request"
11189 statement will apply, using "if" or "unless" followed by a condition. This
11190 condition will be evaluated while parsing the request, so any criteria can be
11191 used. See section 7 for ACL based conditions.
11192
11193 There is no limit on the number of "stick store-request" statements, but
11194 there is a limit of 8 simultaneous stores per request or response. This
11195 makes it possible to store up to 8 criteria, all extracted from either the
11196 request or the response, regardless of the number of rules. Only the 8 first
11197 ones which match will be kept. Using this, it is possible to feed multiple
11198 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +010011199 another protocol or access method. Using multiple store-request rules with
11200 the same table is possible and may be used to find the best criterion to rely
11201 on, by arranging the rules by decreasing preference order. Only the first
11202 extracted criterion for a given table will be stored. All subsequent store-
11203 request rules referencing the same table will be skipped and their ACLs will
11204 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011205
11206 The "store-request" rules are evaluated once the server connection has been
11207 established, so that the table will contain the real server that processed
11208 the request.
11209
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011210 Note : Consider not using this feature in multi-process mode (nbproc > 1)
11211 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011212 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011213
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011214 Example :
11215 # forward SMTP users to the same server they just used for POP in the
11216 # last 30 minutes
11217 backend pop
11218 mode tcp
11219 balance roundrobin
11220 stick store-request src
11221 stick-table type ip size 200k expire 30m
11222 server s1 192.168.1.1:110
11223 server s2 192.168.1.1:110
11224
11225 backend smtp
11226 mode tcp
11227 balance roundrobin
11228 stick match src table pop
11229 server s1 192.168.1.1:25
11230 server s2 192.168.1.1:25
11231
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011232 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +020011233 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011234
11235
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011236stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Thayne McCombs92149f92020-11-20 01:28:26 -070011237 size <size> [expire <expire>] [nopurge] [peers <peersect>] [srvkey <srvkey>]
Emeric Brunf099e792010-09-27 12:05:28 +020011238 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +080011239 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011240 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +020011241 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011242
11243 Arguments :
11244 ip a table declared with "type ip" will only store IPv4 addresses.
11245 This form is very compact (about 50 bytes per entry) and allows
11246 very fast entry lookup and stores with almost no overhead. This
11247 is mainly used to store client source IP addresses.
11248
David du Colombier9a6d3c92011-03-17 10:40:24 +010011249 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
11250 This form is very compact (about 60 bytes per entry) and allows
11251 very fast entry lookup and stores with almost no overhead. This
11252 is mainly used to store client source IP addresses.
11253
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011254 integer a table declared with "type integer" will store 32bit integers
11255 which can represent a client identifier found in a request for
11256 instance.
11257
11258 string a table declared with "type string" will store substrings of up
11259 to <len> characters. If the string provided by the pattern
11260 extractor is larger than <len>, it will be truncated before
11261 being stored. During matching, at most <len> characters will be
11262 compared between the string in the table and the extracted
11263 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011264 to 32 characters.
11265
11266 binary a table declared with "type binary" will store binary blocks
11267 of <len> bytes. If the block provided by the pattern
11268 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +020011269 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011270 is shorter than <len>, it will be padded by 0. When not
11271 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011272
11273 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +020011274 "string" type table (See type "string" above). Or the number
11275 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011276 changing this parameter as memory usage will proportionally
11277 increase.
11278
11279 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +010011280 value directly impacts memory usage. Count approximately
11281 50 bytes per entry, plus the size of a string if any. The size
11282 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011283
11284 [nopurge] indicates that we refuse to purge older entries when the table
Daniel Corbett9f0843f2021-05-08 10:50:37 -040011285 is full. When not specified and the table is full when HAProxy
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011286 wants to store an entry in it, it will flush a few of the oldest
11287 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +010011288 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011289 be desirable to refuse new entries instead of purging the older
11290 ones. That may be the case when the amount of data to store is
11291 far above the hardware limits and we prefer not to offer access
11292 to new clients than to reject the ones already connected. When
11293 using this parameter, be sure to properly set the "expire"
11294 parameter (see below).
11295
Emeric Brunf099e792010-09-27 12:05:28 +020011296 <peersect> is the name of the peers section to use for replication. Entries
11297 which associate keys to server IDs are kept synchronized with
11298 the remote peers declared in this section. All entries are also
11299 automatically learned from the local peer (old process) during a
11300 soft restart.
11301
Willy Tarreau1abc6732015-05-01 19:21:02 +020011302 NOTE : each peers section may be referenced only by tables
11303 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +010011304
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011305 <expire> defines the maximum duration of an entry in the table since it
11306 was last created, refreshed or matched. The expiration delay is
11307 defined using the standard time format, similarly as the various
11308 timeouts. The maximum duration is slightly above 24 days. See
Willy Tarreau4b103022021-02-12 17:59:10 +010011309 section 2.5 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +020011310 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011311 be removed once full. Be sure not to use the "nopurge" parameter
11312 if not expiration delay is specified.
11313
Thayne McCombs92149f92020-11-20 01:28:26 -070011314 <srvkey> specifies how each server is identified for the purposes of the
11315 stick table. The valid values are "name" and "addr". If "name" is
11316 given, then <name> argument for the server (may be generated by
11317 a template). If "addr" is given, then the server is identified
11318 by its current network address, including the port. "addr" is
11319 especially useful if you are using service discovery to generate
11320 the addresses for servers with peered stick-tables and want
11321 to consistently use the same host across peers for a stickiness
11322 token.
11323
Willy Tarreau08d5f982010-06-06 13:34:54 +020011324 <data_type> is used to store additional information in the stick-table. This
11325 may be used by ACLs in order to control various criteria related
11326 to the activity of the client matching the stick-table. For each
11327 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +020011328 that the additional data can fit. Several data types may be
11329 stored with an entry. Multiple data types may be specified after
11330 the "store" keyword, as a comma-separated list. Alternatively,
11331 it is possible to repeat the "store" keyword followed by one or
11332 several data types. Except for the "server_id" type which is
11333 automatically detected and enabled, all data types must be
11334 explicitly declared to be stored. If an ACL references a data
11335 type which is not stored, the ACL will simply not match. Some
11336 data types require an argument which must be passed just after
11337 the type between parenthesis. See below for the supported data
11338 types and their arguments.
11339
11340 The data types that can be stored with an entry are the following :
11341 - server_id : this is an integer which holds the numeric ID of the server a
11342 request was assigned to. It is used by the "stick match", "stick store",
11343 and "stick on" rules. It is automatically enabled when referenced.
11344
11345 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
11346 integer which may be used for anything. Most of the time it will be used
11347 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +010011348 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011349
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011350 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
11351 over a period. It is a positive 32-bit integer integer which may be used
11352 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +010011353 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011354 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +010011355 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011356
Frédéric Lécaille6778b272018-01-29 15:22:53 +010011357 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
11358 integer which may be used for anything. Most of the time it will be used
11359 to put a special tag on some entries, for instance to note that a
11360 specific behavior was detected and must be known for future matches.
11361
11362 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
11363 over a period. It is a positive 32-bit integer integer which may be used
11364 for anything. Just like <gpc1>, it counts events, but instead of keeping
11365 a cumulative number, it maintains the rate at which the counter is
11366 incremented. Most of the time it will be used to measure the frequency of
11367 occurrence of certain events (e.g. requests to a specific URL).
11368
Emeric Bruna5d15312021-07-01 18:34:48 +020011369 - gpt0 : first General Purpose Tag. It is a positive 32-bit integer
11370 integer which may be used for anything. Most of the time it will be used
11371 to put a special tag on some entries, for instance to note that a
11372 specific behavior was detected and must be known for future matches
11373
Willy Tarreauc9705a12010-07-27 20:05:50 +020011374 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
11375 the absolute number of connections received from clients which matched
11376 this entry. It does not mean the connections were accepted, just that
11377 they were received.
11378
11379 - conn_cur : Current Connections. It is a positive 32-bit integer which
11380 stores the concurrent connection counts for the entry. It is incremented
11381 once an incoming connection matches the entry, and decremented once the
11382 connection leaves. That way it is possible to know at any time the exact
11383 number of concurrent connections for an entry.
11384
11385 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
11386 integer parameter <period> which indicates in milliseconds the length
11387 of the period over which the average is measured. It reports the average
11388 incoming connection rate over that period, in connections per period. The
11389 result is an integer which can be matched using ACLs.
11390
11391 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
11392 the absolute number of sessions received from clients which matched this
11393 entry. A session is a connection that was accepted by the layer 4 rules.
11394
11395 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
11396 integer parameter <period> which indicates in milliseconds the length
11397 of the period over which the average is measured. It reports the average
11398 incoming session rate over that period, in sessions per period. The
11399 result is an integer which can be matched using ACLs.
11400
11401 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
11402 counts the absolute number of HTTP requests received from clients which
11403 matched this entry. It does not matter whether they are valid requests or
11404 not. Note that this is different from sessions when keep-alive is used on
11405 the client side.
11406
11407 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
11408 integer parameter <period> which indicates in milliseconds the length
11409 of the period over which the average is measured. It reports the average
11410 HTTP request rate over that period, in requests per period. The result is
11411 an integer which can be matched using ACLs. It does not matter whether
11412 they are valid requests or not. Note that this is different from sessions
11413 when keep-alive is used on the client side.
11414
11415 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
11416 counts the absolute number of HTTP requests errors induced by clients
11417 which matched this entry. Errors are counted on invalid and truncated
11418 requests, as well as on denied or tarpitted requests, and on failed
11419 authentications. If the server responds with 4xx, then the request is
11420 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +010011421 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +020011422
11423 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
11424 integer parameter <period> which indicates in milliseconds the length
11425 of the period over which the average is measured. It reports the average
11426 HTTP request error rate over that period, in requests per period (see
11427 http_err_cnt above for what is accounted as an error). The result is an
11428 integer which can be matched using ACLs.
11429
Willy Tarreau826f3ab2021-02-10 12:07:15 +010011430 - http_fail_cnt : HTTP Failure Count. It is a positive 32-bit integer which
11431 counts the absolute number of HTTP response failures induced by servers
11432 which matched this entry. Errors are counted on invalid and truncated
11433 responses, as well as any 5xx response other than 501 or 505. It aims at
11434 being used combined with path or URI to detect service failures.
11435
11436 - http_fail_rate(<period>) : frequency counter (takes 12 bytes). It takes
11437 an integer parameter <period> which indicates in milliseconds the length
11438 of the period over which the average is measured. It reports the average
11439 HTTP response failure rate over that period, in requests per period (see
11440 http_fail_cnt above for what is accounted as a failure). The result is an
11441 integer which can be matched using ACLs.
11442
Willy Tarreauc9705a12010-07-27 20:05:50 +020011443 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +010011444 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +020011445 which matched this entry. Headers are included in the count. This may be
11446 used to limit abuse of upload features on photo or video servers.
11447
11448 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
11449 integer parameter <period> which indicates in milliseconds the length
11450 of the period over which the average is measured. It reports the average
11451 incoming bytes rate over that period, in bytes per period. It may be used
11452 to detect users which upload too much and too fast. Warning: with large
11453 uploads, it is possible that the amount of uploaded data will be counted
11454 once upon termination, thus causing spikes in the average transfer speed
11455 instead of having a smooth one. This may partially be smoothed with
11456 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
11457 recommended for better fairness.
11458
11459 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +010011460 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +020011461 matched this entry. Headers are included in the count. This may be used
11462 to limit abuse of bots sucking the whole site.
11463
11464 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
11465 an integer parameter <period> which indicates in milliseconds the length
11466 of the period over which the average is measured. It reports the average
11467 outgoing bytes rate over that period, in bytes per period. It may be used
11468 to detect users which download too much and too fast. Warning: with large
11469 transfers, it is possible that the amount of transferred data will be
11470 counted once upon termination, thus causing spikes in the average
11471 transfer speed instead of having a smooth one. This may partially be
11472 smoothed with "option contstats" though this is not perfect yet. Use of
11473 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +020011474
Willy Tarreauc00cdc22010-06-06 16:48:26 +020011475 There is only one stick-table per proxy. At the moment of writing this doc,
11476 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011477 to be required, simply create a dummy backend with a stick-table in it and
11478 reference it.
11479
11480 It is important to understand that stickiness based on learning information
11481 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +010011482 lost upon restart unless peers are properly configured to transfer such
11483 information upon restart (recommended). In general it can be good as a
11484 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011485
Willy Tarreauc9705a12010-07-27 20:05:50 +020011486 Last, memory requirements may be important when storing many data types.
11487 Indeed, storing all indicators above at once in each entry requires 116 bytes
11488 per entry, or 116 MB for a 1-million entries table. This is definitely not
11489 something that can be ignored.
11490
11491 Example:
11492 # Keep track of counters of up to 1 million IP addresses over 5 minutes
11493 # and store a general purpose counter and the average connection rate
11494 # computed over a sliding window of 30 seconds.
11495 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
11496
Willy Tarreau4b103022021-02-12 17:59:10 +010011497 See also : "stick match", "stick on", "stick store-request", section 2.5
David du Colombiera13d1b92011-03-17 10:40:22 +010011498 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +010011499
11500
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011501stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +010011502 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011503 May be used in sections : defaults | frontend | listen | backend
11504 no | no | yes | yes
11505
11506 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +020011507 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011508 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +010011509 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011510 server is selected.
11511
11512 <table> is an optional stickiness table name. If unspecified, the same
11513 backend's table is used. A stickiness table is declared using
11514 the "stick-table" statement.
11515
11516 <cond> is an optional storage condition. It makes it possible to store
11517 certain criteria only when some conditions are met (or not met).
11518 For instance, it could be used to store the SSL session ID only
11519 when the response is a SSL server hello.
11520
11521 Some protocols or applications require complex stickiness rules and cannot
11522 always simply rely on cookies nor hashing. The "stick store-response"
11523 statement describes a rule to decide what to extract from the response and
11524 when to do it, in order to store it into a stickiness table for further
11525 requests to match it using the "stick match" statement. Obviously the
11526 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011527 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011528 See section 7 for a complete list of possible patterns and transformation
11529 rules.
11530
11531 The table has to be declared using the "stick-table" statement. It must be of
11532 a type compatible with the pattern. By default it is the one which is present
11533 in the same backend. It is possible to share a table with other backends by
11534 referencing it using the "table" keyword. If another table is referenced,
11535 the server's ID inside the backends are used. By default, all server IDs
11536 start at 1 in each backend, so the server ordering is enough. But in case of
11537 doubt, it is highly recommended to force server IDs using their "id" setting.
11538
11539 It is possible to restrict the conditions where a "stick store-response"
11540 statement will apply, using "if" or "unless" followed by a condition. This
11541 condition will be evaluated while parsing the response, so any criteria can
11542 be used. See section 7 for ACL based conditions.
11543
11544 There is no limit on the number of "stick store-response" statements, but
11545 there is a limit of 8 simultaneous stores per request or response. This
11546 makes it possible to store up to 8 criteria, all extracted from either the
11547 request or the response, regardless of the number of rules. Only the 8 first
11548 ones which match will be kept. Using this, it is possible to feed multiple
11549 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +010011550 another protocol or access method. Using multiple store-response rules with
11551 the same table is possible and may be used to find the best criterion to rely
11552 on, by arranging the rules by decreasing preference order. Only the first
11553 extracted criterion for a given table will be stored. All subsequent store-
11554 response rules referencing the same table will be skipped and their ACLs will
11555 not be evaluated. However, even if a store-request rule references a table, a
11556 store-response rule may also use the same table. This means that each table
11557 may learn exactly one element from the request and one element from the
11558 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011559
11560 The table will contain the real server that processed the request.
11561
11562 Example :
11563 # Learn SSL session ID from both request and response and create affinity.
11564 backend https
11565 mode tcp
11566 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +020011567 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011568 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011569
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011570 acl clienthello req_ssl_hello_type 1
11571 acl serverhello rep_ssl_hello_type 2
11572
11573 # use tcp content accepts to detects ssl client and server hello.
11574 tcp-request inspect-delay 5s
11575 tcp-request content accept if clienthello
11576
11577 # no timeout on response inspect delay by default.
11578 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011579
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011580 # SSL session ID (SSLID) may be present on a client or server hello.
11581 # Its length is coded on 1 byte at offset 43 and its value starts
11582 # at offset 44.
11583
11584 # Match and learn on request if client hello.
11585 stick on payload_lv(43,1) if clienthello
11586
11587 # Learn on response if server hello.
11588 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +020011589
Emeric Brun6a1cefa2010-09-24 18:15:17 +020011590 server s1 192.168.1.1:443
11591 server s2 192.168.1.1:443
11592
11593 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
11594 extraction.
11595
11596
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011597tcp-check comment <string>
11598 Defines a comment for the following the tcp-check rule, reported in logs if
11599 it fails.
11600 May be used in sections : defaults | frontend | listen | backend
11601 yes | no | yes | yes
11602
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011603 Arguments :
11604 <string> is the comment message to add in logs if the following tcp-check
11605 rule fails.
11606
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011607 It only works for connect, send and expect rules. It is useful to make
11608 user-friendly error reporting.
11609
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011610 See also : "option tcp-check", "tcp-check connect", "tcp-check send" and
11611 "tcp-check expect".
11612
11613
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011614tcp-check connect [default] [port <expr>] [addr <ip>] [send-proxy] [via-socks4]
11615 [ssl] [sni <sni>] [alpn <alpn>] [linger]
Christopher Fauletedc6ed92020-04-23 16:27:59 +020011616 [proto <name>] [comment <msg>]
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011617 Opens a new connection
11618 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011619 yes | no | yes | yes
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011620
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011621 Arguments :
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011622 comment <msg> defines a message to report if the rule evaluation fails.
11623
Christopher Faulet4dce5922020-03-30 13:54:42 +020011624 default Use default options of the server line to do the health
Daniel Corbett67a82712020-07-06 23:01:19 -040011625 checks. The server options are used only if not redefined.
Christopher Faulet4dce5922020-03-30 13:54:42 +020011626
Christopher Fauletb7d30092020-03-30 15:19:03 +020011627 port <expr> if not set, check port or server port is used.
Christopher Faulet5c288742020-03-31 08:15:58 +020011628 It tells HAProxy where to open the connection to.
11629 <port> must be a valid TCP port source integer, from 1 to
Christopher Fauletb7d30092020-03-30 15:19:03 +020011630 65535 or an sample-fetch expression.
Christopher Faulet5c288742020-03-31 08:15:58 +020011631
11632 addr <ip> defines the IP address to do the health check.
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011633
11634 send-proxy send a PROXY protocol string
11635
Christopher Faulet085426a2020-03-30 13:07:02 +020011636 via-socks4 enables outgoing health checks using upstream socks4 proxy.
11637
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011638 ssl opens a ciphered connection
11639
Christopher Faulet79b31d42020-03-30 13:00:05 +020011640 sni <sni> specifies the SNI to use to do health checks over SSL.
11641
Christopher Faulet98572322020-03-30 13:16:44 +020011642 alpn <alpn> defines which protocols to advertise with ALPN. The protocol
11643 list consists in a comma-delimited list of protocol names,
11644 for instance: "http/1.1,http/1.0" (without quotes).
11645 If it is not set, the server ALPN is used.
11646
Christopher Fauletedc6ed92020-04-23 16:27:59 +020011647 proto <name> forces the multiplexer's protocol to use for this connection.
11648 It must be a TCP mux protocol and it must be usable on the
11649 backend side. The list of available protocols is reported in
11650 haproxy -vv.
11651
Christopher Faulet5c288742020-03-31 08:15:58 +020011652 linger cleanly close the connection instead of using a single RST.
Gaetan Rivetf8ba6772020-02-07 15:37:17 +010011653
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011654 When an application lies on more than a single TCP port or when HAProxy
11655 load-balance many services in a single backend, it makes sense to probe all
11656 the services individually before considering a server as operational.
11657
11658 When there are no TCP port configured on the server line neither server port
11659 directive, then the 'tcp-check connect port <port>' must be the first step
11660 of the sequence.
11661
11662 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
11663 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
11664 do.
11665
11666 When a connect must start the ruleset, if may still be preceded by set-var,
11667 unset-var or comment rules.
11668
11669 Examples :
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011670 # check HTTP and HTTPs services on a server.
11671 # first open port 80 thanks to server line port directive, then
11672 # tcp-check opens port 443, ciphered and run a request on it:
11673 option tcp-check
11674 tcp-check connect
11675 tcp-check send GET\ /\ HTTP/1.0\r\n
11676 tcp-check send Host:\ haproxy.1wt.eu\r\n
11677 tcp-check send \r\n
11678 tcp-check expect rstring (2..|3..)
11679 tcp-check connect port 443 ssl
11680 tcp-check send GET\ /\ HTTP/1.0\r\n
11681 tcp-check send Host:\ haproxy.1wt.eu\r\n
11682 tcp-check send \r\n
11683 tcp-check expect rstring (2..|3..)
11684 server www 10.0.0.1 check port 80
11685
11686 # check both POP and IMAP from a single server:
11687 option tcp-check
Gaetan Rivetf8ba6772020-02-07 15:37:17 +010011688 tcp-check connect port 110 linger
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011689 tcp-check expect string +OK\ POP3\ ready
11690 tcp-check connect port 143
11691 tcp-check expect string *\ OK\ IMAP4\ ready
11692 server mail 10.0.0.1 check
11693
11694 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
11695
11696
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011697tcp-check expect [min-recv <int>] [comment <msg>]
Christopher Fauletec07e382020-04-07 14:56:26 +020011698 [ok-status <st>] [error-status <st>] [tout-status <st>]
Christopher Faulet98cc57c2020-04-01 20:52:31 +020011699 [on-success <fmt>] [on-error <fmt>] [status-code <expr>]
Christopher Fauletcf80f2f2020-04-01 11:04:52 +020011700 [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +010011701 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011702 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011703 yes | no | yes | yes
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011704
11705 Arguments :
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011706 comment <msg> defines a message to report if the rule evaluation fails.
11707
Gaetan Rivet1afd8262020-02-07 15:37:17 +010011708 min-recv is optional and can define the minimum amount of data required to
11709 evaluate the current expect rule. If the number of received bytes
11710 is under this limit, the check will wait for more data. This
11711 option can be used to resolve some ambiguous matching rules or to
11712 avoid executing costly regex matches on content known to be still
11713 incomplete. If an exact string (string or binary) is used, the
11714 minimum between the string length and this parameter is used.
11715 This parameter is ignored if it is set to -1. If the expect rule
11716 does not match, the check will wait for more data. If set to 0,
11717 the evaluation result is always conclusive.
11718
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011719 <match> is a keyword indicating how to look for a specific pattern in the
Gaetan Rivetefab6c62020-02-07 15:37:17 +010011720 response. The keyword may be one of "string", "rstring", "binary" or
11721 "rbinary".
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011722 The keyword may be preceded by an exclamation mark ("!") to negate
11723 the match. Spaces are allowed between the exclamation mark and the
11724 keyword. See below for more details on the supported keywords.
11725
Christopher Fauletec07e382020-04-07 14:56:26 +020011726 ok-status <st> is optional and can be used to set the check status if
11727 the expect rule is successfully evaluated and if it is
11728 the last rule in the tcp-check ruleset. "L7OK", "L7OKC",
Christopher Fauletd888f0f2020-05-07 07:40:17 +020011729 "L6OK" and "L4OK" are supported :
11730 - L7OK : check passed on layer 7
Christopher Faulet83662b52020-11-20 17:47:47 +010011731 - L7OKC : check conditionally passed on layer 7, set
11732 server to NOLB state.
Christopher Fauletd888f0f2020-05-07 07:40:17 +020011733 - L6OK : check passed on layer 6
11734 - L4OK : check passed on layer 4
Christopher Fauletec07e382020-04-07 14:56:26 +020011735 By default "L7OK" is used.
11736
Christopher Fauletcf80f2f2020-04-01 11:04:52 +020011737 error-status <st> is optional and can be used to set the check status if
11738 an error occurred during the expect rule evaluation.
Christopher Faulet83662b52020-11-20 17:47:47 +010011739 "L7OKC", "L7RSP", "L7STS", "L6RSP" and "L4CON" are
11740 supported :
11741 - L7OKC : check conditionally passed on layer 7, set
11742 server to NOLB state.
Christopher Fauletd888f0f2020-05-07 07:40:17 +020011743 - L7RSP : layer 7 invalid response - protocol error
11744 - L7STS : layer 7 response error, for example HTTP 5xx
11745 - L6RSP : layer 6 invalid response - protocol error
11746 - L4CON : layer 1-4 connection problem
11747 By default "L7RSP" is used.
Christopher Fauletcf80f2f2020-04-01 11:04:52 +020011748
Christopher Fauletec07e382020-04-07 14:56:26 +020011749 tout-status <st> is optional and can be used to set the check status if
Christopher Fauletcf80f2f2020-04-01 11:04:52 +020011750 a timeout occurred during the expect rule evaluation.
Christopher Fauletd888f0f2020-05-07 07:40:17 +020011751 "L7TOUT", "L6TOUT", and "L4TOUT" are supported :
11752 - L7TOUT : layer 7 (HTTP/SMTP) timeout
11753 - L6TOUT : layer 6 (SSL) timeout
11754 - L4TOUT : layer 1-4 timeout
Christopher Fauletcf80f2f2020-04-01 11:04:52 +020011755 By default "L7TOUT" is used.
11756
Christopher Fauletbe52b4d2020-04-01 16:30:22 +020011757 on-success <fmt> is optional and can be used to customize the
11758 informational message reported in logs if the expect
11759 rule is successfully evaluated and if it is the last rule
11760 in the tcp-check ruleset. <fmt> is a log-format string.
11761
11762 on-error <fmt> is optional and can be used to customize the
11763 informational message reported in logs if an error
11764 occurred during the expect rule evaluation. <fmt> is a
11765 log-format string.
11766
Christopher Faulet98cc57c2020-04-01 20:52:31 +020011767 status-code <expr> is optional and can be used to set the check status code
11768 reported in logs, on success or on error. <expr> is a
11769 standard HAProxy expression formed by a sample-fetch
11770 followed by some converters.
11771
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011772 <pattern> is the pattern to look for. It may be a string or a regular
11773 expression. If the pattern contains spaces, they must be escaped
11774 with the usual backslash ('\').
11775 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +010011776 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011777 two digits will represent a byte. The hexadecimal digits may be
11778 used upper or lower case.
11779
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011780 The available matches are intentionally similar to their http-check cousins :
11781
11782 string <string> : test the exact string matches in the response buffer.
11783 A health check response will be considered valid if the
11784 response's buffer contains this exact string. If the
11785 "string" keyword is prefixed with "!", then the response
11786 will be considered invalid if the body contains this
11787 string. This can be used to look for a mandatory pattern
11788 in a protocol response, or to detect a failure when a
11789 specific error appears in a protocol banner.
11790
11791 rstring <regex> : test a regular expression on the response buffer.
11792 A health check response will be considered valid if the
11793 response's buffer matches this expression. If the
11794 "rstring" keyword is prefixed with "!", then the response
11795 will be considered invalid if the body matches the
11796 expression.
11797
Christopher Fauletaaab0832020-05-05 15:54:22 +020011798 string-lf <fmt> : test a log-format string match in the response's buffer.
11799 A health check response will be considered valid if the
11800 response's buffer contains the string resulting of the
11801 evaluation of <fmt>, which follows the log-format rules.
11802 If prefixed with "!", then the response will be
11803 considered invalid if the buffer contains the string.
11804
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011805 binary <hexstring> : test the exact string in its hexadecimal form matches
11806 in the response buffer. A health check response will
11807 be considered valid if the response's buffer contains
11808 this exact hexadecimal string.
11809 Purpose is to match data on binary protocols.
11810
Gaetan Rivetefab6c62020-02-07 15:37:17 +010011811 rbinary <regex> : test a regular expression on the response buffer, like
11812 "rstring". However, the response buffer is transformed
11813 into its hexadecimal form, including NUL-bytes. This
11814 allows using all regex engines to match any binary
11815 content. The hexadecimal transformation takes twice the
11816 size of the original response. As such, the expected
11817 pattern should work on at-most half the response buffer
11818 size.
11819
Christopher Fauletaaab0832020-05-05 15:54:22 +020011820 binary-lf <hexfmt> : test a log-format string in its hexadecimal form
11821 match in the response's buffer. A health check response
11822 will be considered valid if the response's buffer
11823 contains the hexadecimal string resulting of the
11824 evaluation of <fmt>, which follows the log-format
11825 rules. If prefixed with "!", then the response will be
11826 considered invalid if the buffer contains the
11827 hexadecimal string. The hexadecimal string is converted
11828 in a binary string before matching the response's
11829 buffer.
11830
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011831 It is important to note that the responses will be limited to a certain size
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +010011832 defined by the global "tune.bufsize" option, which defaults to 16384 bytes.
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011833 Thus, too large responses may not contain the mandatory pattern when using
11834 "string", "rstring" or binary. If a large response is absolutely required, it
11835 is possible to change the default max size by setting the global variable.
11836 However, it is worth keeping in mind that parsing very large responses can
11837 waste some CPU cycles, especially when regular expressions are used, and that
11838 it is always better to focus the checks on smaller resources. Also, in its
11839 current state, the check will not find any string nor regex past a null
11840 character in the response. Similarly it is not possible to request matching
11841 the null character.
11842
11843 Examples :
11844 # perform a POP check
11845 option tcp-check
11846 tcp-check expect string +OK\ POP3\ ready
11847
11848 # perform an IMAP check
11849 option tcp-check
11850 tcp-check expect string *\ OK\ IMAP4\ ready
11851
11852 # look for the redis master server
11853 option tcp-check
11854 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +020011855 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011856 tcp-check send info\ replication\r\n
11857 tcp-check expect string role:master
11858 tcp-check send QUIT\r\n
11859 tcp-check expect string +OK
11860
11861
11862 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +010011863 "tcp-check send-binary", "http-check expect", tune.bufsize
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011864
11865
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011866tcp-check send <data> [comment <msg>]
11867tcp-check send-lf <fmt> [comment <msg>]
11868 Specify a string or a log-format string to be sent as a question during a
11869 generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011870 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011871 yes | no | yes | yes
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011872
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011873 Arguments :
11874 comment <msg> defines a message to report if the rule evaluation fails.
11875
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011876 <data> is the string that will be sent during a generic health
11877 check session.
Christopher Faulet16fff672020-04-30 07:50:54 +020011878
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011879 <fmt> is the log-format string that will be sent, once evaluated,
11880 during a generic health check session.
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011881
11882 Examples :
11883 # look for the redis master server
11884 option tcp-check
11885 tcp-check send info\ replication\r\n
11886 tcp-check expect string role:master
11887
11888 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +010011889 "tcp-check send-binary", tune.bufsize
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011890
11891
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011892tcp-check send-binary <hexstring> [comment <msg>]
11893tcp-check send-binary-lf <hexfmt> [comment <msg>]
11894 Specify an hex digits string or an hex digits log-format string to be sent as
11895 a binary question during a raw tcp health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011896 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011897 yes | no | yes | yes
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011898
Christopher Faulet4f5c2e22020-04-23 15:22:33 +020011899 Arguments :
11900 comment <msg> defines a message to report if the rule evaluation fails.
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011901
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011902 <hexstring> is the hexadecimal string that will be send, once converted
11903 to binary, during a generic health check session.
Christopher Faulet16fff672020-04-30 07:50:54 +020011904
Christopher Fauletb50b3e62020-05-05 18:43:43 +020011905 <hexfmt> is the hexadecimal log-format string that will be send, once
11906 evaluated and converted to binary, during a generic health
11907 check session.
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011908
11909 Examples :
11910 # redis check in binary
11911 option tcp-check
11912 tcp-check send-binary 50494e470d0a # PING\r\n
11913 tcp-check expect binary 2b504F4e47 # +PONG
11914
11915
11916 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
Christopher Fauletbb9fb8b2020-11-25 17:20:57 +010011917 "tcp-check send", tune.bufsize
Willy Tarreau938c7fe2014-04-25 14:21:39 +020011918
11919
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011920tcp-check set-var(<var-name>) <expr>
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011921 This operation sets the content of a variable. The variable is declared inline.
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011922 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011923 yes | no | yes | yes
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011924
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011925 Arguments :
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011926 <var-name> The name of the variable starts with an indication about its
11927 scope. The scopes allowed for tcp-check are:
11928 "proc" : the variable is shared with the whole process.
11929 "sess" : the variable is shared with the tcp-check session.
11930 "check": the variable is declared for the lifetime of the tcp-check.
11931 This prefix is followed by a name. The separator is a '.'.
11932 The name may only contain characters 'a-z', 'A-Z', '0-9', '.',
11933 and '-'.
11934
11935 <expr> Is a sample-fetch expression potentially followed by converters.
11936
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011937 Examples :
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011938 tcp-check set-var(check.port) int(1234)
11939
11940
11941tcp-check unset-var(<var-name>)
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011942 Free a reference to a variable within its scope.
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011943 May be used in sections: defaults | frontend | listen | backend
Christopher Faulet404f9192020-04-09 23:13:54 +020011944 yes | no | yes | yes
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011945
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011946 Arguments :
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011947 <var-name> The name of the variable starts with an indication about its
11948 scope. The scopes allowed for tcp-check are:
11949 "proc" : the variable is shared with the whole process.
11950 "sess" : the variable is shared with the tcp-check session.
11951 "check": the variable is declared for the lifetime of the tcp-check.
11952 This prefix is followed by a name. The separator is a '.'.
11953 The name may only contain characters 'a-z', 'A-Z', '0-9', '.',
11954 and '-'.
11955
Christopher Fauletc52ea4d2020-04-23 15:43:35 +020011956 Examples :
Gaetan Rivet0c39ecc2020-02-24 17:34:11 +010011957 tcp-check unset-var(check.port)
11958
11959
Willy Tarreaue9656522010-08-17 15:40:09 +020011960tcp-request connection <action> [{if | unless} <condition>]
11961 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +020011962 May be used in sections : defaults | frontend | listen | backend
11963 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +020011964 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020011965 <action> defines the action to perform if the condition applies. See
11966 below.
Willy Tarreau1a687942010-05-23 22:40:30 +020011967
Willy Tarreaue9656522010-08-17 15:40:09 +020011968 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +020011969
11970 Immediately after acceptance of a new incoming connection, it is possible to
11971 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +020011972 or dropped or have its counters tracked. Those conditions cannot make use of
11973 any data contents because the connection has not been read from yet, and the
11974 buffers are not yet allocated. This is used to selectively and very quickly
11975 accept or drop connections from various sources with a very low overhead. If
11976 some contents need to be inspected in order to take the decision, the
11977 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020011978
Willy Tarreaue9656522010-08-17 15:40:09 +020011979 The "tcp-request connection" rules are evaluated in their exact declaration
11980 order. If no rule matches or if there is no rule, the default action is to
11981 accept the incoming connection. There is no specific limit to the number of
11982 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020011983
Willy Tarreaua9083d02015-05-08 15:27:59 +020011984 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +020011985 - accept :
11986 accepts the connection if the condition is true (when used with "if")
11987 or false (when used with "unless"). The first such rule executed ends
11988 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020011989
Willy Tarreaue9656522010-08-17 15:40:09 +020011990 - reject :
11991 rejects the connection if the condition is true (when used with "if")
11992 or false (when used with "unless"). The first such rule executed ends
11993 the rules evaluation. Rejected connections do not even become a
11994 session, which is why they are accounted separately for in the stats,
11995 as "denied connections". They are not considered for the session
11996 rate-limit and are not logged either. The reason is that these rules
11997 should only be used to filter extremely high connection rates such as
11998 the ones encountered during a massive DDoS attack. Under these extreme
11999 conditions, the simple action of logging each event would make the
12000 system collapse and would considerably lower the filtering capacity. If
12001 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +020012002 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012003
Willy Tarreau4f0d9192013-06-11 20:40:55 +020012004 - expect-proxy layer4 :
12005 configures the client-facing connection to receive a PROXY protocol
12006 header before any byte is read from the socket. This is equivalent to
12007 having the "accept-proxy" keyword on the "bind" line, except that using
12008 the TCP rule allows the PROXY protocol to be accepted only for certain
12009 IP address ranges using an ACL. This is convenient when multiple layers
12010 of load balancers are passed through by traffic coming from public
12011 hosts.
12012
Bertrand Jacquin90759682016-06-06 15:35:39 +010012013 - expect-netscaler-cip layer4 :
12014 configures the client-facing connection to receive a NetScaler Client
12015 IP insertion protocol header before any byte is read from the socket.
12016 This is equivalent to having the "accept-netscaler-cip" keyword on the
12017 "bind" line, except that using the TCP rule allows the PROXY protocol
12018 to be accepted only for certain IP address ranges using an ACL. This
12019 is convenient when multiple layers of load balancers are passed
12020 through by traffic coming from public hosts.
12021
Willy Tarreau18bf01e2014-06-13 16:18:52 +020012022 - capture <sample> len <length> :
12023 This only applies to "tcp-request content" rules. It captures sample
12024 expression <sample> from the request buffer, and converts it to a
12025 string of at most <len> characters. The resulting string is stored into
12026 the next request "capture" slot, so it will possibly appear next to
12027 some captured HTTP headers. It will then automatically appear in the
12028 logs, and it will be possible to extract it using sample fetch rules to
12029 feed it into headers or anything. The length should be limited given
12030 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +020012031 session life. Please check section 7.3 (Fetching samples) and "capture
12032 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +020012033
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012034 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +020012035 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020012036 rules do not stop evaluation and do not change default action. The
12037 number of counters that may be simultaneously tracked by the same
12038 connection is set in MAX_SESS_STKCTR at build time (reported in
John Roeslerfb2fce12019-07-10 15:45:51 -050012039 haproxy -vv) which defaults to 3, so the track-sc number is between 0
Matteo Contrini1857b8c2020-10-16 17:35:54 +020012040 and (MAX_SESS_STKCTR-1). The first "track-sc0" rule executed enables
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020012041 tracking of the counters of the specified table as the first set. The
12042 first "track-sc1" rule executed enables tracking of the counters of the
12043 specified table as the second set. The first "track-sc2" rule executed
12044 enables tracking of the counters of the specified table as the third
12045 set. It is a recommended practice to use the first set of counters for
12046 the per-frontend counters and the second set for the per-backend ones.
12047 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012048
Willy Tarreaue9656522010-08-17 15:40:09 +020012049 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +020012050 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +020012051 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +010012052 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012053 and used to select which table entry to update the counters.
12054 Note that "tcp-request connection" cannot use content-based
12055 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012056
Willy Tarreaue9656522010-08-17 15:40:09 +020012057 <table> is an optional table to be used instead of the default one,
12058 which is the stick-table declared in the current proxy. All
12059 the counters for the matches and updates for the key will
12060 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012061
Willy Tarreaue9656522010-08-17 15:40:09 +020012062 Once a "track-sc*" rule is executed, the key is looked up in the table
12063 and if it is not found, an entry is allocated for it. Then a pointer to
12064 that entry is kept during all the session's life, and this entry's
12065 counters are updated as often as possible, every time the session's
12066 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012067 Counters are only updated for events that happen after the tracking has
12068 been started. For example, connection counters will not be updated when
12069 tracking layer 7 information, since the connection event happens before
12070 layer7 information is extracted.
12071
Willy Tarreaue9656522010-08-17 15:40:09 +020012072 If the entry tracks concurrent connection counters, one connection is
12073 counted for as long as the entry is tracked, and the entry will not
12074 expire during that time. Tracking counters also provides a performance
12075 advantage over just checking the keys, because only one table lookup is
12076 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012077
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020012078 - sc-inc-gpc0(<sc-id>):
12079 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
12080 counter designated by <sc-id>. If an error occurs, this action silently
12081 fails and the actions evaluation continues.
12082
Frédéric Lécaille6778b272018-01-29 15:22:53 +010012083 - sc-inc-gpc1(<sc-id>):
12084 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
12085 counter designated by <sc-id>. If an error occurs, this action silently
12086 fails and the actions evaluation continues.
12087
Cédric Dufour0d7712d2019-11-06 18:38:53 +010012088 - sc-set-gpt0(<sc-id>) { <int> | <expr> }:
12089 This action sets the 32-bit unsigned GPT0 tag according to the sticky
12090 counter designated by <sc-id> and the value of <int>/<expr>. The
12091 expected result is a boolean. If an error occurs, this action silently
12092 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012093
William Lallemand2e785f22016-05-25 01:48:42 +020012094 - set-src <expr> :
12095 Is used to set the source IP address to the value of specified
12096 expression. Useful if you want to mask source IP for privacy.
12097 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020012098 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +020012099
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020012100 Arguments:
12101 <expr> Is a standard HAProxy expression formed by a sample-fetch
12102 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +020012103
12104 Example:
William Lallemand2e785f22016-05-25 01:48:42 +020012105 tcp-request connection set-src src,ipmask(24)
12106
Willy Tarreau0c630532016-10-21 17:52:58 +020012107 When possible, set-src preserves the original source port as long as the
12108 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +020012109
William Lallemand44be6402016-05-25 01:51:35 +020012110 - set-src-port <expr> :
12111 Is used to set the source port address to the value of specified
12112 expression.
12113
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020012114 Arguments:
12115 <expr> Is a standard HAProxy expression formed by a sample-fetch
12116 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +020012117
12118 Example:
William Lallemand44be6402016-05-25 01:51:35 +020012119 tcp-request connection set-src-port int(4000)
12120
Willy Tarreau0c630532016-10-21 17:52:58 +020012121 When possible, set-src-port preserves the original source address as long
12122 as the address family supports a port, otherwise it forces the source
12123 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +020012124
William Lallemand13e9b0c2016-05-25 02:34:07 +020012125 - set-dst <expr> :
12126 Is used to set the destination IP address to the value of specified
12127 expression. Useful if you want to mask IP for privacy in log.
12128 If you want to provide an IP from a HTTP header use "http-request
12129 set-dst". If you want to connect to the new address/port, use
12130 '0.0.0.0:0' as a server address in the backend.
12131
12132 <expr> Is a standard HAProxy expression formed by a sample-fetch
12133 followed by some converters.
12134
12135 Example:
12136
12137 tcp-request connection set-dst dst,ipmask(24)
12138 tcp-request connection set-dst ipv4(10.0.0.1)
12139
Willy Tarreau0c630532016-10-21 17:52:58 +020012140 When possible, set-dst preserves the original destination port as long as
12141 the address family allows it, otherwise the destination port is set to 0.
12142
William Lallemand13e9b0c2016-05-25 02:34:07 +020012143 - set-dst-port <expr> :
12144 Is used to set the destination port address to the value of specified
12145 expression. If you want to connect to the new address/port, use
12146 '0.0.0.0:0' as a server address in the backend.
12147
12148
12149 <expr> Is a standard HAProxy expression formed by a sample-fetch
12150 followed by some converters.
12151
12152 Example:
12153
12154 tcp-request connection set-dst-port int(4000)
12155
Willy Tarreau0c630532016-10-21 17:52:58 +020012156 When possible, set-dst-port preserves the original destination address as
12157 long as the address family supports a port, otherwise it forces the
12158 destination address to IPv4 "0.0.0.0" before rewriting the port.
12159
Willy Tarreau2d392c22015-08-24 01:43:45 +020012160 - "silent-drop" :
12161 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012162 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020012163 to prevent the client from being notified. The effect it then that the
12164 client still sees an established connection while there's none on
12165 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
12166 except that it doesn't use any local resource at all on the machine
12167 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010012168 slow down stronger attackers. It is important to understand the impact
12169 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020012170 client and HAProxy (firewalls, proxies, load balancers) will also keep
12171 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010012172 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020012173 TCP_REPAIR socket option is used to block the emission of a TCP
12174 reset. On other systems, the socket's TTL is reduced to 1 so that the
12175 TCP reset doesn't pass the first router, though it's still delivered to
12176 local networks. Do not use it unless you fully understand how it works.
12177
Willy Tarreaue9656522010-08-17 15:40:09 +020012178 Note that the "if/unless" condition is optional. If no condition is set on
12179 the action, it is simply performed unconditionally. That can be useful for
12180 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012181
Willy Tarreaue9656522010-08-17 15:40:09 +020012182 Example: accept all connections from white-listed hosts, reject too fast
12183 connection without counting them, and track accepted connections.
12184 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012185
Willy Tarreaue9656522010-08-17 15:40:09 +020012186 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012187 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012188 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012189
Willy Tarreaue9656522010-08-17 15:40:09 +020012190 Example: accept all connections from white-listed hosts, count all other
12191 connections and reject too fast ones. This results in abusive ones
12192 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012193
Willy Tarreaue9656522010-08-17 15:40:09 +020012194 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012195 tcp-request connection track-sc0 src
12196 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012197
Willy Tarreau4f0d9192013-06-11 20:40:55 +020012198 Example: enable the PROXY protocol for traffic coming from all known proxies.
12199
12200 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
12201
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012202 See section 7 about ACL usage.
12203
Willy Tarreau4f614292016-10-21 17:49:36 +020012204 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012205
12206
Willy Tarreaue9656522010-08-17 15:40:09 +020012207tcp-request content <action> [{if | unless} <condition>]
12208 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +020012209 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020012210 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +020012211 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020012212 <action> defines the action to perform if the condition applies. See
12213 below.
Willy Tarreau62644772008-07-16 18:36:06 +020012214
Willy Tarreaue9656522010-08-17 15:40:09 +020012215 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +020012216
Davor Ocelice9ed2812017-12-25 17:49:28 +010012217 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +020012218 called "TCP content inspection". During this stage, ACL-based rules are
12219 evaluated every time the request contents are updated, until either an
Christopher Fauletae863c62021-03-15 12:03:44 +010012220 "accept", a "reject" or a "switch-mode" rule matches, or the TCP request
12221 inspection delay expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +020012222
Willy Tarreaue9656522010-08-17 15:40:09 +020012223 The first difference between these rules and "tcp-request connection" rules
12224 is that "tcp-request content" rules can make use of contents to take a
12225 decision. Most often, these decisions will consider a protocol recognition or
12226 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +010012227 both frontends and backends. In case of HTTP keep-alive with the client, all
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012228 tcp-request content rules are evaluated again, so HAProxy keeps a record of
Willy Tarreauf3338342014-01-28 21:40:28 +010012229 what sticky counters were assigned by a "tcp-request connection" versus a
12230 "tcp-request content" rule, and flushes all the content-related ones after
12231 processing an HTTP request, so that they may be evaluated again by the rules
12232 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012233 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +010012234 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +020012235
Willy Tarreaue9656522010-08-17 15:40:09 +020012236 Content-based rules are evaluated in their exact declaration order. If no
12237 rule matches or if there is no rule, the default action is to accept the
12238 contents. There is no specific limit to the number of rules which may be
12239 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +020012240
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012241 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +020012242 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +010012243 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +020012244 - reject : the request is rejected and the connection is closed
12245 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -040012246 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012247 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020012248 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010012249 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +010012250 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020012251 - set-dst <expr>
12252 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012253 - set-var(<var-name>) <expr>
Christopher Fauletae863c62021-03-15 12:03:44 +010012254 - switch-mode http [ proto <name> ]
Christopher Faulet85d79c92016-11-09 16:54:56 +010012255 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +020012256 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +010012257 - send-spoe-group <engine-name> <group-name>
Christopher Faulet579d83b2019-11-22 15:34:17 +010012258 - use-service <service-name>
Willy Tarreau62644772008-07-16 18:36:06 +020012259
Willy Tarreaue9656522010-08-17 15:40:09 +020012260 They have the same meaning as their counter-parts in "tcp-request connection"
12261 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +010012262 For "do-resolve" action, please check the "http-request do-resolve"
12263 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +020012264
Willy Tarreauf3338342014-01-28 21:40:28 +010012265 While there is nothing mandatory about it, it is recommended to use the
12266 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
12267 content" rules in the frontend, and track-sc2 for "tcp-request content"
12268 rules in the backend, because that makes the configuration more readable
12269 and easier to troubleshoot, but this is just a guideline and all counters
12270 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +020012271
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012272 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +020012273 the action, it is simply performed unconditionally. That can be useful for
12274 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +020012275
Christopher Faulet2079a4a2020-10-02 11:48:57 +020012276 Note also that it is recommended to use a "tcp-request session" rule to track
12277 information that does *not* depend on Layer 7 contents, especially for HTTP
12278 frontends. Some HTTP processing are performed at the session level and may
12279 lead to an early rejection of the requests. Thus, the tracking at the content
12280 level may be disturbed in such case. A warning is emitted during startup to
12281 prevent, as far as possible, such unreliable usage.
12282
Willy Tarreaue9656522010-08-17 15:40:09 +020012283 It is perfectly possible to match layer 7 contents with "tcp-request content"
Christopher Faulet7ea509e2020-10-02 11:38:46 +020012284 rules from a TCP proxy, since HTTP-specific ACL matches are able to
12285 preliminarily parse the contents of a buffer before extracting the required
12286 data. If the buffered contents do not parse as a valid HTTP message, then the
12287 ACL does not match. The parser which is involved there is exactly the same
12288 as for all other HTTP processing, so there is no risk of parsing something
12289 differently. In an HTTP frontend or an HTTP backend, it is guaranteed that
12290 HTTP contents will always be immediately present when the rule is evaluated
12291 first because the HTTP parsing is performed in the early stages of the
12292 connection processing, at the session level. But for such proxies, using
12293 "http-request" rules is much more natural and recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020012294
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012295 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020012296 are present when the rule is processed. The rule processing engine is able to
12297 wait until the inspect delay expires when the data to be tracked is not yet
12298 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012299
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020012300 The "set-dst" and "set-dst-port" are used to set respectively the destination
12301 IP and port. More information on how to use it at "http-request set-dst".
12302
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012303 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +020012304 declared inline. For "tcp-request session" rules, only session-level
12305 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012306
Daniel Schneller0b547052016-03-21 20:46:57 +010012307 <var-name> The name of the variable starts with an indication about
12308 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012309 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012310 "sess" : the variable is shared with the whole session
12311 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012312 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010012313 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012314 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010012315 "res" : the variable is shared only during response
12316 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012317 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010012318 The name may only contain characters 'a-z', 'A-Z', '0-9',
12319 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012320
12321 <expr> Is a standard HAProxy expression formed by a sample-fetch
12322 followed by some converters.
12323
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012324 The "switch-mode" is used to perform a connection upgrade. Only HTTP
Christopher Fauletae863c62021-03-15 12:03:44 +010012325 upgrades are supported for now. The protocol may optionally be
12326 specified. This action is only available for a proxy with the frontend
12327 capability. The connection upgrade is immediately performed, following
12328 "tcp-request content" rules are not evaluated. This upgrade method should be
12329 preferred to the implicit one consisting to rely on the backend mode. When
12330 used, it is possible to set HTTP directives in a frontend without any
Ilya Shipitsin3df59892021-05-10 12:50:00 +050012331 warning. These directives will be conditionally evaluated if the HTTP upgrade
Christopher Fauletae863c62021-03-15 12:03:44 +010012332 is performed. However, an HTTP backend must still be selected. It remains
12333 unsupported to route an HTTP connection (upgraded or not) to a TCP server.
12334
Christopher Faulet4d37e532021-03-26 14:44:00 +010012335 See section 4 about Proxies for more details on HTTP upgrades.
12336
Christopher Faulet85d79c92016-11-09 16:54:56 +010012337 The "unset-var" is used to unset a variable. See above for details about
12338 <var-name>.
12339
Patrick Hemmer268a7072018-05-11 12:52:31 -040012340 The "set-priority-class" is used to set the queue priority class of the
12341 current request. The value must be a sample expression which converts to an
12342 integer in the range -2047..2047. Results outside this range will be
12343 truncated. The priority class determines the order in which queued requests
12344 are processed. Lower values have higher priority.
12345
12346 The "set-priority-offset" is used to set the queue priority timestamp offset
12347 of the current request. The value must be a sample expression which converts
12348 to an integer in the range -524287..524287. Results outside this range will be
12349 truncated. When a request is queued, it is ordered first by the priority
12350 class, then by the current timestamp adjusted by the given offset in
12351 milliseconds. Lower values have higher priority.
12352 Note that the resulting timestamp is is only tracked with enough precision for
12353 524,287ms (8m44s287ms). If the request is queued long enough to where the
12354 adjusted timestamp exceeds this value, it will be misidentified as highest
12355 priority. Thus it is important to set "timeout queue" to a value, where when
12356 combined with the offset, does not exceed this limit.
12357
Christopher Faulet76c09ef2017-09-21 11:03:52 +020012358 The "send-spoe-group" is used to trigger sending of a group of SPOE
12359 messages. To do so, the SPOE engine used to send messages must be defined, as
12360 well as the SPOE group to send. Of course, the SPOE engine must refer to an
12361 existing SPOE filter. If not engine name is provided on the SPOE filter line,
12362 the SPOE agent name must be used.
12363
12364 <engine-name> The SPOE engine name.
12365
12366 <group-name> The SPOE group name as specified in the engine configuration.
12367
Christopher Faulet579d83b2019-11-22 15:34:17 +010012368 The "use-service" is used to executes a TCP service which will reply to the
12369 request and stop the evaluation of the rules. This service may choose to
12370 reply by sending any valid response or it may immediately close the
12371 connection without sending anything. Outside natives services, it is possible
12372 to write your own services in Lua. No further "tcp-request" rules are
12373 evaluated.
12374
12375 Example:
12376 tcp-request content use-service lua.deny { src -f /etc/haproxy/blacklist.lst }
12377
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012378 Example:
12379
12380 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +010012381 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012382
Willy Tarreau62644772008-07-16 18:36:06 +020012383 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +020012384 # Accept HTTP requests containing a Host header saying "example.com"
Christopher Fauletae863c62021-03-15 12:03:44 +010012385 # and reject everything else. (Only works for HTTP/1 connections)
Willy Tarreaue9656522010-08-17 15:40:09 +020012386 acl is_host_com hdr(Host) -i example.com
12387 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +020012388 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +020012389 tcp-request content reject
12390
Christopher Fauletae863c62021-03-15 12:03:44 +010012391 # Accept HTTP requests containing a Host header saying "example.com"
12392 # and reject everything else. (works for HTTP/1 and HTTP/2 connections)
12393 acl is_host_com hdr(Host) -i example.com
12394 tcp-request inspect-delay 5s
12395 tcp-request switch-mode http if HTTP
12396 tcp-request reject # non-HTTP traffic is implicit here
12397 ...
12398 http-request reject unless is_host_com
12399
Willy Tarreaue9656522010-08-17 15:40:09 +020012400 Example:
Willy Tarreau62644772008-07-16 18:36:06 +020012401 # reject SMTP connection if client speaks first
12402 tcp-request inspect-delay 30s
12403 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012404 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +020012405
12406 # Forward HTTPS connection only if client speaks
12407 tcp-request inspect-delay 30s
12408 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020012409 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +020012410 tcp-request content reject
12411
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012412 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030012413 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012414 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020012415 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030012416 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
12417 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012418
12419 Example:
12420 # track request counts per "base" (concatenation of Host+URL)
12421 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020012422 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010012423
Willy Tarreaue9656522010-08-17 15:40:09 +020012424 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030012425 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +020012426
12427 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +010012428 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +020012429 # protecting all our sites
12430 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012431 tcp-request connection track-sc0 src
12432 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +020012433 ...
12434 use_backend http_dynamic if { path_end .php }
12435
12436 backend http_dynamic
12437 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012438 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +020012439 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012440 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030012441 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020012442 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +020012443 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +020012444
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012445 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +020012446
Jarno Huuskonen95b012b2017-04-06 13:59:14 +030012447 See also : "tcp-request connection", "tcp-request session",
12448 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +020012449
12450
12451tcp-request inspect-delay <timeout>
12452 Set the maximum allowed time to wait for data during content inspection
12453 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020012454 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +020012455 Arguments :
12456 <timeout> is the timeout value specified in milliseconds by default, but
12457 can be in any other unit if the number is suffixed by the unit,
12458 as explained at the top of this document.
12459
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012460 People using HAProxy primarily as a TCP relay are often worried about the
Willy Tarreau62644772008-07-16 18:36:06 +020012461 risk of passing any type of protocol to a server without any analysis. In
12462 order to be able to analyze the request contents, we must first withhold
12463 the data then analyze them. This statement simply enables withholding of
12464 data for at most the specified amount of time.
12465
Willy Tarreaufb356202010-08-03 14:02:05 +020012466 TCP content inspection applies very early when a connection reaches a
12467 frontend, then very early when the connection is forwarded to a backend. This
12468 means that a connection may experience a first delay in the frontend and a
12469 second delay in the backend if both have tcp-request rules.
12470
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012471 Note that when performing content inspection, HAProxy will evaluate the whole
Willy Tarreau62644772008-07-16 18:36:06 +020012472 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012473 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +020012474 a last check is performed upon expiration, this time considering that the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012475 contents are definitive. If no delay is set, HAProxy will not wait at all
Willy Tarreaud869b242009-03-15 14:43:58 +010012476 and will immediately apply a verdict based on the available information.
12477 Obviously this is unlikely to be very useful and might even be racy, so such
12478 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020012479
12480 As soon as a rule matches, the request is released and continues as usual. If
12481 the timeout is reached and no rule matches, the default policy will be to let
12482 it pass through unaffected.
12483
12484 For most protocols, it is enough to set it to a few seconds, as most clients
12485 send the full request immediately upon connection. Add 3 or more seconds to
12486 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +010012487 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +010012488 before the server (e.g. SMTP), or to wait for a client to talk before passing
12489 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +020012490 least the inspection delay, otherwise it will expire first. If the client
12491 closes the connection or if the buffer is full, the delay immediately expires
12492 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +020012493
Willy Tarreau55165fe2009-05-10 12:02:55 +020012494 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +020012495 "timeout client".
12496
12497
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012498tcp-response content <action> [{if | unless} <condition>]
12499 Perform an action on a session response depending on a layer 4-7 condition
12500 May be used in sections : defaults | frontend | listen | backend
12501 no | no | yes | yes
12502 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020012503 <action> defines the action to perform if the condition applies. See
12504 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012505
12506 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
12507
Davor Ocelice9ed2812017-12-25 17:49:28 +010012508 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012509 called "TCP content inspection". During this stage, ACL-based rules are
12510 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020012511 "accept", "close" or a "reject" rule matches, or a TCP response inspection
12512 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012513
12514 Most often, these decisions will consider a protocol recognition or validity.
12515
12516 Content-based rules are evaluated in their exact declaration order. If no
12517 rule matches or if there is no rule, the default action is to accept the
12518 contents. There is no specific limit to the number of rules which may be
12519 inserted.
12520
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012521 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012522 - accept :
12523 accepts the response if the condition is true (when used with "if")
12524 or false (when used with "unless"). The first such rule executed ends
12525 the rules evaluation.
12526
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020012527 - close :
12528 immediately closes the connection with the server if the condition is
12529 true (when used with "if"), or false (when used with "unless"). The
12530 first such rule executed ends the rules evaluation. The main purpose of
12531 this action is to force a connection to be finished between a client
12532 and a server after an exchange when the application protocol expects
12533 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012534 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020012535 protocols.
12536
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012537 - reject :
12538 rejects the response if the condition is true (when used with "if")
12539 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012540 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012541
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012542 - set-var(<var-name>) <expr>
Willy Tarreau9de54ba2021-09-02 20:51:21 +020012543 Sets a variable from an expression.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012544
Christopher Faulet85d79c92016-11-09 16:54:56 +010012545 - unset-var(<var-name>)
12546 Unsets a variable.
12547
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020012548 - sc-inc-gpc0(<sc-id>):
12549 This action increments the GPC0 counter according to the sticky
12550 counter designated by <sc-id>. If an error occurs, this action fails
12551 silently and the actions evaluation continues.
12552
Frédéric Lécaille6778b272018-01-29 15:22:53 +010012553 - sc-inc-gpc1(<sc-id>):
12554 This action increments the GPC1 counter according to the sticky
12555 counter designated by <sc-id>. If an error occurs, this action fails
12556 silently and the actions evaluation continues.
12557
Cédric Dufour0d7712d2019-11-06 18:38:53 +010012558 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
12559 This action sets the 32-bit unsigned GPT0 tag according to the sticky
12560 counter designated by <sc-id> and the value of <int>/<expr>. The
12561 expected result is a boolean. If an error occurs, this action silently
12562 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +020012563
Willy Tarreau2d392c22015-08-24 01:43:45 +020012564 - "silent-drop" :
12565 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012566 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020012567 to prevent the client from being notified. The effect it then that the
12568 client still sees an established connection while there's none on
12569 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
12570 except that it doesn't use any local resource at all on the machine
12571 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010012572 slow down stronger attackers. It is important to understand the impact
12573 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020012574 client and HAProxy (firewalls, proxies, load balancers) will also keep
12575 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010012576 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020012577 TCP_REPAIR socket option is used to block the emission of a TCP
12578 reset. On other systems, the socket's TTL is reduced to 1 so that the
12579 TCP reset doesn't pass the first router, though it's still delivered to
12580 local networks. Do not use it unless you fully understand how it works.
12581
Christopher Faulet76c09ef2017-09-21 11:03:52 +020012582 - send-spoe-group <engine-name> <group-name>
12583 Send a group of SPOE messages.
12584
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012585 Note that the "if/unless" condition is optional. If no condition is set on
12586 the action, it is simply performed unconditionally. That can be useful for
12587 for changing the default action to a reject.
12588
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012589 It is perfectly possible to match layer 7 contents with "tcp-response
12590 content" rules, but then it is important to ensure that a full response has
12591 been buffered, otherwise no contents will match. In order to achieve this,
12592 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012593 period.
12594
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012595 The "set-var" is used to set the content of a variable. The variable is
12596 declared inline.
12597
Daniel Schneller0b547052016-03-21 20:46:57 +010012598 <var-name> The name of the variable starts with an indication about
12599 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010012600 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010012601 "sess" : the variable is shared with the whole session
12602 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012603 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010012604 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012605 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010012606 "res" : the variable is shared only during response
12607 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012608 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010012609 The name may only contain characters 'a-z', 'A-Z', '0-9',
12610 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020012611
12612 <expr> Is a standard HAProxy expression formed by a sample-fetch
12613 followed by some converters.
12614
Christopher Faulet85d79c92016-11-09 16:54:56 +010012615 The "unset-var" is used to unset a variable. See above for details about
12616 <var-name>.
12617
Christopher Faulet76c09ef2017-09-21 11:03:52 +020012618 The "send-spoe-group" is used to trigger sending of a group of SPOE
12619 messages. To do so, the SPOE engine used to send messages must be defined, as
12620 well as the SPOE group to send. Of course, the SPOE engine must refer to an
12621 existing SPOE filter. If not engine name is provided on the SPOE filter line,
12622 the SPOE agent name must be used.
12623
12624 <engine-name> The SPOE engine name.
12625
12626 <group-name> The SPOE group name as specified in the engine configuration.
12627
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012628 See section 7 about ACL usage.
12629
12630 See also : "tcp-request content", "tcp-response inspect-delay"
12631
12632
Willy Tarreau4f614292016-10-21 17:49:36 +020012633tcp-request session <action> [{if | unless} <condition>]
12634 Perform an action on a validated session depending on a layer 5 condition
12635 May be used in sections : defaults | frontend | listen | backend
12636 no | yes | yes | no
12637 Arguments :
12638 <action> defines the action to perform if the condition applies. See
12639 below.
12640
12641 <condition> is a standard layer5-only ACL-based condition (see section 7).
12642
Davor Ocelice9ed2812017-12-25 17:49:28 +010012643 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020012644 it is possible to evaluate some conditions to decide whether this session
12645 must be accepted or dropped or have its counters tracked. Those conditions
12646 cannot make use of any data contents because no buffers are allocated yet and
12647 the processing cannot wait at this stage. The main use case it to copy some
12648 early information into variables (since variables are accessible in the
12649 session), or to keep track of some information collected after the handshake,
12650 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010012651 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020012652 extracted information can thus be copied to a variable or tracked using
12653 "track-sc" rules. Of course it is also possible to decide to accept/reject as
12654 with other rulesets. Most operations performed here could also be performed
12655 in "tcp-request content" rules, except that in HTTP these rules are evaluated
12656 for each new request, and that might not always be acceptable. For example a
12657 rule might increment a counter on each evaluation. It would also be possible
12658 that a country is resolved by geolocation from the source IP address,
12659 assigned to a session-wide variable, then the source address rewritten from
12660 an HTTP header for all requests. If some contents need to be inspected in
12661 order to take the decision, the "tcp-request content" statements must be used
12662 instead.
12663
12664 The "tcp-request session" rules are evaluated in their exact declaration
12665 order. If no rule matches or if there is no rule, the default action is to
12666 accept the incoming session. There is no specific limit to the number of
12667 rules which may be inserted.
12668
12669 Several types of actions are supported :
12670 - accept : the request is accepted
12671 - reject : the request is rejected and the connection is closed
12672 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
12673 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010012674 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +010012675 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Christopher Faulet57759f32021-06-23 12:19:25 +020012676 - set-dst <expr>
12677 - set-dst-port <expr>
12678 - set-src <expr>
12679 - set-src-port <expr>
Willy Tarreau4f614292016-10-21 17:49:36 +020012680 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010012681 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020012682 - silent-drop
12683
12684 These actions have the same meaning as their respective counter-parts in
12685 "tcp-request connection" and "tcp-request content", so please refer to these
12686 sections for a complete description.
12687
12688 Note that the "if/unless" condition is optional. If no condition is set on
12689 the action, it is simply performed unconditionally. That can be useful for
12690 "track-sc*" actions as well as for changing the default action to a reject.
12691
12692 Example: track the original source address by default, or the one advertised
12693 in the PROXY protocol header for connection coming from the local
12694 proxies. The first connection-level rule enables receipt of the
12695 PROXY protocol for these ones, the second rule tracks whatever
12696 address we decide to keep after optional decoding.
12697
12698 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
12699 tcp-request session track-sc0 src
12700
12701 Example: accept all sessions from white-listed hosts, reject too fast
12702 sessions without counting them, and track accepted sessions.
12703 This results in session rate being capped from abusive sources.
12704
12705 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
12706 tcp-request session reject if { src_sess_rate gt 10 }
12707 tcp-request session track-sc0 src
12708
12709 Example: accept all sessions from white-listed hosts, count all other
12710 sessions and reject too fast ones. This results in abusive ones
12711 being blocked as long as they don't slow down.
12712
12713 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
12714 tcp-request session track-sc0 src
12715 tcp-request session reject if { sc0_sess_rate gt 10 }
12716
12717 See section 7 about ACL usage.
12718
12719 See also : "tcp-request connection", "tcp-request content", "stick-table"
12720
12721
Emeric Brun0a3b67f2010-09-24 15:34:53 +020012722tcp-response inspect-delay <timeout>
12723 Set the maximum allowed time to wait for a response during content inspection
12724 May be used in sections : defaults | frontend | listen | backend
12725 no | no | yes | yes
12726 Arguments :
12727 <timeout> is the timeout value specified in milliseconds by default, but
12728 can be in any other unit if the number is suffixed by the unit,
12729 as explained at the top of this document.
12730
12731 See also : "tcp-response content", "tcp-request inspect-delay".
12732
12733
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012734timeout check <timeout>
12735 Set additional check timeout, but only after a connection has been already
12736 established.
12737
12738 May be used in sections: defaults | frontend | listen | backend
12739 yes | no | yes | yes
12740 Arguments:
12741 <timeout> is the timeout value specified in milliseconds by default, but
12742 can be in any other unit if the number is suffixed by the unit,
12743 as explained at the top of this document.
12744
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012745 If set, HAProxy uses min("timeout connect", "inter") as a connect timeout
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012746 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010012747 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012748 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010012749 (Please also note that there is no valid reason to have such long connect
12750 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
12751 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012752
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012753 If "timeout check" is not set HAProxy uses "inter" for complete check
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012754 timeout (connect + read) exactly like all <1.3.15 version.
12755
12756 In most cases check request is much simpler and faster to handle than normal
12757 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010012758 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012759
12760 This parameter is specific to backends, but can be specified once for all in
12761 "defaults" sections. This is in fact one of the easiest solutions not to
12762 forget about it.
12763
Willy Tarreau41a340d2008-01-22 12:25:31 +010012764 See also: "timeout connect", "timeout queue", "timeout server",
12765 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012766
12767
Willy Tarreau0ba27502007-12-24 16:55:16 +010012768timeout client <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010012769 Set the maximum inactivity time on the client side.
12770 May be used in sections : defaults | frontend | listen | backend
12771 yes | yes | yes | no
12772 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010012773 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010012774 can be in any other unit if the number is suffixed by the unit,
12775 as explained at the top of this document.
12776
12777 The inactivity timeout applies when the client is expected to acknowledge or
12778 send data. In HTTP mode, this timeout is particularly important to consider
12779 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010012780 response while it is reading data sent by the server. That said, for the
12781 first phase, it is preferable to set the "timeout http-request" to better
12782 protect HAProxy from Slowloris like attacks. The value is specified in
12783 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010012784 suffixed by the unit, as specified at the top of this document. In TCP mode
12785 (and to a lesser extent, in HTTP mode), it is highly recommended that the
12786 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010012787 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010012788 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010012789 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
12790 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020012791 which overrides "timeout client" and "timeout server" for tunnels, as well as
12792 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012793
12794 This parameter is specific to frontends, but can be specified once for all in
12795 "defaults" sections. This is in fact one of the easiest solutions not to
12796 forget about it. An unspecified timeout results in an infinite timeout, which
12797 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050012798 during startup because it may result in accumulation of expired sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010012799 the system if the system's timeouts are not configured either.
12800
Willy Tarreau95c4e142017-11-26 12:18:55 +010012801 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010012802
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020012803 See also : "timeout server", "timeout tunnel", "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012804
Willy Tarreau0ba27502007-12-24 16:55:16 +010012805
Willy Tarreau05cdd962014-05-10 14:30:07 +020012806timeout client-fin <timeout>
12807 Set the inactivity timeout on the client side for half-closed connections.
12808 May be used in sections : defaults | frontend | listen | backend
12809 yes | yes | yes | no
12810 Arguments :
12811 <timeout> is the timeout value specified in milliseconds by default, but
12812 can be in any other unit if the number is suffixed by the unit,
12813 as explained at the top of this document.
12814
12815 The inactivity timeout applies when the client is expected to acknowledge or
12816 send data while one direction is already shut down. This timeout is different
12817 from "timeout client" in that it only applies to connections which are closed
12818 in one direction. This is particularly useful to avoid keeping connections in
12819 FIN_WAIT state for too long when clients do not disconnect cleanly. This
12820 problem is particularly common long connections such as RDP or WebSocket.
12821 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010012822 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
12823 frame was sent, often indicating an expectation that the connection quickly
12824 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020012825
12826 This parameter is specific to frontends, but can be specified once for all in
12827 "defaults" sections. By default it is not set, so half-closed connections
12828 will use the other timeouts (timeout.client or timeout.tunnel).
12829
12830 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
12831
12832
Willy Tarreau0ba27502007-12-24 16:55:16 +010012833timeout connect <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010012834 Set the maximum time to wait for a connection attempt to a server to succeed.
12835 May be used in sections : defaults | frontend | listen | backend
12836 yes | no | yes | yes
12837 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010012838 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010012839 can be in any other unit if the number is suffixed by the unit,
12840 as explained at the top of this document.
12841
Daniel Corbett9f0843f2021-05-08 10:50:37 -040012842 If the server is located on the same LAN as HAProxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010012843 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010012844 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010012845 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010012846 connect timeout also presets both queue and tarpit timeouts to the same value
12847 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012848
12849 This parameter is specific to backends, but can be specified once for all in
12850 "defaults" sections. This is in fact one of the easiest solutions not to
12851 forget about it. An unspecified timeout results in an infinite timeout, which
12852 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050012853 during startup because it may result in accumulation of failed sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010012854 the system if the system's timeouts are not configured either.
12855
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020012856 See also: "timeout check", "timeout queue", "timeout server", "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012857
Willy Tarreau0ba27502007-12-24 16:55:16 +010012858
Willy Tarreaub16a5742010-01-10 14:46:16 +010012859timeout http-keep-alive <timeout>
12860 Set the maximum allowed time to wait for a new HTTP request to appear
12861 May be used in sections : defaults | frontend | listen | backend
12862 yes | yes | yes | yes
12863 Arguments :
12864 <timeout> is the timeout value specified in milliseconds by default, but
12865 can be in any other unit if the number is suffixed by the unit,
12866 as explained at the top of this document.
12867
12868 By default, the time to wait for a new request in case of keep-alive is set
12869 by "timeout http-request". However this is not always convenient because some
12870 people want very short keep-alive timeouts in order to release connections
12871 faster, and others prefer to have larger ones but still have short timeouts
12872 once the request has started to present itself.
12873
12874 The "http-keep-alive" timeout covers these needs. It will define how long to
12875 wait for a new HTTP request to start coming after a response was sent. Once
12876 the first byte of request has been seen, the "http-request" timeout is used
12877 to wait for the complete request to come. Note that empty lines prior to a
12878 new request do not refresh the timeout and are not counted as a new request.
12879
12880 There is also another difference between the two timeouts : when a connection
12881 expires during timeout http-keep-alive, no error is returned, the connection
12882 just closes. If the connection expires in "http-request" while waiting for a
12883 connection to complete, a HTTP 408 error is returned.
12884
12885 In general it is optimal to set this value to a few tens to hundreds of
12886 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010012887 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010012888 1 millisecond) it will probably only accept pipelined requests but not the
12889 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020012890 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010012891
12892 If this parameter is not set, the "http-request" timeout applies, and if both
12893 are not set, "timeout client" still applies at the lower level. It should be
12894 set in the frontend to take effect, unless the frontend is in TCP mode, in
12895 which case the HTTP backend's timeout will be used.
12896
Willy Tarreau95c4e142017-11-26 12:18:55 +010012897 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
12898 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010012899 (where we only have one connection per client and a connection setup).
12900
Willy Tarreaub16a5742010-01-10 14:46:16 +010012901 See also : "timeout http-request", "timeout client".
12902
12903
Willy Tarreau036fae02008-01-06 13:24:40 +010012904timeout http-request <timeout>
12905 Set the maximum allowed time to wait for a complete HTTP request
12906 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020012907 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010012908 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010012909 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010012910 can be in any other unit if the number is suffixed by the unit,
12911 as explained at the top of this document.
12912
12913 In order to offer DoS protection, it may be required to lower the maximum
12914 accepted time to receive a complete HTTP request without affecting the client
12915 timeout. This helps protecting against established connections on which
12916 nothing is sent. The client timeout cannot offer a good protection against
12917 this abuse because it is an inactivity timeout, which means that if the
12918 attacker sends one character every now and then, the timeout will not
12919 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020012920 types, the request will be aborted if it does not complete in time. When the
12921 timeout expires, an HTTP 408 response is sent to the client to inform it
12922 about the problem, and the connection is closed. The logs will report
12923 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010012924 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020012925 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
12926 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010012927
Baptiste Assmanneccdf432015-10-28 13:49:01 +010012928 By default, this timeout only applies to the header part of the request,
12929 and not to any data. As soon as the empty line is received, this timeout is
12930 not used anymore. When combined with "option http-buffer-request", this
12931 timeout also applies to the body of the request..
12932 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010012933 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010012934
12935 Generally it is enough to set it to a few seconds, as most clients send the
12936 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010012937 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010012938 generally work on local networks as long as there are no packet losses. This
12939 will prevent people from sending bare HTTP requests using telnet.
12940
12941 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020012942 chunk of the incoming request. It should be set in the frontend to take
12943 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
12944 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010012945
Willy Tarreau0f228a02015-05-01 15:37:53 +020012946 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010012947 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010012948
Willy Tarreau844e3c52008-01-11 16:28:18 +010012949
12950timeout queue <timeout>
12951 Set the maximum time to wait in the queue for a connection slot to be free
12952 May be used in sections : defaults | frontend | listen | backend
12953 yes | no | yes | yes
12954 Arguments :
12955 <timeout> is the timeout value specified in milliseconds by default, but
12956 can be in any other unit if the number is suffixed by the unit,
12957 as explained at the top of this document.
12958
12959 When a server's maxconn is reached, connections are left pending in a queue
12960 which may be server-specific or global to the backend. In order not to wait
12961 indefinitely, a timeout is applied to requests pending in the queue. If the
12962 timeout is reached, it is considered that the request will almost never be
12963 served, so it is dropped and a 503 error is returned to the client.
12964
12965 The "timeout queue" statement allows to fix the maximum time for a request to
12966 be left pending in a queue. If unspecified, the same value as the backend's
12967 connection timeout ("timeout connect") is used, for backwards compatibility
12968 with older versions with no "timeout queue" parameter.
12969
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020012970 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010012971
12972
12973timeout server <timeout>
Willy Tarreau844e3c52008-01-11 16:28:18 +010012974 Set the maximum inactivity time on the server side.
12975 May be used in sections : defaults | frontend | listen | backend
12976 yes | no | yes | yes
12977 Arguments :
12978 <timeout> is the timeout value specified in milliseconds by default, but
12979 can be in any other unit if the number is suffixed by the unit,
12980 as explained at the top of this document.
12981
12982 The inactivity timeout applies when the server is expected to acknowledge or
12983 send data. In HTTP mode, this timeout is particularly important to consider
12984 during the first phase of the server's response, when it has to send the
12985 headers, as it directly represents the server's processing time for the
12986 request. To find out what value to put there, it's often good to start with
12987 what would be considered as unacceptable response times, then check the logs
12988 to observe the response time distribution, and adjust the value accordingly.
12989
12990 The value is specified in milliseconds by default, but can be in any other
12991 unit if the number is suffixed by the unit, as specified at the top of this
12992 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
12993 recommended that the client timeout remains equal to the server timeout in
12994 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010012995 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010012996 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010012997 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
12998 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020012999 "timeout tunnel", which overrides "timeout client" and "timeout server" for
13000 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013001
13002 This parameter is specific to backends, but can be specified once for all in
13003 "defaults" sections. This is in fact one of the easiest solutions not to
13004 forget about it. An unspecified timeout results in an infinite timeout, which
13005 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050013006 during startup because it may result in accumulation of expired sessions in
Willy Tarreau844e3c52008-01-11 16:28:18 +010013007 the system if the system's timeouts are not configured either.
13008
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020013009 See also : "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010013010
Willy Tarreau05cdd962014-05-10 14:30:07 +020013011
13012timeout server-fin <timeout>
13013 Set the inactivity timeout on the server side for half-closed connections.
13014 May be used in sections : defaults | frontend | listen | backend
13015 yes | no | yes | yes
13016 Arguments :
13017 <timeout> is the timeout value specified in milliseconds by default, but
13018 can be in any other unit if the number is suffixed by the unit,
13019 as explained at the top of this document.
13020
13021 The inactivity timeout applies when the server is expected to acknowledge or
13022 send data while one direction is already shut down. This timeout is different
13023 from "timeout server" in that it only applies to connections which are closed
13024 in one direction. This is particularly useful to avoid keeping connections in
13025 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
13026 This problem is particularly common long connections such as RDP or WebSocket.
13027 Note that this timeout can override "timeout tunnel" when a connection shuts
13028 down in one direction. This setting was provided for completeness, but in most
13029 situations, it should not be needed.
13030
13031 This parameter is specific to backends, but can be specified once for all in
13032 "defaults" sections. By default it is not set, so half-closed connections
13033 will use the other timeouts (timeout.server or timeout.tunnel).
13034
13035 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
13036
Willy Tarreau844e3c52008-01-11 16:28:18 +010013037
13038timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010013039 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010013040 May be used in sections : defaults | frontend | listen | backend
13041 yes | yes | yes | yes
13042 Arguments :
13043 <timeout> is the tarpit duration specified in milliseconds by default, but
13044 can be in any other unit if the number is suffixed by the unit,
13045 as explained at the top of this document.
13046
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020013047 When a connection is tarpitted using "http-request tarpit", it is maintained
13048 open with no activity for a certain amount of time, then closed. "timeout
13049 tarpit" defines how long it will be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013050
13051 The value is specified in milliseconds by default, but can be in any other
13052 unit if the number is suffixed by the unit, as specified at the top of this
13053 document. If unspecified, the same value as the backend's connection timeout
13054 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010013055 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013056
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020013057 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010013058
13059
Willy Tarreauce887fd2012-05-12 12:50:00 +020013060timeout tunnel <timeout>
13061 Set the maximum inactivity time on the client and server side for tunnels.
13062 May be used in sections : defaults | frontend | listen | backend
13063 yes | no | yes | yes
13064 Arguments :
13065 <timeout> is the timeout value specified in milliseconds by default, but
13066 can be in any other unit if the number is suffixed by the unit,
13067 as explained at the top of this document.
13068
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013069 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020013070 between a client and a server, and the connection remains inactive in both
13071 directions. This timeout supersedes both the client and server timeouts once
13072 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010013073 analyzer remains attached to either connection (e.g. tcp content rules are
13074 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020013075 when switching to the WebSocket protocol, or forwarding a CONNECT request
13076 to a proxy), or after the first response when no keepalive/close option is
13077 specified.
13078
Willy Tarreau05cdd962014-05-10 14:30:07 +020013079 Since this timeout is usually used in conjunction with long-lived connections,
13080 it usually is a good idea to also set "timeout client-fin" to handle the
13081 situation where a client suddenly disappears from the net and does not
13082 acknowledge a close, or sends a shutdown and does not acknowledge pending
13083 data anymore. This can happen in lossy networks where firewalls are present,
13084 and is detected by the presence of large amounts of sessions in a FIN_WAIT
13085 state.
13086
Willy Tarreauce887fd2012-05-12 12:50:00 +020013087 The value is specified in milliseconds by default, but can be in any other
13088 unit if the number is suffixed by the unit, as specified at the top of this
13089 document. Whatever the expected normal idle time, it is a good practice to
13090 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010013091 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020013092
13093 This parameter is specific to backends, but can be specified once for all in
13094 "defaults" sections. This is in fact one of the easiest solutions not to
13095 forget about it.
13096
13097 Example :
13098 defaults http
13099 option http-server-close
13100 timeout connect 5s
13101 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020013102 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020013103 timeout server 30s
13104 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
13105
Willy Tarreau05cdd962014-05-10 14:30:07 +020013106 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020013107
13108
Willy Tarreau844e3c52008-01-11 16:28:18 +010013109transparent (deprecated)
13110 Enable client-side transparent proxying
13111 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010013112 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010013113 Arguments : none
13114
13115 This keyword was introduced in order to provide layer 7 persistence to layer
13116 3 load balancers. The idea is to use the OS's ability to redirect an incoming
13117 connection for a remote address to a local process (here HAProxy), and let
13118 this process know what address was initially requested. When this option is
13119 used, sessions without cookies will be forwarded to the original destination
13120 IP address of the incoming request (which should match that of another
13121 equipment), while requests with cookies will still be forwarded to the
13122 appropriate server.
13123
13124 The "transparent" keyword is deprecated, use "option transparent" instead.
13125
13126 Note that contrary to a common belief, this option does NOT make HAProxy
13127 present the client's IP to the server when establishing the connection.
13128
Willy Tarreau844e3c52008-01-11 16:28:18 +010013129 See also: "option transparent"
13130
William Lallemanda73203e2012-03-12 12:48:57 +010013131unique-id-format <string>
13132 Generate a unique ID for each request.
13133 May be used in sections : defaults | frontend | listen | backend
13134 yes | yes | yes | no
13135 Arguments :
13136 <string> is a log-format string.
13137
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013138 This keyword creates a ID for each request using the custom log format. A
13139 unique ID is useful to trace a request passing through many components of
13140 a complex infrastructure. The newly created ID may also be logged using the
13141 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010013142
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013143 The format should be composed from elements that are guaranteed to be
Daniel Corbett9f0843f2021-05-08 10:50:37 -040013144 unique when combined together. For instance, if multiple HAProxy instances
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013145 are involved, it might be important to include the node name. It is often
13146 needed to log the incoming connection's source and destination addresses
13147 and ports. Note that since multiple requests may be performed over the same
13148 connection, including a request counter may help differentiate them.
13149 Similarly, a timestamp may protect against a rollover of the counter.
13150 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010013151
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013152 It is recommended to use hexadecimal notation for many fields since it
13153 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010013154
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013155 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010013156
Julien Vehentf21be322014-03-07 08:27:34 -050013157 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010013158
13159 will generate:
13160
13161 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
13162
13163 See also: "unique-id-header"
13164
13165unique-id-header <name>
13166 Add a unique ID header in the HTTP request.
13167 May be used in sections : defaults | frontend | listen | backend
13168 yes | yes | yes | no
13169 Arguments :
13170 <name> is the name of the header.
13171
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013172 Add a unique-id header in the HTTP request sent to the server, using the
13173 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010013174
Cyril Bonté108cf6e2012-04-21 23:30:29 +020013175 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010013176
Julien Vehentf21be322014-03-07 08:27:34 -050013177 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010013178 unique-id-header X-Unique-ID
13179
13180 will generate:
13181
13182 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
13183
13184 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010013185
Willy Tarreauf51658d2014-04-23 01:21:56 +020013186use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020013187 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013188 May be used in sections : defaults | frontend | listen | backend
13189 no | yes | yes | no
13190 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010013191 <backend> is the name of a valid backend or "listen" section, or a
13192 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013193
Willy Tarreauf51658d2014-04-23 01:21:56 +020013194 <condition> is a condition composed of ACLs, as described in section 7. If
13195 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013196
13197 When doing content-switching, connections arrive on a frontend and are then
13198 dispatched to various backends depending on a number of conditions. The
13199 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020013200 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010013201 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020013202 source address validation) or combined with a "tcp-request" rule to wait for
13203 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013204
13205 There may be as many "use_backend" rules as desired. All of these rules are
13206 evaluated in their declaration order, and the first one which matches will
13207 assign the backend.
13208
13209 In the first form, the backend will be used if the condition is met. In the
13210 second form, the backend will be used if the condition is not met. If no
13211 condition is valid, the backend defined with "default_backend" will be used.
13212 If no default backend is defined, either the servers in the same section are
13213 used (in case of a "listen" section) or, in case of a frontend, no server is
13214 used and a 503 service unavailable response is returned.
13215
Willy Tarreau51aecc72009-07-12 09:47:04 +020013216 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013217 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020013218 and backend processing will immediately follow, or the backend will wait for
13219 a complete HTTP request to get in. This feature is useful when a frontend
13220 must decode several protocols on a unique port, one of them being HTTP.
13221
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010013222 When <backend> is a simple name, it is resolved at configuration time, and an
13223 error is reported if the specified backend does not exist. If <backend> is
13224 a log-format string instead, no check may be done at configuration time, so
13225 the backend name is resolved dynamically at run time. If the resulting
13226 backend name does not correspond to any valid backend, no other rule is
13227 evaluated, and the default_backend directive is applied instead. Note that
13228 when using dynamic backend names, it is highly recommended to use a prefix
13229 that no other backend uses in order to ensure that an unauthorized backend
13230 cannot be forced from the request.
13231
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013232 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010013233 used to detect the association between frontends and backends to compute the
13234 backend's "fullconn" setting. This cannot be done for dynamic names.
13235
13236 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
13237 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010013238
Christopher Fauletb30b3102019-09-12 23:03:09 +020013239use-fcgi-app <name>
13240 Defines the FastCGI application to use for the backend.
13241 May be used in sections : defaults | frontend | listen | backend
13242 no | no | yes | yes
13243 Arguments :
13244 <name> is the name of the FastCGI application to use.
13245
13246 See section 10.1 about FastCGI application setup for details.
Willy Tarreau036fae02008-01-06 13:24:40 +010013247
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013248use-server <server> if <condition>
13249use-server <server> unless <condition>
13250 Only use a specific server if/unless an ACL-based condition is matched.
13251 May be used in sections : defaults | frontend | listen | backend
13252 no | no | yes | yes
13253 Arguments :
Jerome Magnin824186b2020-03-29 09:37:12 +020013254 <server> is the name of a valid server in the same backend section
13255 or a "log-format" string resolving to a server name.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013256
13257 <condition> is a condition composed of ACLs, as described in section 7.
13258
13259 By default, connections which arrive to a backend are load-balanced across
13260 the available servers according to the configured algorithm, unless a
13261 persistence mechanism such as a cookie is used and found in the request.
13262
13263 Sometimes it is desirable to forward a particular request to a specific
13264 server without having to declare a dedicated backend for this server. This
13265 can be achieved using the "use-server" rules. These rules are evaluated after
13266 the "redirect" rules and before evaluating cookies, and they have precedence
13267 on them. There may be as many "use-server" rules as desired. All of these
13268 rules are evaluated in their declaration order, and the first one which
13269 matches will assign the server.
13270
13271 If a rule designates a server which is down, and "option persist" is not used
13272 and no force-persist rule was validated, it is ignored and evaluation goes on
13273 with the next rules until one matches.
13274
13275 In the first form, the server will be used if the condition is met. In the
13276 second form, the server will be used if the condition is not met. If no
13277 condition is valid, the processing continues and the server will be assigned
13278 according to other persistence mechanisms.
13279
13280 Note that even if a rule is matched, cookie processing is still performed but
13281 does not assign the server. This allows prefixed cookies to have their prefix
13282 stripped.
13283
13284 The "use-server" statement works both in HTTP and TCP mode. This makes it
13285 suitable for use with content-based inspection. For instance, a server could
Lukas Tribusa267b5d2020-07-19 00:25:06 +020013286 be selected in a farm according to the TLS SNI field when using protocols with
Alex18c88772021-06-05 13:23:08 +020013287 implicit TLS (also see "req.ssl_sni"). And if these servers have their weight
Lukas Tribusa267b5d2020-07-19 00:25:06 +020013288 set to zero, they will not be used for other traffic.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013289
13290 Example :
13291 # intercept incoming TLS requests based on the SNI field
Alex18c88772021-06-05 13:23:08 +020013292 use-server www if { req.ssl_sni -i www.example.com }
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013293 server www 192.168.0.1:443 weight 0
Alex18c88772021-06-05 13:23:08 +020013294 use-server mail if { req.ssl_sni -i mail.example.com }
Lukas Tribusa267b5d2020-07-19 00:25:06 +020013295 server mail 192.168.0.1:465 weight 0
Alex18c88772021-06-05 13:23:08 +020013296 use-server imap if { req.ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000013297 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013298 # all the rest is forwarded to this server
13299 server default 192.168.0.2:443 check
13300
Jerome Magnin824186b2020-03-29 09:37:12 +020013301 When <server> is a simple name, it is checked against existing servers in the
13302 configuration and an error is reported if the specified server does not exist.
13303 If it is a log-format, no check is performed when parsing the configuration,
13304 and if we can't resolve a valid server name at runtime but the use-server rule
Ilya Shipitsin11057a32020-06-21 21:18:27 +050013305 was conditioned by an ACL returning true, no other use-server rule is applied
Jerome Magnin824186b2020-03-29 09:37:12 +020013306 and we fall back to load balancing.
13307
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013308 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013309
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013310
Davor Ocelice9ed2812017-12-25 17:49:28 +0100133115. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013312--------------------------
13313
13314The "bind", "server" and "default-server" keywords support a number of settings
13315depending on some build options and on the system HAProxy was built on. These
13316settings generally each consist in one word sometimes followed by a value,
13317written on the same line as the "bind" or "server" line. All these options are
13318described in this section.
13319
13320
133215.1. Bind options
13322-----------------
13323
13324The "bind" keyword supports a certain number of settings which are all passed
13325as arguments on the same line. The order in which those arguments appear makes
13326no importance, provided that they appear after the bind address. All of these
13327parameters are optional. Some of them consist in a single words (booleans),
13328while other ones expect a value after them. In this case, the value must be
13329provided immediately after the setting name.
13330
13331The currently supported settings are the following ones.
13332
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010013333accept-netscaler-cip <magic number>
13334 Enforces the use of the NetScaler Client IP insertion protocol over any
13335 connection accepted by any of the TCP sockets declared on the same line. The
13336 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
13337 the incoming connection to be used everywhere an address is used, with the
13338 only exception of "tcp-request connection" rules which will only see the
13339 real connection address. Logs will reflect the addresses indicated in the
13340 protocol, unless it is violated, in which case the real address will still
13341 be used. This keyword combined with support from external components can be
13342 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010013343 mechanism which is not always reliable and not even always usable. See also
13344 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
13345 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010013346
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013347accept-proxy
13348 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020013349 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
13350 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013351 3/4 addresses of the incoming connection to be used everywhere an address is
13352 used, with the only exception of "tcp-request connection" rules which will
13353 only see the real connection address. Logs will reflect the addresses
13354 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010013355 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013356 components can be used as an efficient and reliable alternative to the
13357 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020013358 usable. See also "tcp-request connection expect-proxy" for a finer-grained
13359 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013360
Olivier Houchardc2aae742017-09-22 18:26:28 +020013361allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010013362 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010013363 due to security considerations. Because it is vulnerable to replay attacks,
John Roeslerfb2fce12019-07-10 15:45:51 -050013364 you should only allow if for requests that are safe to replay, i.e. requests
Olivier Houchard69752962019-01-08 15:35:32 +010013365 that are idempotent. You can use the "wait-for-handshake" action for any
13366 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020013367
Willy Tarreauab861d32013-04-02 02:30:41 +020013368alpn <protocols>
13369 This enables the TLS ALPN extension and advertises the specified protocol
13370 list as supported on top of ALPN. The protocol list consists in a comma-
13371 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050013372 quotes). This requires that the SSL library is built with support for TLS
Willy Tarreauab861d32013-04-02 02:30:41 +020013373 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010013374 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
13375 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
13376 now obsolete NPN extension. At the time of writing this, most browsers still
13377 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
13378 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
13379 are expected to be supported, both versions can be advertised, in order of
13380 preference, like below :
13381
13382 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020013383
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013384backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010013385 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013386 backlog is used instead, which generally defaults to the maxconn value.
13387
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010013388curves <curves>
13389 This setting is only available when support for OpenSSL was built in. It sets
13390 the string describing the list of elliptic curves algorithms ("curve suite")
13391 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
13392 string is a colon-delimited list of curve name.
13393 Example: "X25519:P-256" (without quote)
13394 When "curves" is set, "ecdhe" parameter is ignored.
13395
Emeric Brun7fb34422012-09-28 15:26:15 +020013396ecdhe <named curve>
13397 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010013398 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
13399 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020013400
Emeric Brunfd33a262012-10-11 16:28:27 +020013401ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020013402 This setting is only available when support for OpenSSL was built in. It
13403 designates a PEM file from which to load CA certificates used to verify
13404 client's certificate.
13405
Emeric Brunb6dc9342012-09-28 17:55:37 +020013406ca-ignore-err [all|<errorID>,...]
13407 This setting is only available when support for OpenSSL was built in.
13408 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
13409 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
13410 error is ignored.
13411
Christopher Faulet31af49d2015-06-09 17:29:50 +020013412ca-sign-file <cafile>
13413 This setting is only available when support for OpenSSL was built in. It
13414 designates a PEM file containing both the CA certificate and the CA private
13415 key used to create and sign server's certificates. This is a mandatory
13416 setting when the dynamic generation of certificates is enabled. See
13417 'generate-certificates' for details.
13418
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000013419ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020013420 This setting is only available when support for OpenSSL was built in. It is
13421 the CA private key passphrase. This setting is optional and used only when
13422 the dynamic generation of certificates is enabled. See
13423 'generate-certificates' for details.
13424
Emmanuel Hocdet842e94e2019-12-16 16:39:17 +010013425ca-verify-file <cafile>
13426 This setting designates a PEM file from which to load CA certificates used to
13427 verify client's certificate. It designates CA certificates which must not be
13428 included in CA names sent in server hello message. Typically, "ca-file" must
13429 be defined with intermediate certificates, and "ca-verify-file" with
13430 certificates to ending the chain, like root CA.
13431
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013432ciphers <ciphers>
13433 This setting is only available when support for OpenSSL was built in. It sets
13434 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000013435 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000013436 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020013437 information and recommendations see e.g.
13438 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
13439 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
13440 cipher configuration, please check the "ciphersuites" keyword.
13441
13442ciphersuites <ciphersuites>
13443 This setting is only available when support for OpenSSL was built in and
13444 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
13445 the list of cipher algorithms ("cipher suite") that are negotiated during the
13446 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000013447 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
13448 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013449
Emeric Brunfd33a262012-10-11 16:28:27 +020013450crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020013451 This setting is only available when support for OpenSSL was built in. It
13452 designates a PEM file from which to load certificate revocation list used
Remi Tricot-Le Breton02bd6842021-05-04 12:22:34 +020013453 to verify client's certificate. You need to provide a certificate revocation
13454 list for every certificate of your certificate authority chain.
Emeric Brun1a073b42012-09-28 17:07:34 +020013455
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013456crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000013457 This setting is only available when support for OpenSSL was built in. It
13458 designates a PEM file containing both the required certificates and any
13459 associated private keys. This file can be built by concatenating multiple
13460 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
13461 requires an intermediate certificate, this can also be concatenated into this
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +010013462 file. Intermediate certificate can also be shared in a directory via
13463 "issuers-chain-path" directive.
Alex Davies0fbf0162013-03-02 16:04:50 +000013464
William Lallemand4c5adbf2020-02-24 14:23:22 +010013465 If the file does not contain a private key, HAProxy will try to load
13466 the key at the same path suffixed by a ".key".
13467
Alex Davies0fbf0162013-03-02 16:04:50 +000013468 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
13469 are loaded.
13470
13471 If a directory name is used instead of a PEM file, then all files found in
William Lallemand3f25ae32020-02-24 16:30:12 +010013472 that directory will be loaded in alphabetic order unless their name ends
13473 with '.key', '.issuer', '.ocsp' or '.sctl' (reserved extensions). This
13474 directive may be specified multiple times in order to load certificates from
13475 multiple files or directories. The certificates will be presented to clients
13476 who provide a valid TLS Server Name Indication field matching one of their
13477 CN or alt subjects. Wildcards are supported, where a wildcard character '*'
13478 is used instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010013479 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000013480
13481 If no SNI is provided by the client or if the SSL library does not support
13482 TLS extensions, or if the client provides an SNI hostname which does not
13483 match any certificate, then the first loaded certificate will be presented.
13484 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010013485 recommended to load the default one first as a file or to ensure that it will
13486 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000013487
Emeric Brune032bfa2012-09-28 13:01:45 +020013488 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013489
Davor Ocelice9ed2812017-12-25 17:49:28 +010013490 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000013491 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010013492 choose a web server that the CA believes requires an intermediate CA (for
13493 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000013494 others, e.g. nginx, result in a wrong bundle that will not work for some
13495 clients).
13496
Daniel Corbett9f0843f2021-05-08 10:50:37 -040013497 For each PEM file, HAProxy checks for the presence of file at the same path
Emeric Brun4147b2e2014-06-16 18:36:30 +020013498 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
13499 Status Request extension (also known as "OCSP stapling") is automatically
13500 enabled. The content of this file is optional. If not empty, it must contain
13501 a valid OCSP Response in DER format. In order to be valid an OCSP Response
13502 must comply with the following rules: it has to indicate a good status,
13503 it has to be a single response for the certificate of the PEM file, and it
13504 has to be valid at the moment of addition. If these rules are not respected
13505 the OCSP Response is ignored and a warning is emitted. In order to identify
13506 which certificate an OCSP Response applies to, the issuer's certificate is
13507 necessary. If the issuer's certificate is not found in the PEM file, it will
13508 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
13509 if it exists otherwise it will fail with an error.
13510
Daniel Corbett9f0843f2021-05-08 10:50:37 -040013511 For each PEM file, HAProxy also checks for the presence of file at the same
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010013512 path suffixed by ".sctl". If such file is found, support for Certificate
13513 Transparency (RFC6962) TLS extension is enabled. The file must contain a
13514 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
13515 to check basic syntax, but no signatures are verified.
13516
yanbzhu6c25e9e2016-01-05 12:52:02 -050013517 There are cases where it is desirable to support multiple key types, e.g. RSA
13518 and ECDSA in the cipher suites offered to the clients. This allows clients
13519 that support EC certificates to be able to use EC ciphers, while
13520 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050013521
William Lallemandf9ff3ec2020-10-02 17:57:44 +020013522 To achieve this, OpenSSL 1.1.1 is required, you can configure this behavior
13523 by providing one crt entry per certificate type, or by configuring a "cert
13524 bundle" like it was required before HAProxy 1.8. See "ssl-load-extra-files".
yanbzhud19630c2015-12-14 15:10:25 -050013525
Emeric Brunb6dc9342012-09-28 17:55:37 +020013526crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000013527 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010013528 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013529 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000013530 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020013531
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010013532crt-list <file>
13533 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010013534 designates a list of PEM file with an optional ssl configuration and a SNI
13535 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010013536
Emmanuel Hocdet98263292016-12-29 18:26:15 +010013537 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
13538
William Lallemand5d036392020-06-30 16:11:36 +020013539 sslbindconf supports "allow-0rtt", "alpn", "ca-file", "ca-verify-file",
13540 "ciphers", "ciphersuites", "crl-file", "curves", "ecdhe", "no-ca-names",
13541 "npn", "verify" configuration. With BoringSSL and Openssl >= 1.1.1
13542 "ssl-min-ver" and "ssl-max-ver" are also supported. It overrides the
13543 configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010013544
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020013545 Wildcards are supported in the SNI filter. Negative filter are also supported,
Joao Moraise51fab02020-11-21 07:42:20 -030013546 useful in combination with a wildcard filter to exclude a particular SNI, or
13547 after the first certificate to exclude a pattern from its CN or Subject Alt
13548 Name (SAN). The certificates will be presented to clients who provide a valid
13549 TLS Server Name Indication field matching one of the SNI filters. If no SNI
13550 filter is specified, the CN and SAN are used. This directive may be specified
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020013551 multiple times. See the "crt" option for more information. The default
13552 certificate is still needed to meet OpenSSL expectations. If it is not used,
13553 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010013554
William Lallemandf9ff3ec2020-10-02 17:57:44 +020013555 Multi-cert bundling (see "ssl-load-extra-files") is supported with crt-list,
13556 as long as only the base name is given in the crt-list. SNI filter will do
13557 the same work on all bundled certificates.
yanbzhud19630c2015-12-14 15:10:25 -050013558
William Lallemand7c26ed72020-06-03 17:34:48 +020013559 Empty lines as well as lines beginning with a hash ('#') will be ignored.
13560
Joao Moraisaa8fcc42020-11-24 08:24:30 -030013561 The first declared certificate of a bind line is used as the default
Daniel Corbett9f0843f2021-05-08 10:50:37 -040013562 certificate, either from crt or crt-list option, which HAProxy should use in
Joao Moraisaa8fcc42020-11-24 08:24:30 -030013563 the TLS handshake if no other certificate matches. This certificate will also
13564 be used if the provided SNI matches its CN or SAN, even if a matching SNI
13565 filter is found on any crt-list. The SNI filter !* can be used after the first
13566 declared certificate to not include its CN and SAN in the SNI tree, so it will
13567 never match except if no other certificate matches. This way the first
13568 declared certificate act as a fallback.
Joao Moraise51fab02020-11-21 07:42:20 -030013569
Emmanuel Hocdet98263292016-12-29 18:26:15 +010013570 crt-list file example:
Joao Moraise51fab02020-11-21 07:42:20 -030013571 cert1.pem !*
William Lallemand7c26ed72020-06-03 17:34:48 +020013572 # comment
Emmanuel Hocdet05942112017-02-20 16:11:50 +010013573 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010013574 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010013575 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010013576
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013577defer-accept
13578 Is an optional keyword which is supported only on certain Linux kernels. It
13579 states that a connection will only be accepted once some data arrive on it,
13580 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010013581 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013582 performance by ensuring that most of the request is already available when
13583 the connection is accepted. On the other hand, it will not be able to detect
13584 connections which don't talk. It is important to note that this option is
13585 broken in all kernels up to 2.6.31, as the connection is never accepted until
13586 the client talks. This can cause issues with front firewalls which would see
13587 an established connection while the proxy will only see it in SYN_RECV. This
13588 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
13589
William Lallemandf6975e92017-05-26 17:42:10 +020013590expose-fd listeners
13591 This option is only usable with the stats socket. It gives your stats socket
13592 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020013593 During a reload with the master-worker mode, the process is automatically
13594 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010013595 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020013596
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013597force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013598 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013599 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013600 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013601 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013602
13603force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013604 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013605 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013606 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013607
13608force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013609 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013610 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013611 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013612
13613force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013614 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013615 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013616 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013617
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020013618force-tlsv13
13619 This option enforces use of TLSv1.3 only on SSL connections instantiated from
13620 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013621 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020013622
Christopher Faulet31af49d2015-06-09 17:29:50 +020013623generate-certificates
13624 This setting is only available when support for OpenSSL was built in. It
13625 enables the dynamic SSL certificates generation. A CA certificate and its
13626 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
13627 a transparent forward proxy, SSL requests generate errors because of a common
13628 name mismatch on the certificate presented to the client. With this option
13629 enabled, HAProxy will try to forge a certificate using the SNI hostname
13630 indicated by the client. This is done only if no certificate matches the SNI
13631 hostname (see 'crt-list'). If an error occurs, the default certificate is
13632 used, else the 'strict-sni' option is set.
13633 It can also be used when HAProxy is configured as a reverse proxy to ease the
13634 deployment of an architecture with many backends.
13635
13636 Creating a SSL certificate is an expensive operation, so a LRU cache is used
13637 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010013638 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020013639 certificate is used many times.
13640
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013641gid <gid>
13642 Sets the group of the UNIX sockets to the designated system gid. It can also
13643 be set by default in the global section's "unix-bind" statement. Note that
13644 some platforms simply ignore this. This setting is equivalent to the "group"
13645 setting except that the group ID is used instead of its name. This setting is
13646 ignored by non UNIX sockets.
13647
13648group <group>
13649 Sets the group of the UNIX sockets to the designated system group. It can
13650 also be set by default in the global section's "unix-bind" statement. Note
13651 that some platforms simply ignore this. This setting is equivalent to the
13652 "gid" setting except that the group name is used instead of its gid. This
13653 setting is ignored by non UNIX sockets.
13654
13655id <id>
13656 Fixes the socket ID. By default, socket IDs are automatically assigned, but
13657 sometimes it is more convenient to fix them to ease monitoring. This value
13658 must be strictly positive and unique within the listener/frontend. This
13659 option can only be used when defining only a single socket.
13660
13661interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010013662 Restricts the socket to a specific interface. When specified, only packets
13663 received from that particular interface are processed by the socket. This is
13664 currently only supported on Linux. The interface must be a primary system
13665 interface, not an aliased interface. It is also possible to bind multiple
13666 frontends to the same address if they are bound to different interfaces. Note
13667 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010013668 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
13669 uses the same interface as inbound traffic, and its associated routing table,
13670 even if there are explicit routes through different interfaces configured.
13671 This can prove useful to address asymmetric routing issues when the same
13672 client IP addresses need to be able to reach frontends hosted on different
13673 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013674
Willy Tarreauabb175f2012-09-24 12:43:26 +020013675level <level>
13676 This setting is used with the stats sockets only to restrict the nature of
13677 the commands that can be issued on the socket. It is ignored by other
13678 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013679 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020013680 read, and no change is allowed. It would make sense on systems where it
13681 is not easy to restrict access to the socket.
13682 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010013683 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020013684 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010013685 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020013686 all counters).
13687
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020013688severity-output <format>
13689 This setting is used with the stats sockets only to configure severity
13690 level output prepended to informational feedback messages. Severity
13691 level of messages can range between 0 and 7, conforming to syslog
13692 rfc5424. Valid and successful socket commands requesting data
13693 (i.e. "show map", "get acl foo" etc.) will never have a severity level
13694 prepended. It is ignored by other sockets. <format> can be one of :
13695 - "none" (default) no severity level is prepended to feedback messages.
13696 - "number" severity level is prepended as a number.
13697 - "string" severity level is prepended as a string following the
13698 rfc5424 convention.
13699
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013700maxconn <maxconn>
13701 Limits the sockets to this number of concurrent connections. Extraneous
13702 connections will remain in the system's backlog until a connection is
13703 released. If unspecified, the limit will be the same as the frontend's
13704 maxconn. Note that in case of port ranges or multiple addresses, the same
13705 value will be applied to each socket. This setting enables different
13706 limitations on expensive sockets, for instance SSL entries which may easily
13707 eat all memory.
13708
13709mode <mode>
13710 Sets the octal mode used to define access permissions on the UNIX socket. It
13711 can also be set by default in the global section's "unix-bind" statement.
13712 Note that some platforms simply ignore this. This setting is ignored by non
13713 UNIX sockets.
13714
13715mss <maxseg>
13716 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
13717 connections. This can be used to force a lower MSS for certain specific
13718 ports, for instance for connections passing through a VPN. Note that this
13719 relies on a kernel feature which is theoretically supported under Linux but
13720 was buggy in all versions prior to 2.6.28. It may or may not work on other
13721 operating systems. It may also not change the advertised value but change the
13722 effective size of outgoing segments. The commonly advertised value for TCPv4
13723 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
13724 positive, it will be used as the advertised MSS. If it is negative, it will
13725 indicate by how much to reduce the incoming connection's advertised MSS for
13726 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
13727
13728name <name>
13729 Sets an optional name for these sockets, which will be reported on the stats
13730 page.
13731
Willy Tarreaud72f0f32015-10-13 14:50:22 +020013732namespace <name>
13733 On Linux, it is possible to specify which network namespace a socket will
13734 belong to. This directive makes it possible to explicitly bind a listener to
13735 a namespace different from the default one. Please refer to your operating
13736 system's documentation to find more details about network namespaces.
13737
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013738nice <nice>
13739 Sets the 'niceness' of connections initiated from the socket. Value must be
13740 in the range -1024..1024 inclusive, and defaults to zero. Positive values
13741 means that such connections are more friendly to others and easily offer
13742 their place in the scheduler. On the opposite, negative values mean that
13743 connections want to run with a higher priority than others. The difference
13744 only happens under high loads when the system is close to saturation.
13745 Negative values are appropriate for low-latency or administration services,
13746 and high values are generally recommended for CPU intensive tasks such as SSL
13747 processing or bulk transfers which are less sensible to latency. For example,
13748 it may make sense to use a positive value for an SMTP socket and a negative
13749 one for an RDP socket.
13750
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020013751no-ca-names
13752 This setting is only available when support for OpenSSL was built in. It
13753 prevents from send CA names in server hello message when ca-file is used.
Emmanuel Hocdet842e94e2019-12-16 16:39:17 +010013754 Use "ca-verify-file" instead of "ca-file" with "no-ca-names".
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020013755
Emeric Brun9b3009b2012-10-05 11:55:06 +020013756no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013757 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013758 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013759 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013760 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013761 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
13762 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013763
Emeric Brun90ad8722012-10-02 14:00:59 +020013764no-tls-tickets
13765 This setting is only available when support for OpenSSL was built in. It
13766 disables the stateless session resumption (RFC 5077 TLS Ticket
13767 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013768 session resumption is more expensive in CPU usage. This option is also
13769 available on global statement "ssl-default-bind-options".
Lukas Tribusbdb386d2020-03-10 00:56:09 +010013770 The TLS ticket mechanism is only used up to TLS 1.2.
13771 Forward Secrecy is compromised with TLS tickets, unless ticket keys
13772 are periodically rotated (via reload or by using "tls-ticket-keys").
Emeric Brun90ad8722012-10-02 14:00:59 +020013773
Emeric Brun9b3009b2012-10-05 11:55:06 +020013774no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013775 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013776 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013777 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013778 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013779 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
13780 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013781
Emeric Brun9b3009b2012-10-05 11:55:06 +020013782no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020013783 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013784 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013785 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013786 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013787 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
13788 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020013789
Emeric Brun9b3009b2012-10-05 11:55:06 +020013790no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020013791 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013792 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020013793 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010013794 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013795 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
13796 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020013797
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020013798no-tlsv13
13799 This setting is only available when support for OpenSSL was built in. It
13800 disables support for TLSv1.3 on any sockets instantiated from the listener
13801 when SSL is supported. Note that SSLv2 is forced disabled in the code and
13802 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013803 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
13804 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020013805
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020013806npn <protocols>
13807 This enables the NPN TLS extension and advertises the specified protocol list
13808 as supported on top of NPN. The protocol list consists in a comma-delimited
13809 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050013810 This requires that the SSL library is built with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020013811 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010013812 replaced with the ALPN extension (see the "alpn" keyword), though this one is
13813 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
13814 version of OpenSSL, NPN might still be used as most clients still support it
13815 at the time of writing this. It is possible to enable both NPN and ALPN
13816 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020013817
Lukas Tribus53ae85c2017-05-04 15:45:40 +000013818prefer-client-ciphers
13819 Use the client's preference when selecting the cipher suite, by default
13820 the server's preference is enforced. This option is also available on
13821 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020013822 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
13823 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
13824 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000013825
Christopher Fauletc644fa92017-11-23 22:44:11 +010013826process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010013827 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010013828 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010013829 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010013830 between the two is applied. If in the end the listener is not allowed to run
13831 on any remaining process, a warning is emitted, and the listener will either
13832 run on the first process of the listener if a single process was specified,
13833 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010013834 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010013835 connections for this listener, for the the process set. If multiple processes
13836 and threads are configured, a warning is emitted, as it either results from a
13837 configuration error or a misunderstanding of these models. For the unlikely
13838 case where several ranges are needed, this directive may be repeated.
13839 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010013840
13841 all | odd | even | number[-[number]]
13842
13843 Ranges can be partially defined. The higher bound can be omitted. In such
13844 case, it is replaced by the corresponding maximum value. The main purpose of
13845 this directive is to be used with the stats sockets and have one different
13846 socket per process. The second purpose is to have multiple bind lines sharing
13847 the same IP:port but not the same process in a listener, so that the system
13848 can distribute the incoming connections into multiple queues and allow a
13849 smoother inter-process load balancing. Currently Linux 3.9 and above is known
13850 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020013851
Christopher Fauleta717b992018-04-10 14:43:00 +020013852proto <name>
13853 Forces the multiplexer's protocol to use for the incoming connections. It
13854 must be compatible with the mode of the frontend (TCP or HTTP). It must also
13855 be usable on the frontend side. The list of available protocols is reported
Christopher Faulet982e17d2021-03-26 14:44:18 +010013856 in haproxy -vv. The protocols properties are reported : the mode (TCP/HTTP),
13857 the side (FE/BE), the mux name and its flags.
13858
13859 Some protocols report errors on aborts (flag=CLEAN_ABRT). Some others are
13860 subject to the head-of-line blocking on server side (flag=HOL_RISK). Finally
13861 some protocols don't support upgrades (flag=NO_UPG). The HTX compatibility is
13862 also reported (flag=HTX).
13863
13864 Here are the protocols that may be used as argument to a "proto" directive on
13865 a bind line :
13866
13867 h2 : mode=HTTP side=FE|BE mux=H2 flags=HTX|CLEAN_ABRT|HOL_RISK|NO_UPG
13868 h1 : mode=HTTP side=FE|BE mux=H1 flags=HTX|NO_UPG
13869 none : mode=TCP side=FE|BE mux=PASS flags=NO_UPG
13870
Daniel Corbett67a82712020-07-06 23:01:19 -040013871 Idea behind this option is to bypass the selection of the best multiplexer's
Christopher Fauleta717b992018-04-10 14:43:00 +020013872 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080013873 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020013874 h2" on the bind line.
13875
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013876ssl
13877 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013878 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013879 certificate is necessary (see "crt" above). All contents in the buffers will
13880 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020013881 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
13882 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013883
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013884ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
13885 This option enforces use of <version> or lower on SSL connections instantiated
William Lallemand50df1cb2020-06-02 10:52:24 +020013886 from this listener. Using this setting without "ssl-min-ver" can be
13887 ambiguous because the default ssl-min-ver value could change in future HAProxy
13888 versions. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013889 "ssl-default-bind-options". See also "ssl-min-ver".
13890
13891ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
William Lallemand50df1cb2020-06-02 10:52:24 +020013892 This option enforces use of <version> or upper on SSL connections
13893 instantiated from this listener. The default value is "TLSv1.2". This option
13894 is also available on global statement "ssl-default-bind-options".
13895 See also "ssl-max-ver".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020013896
Emmanuel Hocdet65623372013-01-24 17:17:15 +010013897strict-sni
13898 This setting is only available when support for OpenSSL was built in. The
13899 SSL/TLS negotiation is allow only if the client provided an SNI which match
13900 a certificate. The default certificate is not used.
13901 See the "crt" option for more information.
13902
Willy Tarreau2af207a2015-02-04 00:45:58 +010013903tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010013904 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010013905 listening socket. This option is available on Linux since version 2.6.37. It
Daniel Corbett9f0843f2021-05-08 10:50:37 -040013906 allows HAProxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010013907 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010013908 useful on long-lived connections experiencing long idle periods such as
13909 remote terminals or database connection pools, where the client and server
13910 timeouts must remain high to allow a long period of idle, but where it is
13911 important to detect that the client has disappeared in order to release all
13912 resources associated with its connection (and the server's session). The
13913 argument is a delay expressed in milliseconds by default. This only works
13914 for regular TCP connections, and is ignored for other protocols.
13915
Willy Tarreau1c862c52012-10-05 16:21:00 +020013916tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010013917 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020013918 enables TCP Fast Open on the listening socket, which means that clients which
13919 support this feature will be able to send a request and receive a response
13920 during the 3-way handshake starting from second connection, thus saving one
13921 round-trip after the first connection. This only makes sense with protocols
13922 that use high connection rates and where each round trip matters. This can
13923 possibly cause issues with many firewalls which do not accept data on SYN
13924 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020013925 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
13926 need to build HAProxy with USE_TFO=1 if your libc doesn't define
13927 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020013928
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010013929tls-ticket-keys <keyfile>
13930 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010013931 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
13932 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
13933 The first key determines the key length used for next keys: you can't mix
13934 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
13935 build option (default 3) and at least as many keys need to be present in
13936 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
13937 penultimate one for encryption. This enables easy key rotation by just
13938 appending new key to the file and reloading the process. Keys must be
13939 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
13940 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010013941 storage such as hard drives (hint: use tmpfs and don't swap those files).
13942 Lifetime hint can be changed using tune.ssl.timeout.
13943
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013944transparent
13945 Is an optional keyword which is supported only on certain Linux kernels. It
13946 indicates that the addresses will be bound even if they do not belong to the
13947 local machine, and that packets targeting any of these addresses will be
13948 intercepted just as if the addresses were locally configured. This normally
13949 requires that IP forwarding is enabled. Caution! do not use this with the
13950 default address '*', as it would redirect any traffic for the specified port.
13951 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
13952 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
13953 kernel version. Some distribution kernels include backports of the feature,
13954 so check for support with your vendor.
13955
Willy Tarreau77e3af92012-11-24 15:07:23 +010013956v4v6
13957 Is an optional keyword which is supported only on most recent systems
13958 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
13959 and IPv6 when it uses the default address. Doing so is sometimes necessary
13960 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013961 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010013962
Willy Tarreau9b6700f2012-11-24 11:55:28 +010013963v6only
13964 Is an optional keyword which is supported only on most recent systems
13965 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
13966 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010013967 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
13968 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010013969
Willy Tarreaub6205fd2012-09-24 12:27:33 +020013970uid <uid>
13971 Sets the owner of the UNIX sockets to the designated system uid. It can also
13972 be set by default in the global section's "unix-bind" statement. Note that
13973 some platforms simply ignore this. This setting is equivalent to the "user"
13974 setting except that the user numeric ID is used instead of its name. This
13975 setting is ignored by non UNIX sockets.
13976
13977user <user>
13978 Sets the owner of the UNIX sockets to the designated system user. It can also
13979 be set by default in the global section's "unix-bind" statement. Note that
13980 some platforms simply ignore this. This setting is equivalent to the "uid"
13981 setting except that the user name is used instead of its uid. This setting is
13982 ignored by non UNIX sockets.
13983
Emeric Brun1a073b42012-09-28 17:07:34 +020013984verify [none|optional|required]
13985 This setting is only available when support for OpenSSL was built in. If set
13986 to 'none', client certificate is not requested. This is the default. In other
13987 cases, a client certificate is requested. If the client does not provide a
13988 certificate after the request and if 'verify' is set to 'required', then the
13989 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020013990 certificate provided by the client is always verified using CAs from
13991 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
13992 is aborted, regardless of the 'verify' option, unless the error code exactly
13993 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020013994
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200139955.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010013996------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013997
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010013998The "server" and "default-server" keywords support a certain number of settings
13999which are all passed as arguments on the server line. The order in which those
14000arguments appear does not count, and they are all optional. Some of those
14001settings are single words (booleans) while others expect one or several values
14002after them. In this case, the values must immediately follow the setting name.
14003Except default-server, all those settings must be specified after the server's
14004address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020014005
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014006 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010014007 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020014008
Frédéric Lécailled2376272017-03-21 18:52:12 +010014009Note that all these settings are supported both by "server" and "default-server"
14010keywords, except "id" which is only supported by "server".
14011
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014012The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010014013
Willy Tarreauceb4ac92012-04-28 00:41:46 +020014014addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014015 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010014016 to send health-checks or to probe the agent-check. On some servers, it may be
14017 desirable to dedicate an IP address to specific component able to perform
14018 complex tests which are more suitable to health-checks than the application.
14019 This parameter is ignored if the "check" parameter is not set. See also the
14020 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014021
Simon Hormand60d6912013-11-25 10:46:36 +090014022agent-check
14023 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010014024 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010014025 to the port set by the "agent-port" parameter and reading an ASCII string
14026 terminated by the first '\r' or '\n' met. The string is made of a series of
14027 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090014028
Willy Tarreau81f5d942013-12-09 20:51:51 +010014029 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090014030 Values in this format will set the weight proportional to the initial
Daniel Corbett9f0843f2021-05-08 10:50:37 -040014031 weight of a server as configured when HAProxy starts. Note that a zero
Willy Tarreauc5af3a62014-10-07 15:27:33 +020014032 weight is reported on the stats page as "DRAIN" since it has the same
14033 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090014034
Davor Ocelice9ed2812017-12-25 17:49:28 +010014035 - The string "maxconn:" followed by an integer (no space between). Values
14036 in this format will set the maxconn of a server. The maximum number of
14037 connections advertised needs to be multiplied by the number of load
14038 balancers and different backends that use this health check to get the
14039 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020014040
Willy Tarreau81f5d942013-12-09 20:51:51 +010014041 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010014042 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090014043
Willy Tarreau81f5d942013-12-09 20:51:51 +010014044 - The word "drain". This will turn the server's administrative state to the
14045 DRAIN mode, thus it will not accept any new connections other than those
14046 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090014047
Willy Tarreau81f5d942013-12-09 20:51:51 +010014048 - The word "maint". This will turn the server's administrative state to the
14049 MAINT mode, thus it will not accept any new connections at all, and health
14050 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090014051
William Dauchyf8e795c2020-09-26 13:35:51 +020014052 - The words "down", "fail", or "stopped", optionally followed by a
Willy Tarreau81f5d942013-12-09 20:51:51 +010014053 description string after a sharp ('#'). All of these mark the server's
14054 operating state as DOWN, but since the word itself is reported on the stats
14055 page, the difference allows an administrator to know if the situation was
14056 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010014057 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010014058 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090014059
Willy Tarreau81f5d942013-12-09 20:51:51 +010014060 - The word "up" sets back the server's operating state as UP if health checks
14061 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090014062
Willy Tarreau81f5d942013-12-09 20:51:51 +010014063 Parameters which are not advertised by the agent are not changed. For
14064 example, an agent might be designed to monitor CPU usage and only report a
14065 relative weight and never interact with the operating status. Similarly, an
14066 agent could be designed as an end-user interface with 3 radio buttons
14067 allowing an administrator to change only the administrative state. However,
14068 it is important to consider that only the agent may revert its own actions,
14069 so if a server is set to DRAIN mode or to DOWN state using the agent, the
14070 agent must implement the other equivalent actions to bring the service into
14071 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090014072
Simon Horman2f1f9552013-11-25 10:46:37 +090014073 Failure to connect to the agent is not considered an error as connectivity
14074 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010014075 parameter. Warning though, it is not a good idea to stop an agent after it
14076 reports "down", since only an agent reporting "up" will be able to turn the
14077 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010014078 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090014079
Willy Tarreau81f5d942013-12-09 20:51:51 +010014080 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010014081 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090014082
James Brown55f9ff12015-10-21 18:19:05 -070014083agent-send <string>
Daniel Corbett9f0843f2021-05-08 10:50:37 -040014084 If this option is specified, HAProxy will send the given string (verbatim)
James Brown55f9ff12015-10-21 18:19:05 -070014085 to the agent server upon connection. You could, for example, encode
14086 the backend name into this string, which would enable your agent to send
14087 different responses based on the backend. Make sure to include a '\n' if
14088 you want to terminate your request with a newline.
14089
Simon Hormand60d6912013-11-25 10:46:36 +090014090agent-inter <delay>
14091 The "agent-inter" parameter sets the interval between two agent checks
14092 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
14093
14094 Just as with every other time-based parameter, it may be entered in any
14095 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
14096 parameter also serves as a timeout for agent checks "timeout check" is
14097 not set. In order to reduce "resonance" effects when multiple servers are
14098 hosted on the same hardware, the agent and health checks of all servers
14099 are started with a small time offset between them. It is also possible to
14100 add some random noise in the agent and health checks interval using the
14101 global "spread-checks" keyword. This makes sense for instance when a lot
14102 of backends use the same servers.
14103
14104 See also the "agent-check" and "agent-port" parameters.
14105
Misiek768d8602017-01-09 09:52:43 +010014106agent-addr <addr>
14107 The "agent-addr" parameter sets address for agent check.
14108
14109 You can offload agent-check to another target, so you can make single place
Daniel Corbett9f0843f2021-05-08 10:50:37 -040014110 managing status and weights of servers defined in HAProxy in case you can't
Misiek768d8602017-01-09 09:52:43 +010014111 make self-aware and self-managing services. You can specify both IP or
14112 hostname, it will be resolved.
14113
Simon Hormand60d6912013-11-25 10:46:36 +090014114agent-port <port>
14115 The "agent-port" parameter sets the TCP port used for agent checks.
14116
14117 See also the "agent-check" and "agent-inter" parameters.
14118
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020014119allow-0rtt
14120 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020014121 Note that early data will be sent only if the client used early data, or
14122 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020014123
Olivier Houchardc7566002018-11-20 23:33:50 +010014124alpn <protocols>
14125 This enables the TLS ALPN extension and advertises the specified protocol
14126 list as supported on top of ALPN. The protocol list consists in a comma-
14127 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050014128 quotes). This requires that the SSL library is built with support for TLS
Olivier Houchardc7566002018-11-20 23:33:50 +010014129 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
14130 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
14131 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
14132 now obsolete NPN extension.
14133 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
14134 be advertised, in order of preference, like below :
14135
14136 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
14137
Amaury Denoyelle69352ec2021-10-18 14:40:29 +020014138 See also "ws" to use an alternative ALPN for websocket streams.
14139
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014140backup
14141 When "backup" is present on a server line, the server is only used in load
14142 balancing when all other non-backup servers are unavailable. Requests coming
14143 with a persistence cookie referencing the server will always be served
14144 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010014145 the "allbackups" option is set in the backend. See also the "no-backup" and
14146 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014147
Emeric Brunef42d922012-10-11 16:11:36 +020014148ca-file <cafile>
14149 This setting is only available when support for OpenSSL was built in. It
14150 designates a PEM file from which to load CA certificates used to verify
14151 server's certificate.
14152
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014153check
Jerome Magnin90702bc2020-04-26 14:23:04 +020014154 This option enables health checks on a server:
14155 - when not set, no health checking is performed, and the server is always
14156 considered available.
14157 - when set and no other check method is configured, the server is considered
14158 available when a connection can be established at the highest configured
14159 transport layer. This means TCP by default, or SSL/TLS when "ssl" or
14160 "check-ssl" are set, both possibly combined with connection prefixes such
14161 as a PROXY protocol header when "send-proxy" or "check-send-proxy" are
14162 set.
14163 - when set and an application-level health check is defined, the
14164 application-level exchanges are performed on top of the configured
14165 transport layer and the server is considered available if all of the
14166 exchanges succeed.
14167
14168 By default, health checks are performed on the same address and port as
14169 configured on the server, using the same encapsulation parameters (SSL/TLS,
14170 proxy-protocol header, etc... ). It is possible to change the destination
14171 address using "addr" and the port using "port". When done, it is assumed the
14172 server isn't checked on the service port, and configured encapsulation
Ilya Shipitsin4329a9a2020-05-05 21:17:10 +050014173 parameters are not reused. One must explicitly set "check-send-proxy" to send
Jerome Magnin90702bc2020-04-26 14:23:04 +020014174 connection headers, "check-ssl" to use SSL/TLS.
14175
14176 When "sni" or "alpn" are set on the server line, their value is not used for
14177 health checks and one must use "check-sni" or "check-alpn".
14178
14179 The default source address for health check traffic is the same as the one
14180 defined in the backend. It can be changed with the "source" keyword.
14181
14182 The interval between checks can be set using the "inter" keyword, and the
14183 "rise" and "fall" keywords can be used to define how many successful or
14184 failed health checks are required to flag a server available or not
14185 available.
14186
14187 Optional application-level health checks can be configured with "option
14188 httpchk", "option mysql-check" "option smtpchk", "option pgsql-check",
14189 "option ldap-check", or "option redis-check".
14190
14191 Example:
14192 # simple tcp check
14193 backend foo
14194 server s1 192.168.0.1:80 check
14195 # this does a tcp connect + tls handshake
14196 backend foo
14197 server s1 192.168.0.1:443 ssl check
14198 # simple tcp check is enough for check success
14199 backend foo
14200 option tcp-check
14201 tcp-check connect
14202 server s1 192.168.0.1:443 ssl check
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014203
Willy Tarreau6c16adc2012-10-05 00:04:16 +020014204check-send-proxy
14205 This option forces emission of a PROXY protocol line with outgoing health
14206 checks, regardless of whether the server uses send-proxy or not for the
14207 normal traffic. By default, the PROXY protocol is enabled for health checks
14208 if it is already enabled for normal traffic and if no "port" nor "addr"
14209 directive is present. However, if such a directive is present, the
14210 "check-send-proxy" option needs to be used to force the use of the
14211 protocol. See also the "send-proxy" option for more information.
14212
Olivier Houchard92150142018-12-21 19:47:01 +010014213check-alpn <protocols>
14214 Defines which protocols to advertise with ALPN. The protocol list consists in
14215 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
14216 (without quotes). If it is not set, the server ALPN is used.
14217
Christopher Fauletedc6ed92020-04-23 16:27:59 +020014218check-proto <name>
14219 Forces the multiplexer's protocol to use for the server's health-check
14220 connections. It must be compatible with the health-check type (TCP or
14221 HTTP). It must also be usable on the backend side. The list of available
Christopher Faulet982e17d2021-03-26 14:44:18 +010014222 protocols is reported in haproxy -vv. The protocols properties are
14223 reported : the mode (TCP/HTTP), the side (FE/BE), the mux name and its flags.
14224
14225 Some protocols report errors on aborts (flag=CLEAN_ABRT). Some others are
14226 subject to the head-of-line blocking on server side (flag=HOL_RISK). Finally
14227 some protocols don't support upgrades (flag=NO_UPG). The HTX compatibility is
14228 also reported (flag=HTX).
14229
14230 Here are the protocols that may be used as argument to a "check-proto"
14231 directive on a server line:
14232
14233 h2 : mode=HTTP side=FE|BE mux=H2 flags=HTX|CLEAN_ABRT|HOL_RISK|NO_UPG
14234 fcgi : mode=HTTP side=BE mux=FCGI flags=HTX|HOL_RISK|NO_UPG
14235 h1 : mode=HTTP side=FE|BE mux=H1 flags=HTX|NO_UPG
14236 none : mode=TCP side=FE|BE mux=PASS flags=NO_UPG
14237
Daniel Corbett67a82712020-07-06 23:01:19 -040014238 Idea behind this option is to bypass the selection of the best multiplexer's
Christopher Fauletedc6ed92020-04-23 16:27:59 +020014239 protocol for health-check connections established to this server.
14240 If not defined, the server one will be used, if set.
14241
Jérôme Magninae9bb762018-12-09 16:08:26 +010014242check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020014243 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010014244 over SSL. It is only possible to use a string to set <sni>. If you want to
14245 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020014246
Willy Tarreau763a95b2012-10-04 23:15:39 +020014247check-ssl
14248 This option forces encryption of all health checks over SSL, regardless of
14249 whether the server uses SSL or not for the normal traffic. This is generally
14250 used when an explicit "port" or "addr" directive is specified and SSL health
14251 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014252 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020014253 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
14254 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010014255 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010014256 See the "ssl" option for more information and "no-check-ssl" to disable
14257 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020014258
Alexander Liu2a54bb72019-05-22 19:44:48 +080014259check-via-socks4
John Roeslerfb2fce12019-07-10 15:45:51 -050014260 This option enables outgoing health checks using upstream socks4 proxy. By
Alexander Liu2a54bb72019-05-22 19:44:48 +080014261 default, the health checks won't go through socks tunnel even it was enabled
14262 for normal traffic.
14263
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014264ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020014265 This setting is only available when support for OpenSSL was built in. This
14266 option sets the string describing the list of cipher algorithms that is
14267 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000014268 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
14269 information and recommendations see e.g.
14270 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
14271 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
14272 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014273
Dirkjan Bussink415150f2018-09-14 11:14:21 +020014274ciphersuites <ciphersuites>
14275 This setting is only available when support for OpenSSL was built in and
14276 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
14277 describing the list of cipher algorithms that is negotiated during the TLS
14278 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000014279 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
14280 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
14281 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020014282
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014283cookie <value>
14284 The "cookie" parameter sets the cookie value assigned to the server to
14285 <value>. This value will be checked in incoming requests, and the first
14286 operational server possessing the same value will be selected. In return, in
14287 cookie insertion or rewrite modes, this value will be assigned to the cookie
14288 sent to the client. There is nothing wrong in having several servers sharing
14289 the same cookie value, and it is in fact somewhat common between normal and
14290 backup servers. See also the "cookie" keyword in backend section.
14291
Emeric Brunef42d922012-10-11 16:11:36 +020014292crl-file <crlfile>
14293 This setting is only available when support for OpenSSL was built in. It
14294 designates a PEM file from which to load certificate revocation list used
14295 to verify server's certificate.
14296
Emeric Bruna7aa3092012-10-26 12:58:00 +020014297crt <cert>
14298 This setting is only available when support for OpenSSL was built in.
14299 It designates a PEM file from which to load both a certificate and the
14300 associated private key. This file can be built by concatenating both PEM
14301 files into one. This certificate will be sent if the server send a client
14302 certificate request.
14303
Remi Tricot-Le Breton7c980df2021-05-07 15:28:08 +020014304 If the file does not contain a private key, HAProxy will try to load the key
14305 at the same path suffixed by a ".key" (provided the "ssl-load-extra-files"
14306 option is set accordingly).
14307
Willy Tarreau96839092010-03-29 10:02:24 +020014308disabled
14309 The "disabled" keyword starts the server in the "disabled" state. That means
14310 that it is marked down in maintenance mode, and no connection other than the
14311 ones allowed by persist mode will reach it. It is very well suited to setup
14312 new servers, because normal traffic will never reach them, while it is still
14313 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010014314 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020014315
Frédéric Lécailled2376272017-03-21 18:52:12 +010014316enabled
14317 This option may be used as 'server' setting to reset any 'disabled'
14318 setting which would have been inherited from 'default-server' directive as
14319 default value.
14320 It may also be used as 'default-server' setting to reset any previous
14321 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020014322
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014323error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010014324 If health observing is enabled, the "error-limit" parameter specifies the
14325 number of consecutive errors that triggers event selected by the "on-error"
14326 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014327
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014328 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014329
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014330fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014331 The "fall" parameter states that a server will be considered as dead after
14332 <count> consecutive unsuccessful health checks. This value defaults to 3 if
14333 unspecified. See also the "check", "inter" and "rise" parameters.
14334
Emeric Brun8694b9a2012-10-05 14:39:07 +020014335force-sslv3
14336 This option enforces use of SSLv3 only when SSL is used to communicate with
14337 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014338 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014339 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020014340
14341force-tlsv10
14342 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014343 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014344 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020014345
14346force-tlsv11
14347 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014348 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014349 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020014350
14351force-tlsv12
14352 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014353 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014354 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020014355
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020014356force-tlsv13
14357 This option enforces use of TLSv1.3 only when SSL is used to communicate with
14358 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014359 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020014360
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014361id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020014362 Set a persistent ID for the server. This ID must be positive and unique for
14363 the proxy. An unused ID will automatically be assigned if unset. The first
14364 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014365
Willy Tarreau6a031d12016-11-07 19:42:35 +010014366init-addr {last | libc | none | <ip>},[...]*
14367 Indicate in what order the server's address should be resolved upon startup
14368 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010014369 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010014370 method which succeeds is used. If the end of the list is reached without
14371 finding a working method, an error is thrown. Method "last" suggests to pick
14372 the address which appears in the state file (see "server-state-file"). Method
14373 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
14374 depending on the operating system and build options). Method "none"
14375 specifically indicates that the server should start without any valid IP
14376 address in a down state. It can be useful to ignore some DNS issues upon
14377 startup, waiting for the situation to get fixed later. Finally, an IP address
14378 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010014379 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010014380 server used to catch old sessions and present them with a decent error
14381 message for example. When the "first" load balancing algorithm is used, this
14382 IP address could point to a fake server used to trigger the creation of new
14383 instances on the fly. This option defaults to "last,libc" indicating that the
14384 previous address found in the state file (if any) is used first, otherwise
14385 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010014386 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010014387
14388 Example:
14389 defaults
14390 # never fail on address resolution
14391 default-server init-addr last,libc,none
14392
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014393inter <delay>
14394fastinter <delay>
14395downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014396 The "inter" parameter sets the interval between two consecutive health checks
14397 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
14398 It is also possible to use "fastinter" and "downinter" to optimize delays
14399 between checks depending on the server state :
14400
Pieter Baauw44fc9df2015-09-17 21:30:46 +020014401 Server state | Interval used
14402 ----------------------------------------+----------------------------------
14403 UP 100% (non-transitional) | "inter"
14404 ----------------------------------------+----------------------------------
14405 Transitionally UP (going down "fall"), | "fastinter" if set,
14406 Transitionally DOWN (going up "rise"), | "inter" otherwise.
14407 or yet unchecked. |
14408 ----------------------------------------+----------------------------------
14409 DOWN 100% (non-transitional) | "downinter" if set,
14410 | "inter" otherwise.
14411 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010014412
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014413 Just as with every other time-based parameter, they can be entered in any
14414 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
14415 serves as a timeout for health checks sent to servers if "timeout check" is
14416 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090014417 hosted on the same hardware, the agent and health checks of all servers
14418 are started with a small time offset between them. It is also possible to
14419 add some random noise in the agent and health checks interval using the
14420 global "spread-checks" keyword. This makes sense for instance when a lot
14421 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014422
Emeric Brun97556472020-05-30 01:42:45 +020014423log-proto <logproto>
14424 The "log-proto" specifies the protocol used to forward event messages to
14425 a server configured in a ring section. Possible values are "legacy"
14426 and "octet-count" corresponding respectively to "Non-transparent-framing"
14427 and "Octet counting" in rfc6587. "legacy" is the default.
14428
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014429maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014430 The "maxconn" parameter specifies the maximal number of concurrent
14431 connections that will be sent to this server. If the number of incoming
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010014432 concurrent connections goes higher than this value, they will be queued,
14433 waiting for a slot to be released. This parameter is very important as it can
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014434 save fragile servers from going down under extreme loads. If a "minconn"
14435 parameter is specified, the limit becomes dynamic. The default value is "0"
14436 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
14437 the backend's "fullconn" keyword.
14438
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010014439 In HTTP mode this parameter limits the number of concurrent requests instead
14440 of the number of connections. Multiple requests might be multiplexed over a
14441 single TCP connection to the server. As an example if you specify a maxconn
14442 of 50 you might see between 1 and 50 actual server connections, but no more
14443 than 50 concurrent requests.
14444
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014445maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014446 The "maxqueue" parameter specifies the maximal number of connections which
14447 will wait in the queue for this server. If this limit is reached, next
14448 requests will be redispatched to other servers instead of indefinitely
14449 waiting to be served. This will break persistence but may allow people to
Willy Tarreau8ae8c482020-10-22 17:19:07 +020014450 quickly re-log in when the server they try to connect to is dying. Some load
14451 balancing algorithms such as leastconn take this into account and accept to
14452 add requests into a server's queue up to this value if it is explicitly set
14453 to a value greater than zero, which often allows to better smooth the load
14454 when dealing with single-digit maxconn values. The default value is "0" which
14455 means the queue is unlimited. See also the "maxconn" and "minconn" parameters
14456 and "balance leastconn".
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014457
Willy Tarreau9c538e02019-01-23 10:21:49 +010014458max-reuse <count>
14459 The "max-reuse" argument indicates the HTTP connection processors that they
14460 should not reuse a server connection more than this number of times to send
14461 new requests. Permitted values are -1 (the default), which disables this
14462 limit, or any positive value. Value zero will effectively disable keep-alive.
14463 This is only used to work around certain server bugs which cause them to leak
14464 resources over time. The argument is not necessarily respected by the lower
14465 layers as there might be technical limitations making it impossible to
14466 enforce. At least HTTP/2 connections to servers will respect it.
14467
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014468minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014469 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
14470 limit following the backend's load. The server will always accept at least
14471 <minconn> connections, never more than <maxconn>, and the limit will be on
14472 the ramp between both values when the backend has less than <fullconn>
14473 concurrent connections. This makes it possible to limit the load on the
14474 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014475 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014476 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014477
Willy Tarreaud72f0f32015-10-13 14:50:22 +020014478namespace <name>
14479 On Linux, it is possible to specify which network namespace a socket will
14480 belong to. This directive makes it possible to explicitly bind a server to
14481 a namespace different from the default one. Please refer to your operating
14482 system's documentation to find more details about network namespaces.
14483
Frédéric Lécailled2376272017-03-21 18:52:12 +010014484no-agent-check
14485 This option may be used as "server" setting to reset any "agent-check"
14486 setting which would have been inherited from "default-server" directive as
14487 default value.
14488 It may also be used as "default-server" setting to reset any previous
14489 "default-server" "agent-check" setting.
14490
14491no-backup
14492 This option may be used as "server" setting to reset any "backup"
14493 setting which would have been inherited from "default-server" directive as
14494 default value.
14495 It may also be used as "default-server" setting to reset any previous
14496 "default-server" "backup" setting.
14497
14498no-check
14499 This option may be used as "server" setting to reset any "check"
14500 setting which would have been inherited from "default-server" directive as
14501 default value.
14502 It may also be used as "default-server" setting to reset any previous
14503 "default-server" "check" setting.
14504
14505no-check-ssl
14506 This option may be used as "server" setting to reset any "check-ssl"
14507 setting which would have been inherited from "default-server" directive as
14508 default value.
14509 It may also be used as "default-server" setting to reset any previous
14510 "default-server" "check-ssl" setting.
14511
Frédéric Lécailled2376272017-03-21 18:52:12 +010014512no-send-proxy
14513 This option may be used as "server" setting to reset any "send-proxy"
14514 setting which would have been inherited from "default-server" directive as
14515 default value.
14516 It may also be used as "default-server" setting to reset any previous
14517 "default-server" "send-proxy" setting.
14518
14519no-send-proxy-v2
14520 This option may be used as "server" setting to reset any "send-proxy-v2"
14521 setting which would have been inherited from "default-server" directive as
14522 default value.
14523 It may also be used as "default-server" setting to reset any previous
14524 "default-server" "send-proxy-v2" setting.
14525
14526no-send-proxy-v2-ssl
14527 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
14528 setting which would have been inherited from "default-server" directive as
14529 default value.
14530 It may also be used as "default-server" setting to reset any previous
14531 "default-server" "send-proxy-v2-ssl" setting.
14532
14533no-send-proxy-v2-ssl-cn
14534 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
14535 setting which would have been inherited from "default-server" directive as
14536 default value.
14537 It may also be used as "default-server" setting to reset any previous
14538 "default-server" "send-proxy-v2-ssl-cn" setting.
14539
14540no-ssl
14541 This option may be used as "server" setting to reset any "ssl"
14542 setting which would have been inherited from "default-server" directive as
14543 default value.
14544 It may also be used as "default-server" setting to reset any previous
14545 "default-server" "ssl" setting.
14546
William Dauchyf6370442020-11-14 19:25:33 +010014547 Note that using `default-server ssl` setting and `no-ssl` on server will
14548 however init SSL connection, so it can be later be enabled through the
14549 runtime API: see `set server` commands in management doc.
14550
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010014551no-ssl-reuse
14552 This option disables SSL session reuse when SSL is used to communicate with
14553 the server. It will force the server to perform a full handshake for every
14554 new connection. It's probably only useful for benchmarking, troubleshooting,
14555 and for paranoid users.
14556
Emeric Brun9b3009b2012-10-05 11:55:06 +020014557no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014558 This option disables support for SSLv3 when SSL is used to communicate with
14559 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014560 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014561
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020014562 Supported in default-server: No
14563
Emeric Brunf9c5c472012-10-11 15:28:34 +020014564no-tls-tickets
14565 This setting is only available when support for OpenSSL was built in. It
14566 disables the stateless session resumption (RFC 5077 TLS Ticket
14567 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014568 session resumption is more expensive in CPU usage for servers. This option
14569 is also available on global statement "ssl-default-server-options".
Lukas Tribusbdb386d2020-03-10 00:56:09 +010014570 The TLS ticket mechanism is only used up to TLS 1.2.
14571 Forward Secrecy is compromised with TLS tickets, unless ticket keys
14572 are periodically rotated (via reload or by using "tls-ticket-keys").
Frédéric Lécailled2376272017-03-21 18:52:12 +010014573 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020014574
Emeric Brun9b3009b2012-10-05 11:55:06 +020014575no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020014576 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020014577 the server. Note that SSLv2 is disabled in the code and cannot be enabled
14578 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014579 often makes sense to disable it when communicating with local servers. This
14580 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014581 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020014582
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020014583 Supported in default-server: No
14584
Emeric Brun9b3009b2012-10-05 11:55:06 +020014585no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020014586 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020014587 the server. Note that SSLv2 is disabled in the code and cannot be enabled
14588 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014589 often makes sense to disable it when communicating with local servers. This
14590 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014591 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020014592
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020014593 Supported in default-server: No
14594
Emeric Brun9b3009b2012-10-05 11:55:06 +020014595no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020014596 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014597 the server. Note that SSLv2 is disabled in the code and cannot be enabled
14598 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010014599 often makes sense to disable it when communicating with local servers. This
14600 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014601 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020014602
14603 Supported in default-server: No
14604
14605no-tlsv13
14606 This option disables support for TLSv1.3 when SSL is used to communicate with
14607 the server. Note that SSLv2 is disabled in the code and cannot be enabled
14608 using any configuration option. TLSv1 is more expensive than SSLv3 so it
14609 often makes sense to disable it when communicating with local servers. This
14610 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014611 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014612
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020014613 Supported in default-server: No
14614
Frédéric Lécailled2376272017-03-21 18:52:12 +010014615no-verifyhost
14616 This option may be used as "server" setting to reset any "verifyhost"
14617 setting which would have been inherited from "default-server" directive as
14618 default value.
14619 It may also be used as "default-server" setting to reset any previous
14620 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020014621
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020014622no-tfo
14623 This option may be used as "server" setting to reset any "tfo"
14624 setting which would have been inherited from "default-server" directive as
14625 default value.
14626 It may also be used as "default-server" setting to reset any previous
14627 "default-server" "tfo" setting.
14628
Simon Hormanfa461682011-06-25 09:39:49 +090014629non-stick
14630 Never add connections allocated to this sever to a stick-table.
14631 This may be used in conjunction with backup to ensure that
14632 stick-table persistence is disabled for backup servers.
14633
Olivier Houchardc7566002018-11-20 23:33:50 +010014634npn <protocols>
14635 This enables the NPN TLS extension and advertises the specified protocol list
14636 as supported on top of NPN. The protocol list consists in a comma-delimited
14637 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050014638 This requires that the SSL library is built with support for TLS extensions
Olivier Houchardc7566002018-11-20 23:33:50 +010014639 enabled (check with haproxy -vv). Note that the NPN extension has been
14640 replaced with the ALPN extension (see the "alpn" keyword), though this one is
14641 only available starting with OpenSSL 1.0.2.
14642
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014643observe <mode>
14644 This option enables health adjusting based on observing communication with
14645 the server. By default this functionality is disabled and enabling it also
14646 requires to enable health checks. There are two supported modes: "layer4" and
14647 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
14648 significant. In layer7, which is only allowed for http proxies, responses
14649 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010014650 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014651
14652 See also the "check", "on-error" and "error-limit".
14653
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014654on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010014655 Select what should happen when enough consecutive errors are detected.
14656 Currently, four modes are available:
14657 - fastinter: force fastinter
14658 - fail-check: simulate a failed check, also forces fastinter (default)
14659 - sudden-death: simulate a pre-fatal failed health check, one more failed
14660 check will mark a server down, forces fastinter
14661 - mark-down: mark the server immediately down and force fastinter
14662
14663 See also the "check", "observe" and "error-limit".
14664
Simon Hormane0d1bfb2011-06-21 14:34:58 +090014665on-marked-down <action>
14666 Modify what occurs when a server is marked down.
14667 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070014668 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
14669 all connections to the server are immediately terminated when the server
14670 goes down. It might be used if the health check detects more complex cases
14671 than a simple connection status, and long timeouts would cause the service
14672 to remain unresponsive for too long a time. For instance, a health check
14673 might detect that a database is stuck and that there's no chance to reuse
14674 existing connections anymore. Connections killed this way are logged with
14675 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090014676
14677 Actions are disabled by default
14678
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070014679on-marked-up <action>
14680 Modify what occurs when a server is marked up.
14681 Currently one action is available:
14682 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
14683 done only if the server is not in backup state and if it is not disabled
14684 (it must have an effective weight > 0). This can be used sometimes to force
14685 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014686 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
14687 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070014688 with extreme care. Sessions killed because a server comes up are logged
14689 with an 'U' termination code (for "Up").
14690
14691 Actions are disabled by default
14692
Willy Tarreau2f3f4d32020-07-01 07:43:51 +020014693pool-low-conn <max>
14694 Set a low threshold on the number of idling connections for a server, below
14695 which a thread will not try to steal a connection from another thread. This
14696 can be useful to improve CPU usage patterns in scenarios involving many very
14697 fast servers, in order to ensure all threads will keep a few idle connections
14698 all the time instead of letting them accumulate over one thread and migrating
14699 them from thread to thread. Typical values of twice the number of threads
14700 seem to show very good performance already with sub-millisecond response
14701 times. The default is zero, indicating that any idle connection can be used
14702 at any time. It is the recommended setting for normal use. This only applies
14703 to connections that can be shared according to the same principles as those
Willy Tarreau0784db82021-02-19 11:45:22 +010014704 applying to "http-reuse". In case connection sharing between threads would
14705 be disabled via "tune.idle-pool.shared", it can become very important to use
14706 this setting to make sure each thread always has a few connections, or the
14707 connection reuse rate will decrease as thread count increases.
Willy Tarreau2f3f4d32020-07-01 07:43:51 +020014708
Olivier Houchard006e3102018-12-10 18:30:32 +010014709pool-max-conn <max>
14710 Set the maximum number of idling connections for a server. -1 means unlimited
14711 connections, 0 means no idle connections. The default is -1. When idle
14712 connections are enabled, orphaned idle connections which do not belong to any
14713 client session anymore are moved to a dedicated pool so that they remain
14714 usable by future clients. This only applies to connections that can be shared
14715 according to the same principles as those applying to "http-reuse".
14716
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010014717pool-purge-delay <delay>
14718 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010014719 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020014720 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010014721
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014722port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014723 Using the "port" parameter, it becomes possible to use a different port to
William Dauchy4858fb22021-02-03 22:30:09 +010014724 send health-checks or to probe the agent-check. On some servers, it may be
14725 desirable to dedicate a port to a specific component able to perform complex
14726 tests which are more suitable to health-checks than the application. It is
14727 common to run a simple script in inetd for instance. This parameter is
14728 ignored if the "check" parameter is not set. See also the "addr" parameter.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014729
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020014730proto <name>
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020014731 Forces the multiplexer's protocol to use for the outgoing connections to this
14732 server. It must be compatible with the mode of the backend (TCP or HTTP). It
14733 must also be usable on the backend side. The list of available protocols is
Christopher Faulet982e17d2021-03-26 14:44:18 +010014734 reported in haproxy -vv.The protocols properties are reported : the mode
14735 (TCP/HTTP), the side (FE/BE), the mux name and its flags.
14736
14737 Some protocols report errors on aborts (flag=CLEAN_ABRT). Some others are
14738 subject to the head-of-line blocking on server side (flag=HOL_RISK). Finally
14739 some protocols don't support upgrades (flag=NO_UPG). The HTX compatibility is
14740 also reported (flag=HTX).
14741
14742 Here are the protocols that may be used as argument to a "proto" directive on
14743 a server line :
14744
14745 h2 : mode=HTTP side=FE|BE mux=H2 flags=HTX|CLEAN_ABRT|HOL_RISK|NO_UPG
14746 fcgi : mode=HTTP side=BE mux=FCGI flags=HTX|HOL_RISK|NO_UPG
14747 h1 : mode=HTTP side=FE|BE mux=H1 flags=HTX|NO_UPG
14748 none : mode=TCP side=FE|BE mux=PASS flags=NO_UPG
14749
Daniel Corbett67a82712020-07-06 23:01:19 -040014750 Idea behind this option is to bypass the selection of the best multiplexer's
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020014751 protocol for all connections established to this server.
14752
Amaury Denoyelle69352ec2021-10-18 14:40:29 +020014753 See also "ws" to use an alternative protocol for websocket streams.
14754
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014755redir <prefix>
14756 The "redir" parameter enables the redirection mode for all GET and HEAD
14757 requests addressing this server. This means that instead of having HAProxy
14758 forward the request to the server, it will send an "HTTP 302" response with
14759 the "Location" header composed of this prefix immediately followed by the
14760 requested URI beginning at the leading '/' of the path component. That means
14761 that no trailing slash should be used after <prefix>. All invalid requests
14762 will be rejected, and all non-GET or HEAD requests will be normally served by
14763 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014764 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010014765 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014766 users to a remote location in case of local disaster. Main use consists in
14767 increasing bandwidth for static servers by having the clients directly
14768 connect to them. Note: never use a relative location here, it would cause a
14769 loop between the client and HAProxy!
14770
14771 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
14772
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014773rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014774 The "rise" parameter states that a server will be considered as operational
14775 after <count> consecutive successful health checks. This value defaults to 2
14776 if unspecified. See also the "check", "inter" and "fall" parameters.
14777
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020014778resolve-opts <option>,<option>,...
14779 Comma separated list of options to apply to DNS resolution linked to this
14780 server.
14781
14782 Available options:
14783
14784 * allow-dup-ip
14785 By default, HAProxy prevents IP address duplication in a backend when DNS
14786 resolution at runtime is in operation.
14787 That said, for some cases, it makes sense that two servers (in the same
14788 backend, being resolved by the same FQDN) have the same IP address.
14789 For such case, simply enable this option.
14790 This is the opposite of prevent-dup-ip.
14791
Daniel Corbettf8716912019-11-17 09:48:56 -050014792 * ignore-weight
14793 Ignore any weight that is set within an SRV record. This is useful when
14794 you would like to control the weights using an alternate method, such as
14795 using an "agent-check" or through the runtime api.
14796
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020014797 * prevent-dup-ip
14798 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
14799 an IP address already set to a server in the same backend and sharing the
14800 same fqdn.
14801 This is the opposite of allow-dup-ip.
14802
14803 Example:
14804 backend b_myapp
14805 default-server init-addr none resolvers dns
14806 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
14807 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
14808
14809 With the option allow-dup-ip set:
14810 * if the nameserver returns a single IP address, then both servers will use
14811 it
14812 * If the nameserver returns 2 IP addresses, then each server will pick up a
14813 different address
14814
14815 Default value: not set
14816
Baptiste Assmann1fa66662015-04-14 00:28:47 +020014817resolve-prefer <family>
14818 When DNS resolution is enabled for a server and multiple IP addresses from
14819 different families are returned, HAProxy will prefer using an IP address
14820 from the family mentioned in the "resolve-prefer" parameter.
14821 Available families: "ipv4" and "ipv6"
14822
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020014823 Default value: ipv6
14824
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020014825 Example:
14826
14827 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020014828
Thierry Fournierac88cfe2016-02-17 22:05:30 +010014829resolve-net <network>[,<network[,...]]
John Roeslerfb2fce12019-07-10 15:45:51 -050014830 This option prioritizes the choice of an ip address matching a network. This is
Thierry Fournierac88cfe2016-02-17 22:05:30 +010014831 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010014832 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010014833 different datacenters. The latency between datacenter is not negligible, so
14834 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010014835 configured network, another address is selected.
14836
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020014837 Example:
14838
14839 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010014840
Baptiste Assmann1fa66662015-04-14 00:28:47 +020014841resolvers <id>
14842 Points to an existing "resolvers" section to resolve current server's
14843 hostname.
14844
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020014845 Example:
14846
14847 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020014848
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020014849 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020014850
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010014851send-proxy
14852 The "send-proxy" parameter enforces use of the PROXY protocol over any
14853 connection established to this server. The PROXY protocol informs the other
14854 end about the layer 3/4 addresses of the incoming connection, so that it can
14855 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014856 upper layer protocol. For connections accepted by an "accept-proxy" or
14857 "accept-netscaler-cip" listener, the advertised address will be used. Only
14858 TCPv4 and TCPv6 address families are supported. Other families such as
14859 Unix sockets, will report an UNKNOWN family. Servers using this option can
Daniel Corbett9f0843f2021-05-08 10:50:37 -040014860 fully be chained to another instance of HAProxy listening with an
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014861 "accept-proxy" setting. This setting must not be used if the server isn't
14862 aware of the protocol. When health checks are sent to the server, the PROXY
14863 protocol is automatically used when this option is set, unless there is an
14864 explicit "port" or "addr" directive, in which case an explicit
14865 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010014866 See also the "no-send-proxy" option of this section and "accept-proxy" and
14867 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010014868
David Safb76832014-05-08 23:42:08 -040014869send-proxy-v2
14870 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
14871 over any connection established to this server. The PROXY protocol informs
14872 the other end about the layer 3/4 addresses of the incoming connection, so
14873 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020014874 whatever the upper layer protocol. It also send ALPN information if an alpn
14875 have been negotiated. This setting must not be used if the server isn't aware
14876 of this version of the protocol. See also the "no-send-proxy-v2" option of
14877 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040014878
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010014879proxy-v2-options <option>[,<option>]*
Tim Duesterhuscf6e0c82020-03-13 12:34:24 +010014880 The "proxy-v2-options" parameter add options to send in PROXY protocol
14881 version 2 when "send-proxy-v2" is used. Options available are:
14882
14883 - ssl : See also "send-proxy-v2-ssl".
14884 - cert-cn : See also "send-proxy-v2-ssl-cn".
14885 - ssl-cipher: Name of the used cipher.
14886 - cert-sig : Signature algorithm of the used certificate.
14887 - cert-key : Key algorithm of the used certificate
14888 - authority : Host name value passed by the client (only SNI from a TLS
14889 connection is supported).
14890 - crc32c : Checksum of the PROXYv2 header.
14891 - unique-id : Send a unique ID generated using the frontend's
14892 "unique-id-format" within the PROXYv2 header.
14893 This unique-id is primarily meant for "mode tcp". It can
14894 lead to unexpected results in "mode http", because the
14895 generated unique ID is also used for the first HTTP request
14896 within a Keep-Alive connection.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010014897
David Safb76832014-05-08 23:42:08 -040014898send-proxy-v2-ssl
14899 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
14900 2 over any connection established to this server. The PROXY protocol informs
14901 the other end about the layer 3/4 addresses of the incoming connection, so
14902 that it can know the client's address or the public address it accessed to,
14903 whatever the upper layer protocol. In addition, the SSL information extension
14904 of the PROXY protocol is added to the PROXY protocol header. This setting
14905 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010014906 See also the "no-send-proxy-v2-ssl" option of this section and the
14907 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040014908
14909send-proxy-v2-ssl-cn
14910 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
14911 2 over any connection established to this server. The PROXY protocol informs
14912 the other end about the layer 3/4 addresses of the incoming connection, so
14913 that it can know the client's address or the public address it accessed to,
14914 whatever the upper layer protocol. In addition, the SSL information extension
14915 of the PROXY protocol, along along with the Common Name from the subject of
14916 the client certificate (if any), is added to the PROXY protocol header. This
14917 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010014918 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
14919 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040014920
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010014921slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014922 The "slowstart" parameter for a server accepts a value in milliseconds which
14923 indicates after how long a server which has just come back up will run at
14924 full speed. Just as with every other time-based parameter, it can be entered
14925 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
14926 linearly from 0 to 100% during this time. The limitation applies to two
14927 parameters :
14928
14929 - maxconn: the number of connections accepted by the server will grow from 1
14930 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
14931
14932 - weight: when the backend uses a dynamic weighted algorithm, the weight
14933 grows linearly from 1 to 100%. In this case, the weight is updated at every
14934 health-check. For this reason, it is important that the "inter" parameter
14935 is smaller than the "slowstart", in order to maximize the number of steps.
14936
Daniel Corbett9f0843f2021-05-08 10:50:37 -040014937 The slowstart never applies when HAProxy starts, otherwise it would cause
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014938 trouble to running servers. It only applies when a server has been previously
14939 seen as failed.
14940
Willy Tarreau732eac42015-07-09 11:40:25 +020014941sni <expression>
14942 The "sni" parameter evaluates the sample fetch expression, converts it to a
14943 string and uses the result as the host name sent in the SNI TLS extension to
14944 the server. A typical use case is to send the SNI received from the client in
14945 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020014946 expression, though alternatives such as req.hdr(host) can also make sense. If
14947 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020014948 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010014949 "verify" directive for more details. If you want to set a SNI for health
14950 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020014951
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020014952source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020014953source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020014954source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014955 The "source" parameter sets the source address which will be used when
14956 connecting to the server. It follows the exact same parameters and principle
14957 as the backend "source" keyword, except that it only applies to the server
14958 referencing it. Please consult the "source" keyword for details.
14959
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020014960 Additionally, the "source" statement on a server line allows one to specify a
14961 source port range by indicating the lower and higher bounds delimited by a
14962 dash ('-'). Some operating systems might require a valid IP address when a
14963 source port range is specified. It is permitted to have the same IP/range for
14964 several servers. Doing so makes it possible to bypass the maximum of 64k
14965 total concurrent connections. The limit will then reach 64k connections per
14966 server.
14967
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000014968 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
14969 specifying the source address without port(s).
14970
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020014971ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020014972 This option enables SSL ciphering on outgoing connections to the server. It
14973 is critical to verify server certificates using "verify" when using SSL to
14974 connect to servers, otherwise the communication is prone to trivial man in
14975 the-middle attacks rendering SSL useless. When this option is used, health
14976 checks are automatically sent in SSL too unless there is a "port" or an
14977 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010014978 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
14979 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020014980
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020014981ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
14982 This option enforces use of <version> or lower when SSL is used to communicate
14983 with the server. This option is also available on global statement
14984 "ssl-default-server-options". See also "ssl-min-ver".
14985
14986ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
14987 This option enforces use of <version> or upper when SSL is used to communicate
14988 with the server. This option is also available on global statement
14989 "ssl-default-server-options". See also "ssl-max-ver".
14990
Frédéric Lécailled2376272017-03-21 18:52:12 +010014991ssl-reuse
14992 This option may be used as "server" setting to reset any "no-ssl-reuse"
14993 setting which would have been inherited from "default-server" directive as
14994 default value.
14995 It may also be used as "default-server" setting to reset any previous
14996 "default-server" "no-ssl-reuse" setting.
14997
14998stick
14999 This option may be used as "server" setting to reset any "non-stick"
15000 setting which would have been inherited from "default-server" directive as
15001 default value.
15002 It may also be used as "default-server" setting to reset any previous
15003 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020015004
Alexander Liu2a54bb72019-05-22 19:44:48 +080015005socks4 <addr>:<port>
John Roeslerfb2fce12019-07-10 15:45:51 -050015006 This option enables upstream socks4 tunnel for outgoing connections to the
Alexander Liu2a54bb72019-05-22 19:44:48 +080015007 server. Using this option won't force the health check to go via socks4 by
15008 default. You will have to use the keyword "check-via-socks4" to enable it.
15009
Willy Tarreau163d4622015-10-13 16:16:41 +020015010tcp-ut <delay>
15011 Sets the TCP User Timeout for all outgoing connections to this server. This
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015012 option is available on Linux since version 2.6.37. It allows HAProxy to
Willy Tarreau163d4622015-10-13 16:16:41 +020015013 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010015014 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020015015 long-lived connections experiencing long idle periods such as remote
15016 terminals or database connection pools, where the client and server timeouts
15017 must remain high to allow a long period of idle, but where it is important to
15018 detect that the server has disappeared in order to release all resources
15019 associated with its connection (and the client's session). One typical use
15020 case is also to force dead server connections to die when health checks are
15021 too slow or during a soft reload since health checks are then disabled. The
15022 argument is a delay expressed in milliseconds by default. This only works for
15023 regular TCP connections, and is ignored for other protocols.
15024
Willy Tarreau034c88c2017-01-23 23:36:45 +010015025tfo
15026 This option enables using TCP fast open when connecting to servers, on
15027 systems that support it (currently only the Linux kernel >= 4.11).
15028 See the "tfo" bind option for more information about TCP fast open.
15029 Please note that when using tfo, you should also use the "conn-failure",
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015030 "empty-response" and "response-timeout" keywords for "retry-on", or HAProxy
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020015031 won't be able to retry the connection on failure. See also "no-tfo".
Willy Tarreau034c88c2017-01-23 23:36:45 +010015032
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015033track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020015034 This option enables ability to set the current state of the server by tracking
15035 another one. It is possible to track a server which itself tracks another
15036 server, provided that at the end of the chain, a server has health checks
15037 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015038 used, it has to be enabled on both proxies.
15039
Frédéric Lécailled2376272017-03-21 18:52:12 +010015040tls-tickets
15041 This option may be used as "server" setting to reset any "no-tls-tickets"
15042 setting which would have been inherited from "default-server" directive as
15043 default value.
Lukas Tribusbdb386d2020-03-10 00:56:09 +010015044 The TLS ticket mechanism is only used up to TLS 1.2.
15045 Forward Secrecy is compromised with TLS tickets, unless ticket keys
15046 are periodically rotated (via reload or by using "tls-ticket-keys").
Frédéric Lécailled2376272017-03-21 18:52:12 +010015047 It may also be used as "default-server" setting to reset any previous
Bjoern Jacke5ab7eb62020-02-13 14:16:16 +010015048 "default-server" "no-tls-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010015049
Emeric Brunef42d922012-10-11 16:11:36 +020015050verify [none|required]
15051 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010015052 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020015053 certificate provided by the server is verified using CAs from 'ca-file' and
15054 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010015055 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020015056 the name passed using the "sni" directive, or if not provided, the static
15057 host name passed using the "verifyhost" directive. When no name is found, the
15058 certificate's names are ignored. For this reason, without SNI it's important
15059 to use "verifyhost". On verification failure the handshake is aborted. It is
15060 critically important to verify server certificates when using SSL to connect
15061 to servers, otherwise the communication is prone to trivial man-in-the-middle
15062 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
15063 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020015064
Evan Broderbe554312013-06-27 00:05:25 -070015065verifyhost <hostname>
15066 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020015067 only takes effect if 'verify required' is also specified. This directive sets
15068 a default static hostname to check the server's certificate against when no
15069 SNI was used to connect to the server. If SNI is not used, this is the only
15070 way to enable hostname verification. This static hostname, when set, will
15071 also be used for health checks (which cannot provide an SNI value). If none
15072 of the hostnames in the certificate match the specified hostname, the
15073 handshake is aborted. The hostnames in the server-provided certificate may
15074 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070015075
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010015076weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015077 The "weight" parameter is used to adjust the server's weight relative to
15078 other servers. All servers will receive a load proportional to their weight
15079 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020015080 load. The default weight is 1, and the maximal value is 256. A value of 0
15081 means the server will not participate in load-balancing but will still accept
15082 persistent connections. If this parameter is used to distribute the load
15083 according to server's capacity, it is recommended to start with values which
15084 can both grow and shrink, for instance between 10 and 100 to leave enough
15085 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015086
Amaury Denoyelle69352ec2021-10-18 14:40:29 +020015087ws { auto | h1 | h2 }
15088 This option allows to configure the protocol used when relaying websocket
15089 streams. This is most notably useful when using an HTTP/2 backend without the
15090 support for H2 websockets through the RFC8441.
15091
15092 The default mode is "auto". This will reuse the same protocol as the main
15093 one. The only difference is when using ALPN. In this case, it can try to
15094 downgrade the ALPN to "http/1.1" only for websocket streams if the configured
15095 server ALPN contains it.
15096
15097 The value "h1" is used to force HTTP/1.1 for websockets streams, through ALPN
15098 if SSL ALPN is activated for the server. Similarly, "h2" can be used to
15099 force HTTP/2.0 websockets. Use this value with care : the server must support
15100 RFC8441 or an error will be reported by haproxy when relaying websockets.
15101
15102 Note that NPN is not taken into account as its usage has been deprecated in
15103 favor of the ALPN extension.
15104
15105 See also "alpn" and "proto".
15106
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015107
Cyril Bonté46175dd2015-07-02 22:45:32 +0200151085.3. Server IP address resolution using DNS
15109-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015110
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015111HAProxy allows using a host name on the server line to retrieve its IP address
15112using name servers. By default, HAProxy resolves the name when parsing the
Thayne McCombscdbcca92021-01-07 21:24:41 -070015113configuration file, at startup and cache the result for the process's life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015114This is not sufficient in some cases, such as in Amazon where a server's IP
15115can change after a reboot or an ELB Virtual IP can change based on current
15116workload.
15117This chapter describes how HAProxy can be configured to process server's name
15118resolution at run time.
15119Whether run time server name resolution has been enable or not, HAProxy will
15120carry on doing the first resolution when parsing the configuration.
15121
15122
Cyril Bonté46175dd2015-07-02 22:45:32 +0200151235.3.1. Global overview
15124----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015125
15126As we've seen in introduction, name resolution in HAProxy occurs at two
15127different steps of the process life:
15128
15129 1. when starting up, HAProxy parses the server line definition and matches a
15130 host name. It uses libc functions to get the host name resolved. This
15131 resolution relies on /etc/resolv.conf file.
15132
Christopher Faulet67957bd2017-09-27 11:00:59 +020015133 2. at run time, HAProxy performs periodically name resolutions for servers
15134 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015135
15136A few other events can trigger a name resolution at run time:
15137 - when a server's health check ends up in a connection timeout: this may be
15138 because the server has a new IP address. So we need to trigger a name
15139 resolution to know this new IP.
15140
Christopher Faulet67957bd2017-09-27 11:00:59 +020015141When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010015142HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020015143SRV label is specified, then the corresponding SRV records will be retrieved
15144from the DNS server, and the provided hostnames will be used. The SRV label
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015145will be checked periodically, and if any server are added or removed, HAProxy
Christopher Faulet67957bd2017-09-27 11:00:59 +020015146will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020015147
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015148A few things important to notice:
John Roeslerfb2fce12019-07-10 15:45:51 -050015149 - all the name servers are queried in the meantime. HAProxy will process the
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015150 first valid response.
15151
15152 - a resolution is considered as invalid (NX, timeout, refused), when all the
15153 servers return an error.
15154
15155
Cyril Bonté46175dd2015-07-02 22:45:32 +0200151565.3.2. The resolvers section
15157----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015158
15159This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020015160HAProxy. There can be as many as resolvers section as needed. Each section can
15161contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015162
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015163When multiple name servers are configured in a resolvers section, then HAProxy
15164uses the first valid response. In case of invalid responses, only the last one
15165is treated. Purpose is to give the chance to a slow server to deliver a valid
15166answer after a fast faulty or outdated server.
15167
15168When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020015169used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015170
Christopher Faulet67957bd2017-09-27 11:00:59 +020015171 1. HAProxy retries the same DNS query with a new query type. The A queries are
15172 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
15173 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015174
Christopher Faulet67957bd2017-09-27 11:00:59 +020015175 2. When the fallback on the query type was done (or not applicable), HAProxy
15176 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015177
Thierry Fournierfc13f792021-12-15 19:03:52 +010015178 3. HAProxy retries previous steps <resolve_retries> times. If no valid
Christopher Faulet67957bd2017-09-27 11:00:59 +020015179 response is received after that, it stops the DNS resolution and reports
15180 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015181
Christopher Faulet67957bd2017-09-27 11:00:59 +020015182For example, with 2 name servers configured in a resolvers section, the
15183following scenarios are possible:
15184
15185 - First response is valid and is applied directly, second response is
15186 ignored
15187
15188 - First response is invalid and second one is valid, then second response is
15189 applied
15190
15191 - First response is a NX domain and second one a truncated response, then
15192 HAProxy retries the query with a new type
15193
15194 - First response is a NX domain and second one is a timeout, then HAProxy
15195 retries the query with a new type
15196
15197 - Query timed out for both name servers, then HAProxy retries it with the
15198 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015199
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015200As a DNS server may not answer all the IPs in one DNS request, HAProxy keeps
Olivier Houcharda8c6db82017-07-06 18:46:47 +020015201a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020015202<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020015203
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015204
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015205resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010015206 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015207
15208A resolvers section accept the following parameters:
15209
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020015210accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010015211 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020015212 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020015213 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
15214 by RFC 6891)
15215
Emeric Brun4c751952021-03-08 16:41:29 +010015216 Note: the maximum allowed value is 65535. Recommended value for UDP is
15217 4096 and it is not recommended to exceed 8192 except if you are sure
15218 that your system and network can handle this (over 65507 makes no sense
15219 since is the maximum UDP payload size). If you are using only TCP
15220 nameservers to handle huge DNS responses, you should put this value
15221 to the max: 65535.
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020015222
Emeric Brunc8f3e452021-04-07 16:04:54 +020015223nameserver <name> <address>[:port] [param*]
15224 Used to configure a nameserver. <name> of the nameserver should ne unique.
15225 By default the <address> is considered of type datagram. This means if an
15226 IPv4 or IPv6 is configured without special address prefixes (paragraph 11.)
15227 the UDP protocol will be used. If an stream protocol address prefix is used,
15228 the nameserver will be considered as a stream server (TCP for instance) and
15229 "server" parameters found in 5.2 paragraph which are relevant for DNS
15230 resolving will be considered. Note: currently, in TCP mode, 4 queries are
15231 pipelined on the same connections. A batch of idle connections are removed
15232 every 5 seconds. "maxconn" can be configured to limit the amount of those
Emeric Brun56fc5d92021-02-12 20:05:45 +010015233 concurrent connections and TLS should also usable if the server supports.
15234
Ben Draut44e609b2018-05-29 15:40:08 -060015235parse-resolv-conf
15236 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
15237 list. Ordered as if each nameserver in /etc/resolv.conf was individually
15238 placed in the resolvers section in place of this directive.
15239
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015240hold <status> <period>
15241 Defines <period> during which the last name resolution should be kept based
15242 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010015243 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020015244 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015245 <period> : interval between two successive name resolution when the last
15246 answer was in <status>. It follows the HAProxy time format.
15247 <period> is in milliseconds by default.
15248
Baptiste Assmann686408b2017-08-18 10:15:42 +020015249 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015250
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015251resolve_retries <nb>
15252 Defines the number <nb> of queries to send to resolve a server name before
15253 giving up.
15254 Default value: 3
15255
Baptiste Assmann62b75b42015-09-09 01:11:36 +020015256 A retry occurs on name server timeout or when the full sequence of DNS query
15257 type failover is over and we need to start up from the default ANY query
15258 type.
15259
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015260timeout <event> <time>
15261 Defines timeouts related to name resolution
15262 <event> : the event on which the <time> timeout period applies to.
15263 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010015264 - resolve : default time to trigger name resolutions when no
15265 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020015266 Default value: 1s
15267 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010015268 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020015269 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015270 <time> : time related to the event. It follows the HAProxy time format.
15271 <time> is expressed in milliseconds.
15272
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020015273 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015274
15275 resolvers mydns
15276 nameserver dns1 10.0.0.1:53
15277 nameserver dns2 10.0.0.2:53
Emeric Brunc8f3e452021-04-07 16:04:54 +020015278 nameserver dns3 tcp@10.0.0.3:53
Ben Draut44e609b2018-05-29 15:40:08 -060015279 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015280 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020015281 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015282 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010015283 hold other 30s
15284 hold refused 30s
15285 hold nx 30s
15286 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015287 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020015288 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020015289
15290
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200152916. Cache
15292---------
15293
15294HAProxy provides a cache, which was designed to perform cache on small objects
15295(favicon, css...). This is a minimalist low-maintenance cache which runs in
15296RAM.
15297
15298The cache is based on a memory which is shared between processes and threads,
15299this memory is split in blocks of 1k.
15300
15301If an object is not used anymore, it can be deleted to store a new object
15302independently of its expiration date. The oldest objects are deleted first
15303when we try to allocate a new one.
15304
15305The cache uses a hash of the host header and the URI as the key.
15306
15307It's possible to view the status of a cache using the Unix socket command
15308"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
15309for more details.
15310
15311When an object is delivered from the cache, the server name in the log is
15312replaced by "<CACHE>".
15313
15314
153156.1. Limitation
15316----------------
15317
15318The cache won't store and won't deliver objects in these cases:
15319
15320- If the response is not a 200
Remi Tricot-Le Breton4f730832020-11-26 15:51:50 +010015321- If the response contains a Vary header and either the process-vary option is
15322 disabled, or a currently unmanaged header is specified in the Vary value (only
15323 accept-encoding and referer are managed for now)
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020015324- If the Content-Length + the headers size is greater than "max-object-size"
15325- If the response is not cacheable
Remi Tricot-Le Bretond493bc82020-11-26 15:51:29 +010015326- If the response does not have an explicit expiration time (s-maxage or max-age
15327 Cache-Control directives or Expires header) or a validator (ETag or Last-Modified
15328 headers)
Remi Tricot-Le Breton5853c0c2020-12-10 17:58:43 +010015329- If the process-vary option is enabled and there are already max-secondary-entries
15330 entries with the same primary key as the current response
Remi Tricot-Le Breton6ca89162021-01-07 14:50:51 +010015331- If the process-vary option is enabled and the response has an unknown encoding (not
15332 mentioned in https://www.iana.org/assignments/http-parameters/http-parameters.xhtml)
15333 while varying on the accept-encoding client header
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020015334
15335- If the request is not a GET
15336- If the HTTP version of the request is smaller than 1.1
15337- If the request contains an Authorization header
15338
15339
153406.2. Setup
15341-----------
15342
15343To setup a cache, you must define a cache section and use it in a proxy with
15344the corresponding http-request and response actions.
15345
15346
153476.2.1. Cache section
15348---------------------
15349
15350cache <name>
15351 Declare a cache section, allocate a shared cache memory named <name>, the
15352 size of cache is mandatory.
15353
15354total-max-size <megabytes>
15355 Define the size in RAM of the cache in megabytes. This size is split in
15356 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
15357
15358max-object-size <bytes>
15359 Define the maximum size of the objects to be cached. Must not be greater than
15360 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
15361 All objects with sizes larger than "max-object-size" will not be cached.
15362
15363max-age <seconds>
Remi Tricot-Le Breton5853c0c2020-12-10 17:58:43 +010015364 Define the maximum expiration duration. The expiration is set as the lowest
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020015365 value between the s-maxage or max-age (in this order) directive in the
15366 Cache-Control response header and this value. The default value is 60
15367 seconds, which means that you can't cache an object more than 60 seconds by
15368 default.
15369
Remi Tricot-Le Bretone6cc5b52020-12-23 18:13:53 +010015370process-vary <on/off>
15371 Enable or disable the processing of the Vary header. When disabled, a response
Remi Tricot-Le Breton754b2422020-11-16 15:56:10 +010015372 containing such a header will never be cached. When enabled, we need to calculate
15373 a preliminary hash for a subset of request headers on all the incoming requests
15374 (which might come with a cpu cost) which will be used to build a secondary key
Remi Tricot-Le Bretone6cc5b52020-12-23 18:13:53 +010015375 for a given request (see RFC 7234#4.1). The default value is off (disabled).
Remi Tricot-Le Breton754b2422020-11-16 15:56:10 +010015376
Remi Tricot-Le Breton5853c0c2020-12-10 17:58:43 +010015377max-secondary-entries <number>
15378 Define the maximum number of simultaneous secondary entries with the same primary
15379 key in the cache. This needs the vary support to be enabled. Its default value is 10
15380 and should be passed a strictly positive integer.
15381
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020015382
153836.2.2. Proxy section
15384---------------------
15385
15386http-request cache-use <name> [ { if | unless } <condition> ]
15387 Try to deliver a cached object from the cache <name>. This directive is also
15388 mandatory to store the cache as it calculates the cache hash. If you want to
15389 use a condition for both storage and delivering that's a good idea to put it
15390 after this one.
15391
15392http-response cache-store <name> [ { if | unless } <condition> ]
15393 Store an http-response within the cache. The storage of the response headers
15394 is done at this step, which means you can use others http-response actions
15395 to modify headers before or after the storage of the response. This action
15396 is responsible for the setup of the cache storage filter.
15397
15398
15399Example:
15400
15401 backend bck1
15402 mode http
15403
15404 http-request cache-use foobar
15405 http-response cache-store foobar
15406 server srv1 127.0.0.1:80
15407
15408 cache foobar
15409 total-max-size 4
15410 max-age 240
15411
15412
Willy Tarreau74ca5042013-06-11 23:12:07 +0200154137. Using ACLs and fetching samples
15414----------------------------------
15415
Davor Ocelice9ed2812017-12-25 17:49:28 +010015416HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020015417client or server information, from tables, environmental information etc...
15418The action of extracting such data is called fetching a sample. Once retrieved,
15419these samples may be used for various purposes such as a key to a stick-table,
15420but most common usages consist in matching them against predefined constant
15421data called patterns.
15422
15423
154247.1. ACL basics
15425---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015426
15427The use of Access Control Lists (ACL) provides a flexible solution to perform
15428content switching and generally to take decisions based on content extracted
15429from the request, the response or any environmental status. The principle is
15430simple :
15431
Willy Tarreau74ca5042013-06-11 23:12:07 +020015432 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015433 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020015434 - apply one or multiple pattern matching methods on this sample
15435 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015436
Willy Tarreau74ca5042013-06-11 23:12:07 +020015437The actions generally consist in blocking a request, selecting a backend, or
15438adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015439
15440In order to define a test, the "acl" keyword is used. The syntax is :
15441
Willy Tarreau74ca5042013-06-11 23:12:07 +020015442 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015443
15444This creates a new ACL <aclname> or completes an existing one with new tests.
15445Those tests apply to the portion of request/response specified in <criterion>
15446and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015447an operator which may be specified before the set of values. Optionally some
15448conversion operators may be applied to the sample, and they will be specified
15449as a comma-delimited list of keywords just after the first keyword. The values
15450are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015451
15452ACL names must be formed from upper and lower case letters, digits, '-' (dash),
15453'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
15454which means that "my_acl" and "My_Acl" are two different ACLs.
15455
15456There is no enforced limit to the number of ACLs. The unused ones do not affect
15457performance, they just consume a small amount of memory.
15458
Willy Tarreau74ca5042013-06-11 23:12:07 +020015459The criterion generally is the name of a sample fetch method, or one of its ACL
15460specific declinations. The default test method is implied by the output type of
15461this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015462methods of a same sample fetch method. The sample fetch methods are the only
15463ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015464
15465Sample fetch methods return data which can be of the following types :
15466 - boolean
15467 - integer (signed or unsigned)
15468 - IPv4 or IPv6 address
15469 - string
15470 - data block
15471
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015472Converters transform any of these data into any of these. For example, some
15473converters might convert a string to a lower-case string while other ones
15474would turn a string to an IPv4 address, or apply a netmask to an IP address.
15475The resulting sample is of the type of the last converter applied to the list,
15476which defaults to the type of the sample fetch method.
15477
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015478Each sample or converter returns data of a specific type, specified with its
15479keyword in this documentation. When an ACL is declared using a standard sample
15480fetch method, certain types automatically involved a default matching method
15481which are summarized in the table below :
15482
15483 +---------------------+-----------------+
15484 | Sample or converter | Default |
15485 | output type | matching method |
15486 +---------------------+-----------------+
15487 | boolean | bool |
15488 +---------------------+-----------------+
15489 | integer | int |
15490 +---------------------+-----------------+
15491 | ip | ip |
15492 +---------------------+-----------------+
15493 | string | str |
15494 +---------------------+-----------------+
15495 | binary | none, use "-m" |
15496 +---------------------+-----------------+
15497
15498Note that in order to match a binary samples, it is mandatory to specify a
15499matching method, see below.
15500
Willy Tarreau74ca5042013-06-11 23:12:07 +020015501The ACL engine can match these types against patterns of the following types :
15502 - boolean
15503 - integer or integer range
15504 - IP address / network
15505 - string (exact, substring, suffix, prefix, subdir, domain)
15506 - regular expression
15507 - hex block
15508
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015509The following ACL flags are currently supported :
15510
Willy Tarreau2b5285d2010-05-09 23:45:24 +020015511 -i : ignore case during matching of all subsequent patterns.
15512 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015513 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010015514 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010015515 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010015516 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020015517 -- : force end of flags. Useful when a string looks like one of the flags.
15518
Willy Tarreau74ca5042013-06-11 23:12:07 +020015519The "-f" flag is followed by the name of a file from which all lines will be
15520read as individual values. It is even possible to pass multiple "-f" arguments
15521if the patterns are to be loaded from multiple files. Empty lines as well as
15522lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
15523will be stripped. If it is absolutely necessary to insert a valid pattern
15524beginning with a sharp, just prefix it with a space so that it is not taken for
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015525a comment. Depending on the data type and match method, HAProxy may load the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015526lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
15527exact string matching. In this case, duplicates will automatically be removed.
15528
Thierry FOURNIER9860c412014-01-29 14:23:29 +010015529The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
15530parsed as two column file. The first column contains the patterns used by the
15531ACL, and the second column contain the samples. The sample can be used later by
15532a map. This can be useful in some rare cases where an ACL would just be used to
15533check for the existence of a pattern in a map before a mapping is applied.
15534
Thierry FOURNIER3534d882014-01-20 17:01:44 +010015535The "-u" flag forces the unique id of the ACL. This unique id is used with the
15536socket interface to identify ACL and dynamically change its values. Note that a
15537file is always identified by its name even if an id is set.
15538
Willy Tarreau74ca5042013-06-11 23:12:07 +020015539Also, note that the "-i" flag applies to subsequent entries and not to entries
15540loaded from files preceding it. For instance :
15541
15542 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
15543
15544In this example, each line of "exact-ua.lst" will be exactly matched against
15545the "user-agent" header of the request. Then each line of "generic-ua" will be
15546case-insensitively matched. Then the word "test" will be insensitively matched
15547as well.
15548
15549The "-m" flag is used to select a specific pattern matching method on the input
15550sample. All ACL-specific criteria imply a pattern matching method and generally
15551do not need this flag. However, this flag is useful with generic sample fetch
15552methods to describe how they're going to be matched against the patterns. This
15553is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010015554obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020015555followed by a pattern matching method name, this method is used instead of the
15556default one for the criterion. This makes it possible to match contents in ways
15557that were not initially planned, or with sample fetch methods which return a
15558string. The matching method also affects the way the patterns are parsed.
15559
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010015560The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
15561By default, if the parser cannot parse ip address it considers that the parsed
15562string is maybe a domain name and try dns resolution. The flag "-n" disable this
15563resolution. It is useful for detecting malformed ip lists. Note that if the DNS
Daniel Corbett9f0843f2021-05-08 10:50:37 -040015564server is not reachable, the HAProxy configuration parsing may last many minutes
John Roeslerfb2fce12019-07-10 15:45:51 -050015565waiting for the timeout. During this time no error messages are displayed. The
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010015566flag "-n" disable this behavior. Note also that during the runtime, this
15567function is disabled for the dynamic acl modifications.
15568
Willy Tarreau74ca5042013-06-11 23:12:07 +020015569There are some restrictions however. Not all methods can be used with all
15570sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
15571be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020015572
15573 - "found" : only check if the requested sample could be found in the stream,
15574 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020015575 to pass any pattern to avoid confusion. This matching method is
15576 particularly useful to detect presence of certain contents such
15577 as headers, cookies, etc... even if they are empty and without
15578 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015579
15580 - "bool" : check the value as a boolean. It can only be applied to fetches
15581 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015582 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015583
15584 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020015585 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015586
15587 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020015588 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015589
Davor Ocelice9ed2812017-12-25 17:49:28 +010015590 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020015591 binary sequence. This may be used with binary or string samples.
15592
15593 - "len" : match the sample's length as an integer. This may be used with
15594 binary or string samples.
15595
Willy Tarreau74ca5042013-06-11 23:12:07 +020015596 - "str" : exact match : match the contents against a string. This may be
15597 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015598
Willy Tarreau74ca5042013-06-11 23:12:07 +020015599 - "sub" : substring match : check that the contents contain at least one of
15600 the provided string patterns. This may be used with binary or
15601 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015602
Willy Tarreau74ca5042013-06-11 23:12:07 +020015603 - "reg" : regex match : match the contents against a list of regular
15604 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015605
Willy Tarreau74ca5042013-06-11 23:12:07 +020015606 - "beg" : prefix match : check that the contents begin like the provided
15607 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015608
Willy Tarreau74ca5042013-06-11 23:12:07 +020015609 - "end" : suffix match : check that the contents end like the provided
15610 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015611
Willy Tarreau74ca5042013-06-11 23:12:07 +020015612 - "dir" : subdir match : check that a slash-delimited portion of the
15613 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015614 This may be used with binary or string samples.
15615
Willy Tarreau74ca5042013-06-11 23:12:07 +020015616 - "dom" : domain match : check that a dot-delimited portion of the contents
15617 exactly match one of the provided string patterns. This may be
15618 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020015619
15620For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
15621request, it is possible to do :
15622
15623 acl jsess_present cook(JSESSIONID) -m found
15624
15625In order to apply a regular expression on the 500 first bytes of data in the
15626buffer, one would use the following acl :
15627
15628 acl script_tag payload(0,500) -m reg -i <script>
15629
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015630On systems where the regex library is much slower when using "-i", it is
15631possible to convert the sample to lowercase before matching, like this :
15632
15633 acl script_tag payload(0,500),lower -m reg <script>
15634
Willy Tarreau74ca5042013-06-11 23:12:07 +020015635All ACL-specific criteria imply a default matching method. Most often, these
15636criteria are composed by concatenating the name of the original sample fetch
15637method and the matching method. For example, "hdr_beg" applies the "beg" match
15638to samples retrieved using the "hdr" fetch method. Since all ACL-specific
15639criteria rely on a sample fetch method, it is always possible instead to use
15640the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020015641
Willy Tarreau74ca5042013-06-11 23:12:07 +020015642If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015643the matching method is simply applied to the underlying sample fetch method.
15644For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020015645
Willy Tarreau74ca5042013-06-11 23:12:07 +020015646 acl short_form hdr_beg(host) www.
15647 acl alternate1 hdr_beg(host) -m beg www.
15648 acl alternate2 hdr_dom(host) -m beg www.
15649 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020015650
Willy Tarreau2b5285d2010-05-09 23:45:24 +020015651
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015652The table below summarizes the compatibility matrix between sample or converter
15653types and the pattern types to fetch against. It indicates for each compatible
15654combination the name of the matching method to be used, surrounded with angle
15655brackets ">" and "<" when the method is the default one and will work by
15656default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015657
Willy Tarreau74ca5042013-06-11 23:12:07 +020015658 +-------------------------------------------------+
15659 | Input sample type |
15660 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015661 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015662 +----------------------+---------+---------+---------+---------+---------+
15663 | none (presence only) | found | found | found | found | found |
15664 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015665 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015666 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015667 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015668 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015669 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015670 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015671 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015672 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020015673 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015674 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015675 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015676 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015677 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015678 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015679 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015680 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015681 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015682 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015683 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015684 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010015685 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020015686 +----------------------+---------+---------+---------+---------+---------+
15687 | hex block | | | | bin | bin |
15688 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020015689
15690
Willy Tarreau74ca5042013-06-11 23:12:07 +0200156917.1.1. Matching booleans
15692------------------------
15693
15694In order to match a boolean, no value is needed and all values are ignored.
15695Boolean matching is used by default for all fetch methods of type "boolean".
15696When boolean matching is used, the fetched value is returned as-is, which means
15697that a boolean "true" will always match and a boolean "false" will never match.
15698
15699Boolean matching may also be enforced using "-m bool" on fetch methods which
15700return an integer value. Then, integer value 0 is converted to the boolean
15701"false" and all other values are converted to "true".
15702
Willy Tarreau6a06a402007-07-15 20:15:28 +020015703
Willy Tarreau74ca5042013-06-11 23:12:07 +0200157047.1.2. Matching integers
15705------------------------
15706
15707Integer matching applies by default to integer fetch methods. It can also be
15708enforced on boolean fetches using "-m int". In this case, "false" is converted
15709to the integer 0, and "true" is converted to the integer 1.
15710
15711Integer matching also supports integer ranges and operators. Note that integer
15712matching only applies to positive values. A range is a value expressed with a
15713lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015714
15715For instance, "1024:65535" is a valid range to represent a range of
15716unprivileged ports, and "1024:" would also work. "0:1023" is a valid
15717representation of privileged ports, and ":1023" would also work.
15718
Willy Tarreau62644772008-07-16 18:36:06 +020015719As a special case, some ACL functions support decimal numbers which are in fact
15720two integers separated by a dot. This is used with some version checks for
15721instance. All integer properties apply to those decimal numbers, including
15722ranges and operators.
15723
Willy Tarreau6a06a402007-07-15 20:15:28 +020015724For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010015725operators with ranges does not make much sense and is strongly discouraged.
15726Similarly, it does not make much sense to perform order comparisons with a set
15727of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015728
Willy Tarreau0ba27502007-12-24 16:55:16 +010015729Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020015730
15731 eq : true if the tested value equals at least one value
15732 ge : true if the tested value is greater than or equal to at least one value
15733 gt : true if the tested value is greater than at least one value
15734 le : true if the tested value is less than or equal to at least one value
15735 lt : true if the tested value is less than at least one value
15736
Willy Tarreau0ba27502007-12-24 16:55:16 +010015737For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020015738
15739 acl negative-length hdr_val(content-length) lt 0
15740
Willy Tarreau62644772008-07-16 18:36:06 +020015741This one matches SSL versions between 3.0 and 3.1 (inclusive) :
15742
15743 acl sslv3 req_ssl_ver 3:3.1
15744
Willy Tarreau6a06a402007-07-15 20:15:28 +020015745
Willy Tarreau74ca5042013-06-11 23:12:07 +0200157467.1.3. Matching strings
15747-----------------------
15748
15749String matching applies to string or binary fetch methods, and exists in 6
15750different forms :
15751
15752 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015753 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020015754
15755 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015756 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020015757
15758 - prefix match (-m beg) : the patterns are compared with the beginning of
15759 the extracted string, and the ACL matches if any of them matches.
15760
15761 - suffix match (-m end) : the patterns are compared with the end of the
15762 extracted string, and the ACL matches if any of them matches.
15763
Baptiste Assmann33db6002016-03-06 23:32:10 +010015764 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020015765 string, delimited with slashes ("/"), and the ACL matches if any of them
15766 matches.
15767
15768 - domain match (-m dom) : the patterns are looked up inside the extracted
15769 string, delimited with dots ("."), and the ACL matches if any of them
15770 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015771
15772String matching applies to verbatim strings as they are passed, with the
15773exception of the backslash ("\") which makes it possible to escape some
15774characters such as the space. If the "-i" flag is passed before the first
15775string, then the matching will be performed ignoring the case. In order
15776to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010015777before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020015778
Mathias Weiersmuellercb250fc2019-12-02 09:43:40 +010015779Do not use string matches for binary fetches which might contain null bytes
15780(0x00), as the comparison stops at the occurrence of the first null byte.
15781Instead, convert the binary fetch to a hex string with the hex converter first.
15782
15783Example:
15784 # matches if the string <tag> is present in the binary sample
15785 acl tag_found req.payload(0,0),hex -m sub 3C7461673E
15786
Willy Tarreau6a06a402007-07-15 20:15:28 +020015787
Willy Tarreau74ca5042013-06-11 23:12:07 +0200157887.1.4. Matching regular expressions (regexes)
15789---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020015790
15791Just like with string matching, regex matching applies to verbatim strings as
15792they are passed, with the exception of the backslash ("\") which makes it
15793possible to escape some characters such as the space. If the "-i" flag is
15794passed before the first regex, then the matching will be performed ignoring
15795the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010015796the "--" flag before the first string. Same principle applies of course to
15797match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020015798
15799
Willy Tarreau74ca5042013-06-11 23:12:07 +0200158007.1.5. Matching arbitrary data blocks
15801-------------------------------------
15802
15803It is possible to match some extracted samples against a binary block which may
15804not safely be represented as a string. For this, the patterns must be passed as
15805a series of hexadecimal digits in an even number, when the match method is set
15806to binary. Each sequence of two digits will represent a byte. The hexadecimal
15807digits may be used upper or lower case.
15808
15809Example :
15810 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
15811 acl hello payload(0,6) -m bin 48656c6c6f0a
15812
15813
158147.1.6. Matching IPv4 and IPv6 addresses
15815---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020015816
15817IPv4 addresses values can be specified either as plain addresses or with a
15818netmask appended, in which case the IPv4 address matches whenever it is
15819within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010015820host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010015821difficult to read and debug configurations. If hostnames are used, you should
15822at least ensure that they are present in /etc/hosts so that the configuration
15823does not depend on any random DNS match at the moment the configuration is
15824parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020015825
Daniel Schnellereba56342016-04-13 00:26:52 +020015826The dotted IPv4 address notation is supported in both regular as well as the
15827abbreviated form with all-0-octets omitted:
15828
15829 +------------------+------------------+------------------+
15830 | Example 1 | Example 2 | Example 3 |
15831 +------------------+------------------+------------------+
15832 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
15833 | 192.168.1 | 10.12 | 127.1 |
15834 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
15835 | 192.168.1/22 | 10.12/8 | 127.1/8 |
15836 +------------------+------------------+------------------+
15837
15838Notice that this is different from RFC 4632 CIDR address notation in which
15839192.168.42/24 would be equivalent to 192.168.42.0/24.
15840
Willy Tarreauceb4ac92012-04-28 00:41:46 +020015841IPv6 may be entered in their usual form, with or without a netmask appended.
15842Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
15843trouble with randomly resolved IP addresses, host names are never allowed in
15844IPv6 patterns.
15845
15846HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
15847following situations :
15848 - tested address is IPv4, pattern address is IPv4, the match applies
15849 in IPv4 using the supplied mask if any.
15850 - tested address is IPv6, pattern address is IPv6, the match applies
15851 in IPv6 using the supplied mask if any.
15852 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
15853 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
15854 ::IPV4 or ::ffff:IPV4, otherwise it fails.
15855 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
15856 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
15857 applied in IPv6 using the supplied IPv6 mask.
15858
Willy Tarreau74ca5042013-06-11 23:12:07 +020015859
158607.2. Using ACLs to form conditions
15861----------------------------------
15862
15863Some actions are only performed upon a valid condition. A condition is a
15864combination of ACLs with operators. 3 operators are supported :
15865
15866 - AND (implicit)
15867 - OR (explicit with the "or" keyword or the "||" operator)
15868 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020015869
Willy Tarreau74ca5042013-06-11 23:12:07 +020015870A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020015871
Willy Tarreau74ca5042013-06-11 23:12:07 +020015872 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020015873
Willy Tarreau74ca5042013-06-11 23:12:07 +020015874Such conditions are generally used after an "if" or "unless" statement,
15875indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020015876
Willy Tarreau74ca5042013-06-11 23:12:07 +020015877For instance, to block HTTP requests to the "*" URL with methods other than
15878"OPTIONS", as well as POST requests without content-length, and GET or HEAD
15879requests with a content-length greater than 0, and finally every request which
15880is not either GET/HEAD/POST/OPTIONS !
15881
15882 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030015883 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
15884 http-request deny if METH_GET HTTP_CONTENT
15885 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020015886
15887To select a different backend for requests to static contents on the "www" site
15888and to every request on the "img", "video", "download" and "ftp" hosts :
15889
15890 acl url_static path_beg /static /images /img /css
15891 acl url_static path_end .gif .png .jpg .css .js
15892 acl host_www hdr_beg(host) -i www
15893 acl host_static hdr_beg(host) -i img. video. download. ftp.
15894
Davor Ocelice9ed2812017-12-25 17:49:28 +010015895 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020015896 # of host "www". Use backend "www" for the rest.
15897 use_backend static if host_static or host_www url_static
15898 use_backend www if host_www
15899
15900It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
15901expressions that are built on the fly without needing to be declared. They must
15902be enclosed between braces, with a space before and after each brace (because
15903the braces must be seen as independent words). Example :
15904
15905 The following rule :
15906
15907 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030015908 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020015909
15910 Can also be written that way :
15911
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030015912 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020015913
15914It is generally not recommended to use this construct because it's a lot easier
15915to leave errors in the configuration when written that way. However, for very
15916simple rules matching only one source IP address for instance, it can make more
15917sense to use them than to declare ACLs with random names. Another example of
15918good use is the following :
15919
15920 With named ACLs :
15921
15922 acl site_dead nbsrv(dynamic) lt 2
15923 acl site_dead nbsrv(static) lt 2
15924 monitor fail if site_dead
15925
15926 With anonymous ACLs :
15927
15928 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
15929
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030015930See section 4.2 for detailed help on the "http-request deny" and "use_backend"
15931keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015932
15933
159347.3. Fetching samples
15935---------------------
15936
15937Historically, sample fetch methods were only used to retrieve data to match
15938against patterns using ACLs. With the arrival of stick-tables, a new class of
15939sample fetch methods was created, most often sharing the same syntax as their
15940ACL counterpart. These sample fetch methods are also known as "fetches". As
15941of now, ACLs and fetches have converged. All ACL fetch methods have been made
15942available as fetch methods, and ACLs may use any sample fetch method as well.
15943
15944This section details all available sample fetch methods and their output type.
15945Some sample fetch methods have deprecated aliases that are used to maintain
15946compatibility with existing configurations. They are then explicitly marked as
15947deprecated and should not be used in new setups.
15948
15949The ACL derivatives are also indicated when available, with their respective
15950matching methods. These ones all have a well defined default pattern matching
15951method, so it is never necessary (though allowed) to pass the "-m" option to
15952indicate how the sample will be matched using ACLs.
15953
15954As indicated in the sample type versus matching compatibility matrix above,
15955when using a generic sample fetch method in an ACL, the "-m" option is
15956mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
15957the same keyword exists as an ACL keyword and as a standard fetch method, the
15958ACL engine will automatically pick the ACL-only one by default.
15959
15960Some of these keywords support one or multiple mandatory arguments, and one or
15961multiple optional arguments. These arguments are strongly typed and are checked
15962when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010015963incorrect argument (e.g. an unresolved backend name). Fetch function arguments
15964are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020015965is optional, it will be indicated below between square brackets ('[ ]'). When
15966all arguments are optional, the parenthesis may be omitted.
15967
15968Thus, the syntax of a standard sample fetch method is one of the following :
15969 - name
15970 - name(arg1)
15971 - name(arg1,arg2)
15972
Thierry FOURNIER060762e2014-04-23 13:29:15 +020015973
159747.3.1. Converters
15975-----------------
15976
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015977Sample fetch methods may be combined with transformations to be applied on top
15978of the fetched sample (also called "converters"). These combinations form what
15979is called "sample expressions" and the result is a "sample". Initially this
15980was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010015981has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010015982unique-id-format, add-header, ...).
15983
15984These transformations are enumerated as a series of specific keywords after the
15985sample fetch method. These keywords may equally be appended immediately after
15986the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010015987support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015988
Willy Tarreau97707872015-01-27 15:12:13 +010015989A certain category of converters are bitwise and arithmetic operators which
15990support performing basic operations on integers. Some bitwise operations are
15991supported (and, or, xor, cpl) and some arithmetic operations are supported
15992(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
15993bool) which make it possible to report a match without having to write an ACL.
15994
Willy Tarreau74ca5042013-06-11 23:12:07 +020015995The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010015996
Ben Shillitof25e8e52016-12-02 14:25:37 +00001599751d.single(<prop>[,<prop>*])
15998 Returns values for the properties requested as a string, where values are
15999 separated by the delimiter specified with "51degrees-property-separator".
16000 The device is identified using the User-Agent header passed to the
16001 converter. The function can be passed up to five property names, and if a
16002 property name can't be found, the value "NoData" is returned.
16003
16004 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010016005 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
16006 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000016007 # User-Agent passed to the converter.
16008 frontend http-in
16009 bind *:8081
16010 default_backend servers
16011 http-request set-header X-51D-DeviceTypeMobileTablet \
16012 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
16013
Willy Tarreau97707872015-01-27 15:12:13 +010016014add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016015 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016016 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010016017 name. The name of the variable starts with an indication about its scope. The
16018 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016019 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016020 "sess" : the variable is shared with the whole session
16021 "txn" : the variable is shared with the transaction (request and response)
16022 "req" : the variable is shared only during request processing
16023 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016024 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016025 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016026
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010016027aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
16028 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
16029 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
16030 need to be base64 encoded and the returned result is in raw byte format.
16031 If the <aead_tag> validation fails, the converter doesn't return any data.
16032 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
16033 converter requires at least OpenSSL 1.0.1.
16034
16035 Example:
16036 http-response set-header X-Decrypted-Text %[var(txn.enc),\
16037 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
16038
Willy Tarreau97707872015-01-27 15:12:13 +010016039and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016040 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016041 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010016042 numeric value or a variable name. The name of the variable starts with an
16043 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016044 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016045 "sess" : the variable is shared with the whole session
16046 "txn" : the variable is shared with the transaction (request and response)
16047 "req" : the variable is shared only during request processing
16048 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016049 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016050 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016051
Holger Just1bfc24b2017-05-06 00:56:53 +020016052b64dec
16053 Converts (decodes) a base64 encoded input string to its binary
16054 representation. It performs the inverse operation of base64().
Moemen MHEDHBI92f7d432021-04-01 20:53:59 +020016055 For base64url("URL and Filename Safe Alphabet" (RFC 4648)) variant
16056 see "ub64dec".
Holger Just1bfc24b2017-05-06 00:56:53 +020016057
Emeric Brun53d1a982014-04-30 18:21:37 +020016058base64
16059 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010016060 transfer binary content in a way that can be reliably transferred (e.g.
Moemen MHEDHBI92f7d432021-04-01 20:53:59 +020016061 an SSL ID can be copied in a header). For base64url("URL and Filename
16062 Safe Alphabet" (RFC 4648)) variant see "ub64enc".
Emeric Brun53d1a982014-04-30 18:21:37 +020016063
Willy Tarreau97707872015-01-27 15:12:13 +010016064bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016065 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010016066 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016067 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010016068 presence of a flag).
16069
Emeric Brun54c4ac82014-11-03 15:32:43 +010016070bytes(<offset>[,<length>])
16071 Extracts some bytes from an input binary sample. The result is a binary
16072 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010016073 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010016074
Willy Tarreau280f42b2018-02-19 15:34:12 +010016075concat([<start>],[<var>],[<end>])
16076 Concatenates up to 3 fields after the current sample which is then turned to
16077 a string. The first one, <start>, is a constant string, that will be appended
16078 immediately after the existing sample. It may be omitted if not used. The
16079 second one, <var>, is a variable name. The variable will be looked up, its
16080 contents converted to a string, and it will be appended immediately after the
16081 <first> part. If the variable is not found, nothing is appended. It may be
16082 omitted as well. The third field, <end> is a constant string that will be
16083 appended after the variable. It may also be omitted. Together, these elements
16084 allow to concatenate variables with delimiters to an existing set of
16085 variables. This can be used to build new variables made of a succession of
Willy Tarreauef21fac2020-02-14 13:37:20 +010016086 other variables, such as colon-delimited values. If commas or closing
Daniel Corbett67a82712020-07-06 23:01:19 -040016087 parenthesis are needed as delimiters, they must be protected by quotes or
Willy Tarreauef21fac2020-02-14 13:37:20 +010016088 backslashes, themselves protected so that they are not stripped by the first
16089 level parser. See examples below.
Willy Tarreau280f42b2018-02-19 15:34:12 +010016090
16091 Example:
16092 tcp-request session set-var(sess.src) src
16093 tcp-request session set-var(sess.dn) ssl_c_s_dn
16094 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
Willy Tarreauef21fac2020-02-14 13:37:20 +010016095 tcp-request session set-var(txn.ipport) "str(),concat('addr=(',sess.ip),concat(',',sess.port,')')"
Willy Tarreau280f42b2018-02-19 15:34:12 +010016096 http-request set-header x-hap-sig %[var(txn.sig)]
16097
Willy Tarreau97707872015-01-27 15:12:13 +010016098cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016099 Takes the input value of type signed integer, applies a ones-complement
16100 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010016101
Willy Tarreau80599772015-01-20 19:35:24 +010016102crc32([<avalanche>])
16103 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
16104 hash function. Optionally, it is possible to apply a full avalanche hash
16105 function to the output if the optional <avalanche> argument equals 1. This
16106 converter uses the same functions as used by the various hash-based load
16107 balancing algorithms, so it will provide exactly the same results. It is
16108 provided for compatibility with other software which want a CRC32 to be
16109 computed on some input keys, so it follows the most common implementation as
16110 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
16111 but may provide a better or at least less predictable distribution. It must
16112 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010016113 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
16114
16115crc32c([<avalanche>])
16116 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
16117 hash function. Optionally, it is possible to apply a full avalanche hash
16118 function to the output if the optional <avalanche> argument equals 1. This
16119 converter uses the same functions as described in RFC4960, Appendix B [8].
16120 It is provided for compatibility with other software which want a CRC32C to be
16121 computed on some input keys. It is slower than the other algorithms and it must
16122 not be used for security purposes as a 32-bit hash is trivial to break. See
16123 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010016124
Christopher Fauletea159d62020-04-01 16:21:44 +020016125cut_crlf
16126 Cuts the string representation of the input sample on the first carriage
16127 return ('\r') or newline ('\n') character found. Only the string length is
16128 updated.
16129
David Carlier29b3ca32015-09-25 14:09:21 +010016130da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020016131 Asks the DeviceAtlas converter to identify the User Agent string passed on
16132 input, and to emit a string made of the concatenation of the properties
16133 enumerated in argument, delimited by the separator defined by the global
16134 keyword "deviceatlas-property-separator", or by default the pipe character
Daniel Corbett9f0843f2021-05-08 10:50:37 -040016135 ('|'). There's a limit of 12 different properties imposed by the HAProxy
David Carlier4542b102015-06-01 13:54:29 +020016136 configuration language.
16137
16138 Example:
16139 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020016140 bind *:8881
16141 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000016142 http-request set-header X-DeviceAtlas-Data %[req.fhdr(User-Agent),da-csv(primaryHardwareType,osName,osVersion,browserName,browserVersion,browserRenderingEngine)]
David Carlier4542b102015-06-01 13:54:29 +020016143
Willy Tarreau0851fd52019-12-17 10:07:25 +010016144debug([<prefix][,<destination>])
16145 This converter is used as debug tool. It takes a capture of the input sample
16146 and sends it to event sink <destination>, which may designate a ring buffer
16147 such as "buf0", as well as "stdout", or "stderr". Available sinks may be
16148 checked at run time by issuing "show events" on the CLI. When not specified,
16149 the output will be "buf0", which may be consulted via the CLI's "show events"
16150 command. An optional prefix <prefix> may be passed to help distinguish
16151 outputs from multiple expressions. It will then appear before the colon in
16152 the output message. The input sample is passed as-is on the output, so that
16153 it is safe to insert the debug converter anywhere in a chain, even with non-
16154 printable sample types.
16155
16156 Example:
16157 tcp-request connection track-sc0 src,debug(track-sc)
Thierry FOURNIER9687c772015-05-07 15:46:29 +020016158
Patrick Gansterer8e366512020-04-22 16:47:57 +020016159digest(<algorithm>)
16160 Converts a binary input sample to a message digest. The result is a binary
16161 sample. The <algorithm> must be an OpenSSL message digest name (e.g. sha256).
16162
Daniel Corbett9f0843f2021-05-08 10:50:37 -040016163 Please note that this converter is only available when HAProxy has been
Patrick Gansterer8e366512020-04-22 16:47:57 +020016164 compiled with USE_OPENSSL.
16165
Willy Tarreau97707872015-01-27 15:12:13 +010016166div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016167 Divides the input value of type signed integer by <value>, and returns the
16168 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016169 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010016170 variable name. The name of the variable starts with an indication about its
16171 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016172 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016173 "sess" : the variable is shared with the whole session
16174 "txn" : the variable is shared with the transaction (request and response)
16175 "req" : the variable is shared only during request processing
16176 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016177 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016178 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016179
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016180djb2([<avalanche>])
16181 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
16182 hash function. Optionally, it is possible to apply a full avalanche hash
16183 function to the output if the optional <avalanche> argument equals 1. This
16184 converter uses the same functions as used by the various hash-based load
16185 balancing algorithms, so it will provide exactly the same results. It is
16186 mostly intended for debugging, but can be used as a stick-table entry to
16187 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010016188 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
16189 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016190
Willy Tarreau97707872015-01-27 15:12:13 +010016191even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016192 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010016193 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
16194
Marcin Deranek9631a282018-04-16 14:30:46 +020016195field(<index>,<delimiters>[,<count>])
16196 Extracts the substring at the given index counting from the beginning
16197 (positive index) or from the end (negative index) considering given delimiters
16198 from an input string. Indexes start at 1 or -1 and delimiters are a string
16199 formatted list of chars. Optionally you can specify <count> of fields to
16200 extract (default: 1). Value of 0 indicates extraction of all remaining
16201 fields.
16202
16203 Example :
16204 str(f1_f2_f3__f5),field(5,_) # f5
16205 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
16206 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
16207 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
16208 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010016209
Baptiste Assmanne138dda2020-10-22 15:39:03 +020016210fix_is_valid
16211 Parses a binary payload and performs sanity checks regarding FIX (Financial
16212 Information eXchange):
16213
16214 - checks that all tag IDs and values are not empty and the tags IDs are well
16215 numeric
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +050016216 - checks the BeginString tag is the first tag with a valid FIX version
Baptiste Assmanne138dda2020-10-22 15:39:03 +020016217 - checks the BodyLength tag is the second one with the right body length
Christopher Fauleted4bef72021-03-18 17:40:56 +010016218 - checks the MsgType tag is the third tag.
Baptiste Assmanne138dda2020-10-22 15:39:03 +020016219 - checks that last tag in the message is the CheckSum tag with a valid
16220 checksum
16221
16222 Due to current HAProxy design, only the first message sent by the client and
16223 the server can be parsed.
16224
16225 This converter returns a boolean, true if the payload contains a valid FIX
16226 message, false if not.
16227
16228 See also the fix_tag_value converter.
16229
16230 Example:
16231 tcp-request inspect-delay 10s
16232 tcp-request content reject unless { req.payload(0,0),fix_is_valid }
16233
16234fix_tag_value(<tag>)
16235 Parses a FIX (Financial Information eXchange) message and extracts the value
16236 from the tag <tag>. <tag> can be a string or an integer pointing to the
16237 desired tag. Any integer value is accepted, but only the following strings
16238 are translated into their integer equivalent: BeginString, BodyLength,
Daniel Corbettbefef702021-03-09 23:00:34 -050016239 MsgType, SenderCompID, TargetCompID, CheckSum. More tag names can be easily
Baptiste Assmanne138dda2020-10-22 15:39:03 +020016240 added.
16241
16242 Due to current HAProxy design, only the first message sent by the client and
16243 the server can be parsed. No message validation is performed by this
16244 converter. It is highly recommended to validate the message first using
16245 fix_is_valid converter.
16246
16247 See also the fix_is_valid converter.
16248
16249 Example:
16250 tcp-request inspect-delay 10s
16251 tcp-request content reject unless { req.payload(0,0),fix_is_valid }
16252 # MsgType tag ID is 35, so both lines below will return the same content
16253 tcp-request content set-var(txn.foo) req.payload(0,0),fix_tag_value(35)
16254 tcp-request content set-var(txn.bar) req.payload(0,0),fix_tag_value(MsgType)
16255
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016256hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010016257 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016258 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010016259 in a way that can be reliably transferred (e.g. an SSL ID can be copied in a
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016260 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010016261
Dragan Dosen3f957b22017-10-24 09:27:34 +020016262hex2i
16263 Converts a hex string containing two hex digits per input byte to an
John Roeslerfb2fce12019-07-10 15:45:51 -050016264 integer. If the input value cannot be converted, then zero is returned.
Dragan Dosen3f957b22017-10-24 09:27:34 +020016265
Christopher Faulet4ccc12f2020-04-01 09:08:32 +020016266htonl
16267 Converts the input integer value to its 32-bit binary representation in the
16268 network byte order. Because sample fetches own signed 64-bit integer, when
16269 this converter is used, the input integer value is first casted to an
16270 unsigned 32-bit integer.
16271
Tim Duesterhusa3082092021-01-21 17:40:49 +010016272hmac(<algorithm>,<key>)
Patrick Gansterer8e366512020-04-22 16:47:57 +020016273 Converts a binary input sample to a message authentication code with the given
16274 key. The result is a binary sample. The <algorithm> must be one of the
16275 registered OpenSSL message digest names (e.g. sha256). The <key> parameter must
16276 be base64 encoded and can either be a string or a variable.
16277
Daniel Corbett9f0843f2021-05-08 10:50:37 -040016278 Please note that this converter is only available when HAProxy has been
Patrick Gansterer8e366512020-04-22 16:47:57 +020016279 compiled with USE_OPENSSL.
16280
Cyril Bonté6bcd1822019-11-05 23:13:59 +010016281http_date([<offset],[<unit>])
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016282 Converts an integer supposed to contain a date since epoch to a string
16283 representing this date in a format suitable for use in HTTP header fields. If
Damien Claisseae6f1252019-10-30 15:57:28 +000016284 an offset value is specified, then it is added to the date before the
16285 conversion is operated. This is particularly useful to emit Date header fields,
16286 Expires values in responses when combined with a positive offset, or
16287 Last-Modified values when the offset is negative.
16288 If a unit value is specified, then consider the timestamp as either
16289 "s" for seconds (default behavior), "ms" for milliseconds, or "us" for
16290 microseconds since epoch. Offset is assumed to have the same unit as
16291 input timestamp.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016292
Tim Duesterhus3943e4f2020-09-11 14:25:23 +020016293iif(<true>,<false>)
16294 Returns the <true> string if the input value is true. Returns the <false>
16295 string otherwise.
16296
16297 Example:
Tim Duesterhus870713b2020-09-11 17:13:12 +020016298 http-request set-header x-forwarded-proto %[ssl_fc,iif(https,http)]
Tim Duesterhus3943e4f2020-09-11 14:25:23 +020016299
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016300in_table(<table>)
16301 Uses the string representation of the input sample to perform a look up in
16302 the specified table. If the key is not found in the table, a boolean false
16303 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010016304 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016305 or not a source IP address or an Authorization header was already seen).
16306
Tim Duesterhusa3082092021-01-21 17:40:49 +010016307ipmask(<mask4>,[<mask6>])
Tim Duesterhus1478aa72018-01-25 16:24:51 +010016308 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020016309 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010016310 table entries and as such use the same server. The mask4 can be passed in
16311 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
16312 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
16313 If no mask6 is given IPv6 addresses will fail to convert for backwards
16314 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020016315
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016316json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010016317 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016318 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020016319 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016320 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
16321 of errors:
16322 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
16323 bytes, ...)
16324 - invalid range (the decoded value is within a UTF-8 prohibited range),
16325 - code overlong (the value is encoded with more bytes than necessary).
16326
16327 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
16328 character is greater than 0xffff because the JSON string escape specification
16329 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
16330 in 4 variants designated by a combination of two suffix letters : "p" for
16331 "permissive" and "s" for "silently ignore". The behaviors of the decoders
16332 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010016333 - "ascii" : never fails;
16334 - "utf8" : fails on any detected errors;
16335 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016336 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010016337 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016338 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
16339 characters corresponding to the other errors.
16340
16341 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010016342 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016343
16344 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016345 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020016346 capture request header user-agent len 150
16347 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020016348
16349 Input request from client 127.0.0.1:
16350 GET / HTTP/1.0
16351 User-Agent: Very "Ugly" UA 1/2
16352
16353 Output log:
16354 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
16355
Alex51c8ad42021-04-15 16:45:15 +020016356json_query(<json_path>,[<output_type>])
16357 The json_query converter supports the JSON types string, boolean and
16358 number. Floating point numbers will be returned as a string. By
16359 specifying the output_type 'int' the value will be converted to an
16360 Integer. If conversion is not possible the json_query converter fails.
16361
16362 <json_path> must be a valid JSON Path string as defined in
16363 https://datatracker.ietf.org/doc/draft-ietf-jsonpath-base/
16364
16365 Example:
16366 # get a integer value from the request body
16367 # "{"integer":4}" => 5
16368 http-request set-var(txn.pay_int) req.body,json_query('$.integer','int'),add(1)
16369
16370 # get a key with '.' in the name
16371 # {"my.key":"myvalue"} => myvalue
16372 http-request set-var(txn.pay_mykey) req.body,json_query('$.my\\.key')
16373
16374 # {"boolean-false":false} => 0
16375 http-request set-var(txn.pay_boolean_false) req.body,json_query('$.boolean-false')
16376
16377 # get the value of the key 'iss' from a JWT Bearer token
16378 http-request set-var(txn.token_payload) req.hdr(Authorization),word(2,.),ub64dec,json_query('$.iss')
16379
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016380language(<value>[,<default>])
16381 Returns the value with the highest q-factor from a list as extracted from the
16382 "accept-language" header using "req.fhdr". Values with no q-factor have a
16383 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
16384 belong to the list of semi-colon delimited <values> will be considered. The
16385 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
16386 given list and a default value is provided, it is returned. Note that language
16387 names may have a variant after a dash ('-'). If this variant is present in the
16388 list, it will be matched, but if it is not, only the base language is checked.
16389 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010016390 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016391 ordering of the values in the request counts, as the first value among
16392 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020016393
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016394 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020016395
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016396 # this configuration switches to the backend matching a
16397 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020016398
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016399 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
16400 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
16401 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
16402 use_backend spanish if es
16403 use_backend french if fr
16404 use_backend english if en
16405 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020016406
Willy Tarreau60a2ee72017-12-15 07:13:48 +010016407length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010016408 Get the length of the string. This can only be placed after a string
16409 sample fetch function or after a transformation keyword returning a string
16410 type. The result is of type integer.
16411
Willy Tarreauffcb2e42014-07-10 16:29:08 +020016412lower
16413 Convert a string sample to lower case. This can only be placed after a string
16414 sample fetch function or after a transformation keyword returning a string
16415 type. The result is of type string.
16416
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020016417ltime(<format>[,<offset>])
16418 Converts an integer supposed to contain a date since epoch to a string
16419 representing this date in local time using a format defined by the <format>
16420 string using strftime(3). The purpose is to allow any date format to be used
16421 in logs. An optional <offset> in seconds may be applied to the input date
16422 (positive or negative). See the strftime() man page for the format supported
16423 by your operating system. See also the utime converter.
16424
16425 Example :
16426
16427 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010016428 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020016429 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
16430
Christopher Faulet51fc9d12020-04-01 17:24:41 +020016431ltrim(<chars>)
16432 Skips any characters from <chars> from the beginning of the string
16433 representation of the input sample.
16434
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016435map(<map_file>[,<default_value>])
16436map_<match_type>(<map_file>[,<default_value>])
16437map_<match_type>_<output_type>(<map_file>[,<default_value>])
16438 Search the input value from <map_file> using the <match_type> matching method,
16439 and return the associated value converted to the type <output_type>. If the
16440 input value cannot be found in the <map_file>, the converter returns the
16441 <default_value>. If the <default_value> is not set, the converter fails and
16442 acts as if no input value could be fetched. If the <match_type> is not set, it
16443 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
16444 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
16445 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010016446
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016447 It is important to avoid overlapping between the keys : IP addresses and
16448 strings are stored in trees, so the first of the finest match will be used.
16449 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010016450
Tim Düsterhus4896c442016-11-29 02:15:19 +010016451 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016452 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010016453
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016454 input type | match method | output type str | output type int | output type ip
16455 -----------+--------------+-----------------+-----------------+---------------
16456 str | str | map_str | map_str_int | map_str_ip
16457 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020016458 str | beg | map_beg | map_beg_int | map_end_ip
16459 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016460 str | sub | map_sub | map_sub_int | map_sub_ip
16461 -----------+--------------+-----------------+-----------------+---------------
16462 str | dir | map_dir | map_dir_int | map_dir_ip
16463 -----------+--------------+-----------------+-----------------+---------------
16464 str | dom | map_dom | map_dom_int | map_dom_ip
16465 -----------+--------------+-----------------+-----------------+---------------
16466 str | end | map_end | map_end_int | map_end_ip
16467 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080016468 str | reg | map_reg | map_reg_int | map_reg_ip
16469 -----------+--------------+-----------------+-----------------+---------------
16470 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016471 -----------+--------------+-----------------+-----------------+---------------
16472 int | int | map_int | map_int_int | map_int_ip
16473 -----------+--------------+-----------------+-----------------+---------------
16474 ip | ip | map_ip | map_ip_int | map_ip_ip
16475 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010016476
Thierry Fournier8feaa662016-02-10 22:55:20 +010016477 The special map called "map_regm" expect matching zone in the regular
16478 expression and modify the output replacing back reference (like "\1") by
16479 the corresponding match text.
16480
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016481 The file contains one key + value per line. Lines which start with '#' are
16482 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
16483 is then the first "word" (series of non-space/tabs characters), and the value
16484 is what follows this series of space/tab till the end of the line excluding
16485 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010016486
Thierry FOURNIER060762e2014-04-23 13:29:15 +020016487 Example :
16488
16489 # this is a comment and is ignored
16490 2.22.246.0/23 United Kingdom \n
16491 <-><-----------><--><------------><---->
16492 | | | | `- trailing spaces ignored
16493 | | | `---------- value
16494 | | `-------------------- middle spaces ignored
16495 | `---------------------------- key
16496 `------------------------------------ leading spaces ignored
16497
Willy Tarreau97707872015-01-27 15:12:13 +010016498mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016499 Divides the input value of type signed integer by <value>, and returns the
16500 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016501 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010016502 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016503 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016504 "sess" : the variable is shared with the whole session
16505 "txn" : the variable is shared with the transaction (request and response)
16506 "req" : the variable is shared only during request processing
16507 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016508 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016509 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016510
Alexbf1bd5a2021-04-24 13:02:21 +020016511mqtt_field_value(<packettype>,<fieldname_or_property_ID>)
Baptiste Assmanne279ca62020-10-27 18:10:06 +010016512 Returns value of <fieldname> found in input MQTT payload of type
16513 <packettype>.
16514 <packettype> can be either a string (case insensitive matching) or a numeric
16515 value corresponding to the type of packet we're supposed to extract data
16516 from.
16517 Supported string and integers can be found here:
16518 https://docs.oasis-open.org/mqtt/mqtt/v3.1.1/os/mqtt-v3.1.1-os.html#_Toc398718021
16519 https://docs.oasis-open.org/mqtt/mqtt/v5.0/os/mqtt-v5.0-os.html#_Toc3901022
16520
16521 <fieldname> depends on <packettype> and can be any of the following below.
16522 (note that <fieldname> matching is case insensitive).
16523 <property id> can only be found in MQTT v5.0 streams. check this table:
16524 https://docs.oasis-open.org/mqtt/mqtt/v5.0/os/mqtt-v5.0-os.html#_Toc3901029
16525
16526 - CONNECT (or 1): flags, protocol_name, protocol_version, client_identifier,
16527 will_topic, will_payload, username, password, keepalive
16528 OR any property ID as a numeric value (for MQTT v5.0
16529 packets only):
16530 17: Session Expiry Interval
16531 33: Receive Maximum
16532 39: Maximum Packet Size
16533 34: Topic Alias Maximum
16534 25: Request Response Information
16535 23: Request Problem Information
16536 21: Authentication Method
16537 22: Authentication Data
16538 18: Will Delay Interval
16539 1: Payload Format Indicator
16540 2: Message Expiry Interval
16541 3: Content Type
16542 8: Response Topic
16543 9: Correlation Data
16544 Not supported yet:
16545 38: User Property
16546
16547 - CONNACK (or 2): flags, protocol_version, reason_code
16548 OR any property ID as a numeric value (for MQTT v5.0
16549 packets only):
16550 17: Session Expiry Interval
16551 33: Receive Maximum
16552 36: Maximum QoS
16553 37: Retain Available
16554 39: Maximum Packet Size
16555 18: Assigned Client Identifier
16556 34: Topic Alias Maximum
16557 31: Reason String
16558 40; Wildcard Subscription Available
16559 41: Subscription Identifiers Available
16560 42: Shared Subscription Available
16561 19: Server Keep Alive
16562 26: Response Information
16563 28: Server Reference
16564 21: Authentication Method
16565 22: Authentication Data
16566 Not supported yet:
16567 38: User Property
16568
16569 Due to current HAProxy design, only the first message sent by the client and
16570 the server can be parsed. Thus this converter can extract data only from
16571 CONNECT and CONNACK packet types. CONNECT is the first message sent by the
16572 client and CONNACK is the first response sent by the server.
16573
16574 Example:
16575
16576 acl data_in_buffer req.len ge 4
16577 tcp-request content set-var(txn.username) \
16578 req.payload(0,0),mqtt_field_value(connect,protocol_name) \
16579 if data_in_buffer
16580 # do the same as above
16581 tcp-request content set-var(txn.username) \
16582 req.payload(0,0),mqtt_field_value(1,protocol_name) \
16583 if data_in_buffer
16584
16585mqtt_is_valid
16586 Checks that the binary input is a valid MQTT packet. It returns a boolean.
16587
16588 Due to current HAProxy design, only the first message sent by the client and
16589 the server can be parsed. Thus this converter can extract data only from
16590 CONNECT and CONNACK packet types. CONNECT is the first message sent by the
16591 client and CONNACK is the first response sent by the server.
16592
16593 Example:
16594
16595 acl data_in_buffer req.len ge 4
Daniel Corbettcc9d9b02021-05-13 10:46:07 -040016596 tcp-request content reject unless { req.payload(0,0),mqtt_is_valid }
Baptiste Assmanne279ca62020-10-27 18:10:06 +010016597
Willy Tarreau97707872015-01-27 15:12:13 +010016598mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016599 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020016600 the product as an signed integer. In case of overflow, the largest possible
16601 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016602 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010016603 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016604 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016605 "sess" : the variable is shared with the whole session
16606 "txn" : the variable is shared with the transaction (request and response)
16607 "req" : the variable is shared only during request processing
16608 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016609 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016610 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016611
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010016612nbsrv
16613 Takes an input value of type string, interprets it as a backend name and
16614 returns the number of usable servers in that backend. Can be used in places
16615 where we want to look up a backend from a dynamic name, like a result of a
16616 map lookup.
16617
Willy Tarreau97707872015-01-27 15:12:13 +010016618neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016619 Takes the input value of type signed integer, computes the opposite value,
16620 and returns the remainder as an signed integer. 0 is identity. This operator
16621 is provided for reversed subtracts : in order to subtract the input from a
16622 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010016623
16624not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016625 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010016626 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016627 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010016628 absence of a flag).
16629
16630odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016631 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010016632 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
16633
16634or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016635 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016636 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010016637 numeric value or a variable name. The name of the variable starts with an
16638 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016639 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016640 "sess" : the variable is shared with the whole session
16641 "txn" : the variable is shared with the transaction (request and response)
16642 "req" : the variable is shared only during request processing
16643 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010016644 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016645 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016646
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010016647protobuf(<field_number>,[<field_type>])
16648 This extracts the protocol buffers message field in raw mode of an input binary
16649 sample representation of a protocol buffer message with <field_number> as field
16650 number (dotted notation) if <field_type> is not present, or as an integer sample
16651 if this field is present (see also "ungrpc" below).
16652 The list of the authorized types is the following one: "int32", "int64", "uint32",
16653 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
16654 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
16655 "float" for the wire type 5. Note that "string" is considered as a length-delimited
16656 type, so it does not require any <field_type> argument to be extracted.
16657 More information may be found here about the protocol buffers message field types:
16658 https://developers.google.com/protocol-buffers/docs/encoding
16659
Willy Tarreauc4dc3502015-01-23 20:39:28 +010016660regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010016661 Applies a regex-based substitution to the input string. It does the same
16662 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
16663 default it will replace in the input string the first occurrence of the
16664 largest part matching the regular expression <regex> with the substitution
16665 string <subst>. It is possible to replace all occurrences instead by adding
16666 the flag "g" in the third argument <flags>. It is also possible to make the
16667 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
16668 string, it is made up from the concatenation of all desired flags. Thus if
16669 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
Willy Tarreauef21fac2020-02-14 13:37:20 +010016670 The first use of this converter is to replace certain characters or sequence
16671 of characters with other ones.
16672
16673 It is highly recommended to enclose the regex part using protected quotes to
16674 improve clarity and never have a closing parenthesis from the regex mixed up
16675 with the parenthesis from the function. Just like in Bourne shell, the first
16676 level of quotes is processed when delimiting word groups on the line, a
16677 second level is usable for argument. It is recommended to use single quotes
16678 outside since these ones do not try to resolve backslashes nor dollar signs.
Willy Tarreau7eda8492015-01-20 19:47:06 +010016679
Willy Tarreaucd0d2ed2020-02-14 17:33:06 +010016680 Examples:
Willy Tarreau7eda8492015-01-20 19:47:06 +010016681
16682 # de-duplicate "/" in header "x-path".
16683 # input: x-path: /////a///b/c/xzxyz/
16684 # output: x-path: /a/b/c/xzxyz/
Willy Tarreauef21fac2020-02-14 13:37:20 +010016685 http-request set-header x-path "%[hdr(x-path),regsub('/+','/','g')]"
Willy Tarreau7eda8492015-01-20 19:47:06 +010016686
Willy Tarreaucd0d2ed2020-02-14 17:33:06 +010016687 # copy query string to x-query and drop all leading '?', ';' and '&'
16688 http-request set-header x-query "%[query,regsub([?;&]*,'')]"
16689
Jerome Magnin07e1e3c2020-02-16 19:20:19 +010016690 # capture groups and backreferences
16691 # both lines do the same.
Willy Tarreau465dc7d2020-10-08 18:05:56 +020016692 http-request redirect location %[url,'regsub("(foo|bar)([0-9]+)?","\2\1",i)']
Jerome Magnin07e1e3c2020-02-16 19:20:19 +010016693 http-request redirect location %[url,regsub(\"(foo|bar)([0-9]+)?\",\"\2\1\",i)]
16694
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020016695capture-req(<id>)
16696 Capture the string entry in the request slot <id> and returns the entry as
16697 is. If the slot doesn't exist, the capture fails silently.
16698
16699 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020016700 "http-response capture", "capture.req.hdr" and
16701 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020016702
16703capture-res(<id>)
16704 Capture the string entry in the response slot <id> and returns the entry as
16705 is. If the slot doesn't exist, the capture fails silently.
16706
16707 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020016708 "http-response capture", "capture.req.hdr" and
16709 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020016710
Christopher Faulet568415a2020-04-01 17:24:47 +020016711rtrim(<chars>)
16712 Skips any characters from <chars> from the end of the string representation
16713 of the input sample.
16714
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016715sdbm([<avalanche>])
16716 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
16717 hash function. Optionally, it is possible to apply a full avalanche hash
16718 function to the output if the optional <avalanche> argument equals 1. This
16719 converter uses the same functions as used by the various hash-based load
16720 balancing algorithms, so it will provide exactly the same results. It is
16721 mostly intended for debugging, but can be used as a stick-table entry to
16722 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010016723 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
16724 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016725
Tim Duesterhusf38175c2020-06-09 11:48:42 +020016726secure_memcmp(<var>)
16727 Compares the contents of <var> with the input value. Both values are treated
16728 as a binary string. Returns a boolean indicating whether both binary strings
16729 match.
16730
16731 If both binary strings have the same length then the comparison will be
16732 performed in constant time.
16733
Daniel Corbett9f0843f2021-05-08 10:50:37 -040016734 Please note that this converter is only available when HAProxy has been
Tim Duesterhusf38175c2020-06-09 11:48:42 +020016735 compiled with USE_OPENSSL.
16736
16737 Example :
16738
16739 http-request set-var(txn.token) hdr(token)
16740 # Check whether the token sent by the client matches the secret token
16741 # value, without leaking the contents using a timing attack.
16742 acl token_given str(my_secret_token),secure_memcmp(txn.token)
16743
Tim Duesterhusef4e45c2021-01-21 17:40:50 +010016744set-var(<var>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010016745 Sets a variable with the input content and returns the content on the output
16746 as-is. The variable keeps the value and the associated input type. The name of
16747 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016748 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016749 "sess" : the variable is shared with the whole session
16750 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020016751 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010016752 "req" : the variable is shared only during request processing,
16753 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020016754 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016755 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020016756
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020016757sha1
Tim Duesterhusd4376302019-06-17 12:41:44 +020016758 Converts a binary input sample to a SHA-1 digest. The result is a binary
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020016759 sample with length of 20 bytes.
16760
Tim Duesterhusd4376302019-06-17 12:41:44 +020016761sha2([<bits>])
16762 Converts a binary input sample to a digest in the SHA-2 family. The result
16763 is a binary sample with length of <bits>/8 bytes.
16764
16765 Valid values for <bits> are 224, 256, 384, 512, each corresponding to
16766 SHA-<bits>. The default value is 256.
16767
Daniel Corbett9f0843f2021-05-08 10:50:37 -040016768 Please note that this converter is only available when HAProxy has been
Tim Duesterhusd4376302019-06-17 12:41:44 +020016769 compiled with USE_OPENSSL.
16770
Nenad Merdanovic177adc92019-08-27 01:58:13 +020016771srv_queue
16772 Takes an input value of type string, either a server name or <backend>/<server>
16773 format and returns the number of queued sessions on that server. Can be used
16774 in places where we want to look up queued sessions from a dynamic name, like a
16775 cookie value (e.g. req.cook(SRVID),srv_queue) and then make a decision to break
16776 persistence or direct a request elsewhere.
16777
Tim Duesterhusca097c12018-04-27 21:18:45 +020016778strcmp(<var>)
16779 Compares the contents of <var> with the input value of type string. Returns
16780 the result as a signed integer compatible with strcmp(3): 0 if both strings
16781 are identical. A value less than 0 if the left string is lexicographically
16782 smaller than the right string or if the left string is shorter. A value greater
16783 than 0 otherwise (right string greater than left string or the right string is
16784 shorter).
16785
Tim Duesterhusf38175c2020-06-09 11:48:42 +020016786 See also the secure_memcmp converter if you need to compare two binary
16787 strings in constant time.
16788
Tim Duesterhusca097c12018-04-27 21:18:45 +020016789 Example :
16790
16791 http-request set-var(txn.host) hdr(host)
16792 # Check whether the client is attempting domain fronting.
16793 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
16794
16795
Willy Tarreau97707872015-01-27 15:12:13 +010016796sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020016797 Subtracts <value> from the input value of type signed integer, and returns
16798 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016799 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010016800 or a variable name. The name of the variable starts with an indication about
16801 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010016802 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010016803 "sess" : the variable is shared with the whole session
16804 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016805 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010016806 "req" : the variable is shared only during request processing,
16807 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020016808 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010016809 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010016810
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016811table_bytes_in_rate(<table>)
16812 Uses the string representation of the input sample to perform a look up in
16813 the specified table. If the key is not found in the table, integer value zero
16814 is returned. Otherwise the converter returns the average client-to-server
16815 bytes rate associated with the input sample in the designated table, measured
16816 in amount of bytes over the period configured in the table. See also the
16817 sc_bytes_in_rate sample fetch keyword.
16818
16819
16820table_bytes_out_rate(<table>)
16821 Uses the string representation of the input sample to perform a look up in
16822 the specified table. If the key is not found in the table, integer value zero
16823 is returned. Otherwise the converter returns the average server-to-client
16824 bytes rate associated with the input sample in the designated table, measured
16825 in amount of bytes over the period configured in the table. See also the
16826 sc_bytes_out_rate sample fetch keyword.
16827
16828table_conn_cnt(<table>)
16829 Uses the string representation of the input sample to perform a look up in
16830 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016831 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016832 connections associated with the input sample in the designated table. See
16833 also the sc_conn_cnt sample fetch keyword.
16834
16835table_conn_cur(<table>)
16836 Uses the string representation of the input sample to perform a look up in
16837 the specified table. If the key is not found in the table, integer value zero
16838 is returned. Otherwise the converter returns the current amount of concurrent
16839 tracked connections associated with the input sample in the designated table.
16840 See also the sc_conn_cur sample fetch keyword.
16841
16842table_conn_rate(<table>)
16843 Uses the string representation of the input sample to perform a look up in
16844 the specified table. If the key is not found in the table, integer value zero
16845 is returned. Otherwise the converter returns the average incoming connection
16846 rate associated with the input sample in the designated table. See also the
16847 sc_conn_rate sample fetch keyword.
16848
Thierry FOURNIER236657b2015-08-19 08:25:14 +020016849table_gpt0(<table>)
16850 Uses the string representation of the input sample to perform a look up in
16851 the specified table. If the key is not found in the table, boolean value zero
16852 is returned. Otherwise the converter returns the current value of the first
16853 general purpose tag associated with the input sample in the designated table.
16854 See also the sc_get_gpt0 sample fetch keyword.
16855
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016856table_gpc0(<table>)
16857 Uses the string representation of the input sample to perform a look up in
16858 the specified table. If the key is not found in the table, integer value zero
16859 is returned. Otherwise the converter returns the current value of the first
16860 general purpose counter associated with the input sample in the designated
16861 table. See also the sc_get_gpc0 sample fetch keyword.
16862
16863table_gpc0_rate(<table>)
16864 Uses the string representation of the input sample to perform a look up in
16865 the specified table. If the key is not found in the table, integer value zero
16866 is returned. Otherwise the converter returns the frequency which the gpc0
16867 counter was incremented over the configured period in the table, associated
16868 with the input sample in the designated table. See also the sc_get_gpc0_rate
16869 sample fetch keyword.
16870
Frédéric Lécaille6778b272018-01-29 15:22:53 +010016871table_gpc1(<table>)
16872 Uses the string representation of the input sample to perform a look up in
16873 the specified table. If the key is not found in the table, integer value zero
16874 is returned. Otherwise the converter returns the current value of the second
16875 general purpose counter associated with the input sample in the designated
16876 table. See also the sc_get_gpc1 sample fetch keyword.
16877
16878table_gpc1_rate(<table>)
16879 Uses the string representation of the input sample to perform a look up in
16880 the specified table. If the key is not found in the table, integer value zero
16881 is returned. Otherwise the converter returns the frequency which the gpc1
16882 counter was incremented over the configured period in the table, associated
16883 with the input sample in the designated table. See also the sc_get_gpc1_rate
16884 sample fetch keyword.
16885
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016886table_http_err_cnt(<table>)
16887 Uses the string representation of the input sample to perform a look up in
16888 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016889 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016890 errors associated with the input sample in the designated table. See also the
16891 sc_http_err_cnt sample fetch keyword.
16892
16893table_http_err_rate(<table>)
16894 Uses the string representation of the input sample to perform a look up in
16895 the specified table. If the key is not found in the table, integer value zero
16896 is returned. Otherwise the average rate of HTTP errors associated with the
16897 input sample in the designated table, measured in amount of errors over the
16898 period configured in the table. See also the sc_http_err_rate sample fetch
16899 keyword.
16900
Willy Tarreau826f3ab2021-02-10 12:07:15 +010016901table_http_fail_cnt(<table>)
16902 Uses the string representation of the input sample to perform a look up in
16903 the specified table. If the key is not found in the table, integer value zero
16904 is returned. Otherwise the converter returns the cumulative number of HTTP
16905 failures associated with the input sample in the designated table. See also
16906 the sc_http_fail_cnt sample fetch keyword.
16907
16908table_http_fail_rate(<table>)
16909 Uses the string representation of the input sample to perform a look up in
16910 the specified table. If the key is not found in the table, integer value zero
16911 is returned. Otherwise the average rate of HTTP failures associated with the
16912 input sample in the designated table, measured in amount of failures over the
16913 period configured in the table. See also the sc_http_fail_rate sample fetch
16914 keyword.
16915
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016916table_http_req_cnt(<table>)
16917 Uses the string representation of the input sample to perform a look up in
16918 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016919 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016920 requests associated with the input sample in the designated table. See also
16921 the sc_http_req_cnt sample fetch keyword.
16922
16923table_http_req_rate(<table>)
16924 Uses the string representation of the input sample to perform a look up in
16925 the specified table. If the key is not found in the table, integer value zero
16926 is returned. Otherwise the average rate of HTTP requests associated with the
16927 input sample in the designated table, measured in amount of requests over the
16928 period configured in the table. See also the sc_http_req_rate sample fetch
16929 keyword.
16930
16931table_kbytes_in(<table>)
16932 Uses the string representation of the input sample to perform a look up in
16933 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016934 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016935 to-server data associated with the input sample in the designated table,
16936 measured in kilobytes. The test is currently performed on 32-bit integers,
16937 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
16938 keyword.
16939
16940table_kbytes_out(<table>)
16941 Uses the string representation of the input sample to perform a look up in
16942 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016943 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016944 to-client data associated with the input sample in the designated table,
16945 measured in kilobytes. The test is currently performed on 32-bit integers,
16946 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
16947 keyword.
16948
16949table_server_id(<table>)
16950 Uses the string representation of the input sample to perform a look up in
16951 the specified table. If the key is not found in the table, integer value zero
16952 is returned. Otherwise the converter returns the server ID associated with
16953 the input sample in the designated table. A server ID is associated to a
16954 sample by a "stick" rule when a connection to a server succeeds. A server ID
16955 zero means that no server is associated with this key.
16956
16957table_sess_cnt(<table>)
16958 Uses the string representation of the input sample to perform a look up in
16959 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010016960 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020016961 sessions associated with the input sample in the designated table. Note that
16962 a session here refers to an incoming connection being accepted by the
16963 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
16964 keyword.
16965
16966table_sess_rate(<table>)
16967 Uses the string representation of the input sample to perform a look up in
16968 the specified table. If the key is not found in the table, integer value zero
16969 is returned. Otherwise the converter returns the average incoming session
16970 rate associated with the input sample in the designated table. Note that a
16971 session here refers to an incoming connection being accepted by the
16972 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
16973 keyword.
16974
16975table_trackers(<table>)
16976 Uses the string representation of the input sample to perform a look up in
16977 the specified table. If the key is not found in the table, integer value zero
16978 is returned. Otherwise the converter returns the current amount of concurrent
16979 connections tracking the same key as the input sample in the designated
16980 table. It differs from table_conn_cur in that it does not rely on any stored
16981 information but on the table's reference count (the "use" value which is
16982 returned by "show table" on the CLI). This may sometimes be more suited for
16983 layer7 tracking. It can be used to tell a server how many concurrent
16984 connections there are from a given address for example. See also the
16985 sc_trackers sample fetch keyword.
16986
Moemen MHEDHBI92f7d432021-04-01 20:53:59 +020016987ub64dec
16988 This converter is the base64url variant of b64dec converter. base64url
16989 encoding is the "URL and Filename Safe Alphabet" variant of base64 encoding.
16990 It is also the encoding used in JWT (JSON Web Token) standard.
16991
16992 Example:
16993 # Decoding a JWT payload:
16994 http-request set-var(txn.token_payload) req.hdr(Authorization),word(2,.),ub64dec
16995
16996ub64enc
16997 This converter is the base64url variant of base64 converter.
16998
Willy Tarreauffcb2e42014-07-10 16:29:08 +020016999upper
17000 Convert a string sample to upper case. This can only be placed after a string
17001 sample fetch function or after a transformation keyword returning a string
17002 type. The result is of type string.
17003
Willy Tarreau62ba9ba2020-04-23 17:54:47 +020017004url_dec([<in_form>])
17005 Takes an url-encoded string provided as input and returns the decoded version
17006 as output. The input and the output are of type string. If the <in_form>
17007 argument is set to a non-zero integer value, the input string is assumed to
17008 be part of a form or query string and the '+' character will be turned into a
17009 space (' '). Otherwise this will only happen after a question mark indicating
17010 a query string ('?').
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020017011
William Dauchy888b0ae2021-01-06 23:39:50 +010017012url_enc([<enc_type>])
17013 Takes a string provided as input and returns the encoded version as output.
17014 The input and the output are of type string. By default the type of encoding
17015 is meant for `query` type. There is no other type supported for now but the
17016 optional argument is here for future changes.
17017
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017018ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010017019 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010017020 sample representation of a gRPC message with <field_number> as field number
17021 (dotted notation) if <field_type> is not present, or as an integer sample if this
17022 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017023 The list of the authorized types is the following one: "int32", "int64", "uint32",
17024 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
17025 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
17026 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010017027 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017028 More information may be found here about the protocol buffers message field types:
17029 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010017030
17031 Example:
17032 // with such a protocol buffer .proto file content adapted from
17033 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
17034
17035 message Point {
17036 int32 latitude = 1;
17037 int32 longitude = 2;
17038 }
17039
17040 message PPoint {
17041 Point point = 59;
17042 }
17043
17044 message Rectangle {
17045 // One corner of the rectangle.
17046 PPoint lo = 48;
17047 // The other corner of the rectangle.
17048 PPoint hi = 49;
17049 }
17050
Peter Gervaidf4c9d22020-06-11 18:05:11 +020017051 let's say a body request is made of a "Rectangle" object value (two PPoint
17052 protocol buffers messages), the four protocol buffers fields could be
17053 extracted with these "ungrpc" directives:
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010017054
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017055 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
17056 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
John Roeslerfb2fce12019-07-10 15:45:51 -050017057 req.body,ungrpc(49.59.1,int32) # "latitude" of "hi" second PPoint
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017058 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
17059
Peter Gervaidf4c9d22020-06-11 18:05:11 +020017060 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017061
Frédéric Lécaille93d33162019-03-06 09:35:59 +010017062 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010017063
Peter Gervaidf4c9d22020-06-11 18:05:11 +020017064 As a gRPC message is always made of a gRPC header followed by protocol buffers
17065 messages, in the previous example the "latitude" of "lo" first PPoint
17066 could be extracted with these equivalent directives:
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010017067
17068 req.body,ungrpc(48.59),protobuf(1,int32)
17069 req.body,ungrpc(48),protobuf(59.1,int32)
17070 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
17071
Peter Gervaidf4c9d22020-06-11 18:05:11 +020017072 Note that the first convert must be "ungrpc", the remaining ones must be
17073 "protobuf" and only the last one may have or not a second argument to
17074 interpret the previous binary sample.
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010017075
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010017076
Tim Duesterhusef4e45c2021-01-21 17:40:50 +010017077unset-var(<var>)
Christopher Faulet85d79c92016-11-09 16:54:56 +010017078 Unsets a variable if the input content is defined. The name of the variable
17079 starts with an indication about its scope. The scopes allowed are:
17080 "proc" : the variable is shared with the whole process
17081 "sess" : the variable is shared with the whole session
17082 "txn" : the variable is shared with the transaction (request and
17083 response),
17084 "req" : the variable is shared only during request processing,
17085 "res" : the variable is shared only during response processing.
17086 This prefix is followed by a name. The separator is a '.'. The name may only
17087 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
17088
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020017089utime(<format>[,<offset>])
17090 Converts an integer supposed to contain a date since epoch to a string
17091 representing this date in UTC time using a format defined by the <format>
17092 string using strftime(3). The purpose is to allow any date format to be used
17093 in logs. An optional <offset> in seconds may be applied to the input date
17094 (positive or negative). See the strftime() man page for the format supported
17095 by your operating system. See also the ltime converter.
17096
17097 Example :
17098
17099 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010017100 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020017101 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
17102
Marcin Deranek9631a282018-04-16 14:30:46 +020017103word(<index>,<delimiters>[,<count>])
17104 Extracts the nth word counting from the beginning (positive index) or from
17105 the end (negative index) considering given delimiters from an input string.
17106 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
Jerome Magnin88209322020-01-28 13:33:44 +010017107 Delimiters at the beginning or end of the input string are ignored.
Marcin Deranek9631a282018-04-16 14:30:46 +020017108 Optionally you can specify <count> of words to extract (default: 1).
17109 Value of 0 indicates extraction of all remaining words.
17110
17111 Example :
17112 str(f1_f2_f3__f5),word(4,_) # f5
17113 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
17114 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
17115 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
17116 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Jerome Magnin88209322020-01-28 13:33:44 +010017117 str(/f1/f2/f3/f4),word(1,/) # f1
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010017118
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020017119wt6([<avalanche>])
17120 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
17121 hash function. Optionally, it is possible to apply a full avalanche hash
17122 function to the output if the optional <avalanche> argument equals 1. This
17123 converter uses the same functions as used by the various hash-based load
17124 balancing algorithms, so it will provide exactly the same results. It is
17125 mostly intended for debugging, but can be used as a stick-table entry to
17126 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010017127 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
17128 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020017129
Willy Tarreau97707872015-01-27 15:12:13 +010017130xor(<value>)
17131 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020017132 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020017133 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010017134 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010017135 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010017136 "sess" : the variable is shared with the whole session
17137 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020017138 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010017139 "req" : the variable is shared only during request processing,
17140 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020017141 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010017142 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010017143
Dragan Dosen04bf0cc2020-12-22 21:44:33 +010017144xxh3([<seed>])
17145 Hashes a binary input sample into a signed 64-bit quantity using the XXH3
17146 64-bit variant of the XXhash hash function. This hash supports a seed which
17147 defaults to zero but a different value maybe passed as the <seed> argument.
17148 This hash is known to be very good and very fast so it can be used to hash
17149 URLs and/or URL parameters for use as stick-table keys to collect statistics
17150 with a low collision rate, though care must be taken as the algorithm is not
17151 considered as cryptographically secure.
17152
Thierry FOURNIER01e09742016-12-26 11:46:11 +010017153xxh32([<seed>])
17154 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
17155 variant of the XXHash hash function. This hash supports a seed which defaults
17156 to zero but a different value maybe passed as the <seed> argument. This hash
17157 is known to be very good and very fast so it can be used to hash URLs and/or
17158 URL parameters for use as stick-table keys to collect statistics with a low
17159 collision rate, though care must be taken as the algorithm is not considered
17160 as cryptographically secure.
17161
17162xxh64([<seed>])
17163 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
17164 variant of the XXHash hash function. This hash supports a seed which defaults
17165 to zero but a different value maybe passed as the <seed> argument. This hash
17166 is known to be very good and very fast so it can be used to hash URLs and/or
17167 URL parameters for use as stick-table keys to collect statistics with a low
17168 collision rate, though care must be taken as the algorithm is not considered
17169 as cryptographically secure.
17170
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010017171
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200171727.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020017173--------------------------------------------
17174
17175A first set of sample fetch methods applies to internal information which does
17176not even relate to any client information. These ones are sometimes used with
17177"monitor-fail" directives to report an internal status to external watchers.
17178The sample fetch methods described in this section are usable anywhere.
17179
17180always_false : boolean
17181 Always returns the boolean "false" value. It may be used with ACLs as a
17182 temporary replacement for another one when adjusting configurations.
17183
17184always_true : boolean
17185 Always returns the boolean "true" value. It may be used with ACLs as a
17186 temporary replacement for another one when adjusting configurations.
17187
17188avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010017189 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020017190 divided by the number of active servers. The current backend is used if no
17191 backend is specified. This is very similar to "queue" except that the size of
17192 the farm is considered, in order to give a more accurate measurement of the
17193 time it may take for a new connection to be processed. The main usage is with
17194 ACL to return a sorry page to new users when it becomes certain they will get
17195 a degraded service, or to pass to the backend servers in a header so that
17196 they decide to work in degraded mode or to disable some functions to speed up
17197 the processing a bit. Note that in the event there would not be any active
17198 server anymore, twice the number of queued connections would be considered as
17199 the measured value. This is a fair estimate, as we expect one server to get
17200 back soon anyway, but we still prefer to send new traffic to another backend
17201 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
17202 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010017203
Willy Tarreau74ca5042013-06-11 23:12:07 +020017204be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020017205 Applies to the number of currently established connections on the backend,
17206 possibly including the connection being evaluated. If no backend name is
17207 specified, the current one is used. But it is also possible to check another
17208 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040017209 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
17210
17211be_conn_free([<backend>]) : integer
17212 Returns an integer value corresponding to the number of available connections
17213 across available servers in the backend. Queue slots are not included. Backup
17214 servers are also not included, unless all other servers are down. If no
17215 backend name is specified, the current one is used. But it is also possible
17216 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040017217 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
17218 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040017219
17220 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
17221 (meaning unlimited), then this fetch clearly does not make sense, in which
17222 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020017223
Willy Tarreau74ca5042013-06-11 23:12:07 +020017224be_sess_rate([<backend>]) : integer
17225 Returns an integer value corresponding to the sessions creation rate on the
17226 backend, in number of new sessions per second. This is used with ACLs to
17227 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010017228 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020017229 online dictionary). It can also be useful to add this element to logs using a
17230 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010017231
17232 Example :
17233 # Redirect to an error page if the dictionary is requested too often
17234 backend dynamic
17235 mode http
17236 acl being_scanned be_sess_rate gt 100
17237 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010017238
Davor Ocelice9ed2812017-12-25 17:49:28 +010017239bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020017240 Returns a binary chain. The input is the hexadecimal representation
17241 of the string.
17242
17243bool(<bool>) : bool
17244 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
17245 'false' and '0' are the same. 'true' and '1' are the same.
17246
Willy Tarreau74ca5042013-06-11 23:12:07 +020017247connslots([<backend>]) : integer
17248 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017249 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020017250 connections on all servers and the maximum queue size. This is probably only
17251 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050017252
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080017253 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020017254 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080017255 usage; see "use_backend" keyword) can be redirected to a different backend.
17256
Willy Tarreau55165fe2009-05-10 12:02:55 +020017257 'connslots' = number of available server connection slots, + number of
17258 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080017259
Willy Tarreaua36af912009-10-10 12:02:45 +020017260 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020017261 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020017262 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020017263 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010017264 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020017265 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020017266 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080017267
Willy Tarreau55165fe2009-05-10 12:02:55 +020017268 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
17269 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020017270 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020017271 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080017272
Willy Tarreau70fe9442018-11-22 16:07:39 +010017273cpu_calls : integer
17274 Returns the number of calls to the task processing the stream or current
17275 request since it was allocated. This number is reset for each new request on
17276 the same connections in case of HTTP keep-alive. This value should usually be
17277 low and stable (around 2 calls for a typically simple request) but may become
17278 high if some processing (compression, caching or analysis) is performed. This
17279 is purely for performance monitoring purposes.
17280
17281cpu_ns_avg : integer
17282 Returns the average number of nanoseconds spent in each call to the task
17283 processing the stream or current request. This number is reset for each new
17284 request on the same connections in case of HTTP keep-alive. This value
17285 indicates the overall cost of processing the request or the connection for
17286 each call. There is no good nor bad value but the time spent in a call
17287 automatically causes latency for other processing (see lat_ns_avg below),
17288 and may affect other connection's apparent response time. Certain operations
17289 like compression, complex regex matching or heavy Lua operations may directly
17290 affect this value, and having it in the logs will make it easier to spot the
17291 faulty processing that needs to be fixed to recover decent performance.
17292 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
17293
17294cpu_ns_tot : integer
17295 Returns the total number of nanoseconds spent in each call to the task
17296 processing the stream or current request. This number is reset for each new
17297 request on the same connections in case of HTTP keep-alive. This value
17298 indicates the overall cost of processing the request or the connection for
17299 each call. There is no good nor bad value but the time spent in a call
17300 automatically causes latency for other processing (see lat_ns_avg below),
17301 induces CPU costs on the machine, and may affect other connection's apparent
17302 response time. Certain operations like compression, complex regex matching or
17303 heavy Lua operations may directly affect this value, and having it in the
17304 logs will make it easier to spot the faulty processing that needs to be fixed
17305 to recover decent performance. The value may be artificially high due to a
17306 high cpu_calls count, for example when processing many HTTP chunks, and for
17307 this reason it is often preferred to log cpu_ns_avg instead.
17308
Cyril Bonté6bcd1822019-11-05 23:13:59 +010017309date([<offset>],[<unit>]) : integer
Willy Tarreau6236d3a2013-07-25 14:28:25 +020017310 Returns the current date as the epoch (number of seconds since 01/01/1970).
Damien Claisseae6f1252019-10-30 15:57:28 +000017311
17312 If an offset value is specified, then it is added to the current date before
17313 returning the value. This is particularly useful to compute relative dates,
17314 as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020017315 It is useful combined with the http_date converter.
17316
Damien Claisseae6f1252019-10-30 15:57:28 +000017317 <unit> is facultative, and can be set to "s" for seconds (default behavior),
17318 "ms" for milliseconds or "us" for microseconds.
17319 If unit is set, return value is an integer reflecting either seconds,
17320 milliseconds or microseconds since epoch, plus offset.
17321 It is useful when a time resolution of less than a second is needed.
17322
Willy Tarreau276fae92013-07-25 14:36:01 +020017323 Example :
17324
17325 # set an expires header to now+1 hour in every response
17326 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020017327
Damien Claisseae6f1252019-10-30 15:57:28 +000017328 # set an expires header to now+1 hour in every response, with
17329 # millisecond granularity
17330 http-response set-header Expires %[date(3600000,ms),http_date(0,ms)]
17331
Etienne Carrierea792a0a2018-01-17 13:43:24 +010017332date_us : integer
17333 Return the microseconds part of the date (the "second" part is returned by
17334 date sample). This sample is coherent with the date sample as it is comes
17335 from the same timeval structure.
17336
Willy Tarreaud716f9b2017-10-13 11:03:15 +020017337distcc_body(<token>[,<occ>]) : binary
17338 Parses a distcc message and returns the body associated to occurrence #<occ>
17339 of the token <token>. Occurrences start at 1, and when unspecified, any may
17340 match though in practice only the first one is checked for now. This can be
17341 used to extract file names or arguments in files built using distcc through
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017342 HAProxy. Please refer to distcc's protocol documentation for the complete
Willy Tarreaud716f9b2017-10-13 11:03:15 +020017343 list of supported tokens.
17344
17345distcc_param(<token>[,<occ>]) : integer
17346 Parses a distcc message and returns the parameter associated to occurrence
17347 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
17348 any may match though in practice only the first one is checked for now. This
17349 can be used to extract certain information such as the protocol version, the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017350 file size or the argument in files built using distcc through HAProxy.
Willy Tarreaud716f9b2017-10-13 11:03:15 +020017351 Another use case consists in waiting for the start of the preprocessed file
17352 contents before connecting to the server to avoid keeping idle connections.
17353 Please refer to distcc's protocol documentation for the complete list of
17354 supported tokens.
17355
17356 Example :
17357 # wait up to 20s for the pre-processed file to be uploaded
17358 tcp-request inspect-delay 20s
17359 tcp-request content accept if { distcc_param(DOTI) -m found }
17360 # send large files to the big farm
17361 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
17362
Willy Tarreau595ec542013-06-12 21:34:28 +020017363env(<name>) : string
17364 Returns a string containing the value of environment variable <name>. As a
17365 reminder, environment variables are per-process and are sampled when the
17366 process starts. This can be useful to pass some information to a next hop
17367 server, or with ACLs to take specific action when the process is started a
17368 certain way.
17369
17370 Examples :
17371 # Pass the Via header to next hop with the local hostname in it
17372 http-request add-header Via 1.1\ %[env(HOSTNAME)]
17373
17374 # reject cookie-less requests when the STOP environment variable is set
17375 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
17376
Willy Tarreau74ca5042013-06-11 23:12:07 +020017377fe_conn([<frontend>]) : integer
17378 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010017379 possibly including the connection being evaluated. If no frontend name is
17380 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020017381 frontend. It can be used to return a sorry page before hard-blocking, or to
17382 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010017383 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020017384 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
17385 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020017386
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020017387fe_req_rate([<frontend>]) : integer
17388 Returns an integer value corresponding to the number of HTTP requests per
17389 second sent to a frontend. This number can differ from "fe_sess_rate" in
17390 situations where client-side keep-alive is enabled.
17391
Willy Tarreau74ca5042013-06-11 23:12:07 +020017392fe_sess_rate([<frontend>]) : integer
17393 Returns an integer value corresponding to the sessions creation rate on the
17394 frontend, in number of new sessions per second. This is used with ACLs to
17395 limit the incoming session rate to an acceptable range in order to prevent
17396 abuse of service at the earliest moment, for example when combined with other
17397 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
17398 down below the limit. It can also be useful to add this element to logs using
17399 a log-format directive. See also the "rate-limit sessions" directive for use
17400 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010017401
17402 Example :
17403 # This frontend limits incoming mails to 10/s with a max of 100
17404 # concurrent connections. We accept any connection below 10/s, and
17405 # force excess clients to wait for 100 ms. Since clients are limited to
17406 # 100 max, there cannot be more than 10 incoming mails per second.
17407 frontend mail
17408 bind :25
17409 mode tcp
17410 maxconn 100
17411 acl too_fast fe_sess_rate ge 10
17412 tcp-request inspect-delay 100ms
17413 tcp-request content accept if ! too_fast
17414 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010017415
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010017416hostname : string
17417 Returns the system hostname.
17418
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020017419int(<integer>) : signed integer
17420 Returns a signed integer.
17421
Thierry FOURNIERcc103292015-06-06 19:30:17 +020017422ipv4(<ipv4>) : ipv4
17423 Returns an ipv4.
17424
17425ipv6(<ipv6>) : ipv6
17426 Returns an ipv6.
17427
Willy Tarreau70fe9442018-11-22 16:07:39 +010017428lat_ns_avg : integer
17429 Returns the average number of nanoseconds spent between the moment the task
17430 handling the stream is woken up and the moment it is effectively called. This
17431 number is reset for each new request on the same connections in case of HTTP
17432 keep-alive. This value indicates the overall latency inflicted to the current
17433 request by all other requests being processed in parallel, and is a direct
17434 indicator of perceived performance due to noisy neighbours. In order to keep
17435 the value low, it is possible to reduce the scheduler's run queue depth using
17436 "tune.runqueue-depth", to reduce the number of concurrent events processed at
17437 once using "tune.maxpollevents", to decrease the stream's nice value using
Willy Tarreaue7723bd2020-06-24 11:11:02 +020017438 the "nice" option on the "bind" lines or in the frontend, to enable low
17439 latency scheduling using "tune.sched.low-latency", or to look for other heavy
17440 requests in logs (those exhibiting large values of "cpu_ns_avg"), whose
17441 processing needs to be adjusted or fixed. Compression of large buffers could
17442 be a culprit, like heavy regex or long lists of regex. Note: this value is
17443 exactly lat_ns_tot divided by cpu_calls.
Willy Tarreau70fe9442018-11-22 16:07:39 +010017444
17445lat_ns_tot : integer
17446 Returns the total number of nanoseconds spent between the moment the task
17447 handling the stream is woken up and the moment it is effectively called. This
17448 number is reset for each new request on the same connections in case of HTTP
17449 keep-alive. This value indicates the overall latency inflicted to the current
17450 request by all other requests being processed in parallel, and is a direct
17451 indicator of perceived performance due to noisy neighbours. In order to keep
17452 the value low, it is possible to reduce the scheduler's run queue depth using
17453 "tune.runqueue-depth", to reduce the number of concurrent events processed at
17454 once using "tune.maxpollevents", to decrease the stream's nice value using
Willy Tarreaue7723bd2020-06-24 11:11:02 +020017455 the "nice" option on the "bind" lines or in the frontend, to enable low
17456 latency scheduling using "tune.sched.low-latency", or to look for other heavy
17457 requests in logs (those exhibiting large values of "cpu_ns_avg"), whose
17458 processing needs to be adjusted or fixed. Compression of large buffers could
17459 be a culprit, like heavy regex or long lists of regex. Note: while it
Willy Tarreau70fe9442018-11-22 16:07:39 +010017460 may intuitively seem that the total latency adds to a transfer time, it is
17461 almost never true because while a task waits for the CPU, network buffers
17462 continue to fill up and the next call will process more at once. The value
17463 may be artificially high due to a high cpu_calls count, for example when
17464 processing many HTTP chunks, and for this reason it is often preferred to log
17465 lat_ns_avg instead, which is a more relevant performance indicator.
17466
Thierry FOURNIERcc103292015-06-06 19:30:17 +020017467meth(<method>) : method
17468 Returns a method.
17469
Willy Tarreau0f30d262014-11-24 16:02:05 +010017470nbproc : integer
17471 Returns an integer value corresponding to the number of processes that were
17472 started (it equals the global "nbproc" setting). This is useful for logging
17473 and debugging purposes.
17474
Willy Tarreau74ca5042013-06-11 23:12:07 +020017475nbsrv([<backend>]) : integer
17476 Returns an integer value corresponding to the number of usable servers of
17477 either the current backend or the named backend. This is mostly used with
17478 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010017479 switch to an alternate backend when the number of servers is too low to
17480 to handle some load. It is useful to report a failure when combined with
17481 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010017482
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040017483prio_class : integer
17484 Returns the priority class of the current session for http mode or connection
17485 for tcp mode. The value will be that set by the last call to "http-request
17486 set-priority-class" or "tcp-request content set-priority-class".
17487
17488prio_offset : integer
17489 Returns the priority offset of the current session for http mode or
17490 connection for tcp mode. The value will be that set by the last call to
17491 "http-request set-priority-offset" or "tcp-request content
17492 set-priority-offset".
17493
Willy Tarreau0f30d262014-11-24 16:02:05 +010017494proc : integer
17495 Returns an integer value corresponding to the position of the process calling
17496 the function, between 1 and global.nbproc. This is useful for logging and
17497 debugging purposes.
17498
Willy Tarreau74ca5042013-06-11 23:12:07 +020017499queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010017500 Returns the total number of queued connections of the designated backend,
17501 including all the connections in server queues. If no backend name is
17502 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020017503 one. This is useful with ACLs or to pass statistics to backend servers. This
17504 can be used to take actions when queuing goes above a known level, generally
17505 indicating a surge of traffic or a massive slowdown on the servers. One
17506 possible action could be to reject new users but still accept old ones. See
17507 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
17508
Willy Tarreau84310e22014-02-14 11:59:04 +010017509rand([<range>]) : integer
17510 Returns a random integer value within a range of <range> possible values,
17511 starting at zero. If the range is not specified, it defaults to 2^32, which
17512 gives numbers between 0 and 4294967295. It can be useful to pass some values
17513 needed to take some routing decisions for example, or just for debugging
17514 purposes. This random must not be used for security purposes.
17515
Willy Tarreau74ca5042013-06-11 23:12:07 +020017516srv_conn([<backend>/]<server>) : integer
17517 Returns an integer value corresponding to the number of currently established
17518 connections on the designated server, possibly including the connection being
17519 evaluated. If <backend> is omitted, then the server is looked up in the
17520 current backend. It can be used to use a specific farm when one server is
17521 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040017522 connections with it. See also the "fe_conn", "be_conn", "queue", and
17523 "srv_conn_free" fetch methods.
17524
17525srv_conn_free([<backend>/]<server>) : integer
17526 Returns an integer value corresponding to the number of available connections
17527 on the designated server, possibly including the connection being evaluated.
17528 The value does not include queue slots. If <backend> is omitted, then the
17529 server is looked up in the current backend. It can be used to use a specific
17530 farm when one server is full, or to inform the server about our view of the
17531 number of active connections with it. See also the "be_conn_free" and
17532 "srv_conn" fetch methods.
17533
17534 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
17535 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020017536
17537srv_is_up([<backend>/]<server>) : boolean
17538 Returns true when the designated server is UP, and false when it is either
17539 DOWN or in maintenance mode. If <backend> is omitted, then the server is
17540 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010017541 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020017542 availability). Another possible use which is more of a hack consists in
17543 using dummy servers as boolean variables that can be enabled or disabled from
17544 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
17545
Willy Tarreauff2b7af2017-10-13 11:46:26 +020017546srv_queue([<backend>/]<server>) : integer
17547 Returns an integer value corresponding to the number of connections currently
17548 pending in the designated server's queue. If <backend> is omitted, then the
17549 server is looked up in the current backend. It can sometimes be used together
17550 with the "use-server" directive to force to use a known faster server when it
17551 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
17552 fetch methods.
17553
Willy Tarreau74ca5042013-06-11 23:12:07 +020017554srv_sess_rate([<backend>/]<server>) : integer
17555 Returns an integer corresponding to the sessions creation rate on the
17556 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017557 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020017558 used with ACLs but can make sense with logs too. This is used to switch to an
17559 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010017560 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020017561 overloading servers).
17562
17563 Example :
17564 # Redirect to a separate back
17565 acl srv1_full srv_sess_rate(be1/srv1) gt 50
17566 acl srv2_full srv_sess_rate(be1/srv2) gt 50
17567 use_backend be2 if srv1_full or srv2_full
17568
Alexbf1bd5a2021-04-24 13:02:21 +020017569srv_iweight([<backend>/]<server>) : integer
Christopher Faulet1bea8652020-07-10 16:03:45 +020017570 Returns an integer corresponding to the server's initial weight. If <backend>
17571 is omitted, then the server is looked up in the current backend. See also
17572 "srv_weight" and "srv_uweight".
17573
Alexbf1bd5a2021-04-24 13:02:21 +020017574srv_uweight([<backend>/]<server>) : integer
Christopher Faulet1bea8652020-07-10 16:03:45 +020017575 Returns an integer corresponding to the user visible server's weight. If
17576 <backend> is omitted, then the server is looked up in the current
17577 backend. See also "srv_weight" and "srv_iweight".
17578
Alexbf1bd5a2021-04-24 13:02:21 +020017579srv_weight([<backend>/]<server>) : integer
Christopher Faulet1bea8652020-07-10 16:03:45 +020017580 Returns an integer corresponding to the current (or effective) server's
17581 weight. If <backend> is omitted, then the server is looked up in the current
17582 backend. See also "srv_iweight" and "srv_uweight".
17583
Willy Tarreau0f30d262014-11-24 16:02:05 +010017584stopping : boolean
17585 Returns TRUE if the process calling the function is currently stopping. This
17586 can be useful for logging, or for relaxing certain checks or helping close
17587 certain connections upon graceful shutdown.
17588
Thierry FOURNIERcc103292015-06-06 19:30:17 +020017589str(<string>) : string
17590 Returns a string.
17591
Willy Tarreau74ca5042013-06-11 23:12:07 +020017592table_avl([<table>]) : integer
17593 Returns the total number of available entries in the current proxy's
17594 stick-table or in the designated stick-table. See also table_cnt.
17595
17596table_cnt([<table>]) : integer
17597 Returns the total number of entries currently in use in the current proxy's
17598 stick-table or in the designated stick-table. See also src_conn_cnt and
17599 table_avl for other entry counting methods.
17600
Christopher Faulet34adb2a2017-11-21 21:45:38 +010017601thread : integer
17602 Returns an integer value corresponding to the position of the thread calling
17603 the function, between 0 and (global.nbthread-1). This is useful for logging
17604 and debugging purposes.
17605
Alexandar Lazica429ad32021-06-01 00:27:01 +020017606uuid([<version>]) : string
17607 Returns a UUID following the RFC4122 standard. If the version is not
17608 specified, a UUID version 4 (fully random) is returned.
17609 Currently, only version 4 is supported.
17610
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020017611var(<var-name>) : undefined
17612 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010017613 sample fetch fails. The name of the variable starts with an indication
17614 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010017615 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010017616 "sess" : the variable is shared with the whole session
17617 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020017618 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010017619 "req" : the variable is shared only during request processing,
17620 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020017621 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010017622 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020017623
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200176247.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020017625----------------------------------
17626
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017627The layer 4 usually describes just the transport layer which in HAProxy is
Willy Tarreau74ca5042013-06-11 23:12:07 +020017628closest to the connection, where no content is yet made available. The fetch
17629methods described here are usable as low as the "tcp-request connection" rule
17630sets unless they require some future information. Those generally include
17631TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020017632the incoming connection. For retrieving a value from a sticky counters, the
17633counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020017634"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
17635used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
17636can be specified as the first integer argument when using the "sc_" prefix.
17637Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
17638table may be specified with the "sc*" form, in which case the currently
17639tracked key will be looked up into this alternate table instead of the table
17640currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020017641
Christopher Faulet7d081f02021-04-15 09:38:37 +020017642bc_dst : ip
17643 This is the destination ip address of the connection on the server side,
17644 which is the server address HAProxy connected to. It is of type IP and works
17645 on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 address is mapped to its
17646 IPv6 equivalent, according to RFC 4291.
17647
17648bc_dst_port : integer
17649 Returns an integer value corresponding to the destination TCP port of the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017650 connection on the server side, which is the port HAProxy connected to.
Christopher Faulet7d081f02021-04-15 09:38:37 +020017651
Jérôme Magnin35e53a62019-01-16 14:38:37 +010017652bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010017653 Returns the backend connection's HTTP major version encoding, which may be 1
17654 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
17655 encoding and not the version present in the request header.
17656
Christopher Faulet7d081f02021-04-15 09:38:37 +020017657bc_src : ip
17658 This is the source ip address of the connection on the server side, which is
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017659 the server address HAProxy connected from. It is of type IP and works on both
Christopher Faulet7d081f02021-04-15 09:38:37 +020017660 IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are mapped to their IPv6
17661 equivalent, according to RFC 4291.
17662
17663bc_src_port : integer
17664 Returns an integer value corresponding to the TCP source port of the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040017665 connection on the server side, which is the port HAProxy connected from.
Christopher Faulet7d081f02021-04-15 09:38:37 +020017666
Willy Tarreau74ca5042013-06-11 23:12:07 +020017667be_id : integer
17668 Returns an integer containing the current backend's id. It can be used in
Christopher Fauletd1b44642020-04-30 09:51:15 +020017669 frontends with responses to check which backend processed the request. It can
17670 also be used in a tcp-check or an http-check ruleset.
Willy Tarreau74ca5042013-06-11 23:12:07 +020017671
Marcin Deranekd2471c22016-12-12 14:08:05 +010017672be_name : string
17673 Returns a string containing the current backend's name. It can be used in
Christopher Fauletd1b44642020-04-30 09:51:15 +020017674 frontends with responses to check which backend processed the request. It can
17675 also be used in a tcp-check or an http-check ruleset.
Marcin Deranekd2471c22016-12-12 14:08:05 +010017676
Amaury Denoyelled91d7792020-12-10 13:43:56 +010017677be_server_timeout : integer
17678 Returns the configuration value in millisecond for the server timeout of the
17679 current backend. This timeout can be overwritten by a "set-timeout" rule. See
17680 also the "cur_server_timeout".
17681
17682be_tunnel_timeout : integer
17683 Returns the configuration value in millisecond for the tunnel timeout of the
17684 current backend. This timeout can be overwritten by a "set-timeout" rule. See
17685 also the "cur_tunnel_timeout".
17686
Amaury Denoyellef7719a22020-12-10 13:43:58 +010017687cur_server_timeout : integer
17688 Returns the currently applied server timeout in millisecond for the stream.
17689 In the default case, this will be equal to be_server_timeout unless a
17690 "set-timeout" rule has been applied. See also "be_server_timeout".
17691
17692cur_tunnel_timeout : integer
17693 Returns the currently applied tunnel timeout in millisecond for the stream.
17694 In the default case, this will be equal to be_tunnel_timeout unless a
17695 "set-timeout" rule has been applied. See also "be_tunnel_timeout".
17696
Willy Tarreau74ca5042013-06-11 23:12:07 +020017697dst : ip
17698 This is the destination IPv4 address of the connection on the client side,
17699 which is the address the client connected to. It can be useful when running
17700 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
17701 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010017702 RFC 4291. When the incoming connection passed through address translation or
17703 redirection involving connection tracking, the original destination address
17704 before the redirection will be reported. On Linux systems, the source and
17705 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
17706 is set, because a late response may reopen a timed out connection and switch
17707 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020017708
17709dst_conn : integer
17710 Returns an integer value corresponding to the number of currently established
17711 connections on the same socket including the one being evaluated. It is
17712 normally used with ACLs but can as well be used to pass the information to
17713 servers in an HTTP header or in logs. It can be used to either return a sorry
17714 page before hard-blocking, or to use a specific backend to drain new requests
17715 when the socket is considered saturated. This offers the ability to assign
17716 different limits to different listening ports or addresses. See also the
17717 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010017718
Willy Tarreau16e01562016-08-09 16:46:18 +020017719dst_is_local : boolean
17720 Returns true if the destination address of the incoming connection is local
17721 to the system, or false if the address doesn't exist on the system, meaning
17722 that it was intercepted in transparent mode. It can be useful to apply
17723 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010017724 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020017725 be delivered only on this address, or SSH access could be locally redirected.
17726 Please note that the check involves a few system calls, so it's better to do
17727 it only once per connection.
17728
Willy Tarreau74ca5042013-06-11 23:12:07 +020017729dst_port : integer
17730 Returns an integer value corresponding to the destination TCP port of the
17731 connection on the client side, which is the port the client connected to.
17732 This might be used when running in transparent mode, when assigning dynamic
17733 ports to some clients for a whole application session, to stick all users to
17734 a same server, or to pass the destination port information to a server using
17735 an HTTP header.
17736
Christopher Faulet9f074f62021-10-25 16:18:15 +020017737fc_fackets : integer
17738 Returns the fack counter measured by the kernel for the client
17739 connection. If the server connection is not established, if the connection is
17740 not TCP or if the operating system does not support TCP_INFO, for example
17741 Linux kernels before 2.4, the sample fetch fails.
17742
Willy Tarreau60ca10a2017-08-18 15:26:54 +020017743fc_http_major : integer
17744 Reports the front connection's HTTP major version encoding, which may be 1
17745 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
17746 encoding and not on the version present in the request header.
17747
Christopher Faulet9f074f62021-10-25 16:18:15 +020017748fc_lost : integer
17749 Returns the lost counter measured by the kernel for the client
17750 connection. If the server connection is not established, if the connection is
17751 not TCP or if the operating system does not support TCP_INFO, for example
17752 Linux kernels before 2.4, the sample fetch fails.
17753
Geoff Simmons7185b782019-08-27 18:31:16 +020017754fc_pp_authority : string
17755 Returns the authority TLV sent by the client in the PROXY protocol header,
17756 if any.
17757
Tim Duesterhusd1b15b62020-03-13 12:34:23 +010017758fc_pp_unique_id : string
17759 Returns the unique ID TLV sent by the client in the PROXY protocol header,
17760 if any.
17761
Emeric Brun4f603012017-01-05 15:11:44 +010017762fc_rcvd_proxy : boolean
17763 Returns true if the client initiated the connection with a PROXY protocol
17764 header.
17765
Christopher Faulet9f074f62021-10-25 16:18:15 +020017766fc_reordering : integer
17767 Returns the reordering counter measured by the kernel for the client
17768 connection. If the server connection is not established, if the connection is
17769 not TCP or if the operating system does not support TCP_INFO, for example
17770 Linux kernels before 2.4, the sample fetch fails.
17771
17772fc_retrans : integer
17773 Returns the retransmits counter measured by the kernel for the client
17774 connection. If the server connection is not established, if the connection is
17775 not TCP or if the operating system does not support TCP_INFO, for example
17776 Linux kernels before 2.4, the sample fetch fails.
17777
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020017778fc_rtt(<unit>) : integer
17779 Returns the Round Trip Time (RTT) measured by the kernel for the client
17780 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
17781 can be set to "ms" for milliseconds or "us" for microseconds. If the server
17782 connection is not established, if the connection is not TCP or if the
17783 operating system does not support TCP_INFO, for example Linux kernels before
17784 2.4, the sample fetch fails.
17785
17786fc_rttvar(<unit>) : integer
17787 Returns the Round Trip Time (RTT) variance measured by the kernel for the
17788 client connection. <unit> is facultative, by default the unit is milliseconds.
17789 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
17790 server connection is not established, if the connection is not TCP or if the
17791 operating system does not support TCP_INFO, for example Linux kernels before
17792 2.4, the sample fetch fails.
17793
Christopher Fauletba0c53e2019-10-17 14:40:48 +020017794fc_sacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070017795 Returns the sacked counter measured by the kernel for the client connection.
17796 If the server connection is not established, if the connection is not TCP or
17797 if the operating system does not support TCP_INFO, for example Linux kernels
17798 before 2.4, the sample fetch fails.
17799
Joe Williams30fcd392016-08-10 07:06:44 -070017800
Christopher Faulet9f074f62021-10-25 16:18:15 +020017801fc_unacked : integer
17802 Returns the unacked counter measured by the kernel for the client connection.
17803 If the server connection is not established, if the connection is not TCP or
17804 if the operating system does not support TCP_INFO, for example Linux kernels
17805 before 2.4, the sample fetch fails.
Joe Williams30fcd392016-08-10 07:06:44 -070017806
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020017807fe_defbe : string
17808 Returns a string containing the frontend's default backend name. It can be
17809 used in frontends to check which backend will handle requests by default.
17810
Willy Tarreau74ca5042013-06-11 23:12:07 +020017811fe_id : integer
17812 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010017813 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020017814 coming via a same frontend to the same server.
17815
Marcin Deranekd2471c22016-12-12 14:08:05 +010017816fe_name : string
17817 Returns a string containing the current frontend's name. It can be used in
17818 backends to check from which frontend it was called, or to stick all users
17819 coming via a same frontend to the same server.
17820
Amaury Denoyelleda184d52020-12-10 13:43:55 +010017821fe_client_timeout : integer
17822 Returns the configuration value in millisecond for the client timeout of the
17823 current frontend.
17824
Cyril Bonté62ba8702014-04-22 23:52:25 +020017825sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017826sc0_bytes_in_rate([<table>]) : integer
17827sc1_bytes_in_rate([<table>]) : integer
17828sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017829 Returns the average client-to-server bytes rate from the currently tracked
17830 counters, measured in amount of bytes over the period configured in the
17831 table. See also src_bytes_in_rate.
17832
Cyril Bonté62ba8702014-04-22 23:52:25 +020017833sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017834sc0_bytes_out_rate([<table>]) : integer
17835sc1_bytes_out_rate([<table>]) : integer
17836sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017837 Returns the average server-to-client bytes rate from the currently tracked
17838 counters, measured in amount of bytes over the period configured in the
17839 table. See also src_bytes_out_rate.
17840
Cyril Bonté62ba8702014-04-22 23:52:25 +020017841sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017842sc0_clr_gpc0([<table>]) : integer
17843sc1_clr_gpc0([<table>]) : integer
17844sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020017845 Clears the first General Purpose Counter associated to the currently tracked
17846 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010017847 stored value is zero, so first invocation will always return zero. This is
17848 typically used as a second ACL in an expression in order to mark a connection
17849 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020017850
Jarno Huuskonen676f6222017-03-30 09:19:45 +030017851 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020017852 # block if 5 consecutive requests continue to come faster than 10 sess
17853 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020017854 acl abuse sc0_http_req_rate gt 10
17855 acl kill sc0_inc_gpc0 gt 5
17856 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020017857 tcp-request connection accept if !abuse save
17858 tcp-request connection reject if abuse kill
17859
Frédéric Lécaille6778b272018-01-29 15:22:53 +010017860sc_clr_gpc1(<ctr>[,<table>]) : integer
17861sc0_clr_gpc1([<table>]) : integer
17862sc1_clr_gpc1([<table>]) : integer
17863sc2_clr_gpc1([<table>]) : integer
17864 Clears the second General Purpose Counter associated to the currently tracked
17865 counters, and returns its previous value. Before the first invocation, the
17866 stored value is zero, so first invocation will always return zero. This is
17867 typically used as a second ACL in an expression in order to mark a connection
17868 when a first ACL was verified.
17869
Cyril Bonté62ba8702014-04-22 23:52:25 +020017870sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017871sc0_conn_cnt([<table>]) : integer
17872sc1_conn_cnt([<table>]) : integer
17873sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010017874 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020017875 counters. See also src_conn_cnt.
17876
Cyril Bonté62ba8702014-04-22 23:52:25 +020017877sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017878sc0_conn_cur([<table>]) : integer
17879sc1_conn_cur([<table>]) : integer
17880sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017881 Returns the current amount of concurrent connections tracking the same
17882 tracked counters. This number is automatically incremented when tracking
17883 begins and decremented when tracking stops. See also src_conn_cur.
17884
Cyril Bonté62ba8702014-04-22 23:52:25 +020017885sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017886sc0_conn_rate([<table>]) : integer
17887sc1_conn_rate([<table>]) : integer
17888sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017889 Returns the average connection rate from the currently tracked counters,
17890 measured in amount of connections over the period configured in the table.
17891 See also src_conn_rate.
17892
Cyril Bonté62ba8702014-04-22 23:52:25 +020017893sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017894sc0_get_gpc0([<table>]) : integer
17895sc1_get_gpc0([<table>]) : integer
17896sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017897 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020017898 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020017899
Frédéric Lécaille6778b272018-01-29 15:22:53 +010017900sc_get_gpc1(<ctr>[,<table>]) : integer
17901sc0_get_gpc1([<table>]) : integer
17902sc1_get_gpc1([<table>]) : integer
17903sc2_get_gpc1([<table>]) : integer
17904 Returns the value of the second General Purpose Counter associated to the
17905 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
17906
Thierry FOURNIER236657b2015-08-19 08:25:14 +020017907sc_get_gpt0(<ctr>[,<table>]) : integer
17908sc0_get_gpt0([<table>]) : integer
17909sc1_get_gpt0([<table>]) : integer
17910sc2_get_gpt0([<table>]) : integer
17911 Returns the value of the first General Purpose Tag associated to the
17912 currently tracked counters. See also src_get_gpt0.
17913
Cyril Bonté62ba8702014-04-22 23:52:25 +020017914sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017915sc0_gpc0_rate([<table>]) : integer
17916sc1_gpc0_rate([<table>]) : integer
17917sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020017918 Returns the average increment rate of the first General Purpose Counter
17919 associated to the currently tracked counters. It reports the frequency
17920 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020017921 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
17922 that the "gpc0_rate" counter must be stored in the stick-table for a value to
17923 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020017924
Frédéric Lécaille6778b272018-01-29 15:22:53 +010017925sc_gpc1_rate(<ctr>[,<table>]) : integer
17926sc0_gpc1_rate([<table>]) : integer
17927sc1_gpc1_rate([<table>]) : integer
17928sc2_gpc1_rate([<table>]) : integer
17929 Returns the average increment rate of the second General Purpose Counter
17930 associated to the currently tracked counters. It reports the frequency
17931 which the gpc1 counter was incremented over the configured period. See also
17932 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
17933 that the "gpc1_rate" counter must be stored in the stick-table for a value to
17934 be returned, as "gpc1" only holds the event count.
17935
Cyril Bonté62ba8702014-04-22 23:52:25 +020017936sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017937sc0_http_err_cnt([<table>]) : integer
17938sc1_http_err_cnt([<table>]) : integer
17939sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010017940 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020017941 counters. This includes the both request errors and 4xx error responses.
17942 See also src_http_err_cnt.
17943
Cyril Bonté62ba8702014-04-22 23:52:25 +020017944sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017945sc0_http_err_rate([<table>]) : integer
17946sc1_http_err_rate([<table>]) : integer
17947sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017948 Returns the average rate of HTTP errors from the currently tracked counters,
17949 measured in amount of errors over the period configured in the table. This
17950 includes the both request errors and 4xx error responses. See also
17951 src_http_err_rate.
17952
Willy Tarreau826f3ab2021-02-10 12:07:15 +010017953sc_http_fail_cnt(<ctr>[,<table>]) : integer
17954sc0_http_fail_cnt([<table>]) : integer
17955sc1_http_fail_cnt([<table>]) : integer
17956sc2_http_fail_cnt([<table>]) : integer
17957 Returns the cumulative number of HTTP response failures from the currently
17958 tracked counters. This includes the both response errors and 5xx status codes
17959 other than 501 and 505. See also src_http_fail_cnt.
17960
17961sc_http_fail_rate(<ctr>[,<table>]) : integer
17962sc0_http_fail_rate([<table>]) : integer
17963sc1_http_fail_rate([<table>]) : integer
17964sc2_http_fail_rate([<table>]) : integer
17965 Returns the average rate of HTTP response failures from the currently tracked
17966 counters, measured in amount of failures over the period configured in the
17967 table. This includes the both response errors and 5xx status codes other than
17968 501 and 505. See also src_http_fail_rate.
17969
Cyril Bonté62ba8702014-04-22 23:52:25 +020017970sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017971sc0_http_req_cnt([<table>]) : integer
17972sc1_http_req_cnt([<table>]) : integer
17973sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010017974 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020017975 counters. This includes every started request, valid or not. See also
17976 src_http_req_cnt.
17977
Cyril Bonté62ba8702014-04-22 23:52:25 +020017978sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017979sc0_http_req_rate([<table>]) : integer
17980sc1_http_req_rate([<table>]) : integer
17981sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017982 Returns the average rate of HTTP requests from the currently tracked
17983 counters, measured in amount of requests over the period configured in
17984 the table. This includes every started request, valid or not. See also
17985 src_http_req_rate.
17986
Cyril Bonté62ba8702014-04-22 23:52:25 +020017987sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020017988sc0_inc_gpc0([<table>]) : integer
17989sc1_inc_gpc0([<table>]) : integer
17990sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020017991 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010017992 tracked counters, and returns its new value. Before the first invocation,
17993 the stored value is zero, so first invocation will increase it to 1 and will
17994 return 1. This is typically used as a second ACL in an expression in order
17995 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020017996
Jarno Huuskonen676f6222017-03-30 09:19:45 +030017997 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020017998 acl abuse sc0_http_req_rate gt 10
17999 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020018000 tcp-request connection reject if abuse kill
18001
Frédéric Lécaille6778b272018-01-29 15:22:53 +010018002sc_inc_gpc1(<ctr>[,<table>]) : integer
18003sc0_inc_gpc1([<table>]) : integer
18004sc1_inc_gpc1([<table>]) : integer
18005sc2_inc_gpc1([<table>]) : integer
18006 Increments the second General Purpose Counter associated to the currently
18007 tracked counters, and returns its new value. Before the first invocation,
18008 the stored value is zero, so first invocation will increase it to 1 and will
18009 return 1. This is typically used as a second ACL in an expression in order
18010 to mark a connection when a first ACL was verified.
18011
Cyril Bonté62ba8702014-04-22 23:52:25 +020018012sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020018013sc0_kbytes_in([<table>]) : integer
18014sc1_kbytes_in([<table>]) : integer
18015sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020018016 Returns the total amount of client-to-server data from the currently tracked
18017 counters, measured in kilobytes. The test is currently performed on 32-bit
18018 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020018019
Cyril Bonté62ba8702014-04-22 23:52:25 +020018020sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020018021sc0_kbytes_out([<table>]) : integer
18022sc1_kbytes_out([<table>]) : integer
18023sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020018024 Returns the total amount of server-to-client data from the currently tracked
18025 counters, measured in kilobytes. The test is currently performed on 32-bit
18026 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020018027
Cyril Bonté62ba8702014-04-22 23:52:25 +020018028sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020018029sc0_sess_cnt([<table>]) : integer
18030sc1_sess_cnt([<table>]) : integer
18031sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010018032 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020018033 into sessions, which means that they were accepted by a "tcp-request
18034 connection" rule, from the currently tracked counters. A backend may count
18035 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040018036 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020018037 with the client. See also src_sess_cnt.
18038
Cyril Bonté62ba8702014-04-22 23:52:25 +020018039sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020018040sc0_sess_rate([<table>]) : integer
18041sc1_sess_rate([<table>]) : integer
18042sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020018043 Returns the average session rate from the currently tracked counters,
18044 measured in amount of sessions over the period configured in the table. A
18045 session is a connection that got past the early "tcp-request connection"
18046 rules. A backend may count more sessions than connections because each
18047 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040018048 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020018049
Cyril Bonté62ba8702014-04-22 23:52:25 +020018050sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020018051sc0_tracked([<table>]) : boolean
18052sc1_tracked([<table>]) : boolean
18053sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020018054 Returns true if the designated session counter is currently being tracked by
18055 the current session. This can be useful when deciding whether or not we want
18056 to set some values in a header passed to the server.
18057
Cyril Bonté62ba8702014-04-22 23:52:25 +020018058sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020018059sc0_trackers([<table>]) : integer
18060sc1_trackers([<table>]) : integer
18061sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010018062 Returns the current amount of concurrent connections tracking the same
18063 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020018064 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010018065 that it does not rely on any stored information but on the table's reference
18066 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020018067 may sometimes be more suited for layer7 tracking. It can be used to tell a
18068 server how many concurrent connections there are from a given address for
18069 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010018070
Willy Tarreau74ca5042013-06-11 23:12:07 +020018071so_id : integer
18072 Returns an integer containing the current listening socket's id. It is useful
18073 in frontends involving many "bind" lines, or to stick all users coming via a
18074 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010018075
Jerome Magnineb421b22020-03-27 22:08:40 +010018076so_name : string
18077 Returns a string containing the current listening socket's name, as defined
18078 with name on a "bind" line. It can serve the same purposes as so_id but with
18079 strings instead of integers.
18080
Willy Tarreau74ca5042013-06-11 23:12:07 +020018081src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010018082 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020018083 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
18084 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
18085 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010018086 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
18087 directive is used, it can be the address of a client behind another
18088 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010018089 "tcp-request connection" which sees the real address. When the incoming
18090 connection passed through address translation or redirection involving
18091 connection tracking, the original destination address before the redirection
18092 will be reported. On Linux systems, the source and destination may seldom
18093 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
18094 response may reopen a timed out connection and switch what is believed to be
18095 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010018096
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010018097 Example:
18098 # add an HTTP header in requests with the originating address' country
18099 http-request set-header X-Country %[src,map_ip(geoip.lst)]
18100
Willy Tarreau74ca5042013-06-11 23:12:07 +020018101src_bytes_in_rate([<table>]) : integer
18102 Returns the average bytes rate from the incoming connection's source address
18103 in the current proxy's stick-table or in the designated stick-table, measured
18104 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018105 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018106
Willy Tarreau74ca5042013-06-11 23:12:07 +020018107src_bytes_out_rate([<table>]) : integer
18108 Returns the average bytes rate to the incoming connection's source address in
18109 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020018110 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018111 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018112
Willy Tarreau74ca5042013-06-11 23:12:07 +020018113src_clr_gpc0([<table>]) : integer
18114 Clears the first General Purpose Counter associated to the incoming
18115 connection's source address in the current proxy's stick-table or in the
18116 designated stick-table, and returns its previous value. If the address is not
18117 found, an entry is created and 0 is returned. This is typically used as a
18118 second ACL in an expression in order to mark a connection when a first ACL
18119 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020018120
Jarno Huuskonen676f6222017-03-30 09:19:45 +030018121 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020018122 # block if 5 consecutive requests continue to come faster than 10 sess
18123 # per second, and reset the counter as soon as the traffic slows down.
18124 acl abuse src_http_req_rate gt 10
18125 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010018126 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020018127 tcp-request connection accept if !abuse save
18128 tcp-request connection reject if abuse kill
18129
Frédéric Lécaille6778b272018-01-29 15:22:53 +010018130src_clr_gpc1([<table>]) : integer
18131 Clears the second General Purpose Counter associated to the incoming
18132 connection's source address in the current proxy's stick-table or in the
18133 designated stick-table, and returns its previous value. If the address is not
18134 found, an entry is created and 0 is returned. This is typically used as a
18135 second ACL in an expression in order to mark a connection when a first ACL
18136 was verified.
18137
Willy Tarreau74ca5042013-06-11 23:12:07 +020018138src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010018139 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020018140 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020018141 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018142 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018143
Willy Tarreau74ca5042013-06-11 23:12:07 +020018144src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020018145 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020018146 current incoming connection's source address in the current proxy's
18147 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018148 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018149
Willy Tarreau74ca5042013-06-11 23:12:07 +020018150src_conn_rate([<table>]) : integer
18151 Returns the average connection rate from the incoming connection's source
18152 address in the current proxy's stick-table or in the designated stick-table,
18153 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018154 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018155
Willy Tarreau74ca5042013-06-11 23:12:07 +020018156src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020018157 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020018158 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020018159 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018160 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018161
Frédéric Lécaille6778b272018-01-29 15:22:53 +010018162src_get_gpc1([<table>]) : integer
18163 Returns the value of the second General Purpose Counter associated to the
18164 incoming connection's source address in the current proxy's stick-table or in
18165 the designated stick-table. If the address is not found, zero is returned.
18166 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
18167
Thierry FOURNIER236657b2015-08-19 08:25:14 +020018168src_get_gpt0([<table>]) : integer
18169 Returns the value of the first General Purpose Tag associated to the
18170 incoming connection's source address in the current proxy's stick-table or in
18171 the designated stick-table. If the address is not found, zero is returned.
18172 See also sc/sc0/sc1/sc2_get_gpt0.
18173
Willy Tarreau74ca5042013-06-11 23:12:07 +020018174src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020018175 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020018176 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020018177 stick-table or in the designated stick-table. It reports the frequency
18178 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018179 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
18180 that the "gpc0_rate" counter must be stored in the stick-table for a value to
18181 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020018182
Frédéric Lécaille6778b272018-01-29 15:22:53 +010018183src_gpc1_rate([<table>]) : integer
18184 Returns the average increment rate of the second General Purpose Counter
18185 associated to the incoming connection's source address in the current proxy's
18186 stick-table or in the designated stick-table. It reports the frequency
18187 which the gpc1 counter was incremented over the configured period. See also
18188 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
18189 that the "gpc1_rate" counter must be stored in the stick-table for a value to
18190 be returned, as "gpc1" only holds the event count.
18191
Willy Tarreau74ca5042013-06-11 23:12:07 +020018192src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010018193 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020018194 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020018195 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018196 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020018197 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018198
Willy Tarreau74ca5042013-06-11 23:12:07 +020018199src_http_err_rate([<table>]) : integer
18200 Returns the average rate of HTTP errors from the incoming connection's source
18201 address in the current proxy's stick-table or in the designated stick-table,
18202 measured in amount of errors over the period configured in the table. This
18203 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018204 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018205
Willy Tarreau826f3ab2021-02-10 12:07:15 +010018206src_http_fail_cnt([<table>]) : integer
18207 Returns the cumulative number of HTTP response failures triggered by the
18208 incoming connection's source address in the current proxy's stick-table or in
Ilya Shipitsin0de36ad2021-02-20 00:23:36 +050018209 the designated stick-table. This includes the both response errors and 5xx
Willy Tarreau826f3ab2021-02-10 12:07:15 +010018210 status codes other than 501 and 505. See also sc/sc0/sc1/sc2_http_fail_cnt.
18211 If the address is not found, zero is returned.
18212
18213src_http_fail_rate([<table>]) : integer
18214 Returns the average rate of HTTP response failures triggered by the incoming
18215 connection's source address in the current proxy's stick-table or in the
18216 designated stick-table, measured in amount of failures over the period
18217 configured in the table. This includes the both response errors and 5xx
18218 status codes other than 501 and 505. If the address is not found, zero is
18219 returned. See also sc/sc0/sc1/sc2_http_fail_rate.
18220
Willy Tarreau74ca5042013-06-11 23:12:07 +020018221src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010018222 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020018223 source address in the current proxy's stick-table or in the designated stick-
18224 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018225 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018226
Willy Tarreau74ca5042013-06-11 23:12:07 +020018227src_http_req_rate([<table>]) : integer
18228 Returns the average rate of HTTP requests from the incoming connection's
18229 source address in the current proxy's stick-table or in the designated stick-
18230 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020018231 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018232 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018233
Willy Tarreau74ca5042013-06-11 23:12:07 +020018234src_inc_gpc0([<table>]) : integer
18235 Increments the first General Purpose Counter associated to the incoming
18236 connection's source address in the current proxy's stick-table or in the
18237 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020018238 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020018239 This is typically used as a second ACL in an expression in order to mark a
18240 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020018241
Jarno Huuskonen676f6222017-03-30 09:19:45 +030018242 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020018243 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010018244 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020018245 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020018246
Frédéric Lécaille6778b272018-01-29 15:22:53 +010018247src_inc_gpc1([<table>]) : integer
18248 Increments the second General Purpose Counter associated to the incoming
18249 connection's source address in the current proxy's stick-table or in the
18250 designated stick-table, and returns its new value. If the address is not
18251 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
18252 This is typically used as a second ACL in an expression in order to mark a
18253 connection when a first ACL was verified.
18254
Willy Tarreau16e01562016-08-09 16:46:18 +020018255src_is_local : boolean
18256 Returns true if the source address of the incoming connection is local to the
18257 system, or false if the address doesn't exist on the system, meaning that it
18258 comes from a remote machine. Note that UNIX addresses are considered local.
18259 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010018260 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020018261 note that the check involves a few system calls, so it's better to do it only
18262 once per connection.
18263
Willy Tarreau74ca5042013-06-11 23:12:07 +020018264src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020018265 Returns the total amount of data received from the incoming connection's
18266 source address in the current proxy's stick-table or in the designated
18267 stick-table, measured in kilobytes. If the address is not found, zero is
18268 returned. The test is currently performed on 32-bit integers, which limits
18269 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018270
Willy Tarreau74ca5042013-06-11 23:12:07 +020018271src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020018272 Returns the total amount of data sent to the incoming connection's source
18273 address in the current proxy's stick-table or in the designated stick-table,
18274 measured in kilobytes. If the address is not found, zero is returned. The
18275 test is currently performed on 32-bit integers, which limits values to 4
18276 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020018277
Willy Tarreau74ca5042013-06-11 23:12:07 +020018278src_port : integer
18279 Returns an integer value corresponding to the TCP source port of the
18280 connection on the client side, which is the port the client connected from.
18281 Usage of this function is very limited as modern protocols do not care much
18282 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010018283
Willy Tarreau74ca5042013-06-11 23:12:07 +020018284src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010018285 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020018286 connection's source IPv4 address in the current proxy's stick-table or in the
18287 designated stick-table, that were transformed into sessions, which means that
18288 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018289 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018290
Willy Tarreau74ca5042013-06-11 23:12:07 +020018291src_sess_rate([<table>]) : integer
18292 Returns the average session rate from the incoming connection's source
18293 address in the current proxy's stick-table or in the designated stick-table,
18294 measured in amount of sessions over the period configured in the table. A
18295 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020018296 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020018297
Willy Tarreau74ca5042013-06-11 23:12:07 +020018298src_updt_conn_cnt([<table>]) : integer
18299 Creates or updates the entry associated to the incoming connection's source
18300 address in the current proxy's stick-table or in the designated stick-table.
18301 This table must be configured to store the "conn_cnt" data type, otherwise
18302 the match will be ignored. The current count is incremented by one, and the
18303 expiration timer refreshed. The updated count is returned, so this match
18304 can't return zero. This was used to reject service abusers based on their
18305 source address. Note: it is recommended to use the more complete "track-sc*"
18306 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020018307
18308 Example :
18309 # This frontend limits incoming SSH connections to 3 per 10 second for
18310 # each source address, and rejects excess connections until a 10 second
18311 # silence is observed. At most 20 addresses are tracked.
18312 listen ssh
18313 bind :22
18314 mode tcp
18315 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020018316 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020018317 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020018318 server local 127.0.0.1:22
18319
Willy Tarreau74ca5042013-06-11 23:12:07 +020018320srv_id : integer
18321 Returns an integer containing the server's id when processing the response.
18322 While it's almost only used with ACLs, it may be used for logging or
Christopher Fauletd1b44642020-04-30 09:51:15 +020018323 debugging. It can also be used in a tcp-check or an http-check ruleset.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020018324
vkill1dfd1652019-10-30 16:58:14 +080018325srv_name : string
18326 Returns a string containing the server's name when processing the response.
18327 While it's almost only used with ACLs, it may be used for logging or
Christopher Fauletd1b44642020-04-30 09:51:15 +020018328 debugging. It can also be used in a tcp-check or an http-check ruleset.
vkill1dfd1652019-10-30 16:58:14 +080018329
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200183307.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020018331----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020018332
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018333The layer 5 usually describes just the session layer which in HAProxy is
Willy Tarreau74ca5042013-06-11 23:12:07 +020018334closest to the session once all the connection handshakes are finished, but
18335when no content is yet made available. The fetch methods described here are
18336usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018337future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020018338
Ben Shillitof25e8e52016-12-02 14:25:37 +00001833951d.all(<prop>[,<prop>*]) : string
18340 Returns values for the properties requested as a string, where values are
18341 separated by the delimiter specified with "51degrees-property-separator".
18342 The device is identified using all the important HTTP headers from the
18343 request. The function can be passed up to five property names, and if a
18344 property name can't be found, the value "NoData" is returned.
18345
18346 Example :
18347 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
18348 # containing the three properties requested using all relevant headers from
18349 # the request.
18350 frontend http-in
18351 bind *:8081
18352 default_backend servers
18353 http-request set-header X-51D-DeviceTypeMobileTablet \
18354 %[51d.all(DeviceType,IsMobile,IsTablet)]
18355
Emeric Brun645ae792014-04-30 14:21:06 +020018356ssl_bc : boolean
18357 Returns true when the back connection was made via an SSL/TLS transport
18358 layer and is locally deciphered. This means the outgoing connection was made
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018359 other a server with the "ssl" option. It can be used in a tcp-check or an
18360 http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018361
18362ssl_bc_alg_keysize : integer
18363 Returns the symmetric cipher key size supported in bits when the outgoing
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018364 connection was made over an SSL/TLS transport layer. It can be used in a
18365 tcp-check or an http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018366
Olivier Houchard6b77f492018-11-22 18:18:29 +010018367ssl_bc_alpn : string
18368 This extracts the Application Layer Protocol Negotiation field from an
18369 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020018370 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010018371 server. The SSL library must have been built with support for TLS
18372 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
18373 not advertised unless the "alpn" keyword on the "server" line specifies a
18374 protocol list. Also, nothing forces the server to pick a protocol from this
18375 list, any other one may be requested. The TLS ALPN extension is meant to
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018376 replace the TLS NPN extension. See also "ssl_bc_npn". It can be used in a
18377 tcp-check or an http-check ruleset.
Olivier Houchard6b77f492018-11-22 18:18:29 +010018378
Emeric Brun645ae792014-04-30 14:21:06 +020018379ssl_bc_cipher : string
18380 Returns the name of the used cipher when the outgoing connection was made
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018381 over an SSL/TLS transport layer. It can be used in a tcp-check or an
18382 http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018383
Patrick Hemmer65674662019-06-04 08:13:03 -040018384ssl_bc_client_random : binary
18385 Returns the client random of the back connection when the incoming connection
18386 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
18387 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018388 It can be used in a tcp-check or an http-check ruleset.
Patrick Hemmer65674662019-06-04 08:13:03 -040018389
Emeric Brun74f7ffa2018-02-19 16:14:12 +010018390ssl_bc_is_resumed : boolean
18391 Returns true when the back connection was made over an SSL/TLS transport
18392 layer and the newly created SSL session was resumed using a cached
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018393 session or a TLS ticket. It can be used in a tcp-check or an http-check
18394 ruleset.
Emeric Brun74f7ffa2018-02-19 16:14:12 +010018395
Olivier Houchard6b77f492018-11-22 18:18:29 +010018396ssl_bc_npn : string
18397 This extracts the Next Protocol Negotiation field from an outgoing connection
18398 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020018399 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010018400 built with support for TLS extensions enabled (check haproxy -vv). Note that
18401 the TLS NPN extension is not advertised unless the "npn" keyword on the
18402 "server" line specifies a protocol list. Also, nothing forces the server to
18403 pick a protocol from this list, any other one may be used. Please note that
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018404 the TLS NPN extension was replaced with ALPN. It can be used in a tcp-check
18405 or an http-check ruleset.
Olivier Houchard6b77f492018-11-22 18:18:29 +010018406
Emeric Brun645ae792014-04-30 14:21:06 +020018407ssl_bc_protocol : string
18408 Returns the name of the used protocol when the outgoing connection was made
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018409 over an SSL/TLS transport layer. It can be used in a tcp-check or an
18410 http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018411
Emeric Brunb73a9b02014-04-30 18:49:19 +020018412ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020018413 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020018414 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018415 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64". It
18416 can be used in a tcp-check or an http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018417
Patrick Hemmer65674662019-06-04 08:13:03 -040018418ssl_bc_server_random : binary
18419 Returns the server random of the back connection when the incoming connection
18420 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
18421 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018422 It can be used in a tcp-check or an http-check ruleset.
Patrick Hemmer65674662019-06-04 08:13:03 -040018423
Emeric Brun645ae792014-04-30 14:21:06 +020018424ssl_bc_session_id : binary
18425 Returns the SSL ID of the back connection when the outgoing connection was
18426 made over an SSL/TLS transport layer. It is useful to log if we want to know
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018427 if session was reused or not. It can be used in a tcp-check or an http-check
18428 ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018429
Patrick Hemmere0275472018-04-28 19:15:51 -040018430ssl_bc_session_key : binary
18431 Returns the SSL session master key of the back connection when the outgoing
18432 connection was made over an SSL/TLS transport layer. It is useful to decrypt
18433 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018434 BoringSSL. It can be used in a tcp-check or an http-check ruleset.
Patrick Hemmere0275472018-04-28 19:15:51 -040018435
Emeric Brun645ae792014-04-30 14:21:06 +020018436ssl_bc_use_keysize : integer
18437 Returns the symmetric cipher key size used in bits when the outgoing
Christopher Fauletd92ea7f2020-04-30 10:03:55 +020018438 connection was made over an SSL/TLS transport layer. It can be used in a
18439 tcp-check or an http-check ruleset.
Emeric Brun645ae792014-04-30 14:21:06 +020018440
Willy Tarreau74ca5042013-06-11 23:12:07 +020018441ssl_c_ca_err : integer
18442 When the incoming connection was made over an SSL/TLS transport layer,
18443 returns the ID of the first error detected during verification of the client
18444 certificate at depth > 0, or 0 if no error was encountered during this
18445 verification process. Please refer to your SSL library's documentation to
18446 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020018447
Willy Tarreau74ca5042013-06-11 23:12:07 +020018448ssl_c_ca_err_depth : integer
18449 When the incoming connection was made over an SSL/TLS transport layer,
18450 returns the depth in the CA chain of the first error detected during the
18451 verification of the client certificate. If no error is encountered, 0 is
18452 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010018453
Christopher Faulet70d10d12020-11-06 12:10:33 +010018454ssl_c_chain_der : binary
William Dauchya598b502020-08-06 18:11:38 +020018455 Returns the DER formatted chain certificate presented by the client when the
18456 incoming connection was made over an SSL/TLS transport layer. When used for
18457 an ACL, the value(s) to match against can be passed in hexadecimal form. One
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +050018458 can parse the result with any lib accepting ASN.1 DER data. It currently
William Dauchya598b502020-08-06 18:11:38 +020018459 does not support resumed sessions.
18460
Christopher Faulet70d10d12020-11-06 12:10:33 +010018461ssl_c_der : binary
18462 Returns the DER formatted certificate presented by the client when the
18463 incoming connection was made over an SSL/TLS transport layer. When used for
18464 an ACL, the value(s) to match against can be passed in hexadecimal form.
18465
Willy Tarreau74ca5042013-06-11 23:12:07 +020018466ssl_c_err : integer
18467 When the incoming connection was made over an SSL/TLS transport layer,
18468 returns the ID of the first error detected during verification at depth 0, or
18469 0 if no error was encountered during this verification process. Please refer
18470 to your SSL library's documentation to find the exhaustive list of error
18471 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020018472
Elliot Otchet71f82972020-01-15 08:12:14 -050018473ssl_c_i_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020018474 When the incoming connection was made over an SSL/TLS transport layer,
18475 returns the full distinguished name of the issuer of the certificate
18476 presented by the client when no <entry> is specified, or the value of the
18477 first given entry found from the beginning of the DN. If a positive/negative
18478 occurrence number is specified as the optional second argument, it returns
18479 the value of the nth given entry value from the beginning/end of the DN.
18480 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
18481 "ssl_c_i_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050018482 The <format> parameter allows you to receive the DN suitable for
18483 consumption by different protocols. Currently supported is rfc2253 for
18484 LDAP v3.
18485 If you'd like to modify the format only you can specify an empty string
18486 and zero for the first two parameters. Example: ssl_c_i_dn(,0,rfc2253)
Willy Tarreau62644772008-07-16 18:36:06 +020018487
Willy Tarreau74ca5042013-06-11 23:12:07 +020018488ssl_c_key_alg : string
18489 Returns the name of the algorithm used to generate the key of the certificate
18490 presented by the client when the incoming connection was made over an SSL/TLS
18491 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020018492
Willy Tarreau74ca5042013-06-11 23:12:07 +020018493ssl_c_notafter : string
18494 Returns the end date presented by the client as a formatted string
18495 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
18496 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020018497
Willy Tarreau74ca5042013-06-11 23:12:07 +020018498ssl_c_notbefore : string
18499 Returns the start date presented by the client as a formatted string
18500 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
18501 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010018502
Elliot Otchet71f82972020-01-15 08:12:14 -050018503ssl_c_s_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020018504 When the incoming connection was made over an SSL/TLS transport layer,
18505 returns the full distinguished name of the subject of the certificate
18506 presented by the client when no <entry> is specified, or the value of the
18507 first given entry found from the beginning of the DN. If a positive/negative
18508 occurrence number is specified as the optional second argument, it returns
18509 the value of the nth given entry value from the beginning/end of the DN.
18510 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
18511 "ssl_c_s_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050018512 The <format> parameter allows you to receive the DN suitable for
18513 consumption by different protocols. Currently supported is rfc2253 for
18514 LDAP v3.
18515 If you'd like to modify the format only you can specify an empty string
18516 and zero for the first two parameters. Example: ssl_c_s_dn(,0,rfc2253)
Willy Tarreaub6672b52011-12-12 17:23:41 +010018517
Willy Tarreau74ca5042013-06-11 23:12:07 +020018518ssl_c_serial : binary
18519 Returns the serial of the certificate presented by the client when the
18520 incoming connection was made over an SSL/TLS transport layer. When used for
18521 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020018522
Willy Tarreau74ca5042013-06-11 23:12:07 +020018523ssl_c_sha1 : binary
18524 Returns the SHA-1 fingerprint of the certificate presented by the client when
18525 the incoming connection was made over an SSL/TLS transport layer. This can be
18526 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020018527 Note that the output is binary, so if you want to pass that signature to the
18528 server, you need to encode it in hex or base64, such as in the example below:
18529
Jarno Huuskonen676f6222017-03-30 09:19:45 +030018530 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020018531 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020018532
Willy Tarreau74ca5042013-06-11 23:12:07 +020018533ssl_c_sig_alg : string
18534 Returns the name of the algorithm used to sign the certificate presented by
18535 the client when the incoming connection was made over an SSL/TLS transport
18536 layer.
Emeric Brun87855892012-10-17 17:39:35 +020018537
Willy Tarreau74ca5042013-06-11 23:12:07 +020018538ssl_c_used : boolean
18539 Returns true if current SSL session uses a client certificate even if current
18540 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020018541
Willy Tarreau74ca5042013-06-11 23:12:07 +020018542ssl_c_verify : integer
18543 Returns the verify result error ID when the incoming connection was made over
18544 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
18545 refer to your SSL library's documentation for an exhaustive list of error
18546 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020018547
Willy Tarreau74ca5042013-06-11 23:12:07 +020018548ssl_c_version : integer
18549 Returns the version of the certificate presented by the client when the
18550 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020018551
Emeric Brun43e79582014-10-29 19:03:26 +010018552ssl_f_der : binary
18553 Returns the DER formatted certificate presented by the frontend when the
18554 incoming connection was made over an SSL/TLS transport layer. When used for
18555 an ACL, the value(s) to match against can be passed in hexadecimal form.
18556
Elliot Otchet71f82972020-01-15 08:12:14 -050018557ssl_f_i_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020018558 When the incoming connection was made over an SSL/TLS transport layer,
18559 returns the full distinguished name of the issuer of the certificate
18560 presented by the frontend when no <entry> is specified, or the value of the
18561 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020018562 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020018563 the value of the nth given entry value from the beginning/end of the DN.
18564 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
18565 "ssl_f_i_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050018566 The <format> parameter allows you to receive the DN suitable for
18567 consumption by different protocols. Currently supported is rfc2253 for
18568 LDAP v3.
18569 If you'd like to modify the format only you can specify an empty string
18570 and zero for the first two parameters. Example: ssl_f_i_dn(,0,rfc2253)
Emeric Brun87855892012-10-17 17:39:35 +020018571
Willy Tarreau74ca5042013-06-11 23:12:07 +020018572ssl_f_key_alg : string
18573 Returns the name of the algorithm used to generate the key of the certificate
18574 presented by the frontend when the incoming connection was made over an
18575 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020018576
Willy Tarreau74ca5042013-06-11 23:12:07 +020018577ssl_f_notafter : string
18578 Returns the end date presented by the frontend as a formatted string
18579 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
18580 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020018581
Willy Tarreau74ca5042013-06-11 23:12:07 +020018582ssl_f_notbefore : string
18583 Returns the start date presented by the frontend as a formatted string
18584 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
18585 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020018586
Elliot Otchet71f82972020-01-15 08:12:14 -050018587ssl_f_s_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020018588 When the incoming connection was made over an SSL/TLS transport layer,
18589 returns the full distinguished name of the subject of the certificate
18590 presented by the frontend when no <entry> is specified, or the value of the
18591 first given entry found from the beginning of the DN. If a positive/negative
18592 occurrence number is specified as the optional second argument, it returns
18593 the value of the nth given entry value from the beginning/end of the DN.
18594 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
18595 "ssl_f_s_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050018596 The <format> parameter allows you to receive the DN suitable for
18597 consumption by different protocols. Currently supported is rfc2253 for
18598 LDAP v3.
18599 If you'd like to modify the format only you can specify an empty string
18600 and zero for the first two parameters. Example: ssl_f_s_dn(,0,rfc2253)
Emeric Brunce5ad802012-10-22 14:11:22 +020018601
Willy Tarreau74ca5042013-06-11 23:12:07 +020018602ssl_f_serial : binary
18603 Returns the serial of the certificate presented by the frontend when the
18604 incoming connection was made over an SSL/TLS transport layer. When used for
18605 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020018606
Emeric Brun55f4fa82014-04-30 17:11:25 +020018607ssl_f_sha1 : binary
18608 Returns the SHA-1 fingerprint of the certificate presented by the frontend
18609 when the incoming connection was made over an SSL/TLS transport layer. This
18610 can be used to know which certificate was chosen using SNI.
18611
Willy Tarreau74ca5042013-06-11 23:12:07 +020018612ssl_f_sig_alg : string
18613 Returns the name of the algorithm used to sign the certificate presented by
18614 the frontend when the incoming connection was made over an SSL/TLS transport
18615 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020018616
Willy Tarreau74ca5042013-06-11 23:12:07 +020018617ssl_f_version : integer
18618 Returns the version of the certificate presented by the frontend when the
18619 incoming connection was made over an SSL/TLS transport layer.
18620
18621ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020018622 Returns true when the front connection was made via an SSL/TLS transport
18623 layer and is locally deciphered. This means it has matched a socket declared
18624 with a "bind" line having the "ssl" option.
18625
Willy Tarreau74ca5042013-06-11 23:12:07 +020018626 Example :
18627 # This passes "X-Proto: https" to servers when client connects over SSL
18628 listen http-https
18629 bind :80
18630 bind :443 ssl crt /etc/haproxy.pem
18631 http-request add-header X-Proto https if { ssl_fc }
18632
18633ssl_fc_alg_keysize : integer
18634 Returns the symmetric cipher key size supported in bits when the incoming
18635 connection was made over an SSL/TLS transport layer.
18636
18637ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018638 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020018639 incoming connection made via a TLS transport layer and locally deciphered by
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018640 HAProxy. The result is a string containing the protocol name advertised by
Willy Tarreau74ca5042013-06-11 23:12:07 +020018641 the client. The SSL library must have been built with support for TLS
18642 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
18643 not advertised unless the "alpn" keyword on the "bind" line specifies a
18644 protocol list. Also, nothing forces the client to pick a protocol from this
18645 list, any other one may be requested. The TLS ALPN extension is meant to
18646 replace the TLS NPN extension. See also "ssl_fc_npn".
18647
Willy Tarreau74ca5042013-06-11 23:12:07 +020018648ssl_fc_cipher : string
18649 Returns the name of the used cipher when the incoming connection was made
18650 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020018651
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010018652ssl_fc_cipherlist_bin : binary
18653 Returns the binary form of the client hello cipher list. The maximum returned
18654 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010018655 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010018656
18657ssl_fc_cipherlist_hex : string
18658 Returns the binary form of the client hello cipher list encoded as
18659 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010018660 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010018661
18662ssl_fc_cipherlist_str : string
18663 Returns the decoded text form of the client hello cipher list. The maximum
18664 number of ciphers returned is according with the value of
18665 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010018666 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020018667 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010018668
18669ssl_fc_cipherlist_xxh : integer
18670 Returns a xxh64 of the cipher list. This hash can be return only is the value
18671 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010018672 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010018673
Patrick Hemmer65674662019-06-04 08:13:03 -040018674ssl_fc_client_random : binary
18675 Returns the client random of the front connection when the incoming connection
18676 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
18677 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
18678
William Lallemand7d42ef52020-07-06 11:41:30 +020018679ssl_fc_client_early_traffic_secret : string
18680 Return the CLIENT_EARLY_TRAFFIC_SECRET as an hexadecimal string for the
18681 front connection when the incoming connection was made over a TLS 1.3
18682 transport layer.
18683 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18684 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18685 activated with "tune.ssl.keylog on" in the global section. See also
18686 "tune.ssl.keylog"
18687
18688ssl_fc_client_handshake_traffic_secret : string
18689 Return the CLIENT_HANDSHAKE_TRAFFIC_SECRET as an hexadecimal string for the
18690 front connection when the incoming connection was made over a TLS 1.3
18691 transport layer.
18692 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18693 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18694 activated with "tune.ssl.keylog on" in the global section. See also
18695 "tune.ssl.keylog"
18696
18697ssl_fc_client_traffic_secret_0 : string
18698 Return the CLIENT_TRAFFIC_SECRET_0 as an hexadecimal string for the
18699 front connection when the incoming connection was made over a TLS 1.3
18700 transport layer.
18701 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18702 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18703 activated with "tune.ssl.keylog on" in the global section. See also
18704 "tune.ssl.keylog"
18705
18706ssl_fc_exporter_secret : string
18707 Return the EXPORTER_SECRET as an hexadecimal string for the
18708 front connection when the incoming connection was made over a TLS 1.3
18709 transport layer.
18710 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18711 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18712 activated with "tune.ssl.keylog on" in the global section. See also
18713 "tune.ssl.keylog"
18714
18715ssl_fc_early_exporter_secret : string
18716 Return the EARLY_EXPORTER_SECRET as an hexadecimal string for the
18717 front connection when the incoming connection was made over an TLS 1.3
18718 transport layer.
18719 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18720 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18721 activated with "tune.ssl.keylog on" in the global section. See also
18722 "tune.ssl.keylog"
18723
Willy Tarreau74ca5042013-06-11 23:12:07 +020018724ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020018725 Returns true if a client certificate is present in an incoming connection over
18726 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010018727 Note: on SSL session resumption with Session ID or TLS ticket, client
18728 certificate is not present in the current connection but may be retrieved
18729 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
18730 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020018731
Olivier Houchardccaa7de2017-10-02 11:51:03 +020018732ssl_fc_has_early : boolean
18733 Returns true if early data were sent, and the handshake didn't happen yet. As
18734 it has security implications, it is useful to be able to refuse those, or
18735 wait until the handshake happened.
18736
Willy Tarreau74ca5042013-06-11 23:12:07 +020018737ssl_fc_has_sni : boolean
18738 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020018739 in an incoming connection was made over an SSL/TLS transport layer. Returns
18740 true when the incoming connection presents a TLS SNI field. This requires
John Roeslerfb2fce12019-07-10 15:45:51 -050018741 that the SSL library is built with support for TLS extensions enabled (check
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020018742 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020018743
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020018744ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020018745 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010018746 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
18747 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020018748
Willy Tarreau74ca5042013-06-11 23:12:07 +020018749ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018750 This extracts the Next Protocol Negotiation field from an incoming connection
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018751 made via a TLS transport layer and locally deciphered by HAProxy. The result
Willy Tarreau74ca5042013-06-11 23:12:07 +020018752 is a string containing the protocol name advertised by the client. The SSL
18753 library must have been built with support for TLS extensions enabled (check
18754 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
18755 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
18756 forces the client to pick a protocol from this list, any other one may be
18757 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020018758
Willy Tarreau74ca5042013-06-11 23:12:07 +020018759ssl_fc_protocol : string
18760 Returns the name of the used protocol when the incoming connection was made
18761 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020018762
Emeric Brunb73a9b02014-04-30 18:49:19 +020018763ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040018764 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020018765 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
Christopher Fauletc5de4192021-11-09 14:23:36 +010018766 can be encoded to base64 using the converter: "ssl_fc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040018767
William Lallemand7d42ef52020-07-06 11:41:30 +020018768ssl_fc_server_handshake_traffic_secret : string
18769 Return the SERVER_HANDSHAKE_TRAFFIC_SECRET as an hexadecimal string for the
18770 front connection when the incoming connection was made over a TLS 1.3
18771 transport layer.
18772 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18773 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18774 activated with "tune.ssl.keylog on" in the global section. See also
18775 "tune.ssl.keylog"
18776
18777ssl_fc_server_traffic_secret_0 : string
18778 Return the SERVER_TRAFFIC_SECRET_0 as an hexadecimal string for the
18779 front connection when the incoming connection was made over an TLS 1.3
18780 transport layer.
18781 Require OpenSSL >= 1.1.1. This is one of the keys dumped by the OpenSSL
18782 keylog callback to generate the SSLKEYLOGFILE. The SSL Key logging must be
18783 activated with "tune.ssl.keylog on" in the global section. See also
18784 "tune.ssl.keylog"
18785
Patrick Hemmer65674662019-06-04 08:13:03 -040018786ssl_fc_server_random : binary
18787 Returns the server random of the front connection when the incoming connection
18788 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
18789 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
18790
Willy Tarreau74ca5042013-06-11 23:12:07 +020018791ssl_fc_session_id : binary
18792 Returns the SSL ID of the front connection when the incoming connection was
18793 made over an SSL/TLS transport layer. It is useful to stick a given client to
18794 a server. It is important to note that some browsers refresh their session ID
18795 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020018796
Patrick Hemmere0275472018-04-28 19:15:51 -040018797ssl_fc_session_key : binary
18798 Returns the SSL session master key of the front connection when the incoming
18799 connection was made over an SSL/TLS transport layer. It is useful to decrypt
18800 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
18801 BoringSSL.
18802
18803
Willy Tarreau74ca5042013-06-11 23:12:07 +020018804ssl_fc_sni : string
18805 This extracts the Server Name Indication TLS extension (SNI) field from an
18806 incoming connection made via an SSL/TLS transport layer and locally
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018807 deciphered by HAProxy. The result (when present) typically is a string
Willy Tarreau74ca5042013-06-11 23:12:07 +020018808 matching the HTTPS host name (253 chars or less). The SSL library must have
18809 been built with support for TLS extensions enabled (check haproxy -vv).
18810
Alex18c88772021-06-05 13:23:08 +020018811 This fetch is different from "req.ssl_sni" above in that it applies to the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018812 connection being deciphered by HAProxy and not to SSL contents being blindly
Willy Tarreau74ca5042013-06-11 23:12:07 +020018813 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
John Roeslerfb2fce12019-07-10 15:45:51 -050018814 requires that the SSL library is built with support for TLS extensions
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020018815 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020018816
Willy Tarreau74ca5042013-06-11 23:12:07 +020018817 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020018818 ssl_fc_sni_end : suffix match
18819 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020018820
Willy Tarreau74ca5042013-06-11 23:12:07 +020018821ssl_fc_use_keysize : integer
18822 Returns the symmetric cipher key size used in bits when the incoming
18823 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020018824
William Lallemandbfa3e812020-06-25 20:07:18 +020018825ssl_s_der : binary
18826 Returns the DER formatted certificate presented by the server when the
18827 outgoing connection was made over an SSL/TLS transport layer. When used for
18828 an ACL, the value(s) to match against can be passed in hexadecimal form.
18829
William Dauchya598b502020-08-06 18:11:38 +020018830ssl_s_chain_der : binary
18831 Returns the DER formatted chain certificate presented by the server when the
18832 outgoing connection was made over an SSL/TLS transport layer. When used for
18833 an ACL, the value(s) to match against can be passed in hexadecimal form. One
Ilya Shipitsin2272d8a2020-12-21 01:22:40 +050018834 can parse the result with any lib accepting ASN.1 DER data. It currently
William Dauchya598b502020-08-06 18:11:38 +020018835 does not support resumed sessions.
18836
William Lallemandbfa3e812020-06-25 20:07:18 +020018837ssl_s_key_alg : string
18838 Returns the name of the algorithm used to generate the key of the certificate
18839 presented by the server when the outgoing connection was made over an
18840 SSL/TLS transport layer.
18841
18842ssl_s_notafter : string
18843 Returns the end date presented by the server as a formatted string
18844 YYMMDDhhmmss[Z] when the outgoing connection was made over an SSL/TLS
18845 transport layer.
18846
18847ssl_s_notbefore : string
18848 Returns the start date presented by the server as a formatted string
18849 YYMMDDhhmmss[Z] when the outgoing connection was made over an SSL/TLS
18850 transport layer.
18851
18852ssl_s_i_dn([<entry>[,<occ>[,<format>]]]) : string
18853 When the outgoing connection was made over an SSL/TLS transport layer,
18854 returns the full distinguished name of the issuer of the certificate
18855 presented by the server when no <entry> is specified, or the value of the
18856 first given entry found from the beginning of the DN. If a positive/negative
18857 occurrence number is specified as the optional second argument, it returns
18858 the value of the nth given entry value from the beginning/end of the DN.
William Lallemand8f600c82020-06-26 09:55:06 +020018859 For instance, "ssl_s_i_dn(OU,2)" the second organization unit, and
18860 "ssl_s_i_dn(CN)" retrieves the common name.
William Lallemandbfa3e812020-06-25 20:07:18 +020018861 The <format> parameter allows you to receive the DN suitable for
18862 consumption by different protocols. Currently supported is rfc2253 for
18863 LDAP v3.
18864 If you'd like to modify the format only you can specify an empty string
18865 and zero for the first two parameters. Example: ssl_s_i_dn(,0,rfc2253)
18866
18867ssl_s_s_dn([<entry>[,<occ>[,<format>]]]) : string
18868 When the outgoing connection was made over an SSL/TLS transport layer,
18869 returns the full distinguished name of the subject of the certificate
18870 presented by the server when no <entry> is specified, or the value of the
18871 first given entry found from the beginning of the DN. If a positive/negative
18872 occurrence number is specified as the optional second argument, it returns
18873 the value of the nth given entry value from the beginning/end of the DN.
William Lallemand8f600c82020-06-26 09:55:06 +020018874 For instance, "ssl_s_s_dn(OU,2)" the second organization unit, and
18875 "ssl_s_s_dn(CN)" retrieves the common name.
William Lallemandbfa3e812020-06-25 20:07:18 +020018876 The <format> parameter allows you to receive the DN suitable for
18877 consumption by different protocols. Currently supported is rfc2253 for
18878 LDAP v3.
18879 If you'd like to modify the format only you can specify an empty string
18880 and zero for the first two parameters. Example: ssl_s_s_dn(,0,rfc2253)
18881
18882ssl_s_serial : binary
18883 Returns the serial of the certificate presented by the server when the
18884 outgoing connection was made over an SSL/TLS transport layer. When used for
18885 an ACL, the value(s) to match against can be passed in hexadecimal form.
18886
18887ssl_s_sha1 : binary
18888 Returns the SHA-1 fingerprint of the certificate presented by the server
18889 when the outgoing connection was made over an SSL/TLS transport layer. This
18890 can be used to know which certificate was chosen using SNI.
18891
18892ssl_s_sig_alg : string
18893 Returns the name of the algorithm used to sign the certificate presented by
18894 the server when the outgoing connection was made over an SSL/TLS transport
18895 layer.
18896
18897ssl_s_version : integer
18898 Returns the version of the certificate presented by the server when the
18899 outgoing connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020018900
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200189017.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020018902------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020018903
Willy Tarreau74ca5042013-06-11 23:12:07 +020018904Fetching samples from buffer contents is a bit different from the previous
18905sample fetches above because the sampled data are ephemeral. These data can
18906only be used when they're available and will be lost when they're forwarded.
18907For this reason, samples fetched from buffer contents during a request cannot
18908be used in a response for example. Even while the data are being fetched, they
18909can change. Sometimes it is necessary to set some delays or combine multiple
18910sample fetch methods to ensure that the expected data are complete and usable,
18911for example through TCP request content inspection. Please see the "tcp-request
18912content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020018913
Christopher Fauleta434a002021-03-25 11:58:51 +010018914Warning : Following sample fetches are ignored if used from HTTP proxies. They
18915 only deal with raw contents found in the buffers. On their side,
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018916 HTTP proxies use structured content. Thus raw representation of
Christopher Fauleta434a002021-03-25 11:58:51 +010018917 these data are meaningless. A warning is emitted if an ACL relies on
18918 one of the following sample fetches. But it is not possible to detect
18919 all invalid usage (for instance inside a log-format string or a
18920 sample expression). So be careful.
18921
Willy Tarreau74ca5042013-06-11 23:12:07 +020018922payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010018923 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020018924 "stick on", "stick match"), and for "res.payload" when used in the context of
18925 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010018926
Willy Tarreau74ca5042013-06-11 23:12:07 +020018927payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
18928 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010018929 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020018930 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010018931
Willy Tarreau74ca5042013-06-11 23:12:07 +020018932req.len : integer
18933req_len : integer (deprecated)
18934 Returns an integer value corresponding to the number of bytes present in the
18935 request buffer. This is mostly used in ACL. It is important to understand
18936 that this test does not return false as long as the buffer is changing. This
18937 means that a check with equality to zero will almost always immediately match
18938 at the beginning of the session, while a test for more data will wait for
Daniel Corbett9f0843f2021-05-08 10:50:37 -040018939 that data to come in and return false only when HAProxy is certain that no
Willy Tarreau74ca5042013-06-11 23:12:07 +020018940 more data will come in. This test was designed to be used with TCP request
18941 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018942
Willy Tarreau74ca5042013-06-11 23:12:07 +020018943req.payload(<offset>,<length>) : binary
18944 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020018945 in the request buffer. As a special case, if the <length> argument is zero,
18946 the the whole buffer from <offset> to the end is extracted. This can be used
18947 with ACLs in order to check for the presence of some content in a buffer at
18948 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018949
Willy Tarreau74ca5042013-06-11 23:12:07 +020018950 ACL alternatives :
18951 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018952
Willy Tarreau74ca5042013-06-11 23:12:07 +020018953req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
18954 This extracts a binary block whose size is specified at <offset1> for <length>
18955 bytes, and which starts at <offset2> if specified or just after the length in
18956 the request buffer. The <offset2> parameter also supports relative offsets if
18957 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018958
Willy Tarreau74ca5042013-06-11 23:12:07 +020018959 ACL alternatives :
18960 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018961
Willy Tarreau74ca5042013-06-11 23:12:07 +020018962 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018963
Willy Tarreau74ca5042013-06-11 23:12:07 +020018964req.proto_http : boolean
18965req_proto_http : boolean (deprecated)
18966 Returns true when data in the request buffer look like HTTP and correctly
18967 parses as such. It is the same parser as the common HTTP request parser which
18968 is used so there should be no surprises. The test does not match until the
18969 request is complete, failed or timed out. This test may be used to report the
18970 protocol in TCP logs, but the biggest use is to block TCP request analysis
18971 until a complete HTTP request is present in the buffer, for example to track
18972 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018973
Willy Tarreau74ca5042013-06-11 23:12:07 +020018974 Example:
18975 # track request counts per "base" (concatenation of Host+URL)
18976 tcp-request inspect-delay 10s
18977 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020018978 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020018979
Willy Tarreau74ca5042013-06-11 23:12:07 +020018980req.rdp_cookie([<name>]) : string
18981rdp_cookie([<name>]) : string (deprecated)
18982 When the request buffer looks like the RDP protocol, extracts the RDP cookie
18983 <name>, or any cookie if unspecified. The parser only checks for the first
18984 cookie, as illustrated in the RDP protocol specification. The cookie name is
18985 case insensitive. Generally the "MSTS" cookie name will be used, as it can
18986 contain the user name of the client connecting to the server if properly
18987 configured on the client. The "MSTSHASH" cookie is often used as well for
18988 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020018989
Willy Tarreau74ca5042013-06-11 23:12:07 +020018990 This differs from "balance rdp-cookie" in that any balancing algorithm may be
18991 used and thus the distribution of clients to backend servers is not linked to
18992 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
18993 such as "balance roundrobin" or "balance leastconn" will lead to a more even
18994 distribution of clients to backend servers than the hash used by "balance
18995 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020018996
Willy Tarreau74ca5042013-06-11 23:12:07 +020018997 ACL derivatives :
18998 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020018999
Willy Tarreau74ca5042013-06-11 23:12:07 +020019000 Example :
19001 listen tse-farm
19002 bind 0.0.0.0:3389
19003 # wait up to 5s for an RDP cookie in the request
19004 tcp-request inspect-delay 5s
19005 tcp-request content accept if RDP_COOKIE
19006 # apply RDP cookie persistence
19007 persist rdp-cookie
19008 # Persist based on the mstshash cookie
19009 # This is only useful makes sense if
19010 # balance rdp-cookie is not used
19011 stick-table type string size 204800
19012 stick on req.rdp_cookie(mstshash)
19013 server srv1 1.1.1.1:3389
19014 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019015
Willy Tarreau74ca5042013-06-11 23:12:07 +020019016 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
19017 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019018
Willy Tarreau74ca5042013-06-11 23:12:07 +020019019req.rdp_cookie_cnt([name]) : integer
19020rdp_cookie_cnt([name]) : integer (deprecated)
19021 Tries to parse the request buffer as RDP protocol, then returns an integer
19022 corresponding to the number of RDP cookies found. If an optional cookie name
19023 is passed, only cookies matching this name are considered. This is mostly
19024 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019025
Willy Tarreau74ca5042013-06-11 23:12:07 +020019026 ACL derivatives :
19027 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019028
Alex Zorin4afdd132018-12-30 13:56:28 +110019029req.ssl_alpn : string
19030 Returns a string containing the values of the Application-Layer Protocol
19031 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
19032 ClientHello message. Note that this only applies to raw contents found in the
19033 request buffer and not to the contents deciphered via an SSL data layer, so
19034 this will not work with "bind" lines having the "ssl" option. This is useful
19035 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020019036 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110019037
19038 Examples :
19039 # Wait for a client hello for at most 5 seconds
19040 tcp-request inspect-delay 5s
19041 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020019042 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110019043 default_backend bk_default
19044
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020019045req.ssl_ec_ext : boolean
19046 Returns a boolean identifying if client sent the Supported Elliptic Curves
19047 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020019048 message. This can be used to present ECC compatible clients with EC
19049 certificate and to use RSA for all others, on the same IP address. Note that
19050 this only applies to raw contents found in the request buffer and not to
19051 contents deciphered via an SSL data layer, so this will not work with "bind"
19052 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020019053
Willy Tarreau74ca5042013-06-11 23:12:07 +020019054req.ssl_hello_type : integer
19055req_ssl_hello_type : integer (deprecated)
19056 Returns an integer value containing the type of the SSL hello message found
19057 in the request buffer if the buffer contains data that parse as a complete
19058 SSL (v3 or superior) client hello message. Note that this only applies to raw
19059 contents found in the request buffer and not to contents deciphered via an
19060 SSL data layer, so this will not work with "bind" lines having the "ssl"
19061 option. This is mostly used in ACL to detect presence of an SSL hello message
19062 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019063
Willy Tarreau74ca5042013-06-11 23:12:07 +020019064req.ssl_sni : string
19065req_ssl_sni : string (deprecated)
19066 Returns a string containing the value of the Server Name TLS extension sent
19067 by a client in a TLS stream passing through the request buffer if the buffer
19068 contains data that parse as a complete SSL (v3 or superior) client hello
19069 message. Note that this only applies to raw contents found in the request
19070 buffer and not to contents deciphered via an SSL data layer, so this will not
Lukas Tribusa267b5d2020-07-19 00:25:06 +020019071 work with "bind" lines having the "ssl" option. This will only work for actual
19072 implicit TLS based protocols like HTTPS (443), IMAPS (993), SMTPS (465),
19073 however it will not work for explicit TLS based protocols, like SMTP (25/587)
19074 or IMAP (143). SNI normally contains the name of the host the client tries to
19075 connect to (for recent browsers). SNI is useful for allowing or denying access
19076 to certain hosts when SSL/TLS is used by the client. This test was designed to
19077 be used with TCP request content inspection. If content switching is needed,
19078 it is recommended to first wait for a complete client hello (type 1), like in
19079 the example below. See also "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019080
Willy Tarreau74ca5042013-06-11 23:12:07 +020019081 ACL derivatives :
Alex18c88772021-06-05 13:23:08 +020019082 req.ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019083
Willy Tarreau74ca5042013-06-11 23:12:07 +020019084 Examples :
19085 # Wait for a client hello for at most 5 seconds
19086 tcp-request inspect-delay 5s
19087 tcp-request content accept if { req_ssl_hello_type 1 }
Alex18c88772021-06-05 13:23:08 +020019088 use_backend bk_allow if { req.ssl_sni -f allowed_sites }
Willy Tarreau74ca5042013-06-11 23:12:07 +020019089 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020019090
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053019091req.ssl_st_ext : integer
19092 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
19093 Returns 1 if the client sent SessionTicket TLS Extension
19094 Returns 2 if the client also sent non-zero length TLS SessionTicket
19095 Note that this only applies to raw contents found in the request buffer and
19096 not to contents deciphered via an SSL data layer, so this will not work with
19097 "bind" lines having the "ssl" option. This can for example be used to detect
19098 whether the client sent a SessionTicket or not and stick it accordingly, if
19099 no SessionTicket then stick on SessionID or don't stick as there's no server
19100 side state is there when SessionTickets are in use.
19101
Willy Tarreau74ca5042013-06-11 23:12:07 +020019102req.ssl_ver : integer
19103req_ssl_ver : integer (deprecated)
19104 Returns an integer value containing the version of the SSL/TLS protocol of a
19105 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
19106 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
19107 composed of the major version multiplied by 65536, added to the minor
19108 version. Note that this only applies to raw contents found in the request
19109 buffer and not to contents deciphered via an SSL data layer, so this will not
19110 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010019111 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020019112 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019113
Willy Tarreau74ca5042013-06-11 23:12:07 +020019114 ACL derivatives :
19115 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010019116
Willy Tarreau47e8eba2013-09-11 23:28:46 +020019117res.len : integer
19118 Returns an integer value corresponding to the number of bytes present in the
19119 response buffer. This is mostly used in ACL. It is important to understand
19120 that this test does not return false as long as the buffer is changing. This
19121 means that a check with equality to zero will almost always immediately match
19122 at the beginning of the session, while a test for more data will wait for
Daniel Corbett9f0843f2021-05-08 10:50:37 -040019123 that data to come in and return false only when HAProxy is certain that no
Willy Tarreau47e8eba2013-09-11 23:28:46 +020019124 more data will come in. This test was designed to be used with TCP response
Christopher Faulete596d182020-05-05 17:46:34 +020019125 content inspection. But it may also be used in tcp-check based expect rules.
Willy Tarreau47e8eba2013-09-11 23:28:46 +020019126
Willy Tarreau74ca5042013-06-11 23:12:07 +020019127res.payload(<offset>,<length>) : binary
19128 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020019129 in the response buffer. As a special case, if the <length> argument is zero,
Christopher Faulete596d182020-05-05 17:46:34 +020019130 the whole buffer from <offset> to the end is extracted. This can be used
Willy Tarreau00f00842013-08-02 11:07:32 +020019131 with ACLs in order to check for the presence of some content in a buffer at
Christopher Faulete596d182020-05-05 17:46:34 +020019132 any location. It may also be used in tcp-check based expect rules.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019133
Willy Tarreau74ca5042013-06-11 23:12:07 +020019134res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
19135 This extracts a binary block whose size is specified at <offset1> for <length>
19136 bytes, and which starts at <offset2> if specified or just after the length in
19137 the response buffer. The <offset2> parameter also supports relative offsets
Christopher Faulete596d182020-05-05 17:46:34 +020019138 if prepended with a '+' or '-' sign. It may also be used in tcp-check based
19139 expect rules.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019140
Willy Tarreau74ca5042013-06-11 23:12:07 +020019141 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019142
Willy Tarreau971f7b62015-09-29 14:06:59 +020019143res.ssl_hello_type : integer
19144rep_ssl_hello_type : integer (deprecated)
19145 Returns an integer value containing the type of the SSL hello message found
19146 in the response buffer if the buffer contains data that parses as a complete
19147 SSL (v3 or superior) hello message. Note that this only applies to raw
19148 contents found in the response buffer and not to contents deciphered via an
19149 SSL data layer, so this will not work with "server" lines having the "ssl"
19150 option. This is mostly used in ACL to detect presence of an SSL hello message
19151 that is supposed to contain an SSL session ID usable for stickiness.
19152
Willy Tarreau74ca5042013-06-11 23:12:07 +020019153wait_end : boolean
19154 This fetch either returns true when the inspection period is over, or does
19155 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010019156 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020019157 actions, such as a delayed reject for some special addresses. Since it either
19158 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010019159 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020019160 "WAIT_END" is always usable without prior declaration. This test was designed
19161 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019162
Willy Tarreau74ca5042013-06-11 23:12:07 +020019163 Examples :
19164 # delay every incoming request by 2 seconds
19165 tcp-request inspect-delay 2s
19166 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010019167
Willy Tarreau74ca5042013-06-11 23:12:07 +020019168 # don't immediately tell bad guys they are rejected
19169 tcp-request inspect-delay 10s
19170 acl goodguys src 10.0.0.0/24
19171 acl badguys src 10.0.1.0/24
19172 tcp-request content accept if goodguys
19173 tcp-request content reject if badguys WAIT_END
19174 tcp-request content reject
19175
19176
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200191777.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020019178--------------------------------------
19179
19180It is possible to fetch samples from HTTP contents, requests and responses.
19181This application layer is also called layer 7. It is only possible to fetch the
19182data in this section when a full HTTP request or response has been parsed from
19183its respective request or response buffer. This is always the case with all
19184HTTP specific rules and for sections running with "mode http". When using TCP
19185content inspection, it may be necessary to support an inspection delay in order
19186to let the request or response come in first. These fetches may require a bit
19187more CPU resources than the layer 4 ones, but not much since the request and
19188response are indexed.
19189
Christopher Faulet4d37e532021-03-26 14:44:00 +010019190Note : Regarding HTTP processing from the tcp-request content rules, everything
19191 will work as expected from an HTTP proxy. However, from a TCP proxy,
19192 without an HTTP upgrade, it will only work for HTTP/1 content. For
19193 HTTP/2 content, only the preface is visible. Thus, it is only possible
19194 to rely to "req.proto_http", "req.ver" and eventually "method" sample
19195 fetches. All other L7 sample fetches will fail. After an HTTP upgrade,
19196 they will work in the same manner than from an HTTP proxy.
19197
Willy Tarreau74ca5042013-06-11 23:12:07 +020019198base : string
19199 This returns the concatenation of the first Host header and the path part of
19200 the request, which starts at the first slash and ends before the question
19201 mark. It can be useful in virtual hosted environments to detect URL abuses as
19202 well as to improve shared caches efficiency. Using this with a limited size
19203 stick table also allows one to collect statistics about most commonly
19204 requested objects by host/path. With ACLs it can allow simple content
19205 switching rules involving the host and the path at the same time, such as
19206 "www.example.com/favicon.ico". See also "path" and "uri".
19207
19208 ACL derivatives :
19209 base : exact string match
19210 base_beg : prefix match
19211 base_dir : subdir match
19212 base_dom : domain match
19213 base_end : suffix match
19214 base_len : length match
19215 base_reg : regex match
19216 base_sub : substring match
19217
19218base32 : integer
19219 This returns a 32-bit hash of the value returned by the "base" fetch method
19220 above. This is useful to track per-URL activity on high traffic sites without
19221 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020019222 memory. The output type is an unsigned integer. The hash function used is
19223 SDBM with full avalanche on the output. Technically, base32 is exactly equal
19224 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020019225
19226base32+src : binary
19227 This returns the concatenation of the base32 fetch above and the src fetch
19228 below. The resulting type is of type binary, with a size of 8 or 20 bytes
19229 depending on the source address family. This can be used to track per-IP,
19230 per-URL counters.
19231
Yves Lafonb4d37082021-02-11 11:01:28 +010019232baseq : string
19233 This returns the concatenation of the first Host header and the path part of
19234 the request with the query-string, which starts at the first slash. Using this
19235 instead of "base" allows one to properly identify the target resource, for
19236 statistics or caching use cases. See also "path", "pathq" and "base".
19237
William Lallemand65ad6e12014-01-31 15:08:02 +010019238capture.req.hdr(<idx>) : string
19239 This extracts the content of the header captured by the "capture request
19240 header", idx is the position of the capture keyword in the configuration.
19241 The first entry is an index of 0. See also: "capture request header".
19242
19243capture.req.method : string
19244 This extracts the METHOD of an HTTP request. It can be used in both request
19245 and response. Unlike "method", it can be used in both request and response
19246 because it's allocated.
19247
19248capture.req.uri : string
19249 This extracts the request's URI, which starts at the first slash and ends
19250 before the first space in the request (without the host part). Unlike "path"
19251 and "url", it can be used in both request and response because it's
19252 allocated.
19253
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020019254capture.req.ver : string
19255 This extracts the request's HTTP version and returns either "HTTP/1.0" or
19256 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
19257 logs because it relies on a persistent flag.
19258
William Lallemand65ad6e12014-01-31 15:08:02 +010019259capture.res.hdr(<idx>) : string
19260 This extracts the content of the header captured by the "capture response
19261 header", idx is the position of the capture keyword in the configuration.
19262 The first entry is an index of 0.
19263 See also: "capture response header"
19264
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020019265capture.res.ver : string
19266 This extracts the response's HTTP version and returns either "HTTP/1.0" or
19267 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
19268 persistent flag.
19269
Willy Tarreaua5910cc2015-05-02 00:46:08 +020019270req.body : binary
Christopher Fauletaf4dc4c2020-05-05 17:33:25 +020019271 This returns the HTTP request's available body as a block of data. It is
19272 recommended to use "option http-buffer-request" to be sure to wait, as much
19273 as possible, for the request's body.
Willy Tarreaua5910cc2015-05-02 00:46:08 +020019274
Thierry FOURNIER9826c772015-05-20 15:50:54 +020019275req.body_param([<name>) : string
19276 This fetch assumes that the body of the POST request is url-encoded. The user
19277 can check if the "content-type" contains the value
19278 "application/x-www-form-urlencoded". This extracts the first occurrence of the
19279 parameter <name> in the body, which ends before '&'. The parameter name is
19280 case-sensitive. If no name is given, any parameter will match, and the first
19281 one will be returned. The result is a string corresponding to the value of the
19282 parameter <name> as presented in the request body (no URL decoding is
19283 performed). Note that the ACL version of this fetch iterates over multiple
19284 parameters and will iteratively report all parameters values if no name is
19285 given.
19286
Willy Tarreaua5910cc2015-05-02 00:46:08 +020019287req.body_len : integer
19288 This returns the length of the HTTP request's available body in bytes. It may
19289 be lower than the advertised length if the body is larger than the buffer. It
Christopher Fauletaf4dc4c2020-05-05 17:33:25 +020019290 is recommended to use "option http-buffer-request" to be sure to wait, as
19291 much as possible, for the request's body.
Willy Tarreaua5910cc2015-05-02 00:46:08 +020019292
19293req.body_size : integer
19294 This returns the advertised length of the HTTP request's body in bytes. It
Christopher Fauletaf4dc4c2020-05-05 17:33:25 +020019295 will represent the advertised Content-Length header, or the size of the
19296 available data in case of chunked encoding.
Willy Tarreaua5910cc2015-05-02 00:46:08 +020019297
Willy Tarreau74ca5042013-06-11 23:12:07 +020019298req.cook([<name>]) : string
19299cook([<name>]) : string (deprecated)
19300 This extracts the last occurrence of the cookie name <name> on a "Cookie"
19301 header line from the request, and returns its value as string. If no name is
19302 specified, the first cookie value is returned. When used with ACLs, all
19303 matching cookies are evaluated. Spaces around the name and the value are
19304 ignored as requested by the Cookie header specification (RFC6265). The cookie
19305 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
19306 well return an empty value if it is present. Use the "found" match to detect
19307 presence. Use the res.cook() variant for response cookies sent by the server.
19308
19309 ACL derivatives :
19310 cook([<name>]) : exact string match
19311 cook_beg([<name>]) : prefix match
19312 cook_dir([<name>]) : subdir match
19313 cook_dom([<name>]) : domain match
19314 cook_end([<name>]) : suffix match
19315 cook_len([<name>]) : length match
19316 cook_reg([<name>]) : regex match
19317 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010019318
Willy Tarreau74ca5042013-06-11 23:12:07 +020019319req.cook_cnt([<name>]) : integer
19320cook_cnt([<name>]) : integer (deprecated)
19321 Returns an integer value representing the number of occurrences of the cookie
19322 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019323
Willy Tarreau74ca5042013-06-11 23:12:07 +020019324req.cook_val([<name>]) : integer
19325cook_val([<name>]) : integer (deprecated)
19326 This extracts the last occurrence of the cookie name <name> on a "Cookie"
19327 header line from the request, and converts its value to an integer which is
19328 returned. If no name is specified, the first cookie value is returned. When
19329 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020019330
Willy Tarreau74ca5042013-06-11 23:12:07 +020019331cookie([<name>]) : string (deprecated)
19332 This extracts the last occurrence of the cookie name <name> on a "Cookie"
19333 header line from the request, or a "Set-Cookie" header from the response, and
19334 returns its value as a string. A typical use is to get multiple clients
19335 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020019336 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020019337 multi-peer synchronization and state keeping across restarts. If no name is
19338 specified, the first cookie value is returned. This fetch should not be used
19339 anymore and should be replaced by req.cook() or res.cook() instead as it
19340 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019341
Willy Tarreau74ca5042013-06-11 23:12:07 +020019342hdr([<name>[,<occ>]]) : string
19343 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
19344 used on responses. Please refer to these respective fetches for more details.
19345 In case of doubt about the fetch direction, please use the explicit ones.
19346 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030019347 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019348
Willy Tarreau74ca5042013-06-11 23:12:07 +020019349req.fhdr(<name>[,<occ>]) : string
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019350 This returns the full value of the last occurrence of header <name> in an
19351 HTTP request. It differs from req.hdr() in that any commas present in the
19352 value are returned and are not used as delimiters. This is sometimes useful
19353 with headers such as User-Agent.
19354
19355 When used from an ACL, all occurrences are iterated over until a match is
19356 found.
19357
Willy Tarreau74ca5042013-06-11 23:12:07 +020019358 Optionally, a specific occurrence might be specified as a position number.
19359 Positive values indicate a position from the first occurrence, with 1 being
19360 the first one. Negative values indicate positions relative to the last one,
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019361 with -1 being the last one.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019362
Willy Tarreau74ca5042013-06-11 23:12:07 +020019363req.fhdr_cnt([<name>]) : integer
19364 Returns an integer value representing the number of occurrences of request
19365 header field name <name>, or the total number of header fields if <name> is
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019366 not specified. Like req.fhdr() it differs from res.hdr_cnt() by not splitting
19367 headers at commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019368
Willy Tarreau74ca5042013-06-11 23:12:07 +020019369req.hdr([<name>[,<occ>]]) : string
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019370 This returns the last comma-separated value of the header <name> in an HTTP
19371 request. The fetch considers any comma as a delimiter for distinct values.
19372 This is useful if you need to process headers that are defined to be a list
19373 of values, such as Accept, or X-Forwarded-For. If full-line headers are
19374 desired instead, use req.fhdr(). Please carefully check RFC 7231 to know how
19375 certain headers are supposed to be parsed. Also, some of them are case
19376 insensitive (e.g. Connection).
19377
19378 When used from an ACL, all occurrences are iterated over until a match is
19379 found.
19380
Willy Tarreau74ca5042013-06-11 23:12:07 +020019381 Optionally, a specific occurrence might be specified as a position number.
19382 Positive values indicate a position from the first occurrence, with 1 being
19383 the first one. Negative values indicate positions relative to the last one,
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019384 with -1 being the last one.
19385
19386 A typical use is with the X-Forwarded-For header once converted to IP,
19387 associated with an IP stick-table.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019388
Willy Tarreau74ca5042013-06-11 23:12:07 +020019389 ACL derivatives :
19390 hdr([<name>[,<occ>]]) : exact string match
19391 hdr_beg([<name>[,<occ>]]) : prefix match
19392 hdr_dir([<name>[,<occ>]]) : subdir match
19393 hdr_dom([<name>[,<occ>]]) : domain match
19394 hdr_end([<name>[,<occ>]]) : suffix match
19395 hdr_len([<name>[,<occ>]]) : length match
19396 hdr_reg([<name>[,<occ>]]) : regex match
19397 hdr_sub([<name>[,<occ>]]) : substring match
19398
19399req.hdr_cnt([<name>]) : integer
19400hdr_cnt([<header>]) : integer (deprecated)
19401 Returns an integer value representing the number of occurrences of request
19402 header field name <name>, or the total number of header field values if
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019403 <name> is not specified. Like req.hdr() it counts each comma separated
19404 part of the header's value. If counting of full-line headers is desired,
19405 then req.fhdr_cnt() should be used instead.
19406
19407 With ACLs, it can be used to detect presence, absence or abuse of a specific
19408 header, as well as to block request smuggling attacks by rejecting requests
19409 which contain more than one of certain headers.
19410
19411 Refer to req.hdr() for more information on header matching.
Willy Tarreau74ca5042013-06-11 23:12:07 +020019412
19413req.hdr_ip([<name>[,<occ>]]) : ip
19414hdr_ip([<name>[,<occ>]]) : ip (deprecated)
19415 This extracts the last occurrence of header <name> in an HTTP request,
19416 converts it to an IPv4 or IPv6 address and returns this address. When used
19417 with ACLs, all occurrences are checked, and if <name> is omitted, every value
Willy Tarreau7b0e00d2021-03-25 14:12:29 +010019418 of every header is checked. The parser strictly adheres to the format
19419 described in RFC7239, with the extension that IPv4 addresses may optionally
19420 be followed by a colon (':') and a valid decimal port number (0 to 65535),
19421 which will be silently dropped. All other forms will not match and will
19422 cause the address to be ignored.
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019423
19424 The <occ> parameter is processed as with req.hdr().
19425
19426 A typical use is with the X-Forwarded-For and X-Client-IP headers.
Willy Tarreau74ca5042013-06-11 23:12:07 +020019427
19428req.hdr_val([<name>[,<occ>]]) : integer
19429hdr_val([<name>[,<occ>]]) : integer (deprecated)
19430 This extracts the last occurrence of header <name> in an HTTP request, and
19431 converts it to an integer value. When used with ACLs, all occurrences are
19432 checked, and if <name> is omitted, every value of every header is checked.
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019433
19434 The <occ> parameter is processed as with req.hdr().
19435
19436 A typical use is with the X-Forwarded-For header.
Willy Tarreau74ca5042013-06-11 23:12:07 +020019437
Christopher Faulet687a68e2020-11-24 17:13:24 +010019438req.hdrs : string
19439 Returns the current request headers as string including the last empty line
19440 separating headers from the request body. The last empty line can be used to
19441 detect a truncated header block. This sample fetch is useful for some SPOE
19442 headers analyzers and for advanced logging.
19443
19444req.hdrs_bin : binary
19445 Returns the current request headers contained in preparsed binary form. This
19446 is useful for offloading some processing with SPOE. Each string is described
19447 by a length followed by the number of bytes indicated in the length. The
19448 length is represented using the variable integer encoding detailed in the
19449 SPOE documentation. The end of the list is marked by a couple of empty header
19450 names and values (length of 0 for both).
19451
19452 *(<str:header-name><str:header-value>)<empty string><empty string>
Frédéric Lécailleec891192019-02-26 15:02:35 +010019453
Christopher Faulet687a68e2020-11-24 17:13:24 +010019454 int: refer to the SPOE documentation for the encoding
19455 str: <int:length><bytes>
Frédéric Lécailleec891192019-02-26 15:02:35 +010019456
Willy Tarreau74ca5042013-06-11 23:12:07 +020019457http_auth(<userlist>) : boolean
19458 Returns a boolean indicating whether the authentication data received from
19459 the client match a username & password stored in the specified userlist. This
19460 fetch function is not really useful outside of ACLs. Currently only http
19461 basic auth is supported.
19462
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010019463http_auth_group(<userlist>) : string
19464 Returns a string corresponding to the user name found in the authentication
19465 data received from the client if both the user name and password are valid
19466 according to the specified userlist. The main purpose is to use it in ACLs
19467 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020019468 This fetch function is not really useful outside of ACLs. Currently only http
19469 basic auth is supported.
19470
19471 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010019472 http_auth_group(<userlist>) : group ...
19473 Returns true when the user extracted from the request and whose password is
19474 valid according to the specified userlist belongs to at least one of the
19475 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020019476
Christopher Fauleta4063562019-08-02 11:51:37 +020019477http_auth_pass : string
Willy Tarreauc9c6cdb2020-03-05 16:03:58 +010019478 Returns the user's password found in the authentication data received from
19479 the client, as supplied in the Authorization header. Not checks are
19480 performed by this sample fetch. Only Basic authentication is supported.
Christopher Fauleta4063562019-08-02 11:51:37 +020019481
19482http_auth_type : string
Willy Tarreauc9c6cdb2020-03-05 16:03:58 +010019483 Returns the authentication method found in the authentication data received from
19484 the client, as supplied in the Authorization header. Not checks are
19485 performed by this sample fetch. Only Basic authentication is supported.
Christopher Fauleta4063562019-08-02 11:51:37 +020019486
19487http_auth_user : string
Willy Tarreauc9c6cdb2020-03-05 16:03:58 +010019488 Returns the user name found in the authentication data received from the
19489 client, as supplied in the Authorization header. Not checks are performed by
19490 this sample fetch. Only Basic authentication is supported.
Christopher Fauleta4063562019-08-02 11:51:37 +020019491
Willy Tarreau74ca5042013-06-11 23:12:07 +020019492http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020019493 Returns true when the request being processed is the first one of the
19494 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020019495 from some requests when a request is not the first one, or to help grouping
19496 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020019497
Willy Tarreau74ca5042013-06-11 23:12:07 +020019498method : integer + string
19499 Returns an integer value corresponding to the method in the HTTP request. For
19500 example, "GET" equals 1 (check sources to establish the matching). Value 9
19501 means "other method" and may be converted to a string extracted from the
19502 stream. This should not be used directly as a sample, this is only meant to
19503 be used from ACLs, which transparently convert methods from patterns to these
19504 integer + string values. Some predefined ACL already check for most common
19505 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019506
Willy Tarreau74ca5042013-06-11 23:12:07 +020019507 ACL derivatives :
19508 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020019509
Willy Tarreau74ca5042013-06-11 23:12:07 +020019510 Example :
19511 # only accept GET and HEAD requests
19512 acl valid_method method GET HEAD
19513 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020019514
Willy Tarreau74ca5042013-06-11 23:12:07 +020019515path : string
19516 This extracts the request's URL path, which starts at the first slash and
19517 ends before the question mark (without the host part). A typical use is with
19518 prefetch-capable caches, and with portals which need to aggregate multiple
19519 information from databases and keep them in caches. Note that with outgoing
19520 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010019521 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020019522 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019523
Willy Tarreau74ca5042013-06-11 23:12:07 +020019524 ACL derivatives :
19525 path : exact string match
19526 path_beg : prefix match
19527 path_dir : subdir match
19528 path_dom : domain match
19529 path_end : suffix match
19530 path_len : length match
19531 path_reg : regex match
19532 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020019533
Christopher Faulete720c322020-09-02 17:25:18 +020019534pathq : string
19535 This extracts the request's URL path with the query-string, which starts at
19536 the first slash. This sample fetch is pretty handy to always retrieve a
19537 relative URI, excluding the scheme and the authority part, if any. Indeed,
19538 while it is the common representation for an HTTP/1.1 request target, in
19539 HTTP/2, an absolute URI is often used. This sample fetch will return the same
19540 result in both cases.
19541
Willy Tarreau49ad95c2015-01-19 15:06:26 +010019542query : string
19543 This extracts the request's query string, which starts after the first
19544 question mark. If no question mark is present, this fetch returns nothing. If
19545 a question mark is present but nothing follows, it returns an empty string.
19546 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010019547 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010019548 which stops before the question mark.
19549
Willy Tarreaueb27ec72015-02-20 13:55:29 +010019550req.hdr_names([<delim>]) : string
19551 This builds a string made from the concatenation of all header names as they
19552 appear in the request when the rule is evaluated. The default delimiter is
19553 the comma (',') but it may be overridden as an optional argument <delim>. In
19554 this case, only the first character of <delim> is considered.
19555
Willy Tarreau74ca5042013-06-11 23:12:07 +020019556req.ver : string
19557req_ver : string (deprecated)
19558 Returns the version string from the HTTP request, for example "1.1". This can
19559 be useful for logs, but is mostly there for ACL. Some predefined ACL already
19560 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019561
Willy Tarreau74ca5042013-06-11 23:12:07 +020019562 ACL derivatives :
19563 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020019564
Christopher Faulete596d182020-05-05 17:46:34 +020019565res.body : binary
19566 This returns the HTTP response's available body as a block of data. Unlike
19567 the request side, there is no directive to wait for the response's body. This
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019568 sample fetch is really useful (and usable) in the health-check context.
19569
19570 It may be used in tcp-check based expect rules.
Christopher Faulete596d182020-05-05 17:46:34 +020019571
19572res.body_len : integer
19573 This returns the length of the HTTP response available body in bytes. Unlike
19574 the request side, there is no directive to wait for the response's body. This
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019575 sample fetch is really useful (and usable) in the health-check context.
19576
19577 It may be used in tcp-check based expect rules.
Christopher Faulete596d182020-05-05 17:46:34 +020019578
19579res.body_size : integer
19580 This returns the advertised length of the HTTP response body in bytes. It
19581 will represent the advertised Content-Length header, or the size of the
19582 available data in case of chunked encoding. Unlike the request side, there is
19583 no directive to wait for the response body. This sample fetch is really
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019584 useful (and usable) in the health-check context.
19585
19586 It may be used in tcp-check based expect rules.
Christopher Faulete596d182020-05-05 17:46:34 +020019587
Remi Tricot-Le Bretonbf971212020-10-27 11:55:57 +010019588res.cache_hit : boolean
19589 Returns the boolean "true" value if the response has been built out of an
19590 HTTP cache entry, otherwise returns boolean "false".
19591
19592res.cache_name : string
19593 Returns a string containing the name of the HTTP cache that was used to
19594 build the HTTP response if res.cache_hit is true, otherwise returns an
19595 empty string.
19596
Willy Tarreau74ca5042013-06-11 23:12:07 +020019597res.comp : boolean
19598 Returns the boolean "true" value if the response has been compressed by
19599 HAProxy, otherwise returns boolean "false". This may be used to add
19600 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019601
Willy Tarreau74ca5042013-06-11 23:12:07 +020019602res.comp_algo : string
19603 Returns a string containing the name of the algorithm used if the response
19604 was compressed by HAProxy, for example : "deflate". This may be used to add
19605 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019606
Willy Tarreau74ca5042013-06-11 23:12:07 +020019607res.cook([<name>]) : string
19608scook([<name>]) : string (deprecated)
19609 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
19610 header line from the response, and returns its value as string. If no name is
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019611 specified, the first cookie value is returned.
19612
19613 It may be used in tcp-check based expect rules.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020019614
Willy Tarreau74ca5042013-06-11 23:12:07 +020019615 ACL derivatives :
19616 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020019617
Willy Tarreau74ca5042013-06-11 23:12:07 +020019618res.cook_cnt([<name>]) : integer
19619scook_cnt([<name>]) : integer (deprecated)
19620 Returns an integer value representing the number of occurrences of the cookie
19621 <name> in the response, or all cookies if <name> is not specified. This is
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019622 mostly useful when combined with ACLs to detect suspicious responses.
19623
19624 It may be used in tcp-check based expect rules.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019625
Willy Tarreau74ca5042013-06-11 23:12:07 +020019626res.cook_val([<name>]) : integer
19627scook_val([<name>]) : integer (deprecated)
19628 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
19629 header line from the response, and converts its value to an integer which is
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019630 returned. If no name is specified, the first cookie value is returned.
19631
19632 It may be used in tcp-check based expect rules.
Willy Tarreaud63335a2010-02-26 12:56:52 +010019633
Willy Tarreau74ca5042013-06-11 23:12:07 +020019634res.fhdr([<name>[,<occ>]]) : string
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019635 This fetch works like the req.fhdr() fetch with the difference that it acts
19636 on the headers within an HTTP response.
19637
19638 Like req.fhdr() the res.fhdr() fetch returns full values. If the header is
19639 defined to be a list you should use res.hdr().
19640
19641 This fetch is sometimes useful with headers such as Date or Expires.
19642
19643 It may be used in tcp-check based expect rules.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019644
Willy Tarreau74ca5042013-06-11 23:12:07 +020019645res.fhdr_cnt([<name>]) : integer
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019646 This fetch works like the req.fhdr_cnt() fetch with the difference that it
19647 acts on the headers within an HTTP response.
19648
19649 Like req.fhdr_cnt() the res.fhdr_cnt() fetch acts on full values. If the
19650 header is defined to be a list you should use res.hdr_cnt().
19651
19652 It may be used in tcp-check based expect rules.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019653
Willy Tarreau74ca5042013-06-11 23:12:07 +020019654res.hdr([<name>[,<occ>]]) : string
19655shdr([<name>[,<occ>]]) : string (deprecated)
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019656 This fetch works like the req.hdr() fetch with the difference that it acts
19657 on the headers within an HTTP response.
19658
Ilya Shipitsinacf84592021-02-06 22:29:08 +050019659 Like req.hdr() the res.hdr() fetch considers the comma to be a delimiter. If
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019660 this is not desired res.fhdr() should be used.
19661
19662 It may be used in tcp-check based expect rules.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019663
Willy Tarreau74ca5042013-06-11 23:12:07 +020019664 ACL derivatives :
19665 shdr([<name>[,<occ>]]) : exact string match
19666 shdr_beg([<name>[,<occ>]]) : prefix match
19667 shdr_dir([<name>[,<occ>]]) : subdir match
19668 shdr_dom([<name>[,<occ>]]) : domain match
19669 shdr_end([<name>[,<occ>]]) : suffix match
19670 shdr_len([<name>[,<occ>]]) : length match
19671 shdr_reg([<name>[,<occ>]]) : regex match
19672 shdr_sub([<name>[,<occ>]]) : substring match
19673
19674res.hdr_cnt([<name>]) : integer
19675shdr_cnt([<name>]) : integer (deprecated)
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019676 This fetch works like the req.hdr_cnt() fetch with the difference that it
19677 acts on the headers within an HTTP response.
19678
19679 Like req.hdr_cnt() the res.hdr_cnt() fetch considers the comma to be a
Ilya Shipitsinacf84592021-02-06 22:29:08 +050019680 delimiter. If this is not desired res.fhdr_cnt() should be used.
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019681
19682 It may be used in tcp-check based expect rules.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019683
Willy Tarreau74ca5042013-06-11 23:12:07 +020019684res.hdr_ip([<name>[,<occ>]]) : ip
19685shdr_ip([<name>[,<occ>]]) : ip (deprecated)
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019686 This fetch works like the req.hdr_ip() fetch with the difference that it
19687 acts on the headers within an HTTP response.
19688
19689 This can be useful to learn some data into a stick table.
19690
19691 It may be used in tcp-check based expect rules.
Willy Tarreau6a06a402007-07-15 20:15:28 +020019692
Willy Tarreaueb27ec72015-02-20 13:55:29 +010019693res.hdr_names([<delim>]) : string
19694 This builds a string made from the concatenation of all header names as they
19695 appear in the response when the rule is evaluated. The default delimiter is
19696 the comma (',') but it may be overridden as an optional argument <delim>. In
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019697 this case, only the first character of <delim> is considered.
19698
19699 It may be used in tcp-check based expect rules.
Willy Tarreaueb27ec72015-02-20 13:55:29 +010019700
Willy Tarreau74ca5042013-06-11 23:12:07 +020019701res.hdr_val([<name>[,<occ>]]) : integer
19702shdr_val([<name>[,<occ>]]) : integer (deprecated)
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019703 This fetch works like the req.hdr_val() fetch with the difference that it
19704 acts on the headers within an HTTP response.
19705
19706 This can be useful to learn some data into a stick table.
19707
19708 It may be used in tcp-check based expect rules.
Christopher Faulete596d182020-05-05 17:46:34 +020019709
19710res.hdrs : string
19711 Returns the current response headers as string including the last empty line
19712 separating headers from the request body. The last empty line can be used to
19713 detect a truncated header block. This sample fetch is useful for some SPOE
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019714 headers analyzers and for advanced logging.
19715
19716 It may also be used in tcp-check based expect rules.
Christopher Faulete596d182020-05-05 17:46:34 +020019717
19718res.hdrs_bin : binary
19719 Returns the current response headers contained in preparsed binary form. This
19720 is useful for offloading some processing with SPOE. It may be used in
19721 tcp-check based expect rules. Each string is described by a length followed
19722 by the number of bytes indicated in the length. The length is represented
19723 using the variable integer encoding detailed in the SPOE documentation. The
19724 end of the list is marked by a couple of empty header names and values
19725 (length of 0 for both).
19726
19727 *(<str:header-name><str:header-value>)<empty string><empty string>
19728
19729 int: refer to the SPOE documentation for the encoding
19730 str: <int:length><bytes>
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010019731
Willy Tarreau74ca5042013-06-11 23:12:07 +020019732res.ver : string
19733resp_ver : string (deprecated)
19734 Returns the version string from the HTTP response, for example "1.1". This
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019735 can be useful for logs, but is mostly there for ACL.
19736
19737 It may be used in tcp-check based expect rules.
Willy Tarreau0e698542011-09-16 08:32:32 +020019738
Willy Tarreau74ca5042013-06-11 23:12:07 +020019739 ACL derivatives :
19740 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010019741
Willy Tarreau74ca5042013-06-11 23:12:07 +020019742set-cookie([<name>]) : string (deprecated)
19743 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
19744 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020019745 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020019746 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010019747
Willy Tarreau74ca5042013-06-11 23:12:07 +020019748 This fetch function is deprecated and has been superseded by the "res.cook"
19749 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010019750
Willy Tarreau74ca5042013-06-11 23:12:07 +020019751status : integer
19752 Returns an integer containing the HTTP status code in the HTTP response, for
19753 example, 302. It is mostly used within ACLs and integer ranges, for example,
Tim Duesterhus27c70ae2021-01-23 17:50:21 +010019754 to remove any Location header if the response is not a 3xx.
19755
19756 It may be used in tcp-check based expect rules.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019757
Thierry Fournier0e00dca2016-04-07 15:47:40 +020019758unique-id : string
19759 Returns the unique-id attached to the request. The directive
19760 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
19761 fails. Note that the unique-id is usually used with HTTP requests, however this
19762 sample fetch can be used with other protocols. Obviously, if it is used with
19763 other protocols than HTTP, the unique-id-format directive must not contain
19764 HTTP parts. See: unique-id-format and unique-id-header
19765
Willy Tarreau74ca5042013-06-11 23:12:07 +020019766url : string
19767 This extracts the request's URL as presented in the request. A typical use is
19768 with prefetch-capable caches, and with portals which need to aggregate
19769 multiple information from databases and keep them in caches. With ACLs, using
19770 "path" is preferred over using "url", because clients may send a full URL as
19771 is normally done with proxies. The only real use is to match "*" which does
19772 not match in "path", and for which there is already a predefined ACL. See
19773 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019774
Willy Tarreau74ca5042013-06-11 23:12:07 +020019775 ACL derivatives :
19776 url : exact string match
19777 url_beg : prefix match
19778 url_dir : subdir match
19779 url_dom : domain match
19780 url_end : suffix match
19781 url_len : length match
19782 url_reg : regex match
19783 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019784
Willy Tarreau74ca5042013-06-11 23:12:07 +020019785url_ip : ip
19786 This extracts the IP address from the request's URL when the host part is
19787 presented as an IP address. Its use is very limited. For instance, a
19788 monitoring system might use this field as an alternative for the source IP in
19789 order to test what path a given source address would follow, or to force an
19790 entry in a table for a given source address. With ACLs it can be used to
19791 restrict access to certain systems through a proxy, for example when combined
19792 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019793
Willy Tarreau74ca5042013-06-11 23:12:07 +020019794url_port : integer
19795 This extracts the port part from the request's URL. Note that if the port is
19796 not specified in the request, port 80 is assumed. With ACLs it can be used to
19797 restrict access to certain systems through a proxy, for example when combined
19798 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019799
Willy Tarreau1ede1da2015-05-07 16:06:18 +020019800urlp([<name>[,<delim>]]) : string
19801url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020019802 This extracts the first occurrence of the parameter <name> in the query
19803 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020019804 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
19805 any parameter will match, and the first one will be returned. The result is
19806 a string corresponding to the value of the parameter <name> as presented in
19807 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020019808 stickiness based on a client ID, to extract an application cookie passed as a
19809 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020019810 this fetch iterates over multiple parameters and will iteratively report all
19811 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019812
Willy Tarreau74ca5042013-06-11 23:12:07 +020019813 ACL derivatives :
19814 urlp(<name>[,<delim>]) : exact string match
19815 urlp_beg(<name>[,<delim>]) : prefix match
19816 urlp_dir(<name>[,<delim>]) : subdir match
19817 urlp_dom(<name>[,<delim>]) : domain match
19818 urlp_end(<name>[,<delim>]) : suffix match
19819 urlp_len(<name>[,<delim>]) : length match
19820 urlp_reg(<name>[,<delim>]) : regex match
19821 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019822
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019823
Willy Tarreau74ca5042013-06-11 23:12:07 +020019824 Example :
19825 # match http://example.com/foo?PHPSESSIONID=some_id
19826 stick on urlp(PHPSESSIONID)
19827 # match http://example.com/foo;JSESSIONID=some_id
19828 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020019829
Jarno Huuskonen676f6222017-03-30 09:19:45 +030019830urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020019831 See "urlp" above. This one extracts the URL parameter <name> in the request
19832 and converts it to an integer value. This can be used for session stickiness
19833 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020019834
Dragan Dosen0070cd52016-06-16 12:19:49 +020019835url32 : integer
19836 This returns a 32-bit hash of the value obtained by concatenating the first
19837 Host header and the whole URL including parameters (not only the path part of
19838 the request, as in the "base32" fetch above). This is useful to track per-URL
19839 activity. A shorter hash is stored, saving a lot of memory. The output type
19840 is an unsigned integer.
19841
19842url32+src : binary
19843 This returns the concatenation of the "url32" fetch and the "src" fetch. The
19844 resulting type is of type binary, with a size of 8 or 20 bytes depending on
19845 the source address family. This can be used to track per-IP, per-URL counters.
19846
Christopher Faulet16032ab2020-04-30 11:30:00 +020019847
Christopher Faulete596d182020-05-05 17:46:34 +0200198487.3.7. Fetching samples for developers
Christopher Fauletd47941d2020-01-08 14:40:19 +010019849---------------------------------------
19850
19851This set of sample fetch methods is reserved to developers and must never be
19852used on a production environment, except on developer demand, for debugging
19853purposes. Moreover, no special care will be taken on backwards compatibility.
19854There is no warranty the following sample fetches will never change, be renamed
19855or simply removed. So be really careful if you should use one of them. To avoid
19856any ambiguity, these sample fetches are placed in the dedicated scope "internal",
19857for instance "internal.strm.is_htx".
19858
19859internal.htx.data : integer
19860 Returns the size in bytes used by data in the HTX message associated to a
19861 channel. The channel is chosen depending on the sample direction.
19862
19863internal.htx.free : integer
19864 Returns the free space (size - used) in bytes in the HTX message associated
19865 to a channel. The channel is chosen depending on the sample direction.
19866
19867internal.htx.free_data : integer
19868 Returns the free space for the data in bytes in the HTX message associated to
19869 a channel. The channel is chosen depending on the sample direction.
19870
19871internal.htx.has_eom : boolean
Christopher Fauletd1ac2b92020-12-02 19:12:22 +010019872 Returns true if the HTX message associated to a channel contains the
19873 end-of-message flag (EOM). Otherwise, it returns false. The channel is chosen
19874 depending on the sample direction.
Christopher Fauletd47941d2020-01-08 14:40:19 +010019875
19876internal.htx.nbblks : integer
19877 Returns the number of blocks present in the HTX message associated to a
19878 channel. The channel is chosen depending on the sample direction.
19879
19880internal.htx.size : integer
19881 Returns the total size in bytes of the HTX message associated to a
19882 channel. The channel is chosen depending on the sample direction.
19883
19884internal.htx.used : integer
19885 Returns the total size used in bytes (data + metadata) in the HTX message
19886 associated to a channel. The channel is chosen depending on the sample
19887 direction.
19888
19889internal.htx_blk.size(<idx>) : integer
19890 Returns the size of the block at the position <idx> in the HTX message
19891 associated to a channel or 0 if it does not exist. The channel is chosen
19892 depending on the sample direction. <idx> may be any positive integer or one
19893 of the special value :
19894 * head : The oldest inserted block
19895 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019896 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019897
19898internal.htx_blk.type(<idx>) : string
19899 Returns the type of the block at the position <idx> in the HTX message
19900 associated to a channel or "HTX_BLK_UNUSED" if it does not exist. The channel
19901 is chosen depending on the sample direction. <idx> may be any positive
19902 integer or one of the special value :
19903 * head : The oldest inserted block
19904 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019905 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019906
19907internal.htx_blk.data(<idx>) : binary
19908 Returns the value of the DATA block at the position <idx> in the HTX message
19909 associated to a channel or an empty string if it does not exist or if it is
19910 not a DATA block. The channel is chosen depending on the sample direction.
19911 <idx> may be any positive integer or one of the special value :
19912
19913 * head : The oldest inserted block
19914 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019915 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019916
19917internal.htx_blk.hdrname(<idx>) : string
19918 Returns the header name of the HEADER block at the position <idx> in the HTX
19919 message associated to a channel or an empty string if it does not exist or if
19920 it is not an HEADER block. The channel is chosen depending on the sample
19921 direction. <idx> may be any positive integer or one of the special value :
19922
19923 * head : The oldest inserted block
19924 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019925 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019926
19927internal.htx_blk.hdrval(<idx>) : string
19928 Returns the header value of the HEADER block at the position <idx> in the HTX
19929 message associated to a channel or an empty string if it does not exist or if
19930 it is not an HEADER block. The channel is chosen depending on the sample
19931 direction. <idx> may be any positive integer or one of the special value :
19932
19933 * head : The oldest inserted block
19934 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019935 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019936
19937internal.htx_blk.start_line(<idx>) : string
19938 Returns the value of the REQ_SL or RES_SL block at the position <idx> in the
19939 HTX message associated to a channel or an empty string if it does not exist
19940 or if it is not a SL block. The channel is chosen depending on the sample
19941 direction. <idx> may be any positive integer or one of the special value :
19942
19943 * head : The oldest inserted block
19944 * tail : The newest inserted block
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050019945 * first : The first block where to (re)start the analysis
Christopher Fauletd47941d2020-01-08 14:40:19 +010019946
19947internal.strm.is_htx : boolean
19948 Returns true if the current stream is an HTX stream. It means the data in the
19949 channels buffers are stored using the internal HTX representation. Otherwise,
19950 it returns false.
19951
19952
Willy Tarreau74ca5042013-06-11 23:12:07 +0200199537.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019954---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010019955
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019956Some predefined ACLs are hard-coded so that they do not have to be declared in
19957every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020019958order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010019959
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019960ACL name Equivalent to Usage
Christopher Faulet779184e2021-04-01 17:24:04 +020019961---------------+----------------------------------+------------------------------------------------------
19962FALSE always_false never match
19963HTTP req.proto_http match if request protocol is valid HTTP
19964HTTP_1.0 req.ver 1.0 match if HTTP request version is 1.0
19965HTTP_1.1 req.ver 1.1 match if HTTP request version is 1.1
Christopher Faulet8043e832021-03-26 16:00:54 +010019966HTTP_2.0 req.ver 2.0 match if HTTP request version is 2.0
Christopher Faulet779184e2021-04-01 17:24:04 +020019967HTTP_CONTENT req.hdr_val(content-length) gt 0 match an existing content-length in the HTTP request
19968HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
19969HTTP_URL_SLASH url_beg / match URL beginning with "/"
19970HTTP_URL_STAR url * match URL equal to "*"
19971LOCALHOST src 127.0.0.1/8 match connection from local host
19972METH_CONNECT method CONNECT match HTTP CONNECT method
19973METH_DELETE method DELETE match HTTP DELETE method
19974METH_GET method GET HEAD match HTTP GET or HEAD method
19975METH_HEAD method HEAD match HTTP HEAD method
19976METH_OPTIONS method OPTIONS match HTTP OPTIONS method
19977METH_POST method POST match HTTP POST method
19978METH_PUT method PUT match HTTP PUT method
19979METH_TRACE method TRACE match HTTP TRACE method
19980RDP_COOKIE req.rdp_cookie_cnt gt 0 match presence of an RDP cookie in the request buffer
19981REQ_CONTENT req.len gt 0 match data in the request buffer
19982TRUE always_true always match
19983WAIT_END wait_end wait for end of content analysis
19984---------------+----------------------------------+------------------------------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010019985
Willy Tarreaub937b7e2010-01-12 15:27:54 +010019986
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200199878. Logging
19988----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010019989
Willy Tarreaucc6c8912009-02-22 10:53:55 +010019990One of HAProxy's strong points certainly lies is its precise logs. It probably
19991provides the finest level of information available for such a product, which is
19992very important for troubleshooting complex environments. Standard information
19993provided in logs include client ports, TCP/HTTP state timers, precise session
19994state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010019995to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010019996headers.
19997
19998In order to improve administrators reactivity, it offers a great transparency
19999about encountered problems, both internal and external, and it is possible to
20000send logs to different sources at the same time with different level filters :
20001
20002 - global process-level logs (system errors, start/stop, etc..)
20003 - per-instance system and internal errors (lack of resource, bugs, ...)
20004 - per-instance external troubles (servers up/down, max connections)
20005 - per-instance activity (client connections), either at the establishment or
20006 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010020007 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060020008 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020009
20010The ability to distribute different levels of logs to different log servers
20011allow several production teams to interact and to fix their problems as soon
20012as possible. For example, the system team might monitor system-wide errors,
20013while the application team might be monitoring the up/down for their servers in
20014real time, and the security team might analyze the activity logs with one hour
20015delay.
20016
20017
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200200188.1. Log levels
20019---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020020
Simon Hormandf791f52011-05-29 15:01:10 +090020021TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020022source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090020023HTTP request, HTTP return code, number of bytes transmitted, conditions
20024in which the session ended, and even exchanged cookies values. For example
20025track a particular user's problems. All messages may be sent to up to two
20026syslog servers. Check the "log" keyword in section 4.2 for more information
20027about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020028
20029
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200200308.2. Log formats
20031----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020032
William Lallemand48940402012-01-30 16:47:22 +010020033HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090020034and will be detailed in the following sections. A few of them may vary
20035slightly with the configuration, due to indicators specific to certain
20036options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020037
20038 - the default format, which is very basic and very rarely used. It only
20039 provides very basic information about the incoming connection at the moment
20040 it is accepted : source IP:port, destination IP:port, and frontend-name.
20041 This mode will eventually disappear so it will not be described to great
20042 extents.
20043
20044 - the TCP format, which is more advanced. This format is enabled when "option
20045 tcplog" is set on the frontend. HAProxy will then usually wait for the
20046 connection to terminate before logging. This format provides much richer
20047 information, such as timers, connection counts, queue size, etc... This
20048 format is recommended for pure TCP proxies.
20049
20050 - the HTTP format, which is the most advanced for HTTP proxying. This format
20051 is enabled when "option httplog" is set on the frontend. It provides the
20052 same information as the TCP format with some HTTP-specific fields such as
20053 the request, the status code, and captures of headers and cookies. This
20054 format is recommended for HTTP proxies.
20055
Emeric Brun3a058f32009-06-30 18:26:00 +020020056 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
20057 fields arranged in the same order as the CLF format. In this mode, all
20058 timers, captures, flags, etc... appear one per field after the end of the
20059 common fields, in the same order they appear in the standard HTTP format.
20060
William Lallemand48940402012-01-30 16:47:22 +010020061 - the custom log format, allows you to make your own log line.
20062
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020063Next sections will go deeper into details for each of these formats. Format
20064specification will be performed on a "field" basis. Unless stated otherwise, a
20065field is a portion of text delimited by any number of spaces. Since syslog
20066servers are susceptible of inserting fields at the beginning of a line, it is
20067always assumed that the first field is the one containing the process name and
20068identifier.
20069
20070Note : Since log lines may be quite long, the log examples in sections below
20071 might be broken into multiple lines. The example log lines will be
20072 prefixed with 3 closing angle brackets ('>>>') and each time a log is
20073 broken into multiple lines, each non-final line will end with a
20074 backslash ('\') and the next line will start indented by two characters.
20075
20076
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200200778.2.1. Default log format
20078-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020079
20080This format is used when no specific option is set. The log is emitted as soon
20081as the connection is accepted. One should note that this currently is the only
20082format which logs the request's destination IP and ports.
20083
20084 Example :
20085 listen www
20086 mode http
20087 log global
20088 server srv1 127.0.0.1:8000
20089
20090 >>> Feb 6 12:12:09 localhost \
20091 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
20092 (www/HTTP)
20093
20094 Field Format Extract from the example above
20095 1 process_name '[' pid ']:' haproxy[14385]:
20096 2 'Connect from' Connect from
20097 3 source_ip ':' source_port 10.0.1.2:33312
20098 4 'to' to
20099 5 destination_ip ':' destination_port 10.0.3.31:8012
20100 6 '(' frontend_name '/' mode ')' (www/HTTP)
20101
20102Detailed fields description :
20103 - "source_ip" is the IP address of the client which initiated the connection.
20104 - "source_port" is the TCP port of the client which initiated the connection.
20105 - "destination_ip" is the IP address the client connected to.
20106 - "destination_port" is the TCP port the client connected to.
20107 - "frontend_name" is the name of the frontend (or listener) which received
20108 and processed the connection.
20109 - "mode is the mode the frontend is operating (TCP or HTTP).
20110
Willy Tarreauceb24bc2010-11-09 12:46:41 +010020111In case of a UNIX socket, the source and destination addresses are marked as
20112"unix:" and the ports reflect the internal ID of the socket which accepted the
20113connection (the same ID as reported in the stats).
20114
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020115It is advised not to use this deprecated format for newer installations as it
20116will eventually disappear.
20117
20118
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200201198.2.2. TCP log format
20120---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020121
20122The TCP format is used when "option tcplog" is specified in the frontend, and
20123is the recommended format for pure TCP proxies. It provides a lot of precious
20124information for troubleshooting. Since this format includes timers and byte
20125counts, the log is normally emitted at the end of the session. It can be
20126emitted earlier if "option logasap" is specified, which makes sense in most
20127environments with long sessions such as remote terminals. Sessions which match
20128the "monitor" rules are never logged. It is also possible not to emit logs for
20129sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020020130specifying "option dontlognull" in the frontend. Successful connections will
20131not be logged if "option dontlog-normal" is specified in the frontend. A few
20132fields may slightly vary depending on some configuration options, those are
20133marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020134
20135 Example :
20136 frontend fnt
20137 mode tcp
20138 option tcplog
20139 log global
20140 default_backend bck
20141
20142 backend bck
20143 server srv1 127.0.0.1:8000
20144
20145 >>> Feb 6 12:12:56 localhost \
20146 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
20147 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
20148
20149 Field Format Extract from the example above
20150 1 process_name '[' pid ']:' haproxy[14387]:
20151 2 client_ip ':' client_port 10.0.1.2:33313
20152 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
20153 4 frontend_name fnt
20154 5 backend_name '/' server_name bck/srv1
20155 6 Tw '/' Tc '/' Tt* 0/0/5007
20156 7 bytes_read* 212
20157 8 termination_state --
20158 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
20159 10 srv_queue '/' backend_queue 0/0
20160
20161Detailed fields description :
20162 - "client_ip" is the IP address of the client which initiated the TCP
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020163 connection to HAProxy. If the connection was accepted on a UNIX socket
Willy Tarreauceb24bc2010-11-09 12:46:41 +010020164 instead, the IP address would be replaced with the word "unix". Note that
20165 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010020166 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010020167 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010020168 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020169
20170 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010020171 If the connection was accepted on a UNIX socket instead, the port would be
20172 replaced with the ID of the accepting socket, which is also reported in the
20173 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020174
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020175 - "accept_date" is the exact date when the connection was received by HAProxy
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020176 (which might be very slightly different from the date observed on the
20177 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020020178 the same date which may appear in any upstream firewall's log. When used in
20179 HTTP mode, the accept_date field will be reset to the first moment the
20180 connection is ready to receive a new request (end of previous response for
20181 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020182
20183 - "frontend_name" is the name of the frontend (or listener) which received
20184 and processed the connection.
20185
20186 - "backend_name" is the name of the backend (or listener) which was selected
20187 to manage the connection to the server. This will be the same as the
20188 frontend if no switching rule has been applied, which is common for TCP
20189 applications.
20190
20191 - "server_name" is the name of the last server to which the connection was
20192 sent, which might differ from the first one if there were connection errors
20193 and a redispatch occurred. Note that this server belongs to the backend
20194 which processed the request. If the connection was aborted before reaching
20195 a server, "<NOSRV>" is indicated instead of a server name.
20196
20197 - "Tw" is the total time in milliseconds spent waiting in the various queues.
20198 It can be "-1" if the connection was aborted before reaching the queue.
20199 See "Timers" below for more details.
20200
20201 - "Tc" is the total time in milliseconds spent waiting for the connection to
20202 establish to the final server, including retries. It can be "-1" if the
20203 connection was aborted before a connection could be established. See
20204 "Timers" below for more details.
20205
20206 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030020207 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020208 "option logasap" was specified, then the time counting stops at the moment
20209 the log is emitted. In this case, a '+' sign is prepended before the value,
20210 indicating that the final one will be larger. See "Timers" below for more
20211 details.
20212
20213 - "bytes_read" is the total number of bytes transmitted from the server to
20214 the client when the log is emitted. If "option logasap" is specified, the
20215 this value will be prefixed with a '+' sign indicating that the final one
20216 may be larger. Please note that this value is a 64-bit counter, so log
20217 analysis tools must be able to handle it without overflowing.
20218
20219 - "termination_state" is the condition the session was in when the session
20220 ended. This indicates the session state, which side caused the end of
20221 session to happen, and for what reason (timeout, error, ...). The normal
20222 flags should be "--", indicating the session was closed by either end with
20223 no data remaining in buffers. See below "Session state at disconnection"
20224 for more details.
20225
20226 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040020227 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020228 limits have been reached. For instance, if actconn is close to 512 when
20229 multiple connection errors occur, chances are high that the system limits
20230 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020020231 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020232
20233 - "feconn" is the total number of concurrent connections on the frontend when
20234 the session was logged. It is useful to estimate the amount of resource
20235 required to sustain high loads, and to detect when the frontend's "maxconn"
20236 has been reached. Most often when this value increases by huge jumps, it is
20237 because there is congestion on the backend servers, but sometimes it can be
20238 caused by a denial of service attack.
20239
20240 - "beconn" is the total number of concurrent connections handled by the
20241 backend when the session was logged. It includes the total number of
20242 concurrent connections active on servers as well as the number of
20243 connections pending in queues. It is useful to estimate the amount of
20244 additional servers needed to support high loads for a given application.
20245 Most often when this value increases by huge jumps, it is because there is
20246 congestion on the backend servers, but sometimes it can be caused by a
20247 denial of service attack.
20248
20249 - "srv_conn" is the total number of concurrent connections still active on
20250 the server when the session was logged. It can never exceed the server's
20251 configured "maxconn" parameter. If this value is very often close or equal
20252 to the server's "maxconn", it means that traffic regulation is involved a
20253 lot, meaning that either the server's maxconn value is too low, or that
20254 there aren't enough servers to process the load with an optimal response
20255 time. When only one of the server's "srv_conn" is high, it usually means
20256 that this server has some trouble causing the connections to take longer to
20257 be processed than on other servers.
20258
20259 - "retries" is the number of connection retries experienced by this session
20260 when trying to connect to the server. It must normally be zero, unless a
20261 server is being stopped at the same moment the connection was attempted.
20262 Frequent retries generally indicate either a network problem between
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020263 HAProxy and the server, or a misconfigured system backlog on the server
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020264 preventing new connections from being queued. This field may optionally be
20265 prefixed with a '+' sign, indicating that the session has experienced a
20266 redispatch after the maximal retry count has been reached on the initial
20267 server. In this case, the server name appearing in the log is the one the
20268 connection was redispatched to, and not the first one, though both may
20269 sometimes be the same in case of hashing for instance. So as a general rule
20270 of thumb, when a '+' is present in front of the retry count, this count
20271 should not be attributed to the logged server.
20272
20273 - "srv_queue" is the total number of requests which were processed before
20274 this one in the server queue. It is zero when the request has not gone
20275 through the server queue. It makes it possible to estimate the approximate
20276 server's response time by dividing the time spent in queue by the number of
20277 requests in the queue. It is worth noting that if a session experiences a
20278 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010020279 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020280 backend queue unless a redispatch occurs.
20281
20282 - "backend_queue" is the total number of requests which were processed before
20283 this one in the backend's global queue. It is zero when the request has not
20284 gone through the global queue. It makes it possible to estimate the average
20285 queue length, which easily translates into a number of missing servers when
20286 divided by a server's "maxconn" parameter. It is worth noting that if a
20287 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010020288 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020289 through both the server queue and the backend queue unless a redispatch
20290 occurs.
20291
20292
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200202938.2.3. HTTP log format
20294----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020295
20296The HTTP format is the most complete and the best suited for HTTP proxies. It
20297is enabled by when "option httplog" is specified in the frontend. It provides
20298the same level of information as the TCP format with additional features which
20299are specific to the HTTP protocol. Just like the TCP format, the log is usually
20300emitted at the end of the session, unless "option logasap" is specified, which
20301generally only makes sense for download sites. A session which matches the
20302"monitor" rules will never logged. It is also possible not to log sessions for
20303which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020020304frontend. Successful connections will not be logged if "option dontlog-normal"
20305is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020306
20307Most fields are shared with the TCP log, some being different. A few fields may
20308slightly vary depending on some configuration options. Those ones are marked
20309with a star ('*') after the field name below.
20310
20311 Example :
20312 frontend http-in
20313 mode http
20314 option httplog
20315 log global
20316 default_backend bck
20317
20318 backend static
20319 server srv1 127.0.0.1:8000
20320
20321 >>> Feb 6 12:14:14 localhost \
20322 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
20323 static/srv1 10/0/30/69/109 200 2750 - - ---- 1/1/1/1/0 0/0 {1wt.eu} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010020324 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020325
20326 Field Format Extract from the example above
20327 1 process_name '[' pid ']:' haproxy[14389]:
20328 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020329 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020330 4 frontend_name http-in
20331 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020332 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020333 7 status_code 200
20334 8 bytes_read* 2750
20335 9 captured_request_cookie -
20336 10 captured_response_cookie -
20337 11 termination_state ----
20338 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
20339 13 srv_queue '/' backend_queue 0/0
20340 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
20341 15 '{' captured_response_headers* '}' {}
20342 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010020343
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020344Detailed fields description :
20345 - "client_ip" is the IP address of the client which initiated the TCP
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020346 connection to HAProxy. If the connection was accepted on a UNIX socket
Willy Tarreauceb24bc2010-11-09 12:46:41 +010020347 instead, the IP address would be replaced with the word "unix". Note that
20348 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010020349 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010020350 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010020351 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020352
20353 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010020354 If the connection was accepted on a UNIX socket instead, the port would be
20355 replaced with the ID of the accepting socket, which is also reported in the
20356 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020357
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020358 - "request_date" is the exact date when the first byte of the HTTP request
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020359 was received by HAProxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020360
20361 - "frontend_name" is the name of the frontend (or listener) which received
20362 and processed the connection.
20363
20364 - "backend_name" is the name of the backend (or listener) which was selected
20365 to manage the connection to the server. This will be the same as the
20366 frontend if no switching rule has been applied.
20367
20368 - "server_name" is the name of the last server to which the connection was
20369 sent, which might differ from the first one if there were connection errors
20370 and a redispatch occurred. Note that this server belongs to the backend
20371 which processed the request. If the request was aborted before reaching a
20372 server, "<NOSRV>" is indicated instead of a server name. If the request was
20373 intercepted by the stats subsystem, "<STATS>" is indicated instead.
20374
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020375 - "TR" is the total time in milliseconds spent waiting for a full HTTP
20376 request from the client (not counting body) after the first byte was
20377 received. It can be "-1" if the connection was aborted before a complete
John Roeslerfb2fce12019-07-10 15:45:51 -050020378 request could be received or a bad request was received. It should
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020379 always be very small because a request generally fits in one single packet.
20380 Large times here generally indicate network issues between the client and
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020381 HAProxy or requests being typed by hand. See section 8.4 "Timing Events"
Willy Tarreau590a0512018-09-05 11:56:48 +020020382 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020383
20384 - "Tw" is the total time in milliseconds spent waiting in the various queues.
20385 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020020386 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020387
20388 - "Tc" is the total time in milliseconds spent waiting for the connection to
20389 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020020390 request was aborted before a connection could be established. See section
20391 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020392
20393 - "Tr" is the total time in milliseconds spent waiting for the server to send
20394 a full HTTP response, not counting data. It can be "-1" if the request was
20395 aborted before a complete response could be received. It generally matches
20396 the server's processing time for the request, though it may be altered by
20397 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020020398 "GET" requests generally indicate an overloaded server. See section 8.4
20399 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020400
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020401 - "Ta" is the time the request remained active in HAProxy, which is the total
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020402 time in milliseconds elapsed between the first byte of the request was
20403 received and the last byte of response was sent. It covers all possible
20404 processing except the handshake (see Th) and idle time (see Ti). There is
20405 one exception, if "option logasap" was specified, then the time counting
20406 stops at the moment the log is emitted. In this case, a '+' sign is
20407 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020020408 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020409
20410 - "status_code" is the HTTP status code returned to the client. This status
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020411 is generally set by the server, but it might also be set by HAProxy when
20412 the server cannot be reached or when its response is blocked by HAProxy.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020413
20414 - "bytes_read" is the total number of bytes transmitted to the client when
20415 the log is emitted. This does include HTTP headers. If "option logasap" is
John Roeslerfb2fce12019-07-10 15:45:51 -050020416 specified, this value will be prefixed with a '+' sign indicating that
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020417 the final one may be larger. Please note that this value is a 64-bit
20418 counter, so log analysis tools must be able to handle it without
20419 overflowing.
20420
20421 - "captured_request_cookie" is an optional "name=value" entry indicating that
20422 the client had this cookie in the request. The cookie name and its maximum
20423 length are defined by the "capture cookie" statement in the frontend
20424 configuration. The field is a single dash ('-') when the option is not
20425 set. Only one cookie may be captured, it is generally used to track session
20426 ID exchanges between a client and a server to detect session crossing
20427 between clients due to application bugs. For more details, please consult
20428 the section "Capturing HTTP headers and cookies" below.
20429
20430 - "captured_response_cookie" is an optional "name=value" entry indicating
20431 that the server has returned a cookie with its response. The cookie name
20432 and its maximum length are defined by the "capture cookie" statement in the
20433 frontend configuration. The field is a single dash ('-') when the option is
20434 not set. Only one cookie may be captured, it is generally used to track
20435 session ID exchanges between a client and a server to detect session
20436 crossing between clients due to application bugs. For more details, please
20437 consult the section "Capturing HTTP headers and cookies" below.
20438
20439 - "termination_state" is the condition the session was in when the session
20440 ended. This indicates the session state, which side caused the end of
20441 session to happen, for what reason (timeout, error, ...), just like in TCP
20442 logs, and information about persistence operations on cookies in the last
20443 two characters. The normal flags should begin with "--", indicating the
20444 session was closed by either end with no data remaining in buffers. See
20445 below "Session state at disconnection" for more details.
20446
20447 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040020448 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020449 limits have been reached. For instance, if actconn is close to 512 or 1024
20450 when multiple connection errors occur, chances are high that the system
20451 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020020452 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020453 system.
20454
20455 - "feconn" is the total number of concurrent connections on the frontend when
20456 the session was logged. It is useful to estimate the amount of resource
20457 required to sustain high loads, and to detect when the frontend's "maxconn"
20458 has been reached. Most often when this value increases by huge jumps, it is
20459 because there is congestion on the backend servers, but sometimes it can be
20460 caused by a denial of service attack.
20461
20462 - "beconn" is the total number of concurrent connections handled by the
20463 backend when the session was logged. It includes the total number of
20464 concurrent connections active on servers as well as the number of
20465 connections pending in queues. It is useful to estimate the amount of
20466 additional servers needed to support high loads for a given application.
20467 Most often when this value increases by huge jumps, it is because there is
20468 congestion on the backend servers, but sometimes it can be caused by a
20469 denial of service attack.
20470
20471 - "srv_conn" is the total number of concurrent connections still active on
20472 the server when the session was logged. It can never exceed the server's
20473 configured "maxconn" parameter. If this value is very often close or equal
20474 to the server's "maxconn", it means that traffic regulation is involved a
20475 lot, meaning that either the server's maxconn value is too low, or that
20476 there aren't enough servers to process the load with an optimal response
20477 time. When only one of the server's "srv_conn" is high, it usually means
20478 that this server has some trouble causing the requests to take longer to be
20479 processed than on other servers.
20480
20481 - "retries" is the number of connection retries experienced by this session
20482 when trying to connect to the server. It must normally be zero, unless a
20483 server is being stopped at the same moment the connection was attempted.
20484 Frequent retries generally indicate either a network problem between
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020485 HAProxy and the server, or a misconfigured system backlog on the server
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020486 preventing new connections from being queued. This field may optionally be
20487 prefixed with a '+' sign, indicating that the session has experienced a
20488 redispatch after the maximal retry count has been reached on the initial
20489 server. In this case, the server name appearing in the log is the one the
20490 connection was redispatched to, and not the first one, though both may
20491 sometimes be the same in case of hashing for instance. So as a general rule
20492 of thumb, when a '+' is present in front of the retry count, this count
20493 should not be attributed to the logged server.
20494
20495 - "srv_queue" is the total number of requests which were processed before
20496 this one in the server queue. It is zero when the request has not gone
20497 through the server queue. It makes it possible to estimate the approximate
20498 server's response time by dividing the time spent in queue by the number of
20499 requests in the queue. It is worth noting that if a session experiences a
20500 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010020501 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020502 backend queue unless a redispatch occurs.
20503
20504 - "backend_queue" is the total number of requests which were processed before
20505 this one in the backend's global queue. It is zero when the request has not
20506 gone through the global queue. It makes it possible to estimate the average
20507 queue length, which easily translates into a number of missing servers when
20508 divided by a server's "maxconn" parameter. It is worth noting that if a
20509 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010020510 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020511 through both the server queue and the backend queue unless a redispatch
20512 occurs.
20513
20514 - "captured_request_headers" is a list of headers captured in the request due
20515 to the presence of the "capture request header" statement in the frontend.
20516 Multiple headers can be captured, they will be delimited by a vertical bar
20517 ('|'). When no capture is enabled, the braces do not appear, causing a
20518 shift of remaining fields. It is important to note that this field may
20519 contain spaces, and that using it requires a smarter log parser than when
20520 it's not used. Please consult the section "Capturing HTTP headers and
20521 cookies" below for more details.
20522
20523 - "captured_response_headers" is a list of headers captured in the response
20524 due to the presence of the "capture response header" statement in the
20525 frontend. Multiple headers can be captured, they will be delimited by a
20526 vertical bar ('|'). When no capture is enabled, the braces do not appear,
20527 causing a shift of remaining fields. It is important to note that this
20528 field may contain spaces, and that using it requires a smarter log parser
20529 than when it's not used. Please consult the section "Capturing HTTP headers
20530 and cookies" below for more details.
20531
20532 - "http_request" is the complete HTTP request line, including the method,
20533 request and HTTP version string. Non-printable characters are encoded (see
20534 below the section "Non-printable characters"). This is always the last
20535 field, and it is always delimited by quotes and is the only one which can
20536 contain quotes. If new fields are added to the log format, they will be
20537 added before this field. This field might be truncated if the request is
20538 huge and does not fit in the standard syslog buffer (1024 characters). This
20539 is the reason why this field must always remain the last one.
20540
20541
Cyril Bontédc4d9032012-04-08 21:57:39 +0200205428.2.4. Custom log format
20543------------------------
William Lallemand48940402012-01-30 16:47:22 +010020544
Willy Tarreau2beef582012-12-20 17:22:52 +010020545The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010020546mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010020547
Davor Ocelice9ed2812017-12-25 17:49:28 +010020548HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010020549Variables can take arguments using braces ('{}'), and multiple arguments are
20550separated by commas within the braces. Flags may be added or removed by
20551prefixing them with a '+' or '-' sign.
20552
20553Special variable "%o" may be used to propagate its flags to all other
20554variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010020555("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010020556
Willy Tarreauc8368452012-12-21 00:09:23 +010020557If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020020558as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010020559less common information such as the client's SSL certificate's DN, or to log
20560the key that would be used to store an entry into a stick table.
20561
Dragan Dosen1e3b16f2020-06-23 18:16:44 +020020562Note: spaces must be escaped. In configuration directives "log-format",
20563"log-format-sd" and "unique-id-format", spaces are considered as
20564delimiters and are merged. In order to emit a verbatim '%', it must be
20565preceded by another '%' resulting in '%%'.
William Lallemand48940402012-01-30 16:47:22 +010020566
Dragan Dosen835b9212016-02-12 13:23:03 +010020567Note: when using the RFC5424 syslog message format, the characters '"',
20568'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
20569https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
20570such cases, the use of the flag "E" should be considered.
20571
William Lallemand48940402012-01-30 16:47:22 +010020572Flags are :
20573 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040020574 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010020575 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
20576 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010020577
20578 Example:
20579
20580 log-format %T\ %t\ Some\ Text
20581 log-format %{+Q}o\ %t\ %s\ %{-Q}r
20582
Dragan Dosen835b9212016-02-12 13:23:03 +010020583 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
20584
William Lallemand48940402012-01-30 16:47:22 +010020585At the moment, the default HTTP format is defined this way :
20586
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020587 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
20588 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010020589
William Lallemandbddd4fd2012-02-27 11:23:10 +010020590the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010020591
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020592 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
20593 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
20594 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010020595
William Lallemandbddd4fd2012-02-27 11:23:10 +010020596and the default TCP format is defined this way :
20597
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020598 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
20599 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010020600
William Lallemand48940402012-01-30 16:47:22 +010020601Please refer to the table below for currently defined variables :
20602
William Lallemandbddd4fd2012-02-27 11:23:10 +010020603 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020020604 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020605 +---+------+-----------------------------------------------+-------------+
20606 | | %o | special variable, apply flags on all next var | |
20607 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010020608 | | %B | bytes_read (from server to client) | numeric |
20609 | H | %CC | captured_request_cookie | string |
20610 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020020611 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000020612 | H | %HM | HTTP method (ex: POST) | string |
Maciej Zdeb21acc332020-11-26 10:45:52 +000020613 | H | %HP | HTTP request URI without query string | string |
Maciej Zdebfcdfd852020-11-30 18:27:47 +000020614 | H | %HPO | HTTP path only (without host nor query string)| string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000020615 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000020616 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
20617 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010020618 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020020619 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020020620 | | %T | gmt_date_time | date |
Christopher Faulet3f177162021-12-03 10:48:36 +010020621 | H | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020622 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020020623 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080020624 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020625 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
20626 | H | %Ti | idle time before the HTTP request | numeric |
20627 | H | %Tq | Th + Ti + TR | numeric |
20628 | H | %TR | time to receive the full request from 1st byte| numeric |
20629 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020020630 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020631 | | %Tt | Tt | numeric |
Damien Claisse57c8eb92020-04-28 12:09:19 +000020632 | | %Tu | Tu | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020633 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010020634 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020635 | | %ac | actconn | numeric |
20636 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010020637 | | %bc | beconn (backend concurrent connections) | numeric |
20638 | | %bi | backend_source_ip (connecting address) | IP |
20639 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020640 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010020641 | | %ci | client_ip (accepted address) | IP |
20642 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020643 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010020644 | | %fc | feconn (frontend concurrent connections) | numeric |
20645 | | %fi | frontend_ip (accepting address) | IP |
20646 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020020647 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020020648 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020020649 | | %hr | captured_request_headers default style | string |
20650 | | %hrl | captured_request_headers CLF style | string list |
20651 | | %hs | captured_response_headers default style | string |
20652 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020020653 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020020654 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020020655 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020656 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010020657 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020658 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010020659 | | %sc | srv_conn (server concurrent connections) | numeric |
20660 | | %si | server_IP (target address) | IP |
20661 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020662 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020020663 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
20664 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010020665 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020666 | H | %tr | date_time of HTTP request | date |
20667 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020020668 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020669 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020020670 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010020671 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010020672
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020020673 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010020674
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010020675
206768.2.5. Error log format
20677-----------------------
20678
20679When an incoming connection fails due to an SSL handshake or an invalid PROXY
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020680protocol header, HAProxy will log the event using a shorter, fixed line format.
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010020681By default, logs are emitted at the LOG_INFO level, unless the option
20682"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010020683will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010020684logged if the "dontlognull" option is set.
20685
20686The format looks like this :
20687
20688 >>> Dec 3 18:27:14 localhost \
20689 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
20690 Connection error during SSL handshake
20691
20692 Field Format Extract from the example above
20693 1 process_name '[' pid ']:' haproxy[6103]:
20694 2 client_ip ':' client_port 127.0.0.1:56059
20695 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
20696 4 frontend_name "/" bind_name ":" frt/f1:
20697 5 message Connection error during SSL handshake
20698
20699These fields just provide minimal information to help debugging connection
20700failures.
20701
20702
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207038.3. Advanced logging options
20704-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020705
20706Some advanced logging options are often looked for but are not easy to find out
20707just by looking at the various options. Here is an entry point for the few
20708options which can enable better logging. Please refer to the keywords reference
20709for more information about their usage.
20710
20711
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207128.3.1. Disabling logging of external tests
20713------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020714
20715It is quite common to have some monitoring tools perform health checks on
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020716HAProxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020717commercial load-balancer, and sometimes it will simply be a more complete
20718monitoring system such as Nagios. When the tests are very frequent, users often
20719ask how to disable logging for those checks. There are three possibilities :
20720
20721 - if connections come from everywhere and are just TCP probes, it is often
20722 desired to simply disable logging of connections without data exchange, by
20723 setting "option dontlognull" in the frontend. It also disables logging of
20724 port scans, which may or may not be desired.
20725
Willy Tarreau9e9919d2020-10-14 15:55:23 +020020726 - it is possible to use the "http-request set-log-level silent" action using
20727 a variety of conditions (source networks, paths, user-agents, etc).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020728
20729 - if the tests are performed on a known URI, use "monitor-uri" to declare
20730 this URI as dedicated to monitoring. Any host sending this request will
20731 only get the result of a health-check, and the request will not be logged.
20732
20733
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207348.3.2. Logging before waiting for the session to terminate
20735----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020736
20737The problem with logging at end of connection is that you have no clue about
20738what is happening during very long sessions, such as remote terminal sessions
20739or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010020740"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020741just before data transfer begins. This means that in case of TCP, it will still
20742log the connection status to the server, and in case of HTTP, it will log just
20743after processing the server headers. In this case, the number of bytes reported
20744is the number of header bytes sent to the client. In order to avoid confusion
20745with normal logs, the total time field and the number of bytes are prefixed
20746with a '+' sign which means that real numbers are certainly larger.
20747
20748
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207498.3.3. Raising log level upon errors
20750------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020020751
20752Sometimes it is more convenient to separate normal traffic from errors logs,
20753for instance in order to ease error monitoring from log files. When the option
20754"log-separate-errors" is used, connections which experience errors, timeouts,
20755retries, redispatches or HTTP status codes 5xx will see their syslog level
20756raised from "info" to "err". This will help a syslog daemon store the log in
20757a separate file. It is very important to keep the errors in the normal traffic
20758file too, so that log ordering is not altered. You should also be careful if
20759you already have configured your syslog daemon to store all logs higher than
20760"notice" in an "admin" file, because the "err" level is higher than "notice".
20761
20762
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207638.3.4. Disabling logging of successful connections
20764--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020020765
20766Although this may sound strange at first, some large sites have to deal with
20767multiple thousands of logs per second and are experiencing difficulties keeping
20768them intact for a long time or detecting errors within them. If the option
20769"dontlog-normal" is set on the frontend, all normal connections will not be
20770logged. In this regard, a normal connection is defined as one without any
20771error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
20772and a response with a status 5xx is not considered normal and will be logged
20773too. Of course, doing is is really discouraged as it will remove most of the
20774useful information from the logs. Do this only if you have no other
20775alternative.
20776
20777
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207788.4. Timing events
20779------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020780
20781Timers provide a great help in troubleshooting network problems. All values are
20782reported in milliseconds (ms). These timers should be used in conjunction with
20783the session termination flags. In TCP mode with "option tcplog" set on the
20784frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020785mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
20786addition, three other measures are provided, "Th", "Ti", and "Tq".
20787
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010020788Timings events in HTTP mode:
20789
20790 first request 2nd request
20791 |<-------------------------------->|<-------------- ...
20792 t tr t tr ...
20793 ---|----|----|----|----|----|----|----|----|--
20794 : Th Ti TR Tw Tc Tr Td : Ti ...
20795 :<---- Tq ---->: :
20796 :<-------------- Tt -------------->:
Damien Claisse57c8eb92020-04-28 12:09:19 +000020797 :<-- -----Tu--------------->:
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010020798 :<--------- Ta --------->:
20799
20800Timings events in TCP mode:
20801
20802 TCP session
20803 |<----------------->|
20804 t t
20805 ---|----|----|----|----|---
20806 | Th Tw Tc Td |
20807 |<------ Tt ------->|
20808
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020809 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010020810 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020811 only happen once during the whole connection's lifetime. A large time here
20812 may indicate that the client only pre-established the connection without
20813 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010020814 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020020815 SSL handshake was very expensive to compute. Please note that this time is
20816 reported only before the first request, so it is safe to average it over
20817 all request to calculate the amortized value. The second and subsequent
20818 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020819
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020820 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
20821 counts between the end of the handshakes and the first byte of the HTTP
20822 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020020823 to count after the end of the transmission the previous response. When a
20824 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
20825 after the previous request. Some browsers pre-establish connections to a
20826 server in order to reduce the latency of a future request, and keep them
20827 pending until they need it. This delay will be reported as the idle time. A
20828 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020829
20830 - TR: total time to get the client request (HTTP mode only). It's the time
20831 elapsed between the first bytes received and the moment the proxy received
20832 the empty line marking the end of the HTTP headers. The value "-1"
20833 indicates that the end of headers has never been seen. This happens when
20834 the client closes prematurely or times out. This time is usually very short
20835 since most requests fit in a single packet. A large time may indicate a
20836 request typed by hand during a test.
20837
20838 - Tq: total time to get the client request from the accept date or since the
20839 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010020840 exactly equal to Th + Ti + TR unless any of them is -1, in which case it
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020841 returns -1 as well. This timer used to be very useful before the arrival of
20842 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
20843 it in favor of TR nowadays, as the idle time adds a lot of noise to the
20844 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020845
20846 - Tw: total time spent in the queues waiting for a connection slot. It
20847 accounts for backend queue as well as the server queues, and depends on the
20848 queue size, and the time needed for the server to complete previous
20849 requests. The value "-1" means that the request was killed before reaching
20850 the queue, which is generally what happens with invalid or denied requests.
20851
20852 - Tc: total time to establish the TCP connection to the server. It's the time
20853 elapsed between the moment the proxy sent the connection request, and the
20854 moment it was acknowledged by the server, or between the TCP SYN packet and
20855 the matching SYN/ACK packet in return. The value "-1" means that the
20856 connection never established.
20857
20858 - Tr: server response time (HTTP mode only). It's the time elapsed between
20859 the moment the TCP connection was established to the server and the moment
20860 the server sent its complete response headers. It purely shows its request
20861 processing time, without the network overhead due to the data transmission.
20862 It is worth noting that when the client has data to send to the server, for
20863 instance during a POST request, the time already runs, and this can distort
20864 apparent response time. For this reason, it's generally wise not to trust
20865 too much this field for POST requests initiated from clients behind an
20866 untrusted network. A value of "-1" here means that the last the response
20867 header (empty line) was never seen, most likely because the server timeout
20868 stroke before the server managed to process the request.
20869
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020870 - Ta: total active time for the HTTP request, between the moment the proxy
20871 received the first byte of the request header and the emission of the last
20872 byte of the response body. The exception is when the "logasap" option is
20873 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
20874 a '+' sign. From this field, we can deduce "Td", the data transmission time,
20875 by subtracting other timers when valid :
20876
20877 Td = Ta - (TR + Tw + Tc + Tr)
20878
20879 Timers with "-1" values have to be excluded from this equation. Note that
20880 "Ta" can never be negative.
20881
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020882 - Tt: total session duration time, between the moment the proxy accepted it
20883 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020884 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
20885 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030020886 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020887
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020888 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020889
20890 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020891 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
20892 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020893
Damien Claisse57c8eb92020-04-28 12:09:19 +000020894 - Tu: total estimated time as seen from client, between the moment the proxy
20895 accepted it and the moment both ends were closed, without idle time.
20896 This is useful to roughly measure end-to-end time as a user would see it,
20897 without idle time pollution from keep-alive time between requests. This
20898 timer in only an estimation of time seen by user as it assumes network
20899 latency is the same in both directions. The exception is when the "logasap"
20900 option is specified. In this case, it only equals (Th+TR+Tw+Tc+Tr), and is
20901 prefixed with a '+' sign.
20902
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020903These timers provide precious indications on trouble causes. Since the TCP
20904protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
20905that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020906due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
20907"Tt" is close to a timeout value specified in the configuration, it often means
20908that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020909
20910Most common cases :
20911
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020912 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
20913 the client and the proxy. This is very rare on local networks but might
20914 happen when clients are on far remote networks and send large requests. It
20915 may happen that values larger than usual appear here without any network
20916 cause. Sometimes, during an attack or just after a resource starvation has
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020917 ended, HAProxy may accept thousands of connections in a few milliseconds.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020918 The time spent accepting these connections will inevitably slightly delay
20919 processing of other connections, and it can happen that request times in the
20920 order of a few tens of milliseconds are measured after a few thousands of
20921 new connections have been accepted at once. Using one of the keep-alive
20922 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020020923 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020924
20925 - If "Tc" is close to 3000, a packet has probably been lost between the
20926 server and the proxy during the server connection phase. This value should
20927 always be very low, such as 1 ms on local networks and less than a few tens
20928 of ms on remote networks.
20929
Willy Tarreau55165fe2009-05-10 12:02:55 +020020930 - If "Tr" is nearly always lower than 3000 except some rare values which seem
20931 to be the average majored by 3000, there are probably some packets lost
20932 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020933
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020934 - If "Ta" is large even for small byte counts, it generally is because
20935 neither the client nor the server decides to close the connection while
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020936 HAProxy is running in tunnel mode and both have agreed on a keep-alive
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020937 connection mode. In order to solve this issue, it will be needed to specify
20938 one of the HTTP options to manipulate keep-alive or close options on either
20939 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
20940 important when connection regulation is used with the "maxconn" option on
20941 the servers, since no new connection will be sent to the server until
20942 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020943
20944Other noticeable HTTP log cases ('xx' means any value to be ignored) :
20945
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020946 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020947 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020948 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020949
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020950 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020951 or it aborted too early. Check the session termination flags
20952 then "timeout http-request" and "timeout client" settings.
20953
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020954 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020955 servers were out of order, because the request was invalid
20956 or forbidden by ACL rules. Check the session termination
20957 flags.
20958
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020959 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
20960 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020961 Check the session termination flags, then check the
20962 "timeout connect" setting. Note that the tarpit action might
20963 return similar-looking patterns, with "Tw" equal to the time
20964 the client connection was maintained open.
20965
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020966 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030020967 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020020968 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020969 termination flags, then check the "timeout server" setting.
20970
20971
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200209728.5. Session state at disconnection
20973-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020974
20975TCP and HTTP logs provide a session termination indicator in the
20976"termination_state" field, just before the number of active connections. It is
209772-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
20978each of which has a special meaning :
20979
20980 - On the first character, a code reporting the first event which caused the
20981 session to terminate :
20982
20983 C : the TCP session was unexpectedly aborted by the client.
20984
20985 S : the TCP session was unexpectedly aborted by the server, or the
20986 server explicitly refused it.
20987
20988 P : the session was prematurely aborted by the proxy, because of a
20989 connection limit enforcement, because a DENY filter was matched,
20990 because of a security check which detected and blocked a dangerous
20991 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010020992 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020020993
Daniel Corbett9f0843f2021-05-08 10:50:37 -040020994 L : the session was locally processed by HAProxy and was not passed to
Willy Tarreau570f2212013-06-10 16:42:09 +020020995 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010020996
20997 R : a resource on the proxy has been exhausted (memory, sockets, source
20998 ports, ...). Usually, this appears during the connection phase, and
20999 system logs should contain a copy of the precise error. If this
21000 happens, it must be considered as a very serious anomaly which
21001 should be fixed as soon as possible by any means.
21002
21003 I : an internal error was identified by the proxy during a self-check.
21004 This should NEVER happen, and you are encouraged to report any log
21005 containing this, because this would almost certainly be a bug. It
21006 would be wise to preventively restart the process after such an
21007 event too, in case it would be caused by memory corruption.
21008
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021009 D : the session was killed by HAProxy because the server was detected
Simon Horman752dc4a2011-06-21 14:34:59 +090021010 as down and was configured to kill all connections when going down.
21011
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021012 U : the session was killed by HAProxy on this backup server because an
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070021013 active server was detected as up and was configured to kill all
21014 backup connections when going up.
21015
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021016 K : the session was actively killed by an admin operating on HAProxy.
Willy Tarreaua2a64e92011-09-07 23:01:56 +020021017
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021018 c : the client-side timeout expired while waiting for the client to
21019 send or receive data.
21020
21021 s : the server-side timeout expired while waiting for the server to
21022 send or receive data.
21023
21024 - : normal session completion, both the client and the server closed
21025 with nothing left in the buffers.
21026
21027 - on the second character, the TCP or HTTP session state when it was closed :
21028
Willy Tarreauf7b30a92010-12-06 22:59:17 +010021029 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021030 (HTTP mode only). Nothing was sent to any server.
21031
21032 Q : the proxy was waiting in the QUEUE for a connection slot. This can
21033 only happen when servers have a 'maxconn' parameter set. It can
21034 also happen in the global queue after a redispatch consecutive to
21035 a failed attempt to connect to a dying server. If no redispatch is
21036 reported, then no connection attempt was made to any server.
21037
21038 C : the proxy was waiting for the CONNECTION to establish on the
21039 server. The server might at most have noticed a connection attempt.
21040
21041 H : the proxy was waiting for complete, valid response HEADERS from the
21042 server (HTTP only).
21043
21044 D : the session was in the DATA phase.
21045
21046 L : the proxy was still transmitting LAST data to the client while the
21047 server had already finished. This one is very rare as it can only
21048 happen when the client dies while receiving the last packets.
21049
21050 T : the request was tarpitted. It has been held open with the client
21051 during the whole "timeout tarpit" duration or until the client
21052 closed, both of which will be reported in the "Tw" timer.
21053
21054 - : normal session completion after end of data transfer.
21055
21056 - the third character tells whether the persistence cookie was provided by
21057 the client (only in HTTP mode) :
21058
21059 N : the client provided NO cookie. This is usually the case for new
21060 visitors, so counting the number of occurrences of this flag in the
21061 logs generally indicate a valid trend for the site frequentation.
21062
21063 I : the client provided an INVALID cookie matching no known server.
21064 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020021065 cookies between HTTP/HTTPS sites, persistence conditionally
21066 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021067
21068 D : the client provided a cookie designating a server which was DOWN,
21069 so either "option persist" was used and the client was sent to
21070 this server, or it was not set and the client was redispatched to
21071 another server.
21072
Willy Tarreau996a92c2010-10-13 19:30:47 +020021073 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021074 server.
21075
Willy Tarreau996a92c2010-10-13 19:30:47 +020021076 E : the client provided a valid cookie, but with a last date which was
21077 older than what is allowed by the "maxidle" cookie parameter, so
21078 the cookie is consider EXPIRED and is ignored. The request will be
21079 redispatched just as if there was no cookie.
21080
21081 O : the client provided a valid cookie, but with a first date which was
21082 older than what is allowed by the "maxlife" cookie parameter, so
21083 the cookie is consider too OLD and is ignored. The request will be
21084 redispatched just as if there was no cookie.
21085
Willy Tarreauc89ccb62012-04-05 21:18:22 +020021086 U : a cookie was present but was not used to select the server because
21087 some other server selection mechanism was used instead (typically a
21088 "use-server" rule).
21089
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021090 - : does not apply (no cookie set in configuration).
21091
21092 - the last character reports what operations were performed on the persistence
21093 cookie returned by the server (only in HTTP mode) :
21094
21095 N : NO cookie was provided by the server, and none was inserted either.
21096
21097 I : no cookie was provided by the server, and the proxy INSERTED one.
21098 Note that in "cookie insert" mode, if the server provides a cookie,
21099 it will still be overwritten and reported as "I" here.
21100
Willy Tarreau996a92c2010-10-13 19:30:47 +020021101 U : the proxy UPDATED the last date in the cookie that was presented by
21102 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030021103 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020021104 date indicated in the cookie. If any other change happens, such as
21105 a redispatch, then the cookie will be marked as inserted instead.
21106
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021107 P : a cookie was PROVIDED by the server and transmitted as-is.
21108
21109 R : the cookie provided by the server was REWRITTEN by the proxy, which
21110 happens in "cookie rewrite" or "cookie prefix" modes.
21111
21112 D : the cookie provided by the server was DELETED by the proxy.
21113
21114 - : does not apply (no cookie set in configuration).
21115
Willy Tarreau996a92c2010-10-13 19:30:47 +020021116The combination of the two first flags gives a lot of information about what
21117was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021118helpful to detect server saturation, network troubles, local system resource
21119starvation, attacks, etc...
21120
21121The most common termination flags combinations are indicated below. They are
21122alphabetically sorted, with the lowercase set just after the upper case for
21123easier finding and understanding.
21124
21125 Flags Reason
21126
21127 -- Normal termination.
21128
21129 CC The client aborted before the connection could be established to the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021130 server. This can happen when HAProxy tries to connect to a recently
21131 dead (or unchecked) server, and the client aborts while HAProxy is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021132 waiting for the server to respond or for "timeout connect" to expire.
21133
21134 CD The client unexpectedly aborted during data transfer. This can be
21135 caused by a browser crash, by an intermediate equipment between the
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021136 client and HAProxy which decided to actively break the connection,
21137 by network routing issues between the client and HAProxy, or by a
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021138 keep-alive session between the server and the client terminated first
21139 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010021140
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021141 cD The client did not send nor acknowledge any data for as long as the
21142 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020021143 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021144
21145 CH The client aborted while waiting for the server to start responding.
21146 It might be the server taking too long to respond or the client
21147 clicking the 'Stop' button too fast.
21148
21149 cH The "timeout client" stroke while waiting for client data during a
21150 POST request. This is sometimes caused by too large TCP MSS values
21151 for PPPoE networks which cannot transport full-sized packets. It can
21152 also happen when client timeout is smaller than server timeout and
21153 the server takes too long to respond.
21154
21155 CQ The client aborted while its session was queued, waiting for a server
21156 with enough empty slots to accept it. It might be that either all the
21157 servers were saturated or that the assigned server was taking too
21158 long a time to respond.
21159
21160 CR The client aborted before sending a full HTTP request. Most likely
21161 the request was typed by hand using a telnet client, and aborted
21162 too early. The HTTP status code is likely a 400 here. Sometimes this
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021163 might also be caused by an IDS killing the connection between HAProxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020021164 and the client. "option http-ignore-probes" can be used to ignore
21165 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021166
21167 cR The "timeout http-request" stroke before the client sent a full HTTP
21168 request. This is sometimes caused by too large TCP MSS values on the
21169 client side for PPPoE networks which cannot transport full-sized
21170 packets, or by clients sending requests by hand and not typing fast
21171 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020021172 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020021173 some browsers started to implement a "pre-connect" feature consisting
21174 in speculatively connecting to some recently visited web sites just
21175 in case the user would like to visit them. This results in many
21176 connections being established to web sites, which end up in 408
21177 Request Timeout if the timeout strikes first, or 400 Bad Request when
21178 the browser decides to close them first. These ones pollute the log
21179 and feed the error counters. Some versions of some browsers have even
21180 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010021181 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020021182 http-ignore-probes" in the frontend, resulting in connections with
21183 zero data transfer to be totally ignored. This will definitely hide
21184 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021185
21186 CT The client aborted while its session was tarpitted. It is important to
21187 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020021188 wrong tarpit rules have been written. If a lot of them happen, it
21189 might make sense to lower the "timeout tarpit" value to something
21190 closer to the average reported "Tw" timer, in order not to consume
21191 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021192
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021193 LR The request was intercepted and locally handled by HAProxy. Generally
Willy Tarreau570f2212013-06-10 16:42:09 +020021194 it means that this was a redirect or a stats request.
21195
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021196 SC The server or an equipment between it and HAProxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021197 the TCP connection (the proxy received a TCP RST or an ICMP message
21198 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010021199 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021200 or no ARP response on local network). When this happens in HTTP mode,
21201 the status code is likely a 502 or 503 here.
21202
21203 sC The "timeout connect" stroke before a connection to the server could
21204 complete. When this happens in HTTP mode, the status code is likely a
21205 503 or 504 here.
21206
21207 SD The connection to the server died with an error during the data
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021208 transfer. This usually means that HAProxy has received an RST from
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021209 the server or an ICMP message from an intermediate equipment while
21210 exchanging data with the server. This can be caused by a server crash
21211 or by a network issue on an intermediate equipment.
21212
21213 sD The server did not send nor acknowledge any data for as long as the
21214 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010021215 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021216 load-balancers, ...), as well as keep-alive sessions maintained
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021217 between the client and the server expiring first on HAProxy.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021218
21219 SH The server aborted before sending its full HTTP response headers, or
21220 it crashed while processing the request. Since a server aborting at
21221 this moment is very rare, it would be wise to inspect its logs to
21222 control whether it crashed and why. The logged request may indicate a
21223 small set of faulty requests, demonstrating bugs in the application.
21224 Sometimes this might also be caused by an IDS killing the connection
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021225 between HAProxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021226
21227 sH The "timeout server" stroke before the server could return its
21228 response headers. This is the most common anomaly, indicating too
21229 long transactions, probably caused by server or database saturation.
21230 The immediate workaround consists in increasing the "timeout server"
21231 setting, but it is important to keep in mind that the user experience
21232 will suffer from these long response times. The only long term
21233 solution is to fix the application.
21234
21235 sQ The session spent too much time in queue and has been expired. See
21236 the "timeout queue" and "timeout connect" settings to find out how to
21237 fix this if it happens too often. If it often happens massively in
21238 short periods, it may indicate general problems on the affected
21239 servers due to I/O or database congestion, or saturation caused by
21240 external attacks.
21241
21242 PC The proxy refused to establish a connection to the server because the
Thayne McCombscdbcca92021-01-07 21:24:41 -070021243 process's socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020021244 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021245 so that it does not happen anymore. This status is very rare and
21246 might happen when the global "ulimit-n" parameter is forced by hand.
21247
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010021248 PD The proxy blocked an incorrectly formatted chunked encoded message in
21249 a request or a response, after the server has emitted its headers. In
21250 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010021251 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020021252 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010021253
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021254 PH The proxy blocked the server's response, because it was invalid,
21255 incomplete, dangerous (cache control), or matched a security filter.
21256 In any case, an HTTP 502 error is sent to the client. One possible
21257 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010021258 containing unauthorized characters. It is also possible but quite
21259 rare, that the proxy blocked a chunked-encoding request from the
21260 client due to an invalid syntax, before the server responded. In this
21261 case, an HTTP 400 error is sent to the client and reported in the
21262 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021263
21264 PR The proxy blocked the client's HTTP request, either because of an
21265 invalid HTTP syntax, in which case it returned an HTTP 400 error to
21266 the client, or because a deny filter matched, in which case it
21267 returned an HTTP 403 error.
21268
21269 PT The proxy blocked the client's request and has tarpitted its
21270 connection before returning it a 500 server error. Nothing was sent
21271 to the server. The connection was maintained open for as long as
21272 reported by the "Tw" timer field.
21273
21274 RC A local resource has been exhausted (memory, sockets, source ports)
21275 preventing the connection to the server from establishing. The error
21276 logs will tell precisely what was missing. This is very rare and can
21277 only be solved by proper system tuning.
21278
Willy Tarreau996a92c2010-10-13 19:30:47 +020021279The combination of the two last flags gives a lot of information about how
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021280persistence was handled by the client, the server and by HAProxy. This is very
Willy Tarreau996a92c2010-10-13 19:30:47 +020021281important to troubleshoot disconnections, when users complain they have to
21282re-authenticate. The commonly encountered flags are :
21283
21284 -- Persistence cookie is not enabled.
21285
21286 NN No cookie was provided by the client, none was inserted in the
21287 response. For instance, this can be in insert mode with "postonly"
21288 set on a GET request.
21289
21290 II A cookie designating an invalid server was provided by the client,
21291 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040021292 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020021293 value can be presented by a client when no other server knows it.
21294
21295 NI No cookie was provided by the client, one was inserted in the
21296 response. This typically happens for first requests from every user
21297 in "insert" mode, which makes it an easy way to count real users.
21298
21299 VN A cookie was provided by the client, none was inserted in the
21300 response. This happens for most responses for which the client has
21301 already got a cookie.
21302
21303 VU A cookie was provided by the client, with a last visit date which is
21304 not completely up-to-date, so an updated cookie was provided in
21305 response. This can also happen if there was no date at all, or if
21306 there was a date but the "maxidle" parameter was not set, so that the
21307 cookie can be switched to unlimited time.
21308
21309 EI A cookie was provided by the client, with a last visit date which is
21310 too old for the "maxidle" parameter, so the cookie was ignored and a
21311 new cookie was inserted in the response.
21312
21313 OI A cookie was provided by the client, with a first visit date which is
21314 too old for the "maxlife" parameter, so the cookie was ignored and a
21315 new cookie was inserted in the response.
21316
21317 DI The server designated by the cookie was down, a new server was
21318 selected and a new cookie was emitted in the response.
21319
21320 VI The server designated by the cookie was not marked dead but could not
21321 be reached. A redispatch happened and selected another one, which was
21322 then advertised in the response.
21323
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021324
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200213258.6. Non-printable characters
21326-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021327
21328In order not to cause trouble to log analysis tools or terminals during log
21329consulting, non-printable characters are not sent as-is into log files, but are
21330converted to the two-digits hexadecimal representation of their ASCII code,
21331prefixed by the character '#'. The only characters that can be logged without
21332being escaped are comprised between 32 and 126 (inclusive). Obviously, the
21333escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
21334is the same for the character '"' which becomes "#22", as well as '{', '|' and
21335'}' when logging headers.
21336
21337Note that the space character (' ') is not encoded in headers, which can cause
21338issues for tools relying on space count to locate fields. A typical header
21339containing spaces is "User-Agent".
21340
21341Last, it has been observed that some syslog daemons such as syslog-ng escape
21342the quote ('"') with a backslash ('\'). The reverse operation can safely be
21343performed since no quote may appear anywhere else in the logs.
21344
21345
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200213468.7. Capturing HTTP cookies
21347---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021348
21349Cookie capture simplifies the tracking a complete user session. This can be
21350achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021351section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021352cookie will simultaneously be checked in the request ("Cookie:" header) and in
21353the response ("Set-Cookie:" header). The respective values will be reported in
21354the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021355locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021356not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
21357user switches to a new session for example, because the server will reassign it
21358a new cookie. It is also possible to detect if a server unexpectedly sets a
21359wrong cookie to a client, leading to session crossing.
21360
21361 Examples :
21362 # capture the first cookie whose name starts with "ASPSESSION"
21363 capture cookie ASPSESSION len 32
21364
21365 # capture the first cookie whose name is exactly "vgnvisitor"
21366 capture cookie vgnvisitor= len 32
21367
21368
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200213698.8. Capturing HTTP headers
21370---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021371
21372Header captures are useful to track unique request identifiers set by an upper
21373proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
21374the response, one can search for information about the response length, how the
21375server asked the cache to behave, or an object location during a redirection.
21376
21377Header captures are performed using the "capture request header" and "capture
21378response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021379section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021380
21381It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010021382time. Non-existent headers are logged as empty strings, and if one header
21383appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021384are grouped within braces '{' and '}' in the same order as they were declared,
21385and delimited with a vertical bar '|' without any space. Response headers
21386follow the same representation, but are displayed after a space following the
21387request headers block. These blocks are displayed just before the HTTP request
21388in the logs.
21389
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020021390As a special case, it is possible to specify an HTTP header capture in a TCP
21391frontend. The purpose is to enable logging of headers which will be parsed in
21392an HTTP backend if the request is then switched to this HTTP backend.
21393
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021394 Example :
21395 # This instance chains to the outgoing proxy
21396 listen proxy-out
21397 mode http
21398 option httplog
21399 option logasap
21400 log global
21401 server cache1 192.168.1.1:3128
21402
21403 # log the name of the virtual server
21404 capture request header Host len 20
21405
21406 # log the amount of data uploaded during a POST
21407 capture request header Content-Length len 10
21408
21409 # log the beginning of the referrer
21410 capture request header Referer len 20
21411
21412 # server name (useful for outgoing proxies only)
21413 capture response header Server len 20
21414
21415 # logging the content-length is useful with "option logasap"
21416 capture response header Content-Length len 10
21417
Davor Ocelice9ed2812017-12-25 17:49:28 +010021418 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021419 capture response header Cache-Control len 8
21420
21421 # the Via header will report the next proxy's name
21422 capture response header Via len 20
21423
21424 # log the URL location during a redirection
21425 capture response header Location len 20
21426
21427 >>> Aug 9 20:26:09 localhost \
21428 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
21429 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
21430 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
21431 "GET http://fr.adserver.yahoo.com/"
21432
21433 >>> Aug 9 20:30:46 localhost \
21434 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
21435 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
21436 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010021437 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021438
21439 >>> Aug 9 20:30:46 localhost \
21440 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
21441 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
21442 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
21443 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010021444 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021445
21446
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200214478.9. Examples of logs
21448---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021449
21450These are real-world examples of logs accompanied with an explanation. Some of
21451them have been made up by hand. The syslog part has been removed for better
21452reading. Their sole purpose is to explain how to decipher them.
21453
21454 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
21455 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
21456 "HEAD / HTTP/1.0"
21457
21458 => long request (6.5s) entered by hand through 'telnet'. The server replied
21459 in 147 ms, and the session ended normally ('----')
21460
21461 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
21462 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
21463 0/9 "HEAD / HTTP/1.0"
21464
21465 => Idem, but the request was queued in the global queue behind 9 other
21466 requests, and waited there for 1230 ms.
21467
21468 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
21469 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
21470 "GET /image.iso HTTP/1.0"
21471
21472 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010021473 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021474 14 ms, 243 bytes of headers were sent to the client, and total time from
21475 accept to first data byte is 30 ms.
21476
21477 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
21478 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
21479 "GET /cgi-bin/bug.cgi? HTTP/1.0"
21480
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020021481 => the proxy blocked a server response either because of an "http-response
21482 deny" rule, or because the response was improperly formatted and not
21483 HTTP-compliant, or because it blocked sensitive information which risked
21484 being cached. In this case, the response is replaced with a "502 bad
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021485 gateway". The flags ("PH--") tell us that it was HAProxy who decided to
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020021486 return the 502 and not the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021487
21488 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010021489 px-http/<NOSRV> -1/-1/-1/-1/8490 -1 0 - - CR-- 2/2/2/0/0 0/0 ""
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021490
21491 => the client never completed its request and aborted itself ("C---") after
21492 8.5s, while the proxy was waiting for the request headers ("-R--").
21493 Nothing was sent to any server.
21494
21495 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
21496 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
21497
21498 => The client never completed its request, which was aborted by the
21499 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010021500 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021501 send a 408 return code to the client.
21502
21503 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
21504 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
21505
21506 => This log was produced with "option tcplog". The client timed out after
21507 5 seconds ("c----").
21508
21509 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
21510 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010021511 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021512
21513 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021514 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010021515 (config says 'retries 3'), and no redispatch (otherwise we would have
21516 seen "/+3"). Status code 503 was returned to the client. There were 115
21517 connections on this server, 202 connections on this proxy, and 205 on
21518 the global process. It is possible that the server refused the
21519 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010021520
Willy Tarreau52b2d222011-09-07 23:48:48 +020021521
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200215229. Supported filters
21523--------------------
21524
21525Here are listed officially supported filters with the list of parameters they
21526accept. Depending on compile options, some of these filters might be
21527unavailable. The list of available filters is reported in haproxy -vv.
21528
21529See also : "filter"
21530
215319.1. Trace
21532----------
21533
Christopher Fauletc41d8bd2020-11-17 10:43:26 +010021534filter trace [name <name>] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021535
21536 Arguments:
21537 <name> is an arbitrary name that will be reported in
21538 messages. If no name is provided, "TRACE" is used.
21539
Christopher Faulet96a577a2020-11-17 10:45:05 +010021540 <quiet> inhibits trace messages.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021541
Davor Ocelice9ed2812017-12-25 17:49:28 +010021542 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021543 default, this filter forwards all previously parsed
21544 data. With this parameter, it only forwards a random
21545 amount of the parsed data.
21546
Davor Ocelice9ed2812017-12-25 17:49:28 +010021547 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010021548
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021549This filter can be used as a base to develop new filters. It defines all
21550callbacks and print a message on the standard error stream (stderr) with useful
21551information for all of them. It may be useful to debug the activity of other
21552filters or, quite simply, HAProxy's activity.
21553
21554Using <random-parsing> and/or <random-forwarding> parameters is a good way to
21555tests the behavior of a filter that parses data exchanged between a client and
21556a server by adding some latencies in the processing.
21557
21558
215599.2. HTTP compression
21560---------------------
21561
21562filter compression
21563
21564The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
21565keyword must still be used to enable and configure the HTTP compression. And
Christopher Fauletb30b3102019-09-12 23:03:09 +020021566when no other filter is used, it is enough. When used with the cache or the
21567fcgi-app enabled, it is also enough. In this case, the compression is always
21568done after the response is stored in the cache. But it is mandatory to
21569explicitly use a filter line to enable the HTTP compression when at least one
21570filter other than the cache or the fcgi-app is used for the same
21571listener/frontend/backend. This is important to know the filters evaluation
21572order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021573
Christopher Fauletb30b3102019-09-12 23:03:09 +020021574See also : "compression", section 9.4 about the cache filter and section 9.5
21575 about the fcgi-app filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020021576
21577
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200215789.3. Stream Processing Offload Engine (SPOE)
21579--------------------------------------------
21580
21581filter spoe [engine <name>] config <file>
21582
21583 Arguments :
21584
21585 <name> is the engine name that will be used to find the right scope in
21586 the configuration file. If not provided, all the file will be
21587 parsed.
21588
21589 <file> is the path of the engine configuration file. This file can
21590 contain configuration of several engines. In this case, each
21591 part must be placed in its own scope.
21592
21593The Stream Processing Offload Engine (SPOE) is a filter communicating with
21594external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010021595streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020021596exchanged with them are configured in dedicated files, for the main part. It
21597also requires dedicated backends, defined in HAProxy configuration.
21598
21599SPOE communicates with external components using an in-house binary protocol,
21600the Stream Processing Offload Protocol (SPOP).
21601
Tim Düsterhus4896c442016-11-29 02:15:19 +010021602For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020021603"doc/SPOE.txt".
21604
Christopher Faulet99a17a22018-12-11 09:18:27 +0100216059.4. Cache
21606----------
21607
21608filter cache <name>
21609
21610 Arguments :
21611
21612 <name> is name of the cache section this filter will use.
21613
21614The cache uses a filter to store cacheable responses. The HTTP rules
21615"cache-store" and "cache-use" must be used to define how and when to use a
John Roeslerfb2fce12019-07-10 15:45:51 -050021616cache. By default the corresponding filter is implicitly defined. And when no
Christopher Fauletb30b3102019-09-12 23:03:09 +020021617other filters than fcgi-app or compression are used, it is enough. In such
21618case, the compression filter is always evaluated after the cache filter. But it
21619is mandatory to explicitly use a filter line to use a cache when at least one
21620filter other than the compression or the fcgi-app is used for the same
Christopher Faulet27d93c32018-12-15 22:32:02 +010021621listener/frontend/backend. This is important to know the filters evaluation
21622order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010021623
Christopher Fauletb30b3102019-09-12 23:03:09 +020021624See also : section 9.2 about the compression filter, section 9.5 about the
21625 fcgi-app filter and section 6 about cache.
21626
21627
216289.5. Fcgi-app
21629-------------
21630
Daniel Corbett67a82712020-07-06 23:01:19 -040021631filter fcgi-app <name>
Christopher Fauletb30b3102019-09-12 23:03:09 +020021632
21633 Arguments :
21634
21635 <name> is name of the fcgi-app section this filter will use.
21636
21637The FastCGI application uses a filter to evaluate all custom parameters on the
21638request path, and to process the headers on the response path. the <name> must
21639reference an existing fcgi-app section. The directive "use-fcgi-app" should be
21640used to define the application to use. By default the corresponding filter is
21641implicitly defined. And when no other filters than cache or compression are
21642used, it is enough. But it is mandatory to explicitly use a filter line to a
21643fcgi-app when at least one filter other than the compression or the cache is
21644used for the same backend. This is important to know the filters evaluation
21645order.
21646
21647See also: "use-fcgi-app", section 9.2 about the compression filter, section 9.4
21648 about the cache filter and section 10 about FastCGI application.
21649
21650
Miroslav Zagoracdc32cd92020-12-13 18:32:57 +0100216519.6. OpenTracing
21652----------------
21653
21654The OpenTracing filter adds native support for using distributed tracing in
21655HAProxy. This is enabled by sending an OpenTracing compliant request to one
21656of the supported tracers such as Datadog, Jaeger, Lightstep and Zipkin tracers.
21657Please note: tracers are not listed by any preference, but alphabetically.
21658
Daniel Corbett9f0843f2021-05-08 10:50:37 -040021659This feature is only enabled when HAProxy was built with USE_OT=1.
Miroslav Zagoracdc32cd92020-12-13 18:32:57 +010021660
21661The OpenTracing filter activation is done explicitly by specifying it in the
21662HAProxy configuration. If this is not done, the OpenTracing filter in no way
21663participates in the work of HAProxy.
21664
21665filter opentracing [id <id>] config <file>
21666
21667 Arguments :
21668
21669 <id> is the OpenTracing filter id that will be used to find the
21670 right scope in the configuration file. If no filter id is
21671 specified, 'ot-filter' is used as default. If scope is not
21672 specified in the configuration file, it applies to all defined
21673 OpenTracing filters.
21674
21675 <file> is the path of the OpenTracing configuration file. The same
21676 file can contain configurations for multiple OpenTracing
21677 filters simultaneously. In that case we do not need to define
21678 scope so the same configuration applies to all filters or each
21679 filter must have its own scope defined.
21680
21681More detailed documentation related to the operation, configuration and use
Willy Tarreaua63d1a02021-04-02 17:16:46 +020021682of the filter can be found in the addons/ot directory.
Miroslav Zagoracdc32cd92020-12-13 18:32:57 +010021683
21684
Christopher Fauletb30b3102019-09-12 23:03:09 +02002168510. FastCGI applications
21686-------------------------
21687
21688HAProxy is able to send HTTP requests to Responder FastCGI applications. This
21689feature was added in HAProxy 2.1. To do so, servers must be configured to use
21690the FastCGI protocol (using the keyword "proto fcgi" on the server line) and a
21691FastCGI application must be configured and used by the backend managing these
21692servers (using the keyword "use-fcgi-app" into the proxy section). Several
21693FastCGI applications may be defined, but only one can be used at a time by a
21694backend.
21695
21696HAProxy implements all features of the FastCGI specification for Responder
21697application. Especially it is able to multiplex several requests on a simple
21698connection.
21699
2170010.1. Setup
21701-----------
21702
2170310.1.1. Fcgi-app section
21704--------------------------
21705
21706fcgi-app <name>
21707 Declare a FastCGI application named <name>. To be valid, at least the
21708 document root must be defined.
21709
21710acl <aclname> <criterion> [flags] [operator] <value> ...
21711 Declare or complete an access list.
21712
21713 See "acl" keyword in section 4.2 and section 7 about ACL usage for
21714 details. ACLs defined for a FastCGI application are private. They cannot be
21715 used by any other application or by any proxy. In the same way, ACLs defined
21716 in any other section are not usable by a FastCGI application. However,
21717 Pre-defined ACLs are available.
21718
21719docroot <path>
21720 Define the document root on the remote host. <path> will be used to build
21721 the default value of FastCGI parameters SCRIPT_FILENAME and
21722 PATH_TRANSLATED. It is a mandatory setting.
21723
21724index <script-name>
21725 Define the script name that will be appended after an URI that ends with a
21726 slash ("/") to set the default value of the FastCGI parameter SCRIPT_NAME. It
21727 is an optional setting.
21728
21729 Example :
21730 index index.php
21731
21732log-stderr global
21733log-stderr <address> [len <length>] [format <format>]
Jan Wagner3e678602020-12-17 22:22:32 +010021734 [sample <ranges>:<sample_size>] <facility> [<level> [<minlevel>]]
Christopher Fauletb30b3102019-09-12 23:03:09 +020021735 Enable logging of STDERR messages reported by the FastCGI application.
21736
21737 See "log" keyword in section 4.2 for details. It is an optional setting. By
21738 default STDERR messages are ignored.
21739
21740pass-header <name> [ { if | unless } <condition> ]
21741 Specify the name of a request header which will be passed to the FastCGI
21742 application. It may optionally be followed by an ACL-based condition, in
21743 which case it will only be evaluated if the condition is true.
21744
21745 Most request headers are already available to the FastCGI application,
21746 prefixed with "HTTP_". Thus, this directive is only required to pass headers
21747 that are purposefully omitted. Currently, the headers "Authorization",
21748 "Proxy-Authorization" and hop-by-hop headers are omitted.
21749
21750 Note that the headers "Content-type" and "Content-length" are never passed to
21751 the FastCGI application because they are already converted into parameters.
21752
21753path-info <regex>
Christopher Faulet28cb3662020-02-14 14:47:37 +010021754 Define a regular expression to extract the script-name and the path-info from
Christopher Faulet6c57f2d2020-02-14 16:55:52 +010021755 the URL-decoded path. Thus, <regex> may have two captures: the first one to
21756 capture the script name and the second one to capture the path-info. The
21757 first one is mandatory, the second one is optional. This way, it is possible
21758 to extract the script-name from the path ignoring the path-info. It is an
21759 optional setting. If it is not defined, no matching is performed on the
21760 path. and the FastCGI parameters PATH_INFO and PATH_TRANSLATED are not
21761 filled.
Christopher Faulet28cb3662020-02-14 14:47:37 +010021762
21763 For security reason, when this regular expression is defined, the newline and
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050021764 the null characters are forbidden from the path, once URL-decoded. The reason
Christopher Faulet28cb3662020-02-14 14:47:37 +010021765 to such limitation is because otherwise the matching always fails (due to a
21766 limitation one the way regular expression are executed in HAProxy). So if one
21767 of these two characters is found in the URL-decoded path, an error is
21768 returned to the client. The principle of least astonishment is applied here.
Christopher Fauletb30b3102019-09-12 23:03:09 +020021769
21770 Example :
Christopher Faulet6c57f2d2020-02-14 16:55:52 +010021771 path-info ^(/.+\.php)(/.*)?$ # both script-name and path-info may be set
21772 path-info ^(/.+\.php) # the path-info is ignored
Christopher Fauletb30b3102019-09-12 23:03:09 +020021773
21774option get-values
21775no option get-values
21776 Enable or disable the retrieve of variables about connection management.
21777
Daniel Corbett67a82712020-07-06 23:01:19 -040021778 HAProxy is able to send the record FCGI_GET_VALUES on connection
Christopher Fauletb30b3102019-09-12 23:03:09 +020021779 establishment to retrieve the value for following variables:
21780
21781 * FCGI_MAX_REQS The maximum number of concurrent requests this
21782 application will accept.
21783
William Lallemand93e548e2019-09-30 13:54:02 +020021784 * FCGI_MPXS_CONNS "0" if this application does not multiplex connections,
21785 "1" otherwise.
Christopher Fauletb30b3102019-09-12 23:03:09 +020021786
21787 Some FastCGI applications does not support this feature. Some others close
Ilya Shipitsin11057a32020-06-21 21:18:27 +050021788 the connection immediately after sending their response. So, by default, this
Christopher Fauletb30b3102019-09-12 23:03:09 +020021789 option is disabled.
21790
21791 Note that the maximum number of concurrent requests accepted by a FastCGI
21792 application is a connection variable. It only limits the number of streams
21793 per connection. If the global load must be limited on the application, the
21794 server parameters "maxconn" and "pool-max-conn" must be set. In addition, if
21795 an application does not support connection multiplexing, the maximum number
21796 of concurrent requests is automatically set to 1.
21797
21798option keep-conn
21799no option keep-conn
21800 Instruct the FastCGI application to keep the connection open or not after
21801 sending a response.
21802
21803 If disabled, the FastCGI application closes the connection after responding
21804 to this request. By default, this option is enabled.
21805
21806option max-reqs <reqs>
21807 Define the maximum number of concurrent requests this application will
21808 accept.
21809
21810 This option may be overwritten if the variable FCGI_MAX_REQS is retrieved
21811 during connection establishment. Furthermore, if the application does not
21812 support connection multiplexing, this option will be ignored. By default set
21813 to 1.
21814
21815option mpxs-conns
21816no option mpxs-conns
21817 Enable or disable the support of connection multiplexing.
21818
21819 This option may be overwritten if the variable FCGI_MPXS_CONNS is retrieved
21820 during connection establishment. It is disabled by default.
21821
21822set-param <name> <fmt> [ { if | unless } <condition> ]
21823 Set a FastCGI parameter that should be passed to this application. Its
21824 value, defined by <fmt> must follows the log-format rules (see section 8.2.4
21825 "Custom Log format"). It may optionally be followed by an ACL-based
21826 condition, in which case it will only be evaluated if the condition is true.
21827
21828 With this directive, it is possible to overwrite the value of default FastCGI
21829 parameters. If the value is evaluated to an empty string, the rule is
21830 ignored. These directives are evaluated in their declaration order.
21831
21832 Example :
21833 # PHP only, required if PHP was built with --enable-force-cgi-redirect
21834 set-param REDIRECT_STATUS 200
21835
21836 set-param PHP_AUTH_DIGEST %[req.hdr(Authorization)]
21837
21838
2183910.1.2. Proxy section
21840---------------------
21841
21842use-fcgi-app <name>
21843 Define the FastCGI application to use for the backend.
21844
21845 Arguments :
21846 <name> is the name of the FastCGI application to use.
21847
21848 This keyword is only available for HTTP proxies with the backend capability
21849 and with at least one FastCGI server. However, FastCGI servers can be mixed
21850 with HTTP servers. But except there is a good reason to do so, it is not
21851 recommended (see section 10.3 about the limitations for details). Only one
21852 application may be defined at a time per backend.
21853
21854 Note that, once a FastCGI application is referenced for a backend, depending
21855 on the configuration some processing may be done even if the request is not
21856 sent to a FastCGI server. Rules to set parameters or pass headers to an
21857 application are evaluated.
21858
21859
2186010.1.3. Example
21861---------------
21862
21863 frontend front-http
21864 mode http
21865 bind *:80
21866 bind *:
21867
21868 use_backend back-dynamic if { path_reg ^/.+\.php(/.*)?$ }
21869 default_backend back-static
21870
21871 backend back-static
21872 mode http
21873 server www A.B.C.D:80
21874
21875 backend back-dynamic
21876 mode http
21877 use-fcgi-app php-fpm
21878 server php-fpm A.B.C.D:9000 proto fcgi
21879
21880 fcgi-app php-fpm
21881 log-stderr global
21882 option keep-conn
21883
21884 docroot /var/www/my-app
21885 index index.php
21886 path-info ^(/.+\.php)(/.*)?$
21887
21888
2188910.2. Default parameters
21890------------------------
21891
21892A Responder FastCGI application has the same purpose as a CGI/1.1 program. In
21893the CGI/1.1 specification (RFC3875), several variables must be passed to the
Ilya Shipitsin8525fd92020-02-29 12:34:59 +050021894script. So HAProxy set them and some others commonly used by FastCGI
Christopher Fauletb30b3102019-09-12 23:03:09 +020021895applications. All these variables may be overwritten, with caution though.
21896
21897 +-------------------+-----------------------------------------------------+
21898 | AUTH_TYPE | Identifies the mechanism, if any, used by HAProxy |
21899 | | to authenticate the user. Concretely, only the |
21900 | | BASIC authentication mechanism is supported. |
21901 | | |
21902 +-------------------+-----------------------------------------------------+
21903 | CONTENT_LENGTH | Contains the size of the message-body attached to |
21904 | | the request. It means only requests with a known |
21905 | | size are considered as valid and sent to the |
21906 | | application. |
21907 | | |
21908 +-------------------+-----------------------------------------------------+
21909 | CONTENT_TYPE | Contains the type of the message-body attached to |
21910 | | the request. It may not be set. |
21911 | | |
21912 +-------------------+-----------------------------------------------------+
21913 | DOCUMENT_ROOT | Contains the document root on the remote host under |
21914 | | which the script should be executed, as defined in |
21915 | | the application's configuration. |
21916 | | |
21917 +-------------------+-----------------------------------------------------+
21918 | GATEWAY_INTERFACE | Contains the dialect of CGI being used by HAProxy |
21919 | | to communicate with the FastCGI application. |
21920 | | Concretely, it is set to "CGI/1.1". |
21921 | | |
21922 +-------------------+-----------------------------------------------------+
21923 | PATH_INFO | Contains the portion of the URI path hierarchy |
21924 | | following the part that identifies the script |
21925 | | itself. To be set, the directive "path-info" must |
21926 | | be defined. |
21927 | | |
21928 +-------------------+-----------------------------------------------------+
21929 | PATH_TRANSLATED | If PATH_INFO is set, it is its translated version. |
21930 | | It is the concatenation of DOCUMENT_ROOT and |
21931 | | PATH_INFO. If PATH_INFO is not set, this parameters |
21932 | | is not set too. |
21933 | | |
21934 +-------------------+-----------------------------------------------------+
21935 | QUERY_STRING | Contains the request's query string. It may not be |
21936 | | set. |
21937 | | |
21938 +-------------------+-----------------------------------------------------+
21939 | REMOTE_ADDR | Contains the network address of the client sending |
21940 | | the request. |
21941 | | |
21942 +-------------------+-----------------------------------------------------+
21943 | REMOTE_USER | Contains the user identification string supplied by |
21944 | | client as part of user authentication. |
21945 | | |
21946 +-------------------+-----------------------------------------------------+
21947 | REQUEST_METHOD | Contains the method which should be used by the |
21948 | | script to process the request. |
21949 | | |
21950 +-------------------+-----------------------------------------------------+
21951 | REQUEST_URI | Contains the request's URI. |
21952 | | |
21953 +-------------------+-----------------------------------------------------+
21954 | SCRIPT_FILENAME | Contains the absolute pathname of the script. it is |
21955 | | the concatenation of DOCUMENT_ROOT and SCRIPT_NAME. |
21956 | | |
21957 +-------------------+-----------------------------------------------------+
21958 | SCRIPT_NAME | Contains the name of the script. If the directive |
21959 | | "path-info" is defined, it is the first part of the |
21960 | | URI path hierarchy, ending with the script name. |
21961 | | Otherwise, it is the entire URI path. |
21962 | | |
21963 +-------------------+-----------------------------------------------------+
21964 | SERVER_NAME | Contains the name of the server host to which the |
21965 | | client request is directed. It is the value of the |
21966 | | header "Host", if defined. Otherwise, the |
21967 | | destination address of the connection on the client |
21968 | | side. |
21969 | | |
21970 +-------------------+-----------------------------------------------------+
21971 | SERVER_PORT | Contains the destination TCP port of the connection |
21972 | | on the client side, which is the port the client |
21973 | | connected to. |
21974 | | |
21975 +-------------------+-----------------------------------------------------+
21976 | SERVER_PROTOCOL | Contains the request's protocol. |
21977 | | |
21978 +-------------------+-----------------------------------------------------+
Christopher Fauletb2a50292021-06-11 13:34:42 +020021979 | SERVER_SOFTWARE | Contains the string "HAProxy" followed by the |
21980 | | current HAProxy version. |
21981 | | |
21982 +-------------------+-----------------------------------------------------+
Christopher Fauletb30b3102019-09-12 23:03:09 +020021983 | HTTPS | Set to a non-empty value ("on") if the script was |
21984 | | queried through the HTTPS protocol. |
21985 | | |
21986 +-------------------+-----------------------------------------------------+
21987
21988
2198910.3. Limitations
21990------------------
21991
21992The current implementation have some limitations. The first one is about the
21993way some request headers are hidden to the FastCGI applications. This happens
21994during the headers analysis, on the backend side, before the connection
21995establishment. At this stage, HAProxy know the backend is using a FastCGI
21996application but it don't know if the request will be routed to a FastCGI server
21997or not. But to hide request headers, it simply removes them from the HTX
21998message. So, if the request is finally routed to an HTTP server, it never see
21999these headers. For this reason, it is not recommended to mix FastCGI servers
22000and HTTP servers under the same backend.
22001
22002Similarly, the rules "set-param" and "pass-header" are evaluated during the
22003request headers analysis. So the evaluation is always performed, even if the
22004requests is finally forwarded to an HTTP server.
22005
22006About the rules "set-param", when a rule is applied, a pseudo header is added
22007into the HTX message. So, the same way than for HTTP header rewrites, it may
22008fail if the buffer is full. The rules "set-param" will compete with
22009"http-request" ones.
22010
22011Finally, all FastCGI params and HTTP headers are sent into a unique record
22012FCGI_PARAM. Encoding of this record must be done in one pass, otherwise a
22013processing error is returned. It means the record FCGI_PARAM, once encoded,
22014must not exceeds the size of a buffer. However, there is no reserve to respect
22015here.
William Lallemand86d0df02017-11-24 21:36:45 +010022016
Emeric Brunce325c42021-04-02 17:05:09 +020022017
2201811. Address formats
22019-------------------
22020
22021Several statements as "bind, "server", "nameserver" and "log" requires an
22022address.
22023
22024This address can be a host name, an IPv4 address, an IPv6 address, or '*'.
22025The '*' is equal to the special address "0.0.0.0" and can be used, in the case
22026of "bind" or "dgram-bind" to listen on all IPv4 of the system.The IPv6
22027equivalent is '::'.
22028
22029Depending of the statement, a port or port range follows the IP address. This
22030is mandatory on 'bind' statement, optional on 'server'.
22031
22032This address can also begin with a slash '/'. It is considered as the "unix"
22033family, and '/' and following characters must be present the path.
22034
22035Default socket type or transport method "datagram" or "stream" depends on the
22036configuration statement showing the address. Indeed, 'bind' and 'server' will
22037use a "stream" socket type by default whereas 'log', 'nameserver' or
22038'dgram-bind' will use a "datagram".
22039
22040Optionally, a prefix could be used to force the address family and/or the
22041socket type and the transport method.
22042
22043
2204411.1 Address family prefixes
22045----------------------------
22046
22047'abns@<name>' following <name> is an abstract namespace (Linux only).
22048
22049'fd@<n>' following address is a file descriptor <n> inherited from the
22050 parent. The fd must be bound and may or may not already be
22051 listening.
22052
22053'ip@<address>[:port1[-port2]]' following <address> is considered as an IPv4 or
22054 IPv6 address depending on the syntax. Depending
22055 on the statement using this address, a port or
22056 a port range may or must be specified.
22057
22058'ipv4@<address>[:port1[-port2]]' following <address> is always considered as
22059 an IPv4 address. Depending on the statement
22060 using this address, a port or a port range
22061 may or must be specified.
22062
22063'ipv6@<address>[:port1[-port2]]' following <address> is always considered as
22064 an IPv6 address. Depending on the statement
22065 using this address, a port or a port range
22066 may or must be specified.
22067
22068'sockpair@<n>' following address is the file descriptor of a connected unix
22069 socket or of a socketpair. During a connection, the initiator
22070 creates a pair of connected sockets, and passes one of them
22071 over the FD to the other end. The listener waits to receive
22072 the FD from the unix socket and uses it as if it were the FD
22073 of an accept(). Should be used carefully.
22074
22075'unix@<path>' following string is considered as a UNIX socket <path>. this
22076 prefix is useful to declare an UNIX socket path which don't
22077 start by slash '/'.
22078
22079
2208011.2 Socket type prefixes
22081-------------------------
22082
22083Previous "Address family prefixes" can also be prefixed to force the socket
22084type and the transport method. The default depends of the statement using
22085this address but in some cases the user may force it to a different one.
22086This is the case for "log" statement where the default is syslog over UDP
22087but we could force to use syslog over TCP.
22088
22089Those prefixes were designed for internal purpose and users should
22090instead use aliases of the next section "11.5.3 Protocol prefixes".
22091
22092If users need one those prefixes to perform what they expect because
22093they can not configure the same using the protocol prefixes, they should
22094report this to the maintainers.
22095
22096'stream+<family>@<address>' forces socket type and transport method
22097 to "stream"
22098
22099'dgram+<family>@<address>' forces socket type and transport method
22100 to "datagram".
22101
22102
2210311.3 Protocol prefixes
22104----------------------
22105
22106'tcp@<address>[:port1[-port2]]' following <address> is considered as an IPv4
22107 or IPv6 address depending of the syntax but
22108 socket type and transport method is forced to
22109 "stream". Depending on the statement using
22110 this address, a port or a port range can or
22111 must be specified. It is considered as an alias
22112 of 'stream+ip@'.
22113
22114'tcp4@<address>[:port1[-port2]]' following <address> is always considered as
22115 an IPv4 address but socket type and transport
22116 method is forced to "stream". Depending on the
22117 statement using this address, a port or port
22118 range can or must be specified.
22119 It is considered as an alias of 'stream+ipv4@'.
22120
22121'tcp6@<address>[:port1[-port2]]' following <address> is always considered as
22122 an IPv6 address but socket type and transport
22123 method is forced to "stream". Depending on the
22124 statement using this address, a port or port
22125 range can or must be specified.
22126 It is considered as an alias of 'stream+ipv4@'.
22127
22128'udp@<address>[:port1[-port2]]' following <address> is considered as an IPv4
22129 or IPv6 address depending of the syntax but
22130 socket type and transport method is forced to
22131 "datagram". Depending on the statement using
22132 this address, a port or a port range can or
22133 must be specified. It is considered as an alias
22134 of 'dgram+ip@'.
22135
22136'udp4@<address>[:port1[-port2]]' following <address> is always considered as
22137 an IPv4 address but socket type and transport
22138 method is forced to "datagram". Depending on
22139 the statement using this address, a port or
22140 port range can or must be specified.
22141 It is considered as an alias of 'stream+ipv4@'.
22142
22143'udp6@<address>[:port1[-port2]]' following <address> is always considered as
22144 an IPv6 address but socket type and transport
22145 method is forced to "datagram". Depending on
22146 the statement using this address, a port or
22147 port range can or must be specified.
22148 It is considered as an alias of 'stream+ipv4@'.
22149
22150'uxdg@<path>' following string is considered as a unix socket <path> but
22151 transport method is forced to "datagram". It is considered as
22152 an alias of 'dgram+unix@'.
22153
22154'uxst@<path>' following string is considered as a unix socket <path> but
22155 transport method is forced to "stream". It is considered as
22156 an alias of 'stream+unix@'.
22157
22158In future versions, other prefixes could be used to specify protocols like
22159QUIC which proposes stream transport based on socket of type "datagram".
22160
Willy Tarreau0ba27502007-12-24 16:55:16 +010022161/*
22162 * Local variables:
22163 * fill-column: 79
22164 * End:
22165 */