blob: b60a8d45ca2d760052211790950f77f8c730a1ca [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 Tarreau2e077f82019-11-25 20:36:16 +01005 version 2.2
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau32bf97f2020-02-25 18:14:02 +01007 2020/02/25
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
442.4. Time format
452.5. Examples
Willy Tarreauc57f0e22009-05-10 13:12:33 +020046
473. Global parameters
483.1. Process management and security
493.2. Performance tuning
503.3. Debugging
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100513.4. Userlists
Cyril Bontédc4d9032012-04-08 21:57:39 +0200523.5. Peers
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200533.6. Mailers
William Lallemandc9515522019-06-12 16:32:11 +0200543.7. Programs
Christopher Faulet76edc0f2020-01-13 15:52:01 +0100553.8. HTTP-errors
Willy Tarreauc57f0e22009-05-10 13:12:33 +020056
574. Proxies
584.1. Proxy keywords matrix
594.2. Alphabetically sorted keywords reference
60
Davor Ocelice9ed2812017-12-25 17:49:28 +0100615. Bind and server options
Willy Tarreau086fbf52012-09-24 20:34:51 +0200625.1. Bind options
635.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200645.3. Server DNS resolution
655.3.1. Global overview
665.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020067
Julien Pivotto6ccee412019-11-27 15:49:54 +0100686. Cache
696.1. Limitation
706.2. Setup
716.2.1. Cache section
726.2.2. Proxy section
73
Willy Tarreau74ca5042013-06-11 23:12:07 +0200747. Using ACLs and fetching samples
757.1. ACL basics
767.1.1. Matching booleans
777.1.2. Matching integers
787.1.3. Matching strings
797.1.4. Matching regular expressions (regexes)
807.1.5. Matching arbitrary data blocks
817.1.6. Matching IPv4 and IPv6 addresses
827.2. Using ACLs to form conditions
837.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200847.3.1. Converters
857.3.2. Fetching samples from internal states
867.3.3. Fetching samples at Layer 4
877.3.4. Fetching samples at Layer 5
887.3.5. Fetching samples from buffer contents (Layer 6)
897.3.6. Fetching HTTP samples (Layer 7)
Christopher Fauletd47941d2020-01-08 14:40:19 +0100907.3.7. Fetching samples for developers
Willy Tarreau74ca5042013-06-11 23:12:07 +0200917.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020092
938. Logging
948.1. Log levels
958.2. Log formats
968.2.1. Default log format
978.2.2. TCP log format
988.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100998.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +01001008.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001018.3. Advanced logging options
1028.3.1. Disabling logging of external tests
1038.3.2. Logging before waiting for the session to terminate
1048.3.3. Raising log level upon errors
1058.3.4. Disabling logging of successful connections
1068.4. Timing events
1078.5. Session state at disconnection
1088.6. Non-printable characters
1098.7. Capturing HTTP cookies
1108.8. Capturing HTTP headers
1118.9. Examples of logs
112
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001139. Supported filters
1149.1. Trace
1159.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001169.3. Stream Processing Offload Engine (SPOE)
Christopher Faulet99a17a22018-12-11 09:18:27 +01001179.4. Cache
Christopher Fauletb30b3102019-09-12 23:03:09 +02001189.5. fcgi-app
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200119
Christopher Fauletb30b3102019-09-12 23:03:09 +020012010. FastCGI applications
12110.1. Setup
12210.1.1. Fcgi-app section
12310.1.2. Proxy section
12410.1.3. Example
12510.2. Default parameters
12610.3. Limitations
127
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128
1291. Quick reminder about HTTP
130----------------------------
131
Davor Ocelice9ed2812017-12-25 17:49:28 +0100132When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200133fully analyzed and indexed, thus it becomes possible to build matching criteria
134on almost anything found in the contents.
135
136However, it is important to understand how HTTP requests and responses are
137formed, and how HAProxy decomposes them. It will then become easier to write
138correct rules and to debug existing configurations.
139
140
1411.1. The HTTP transaction model
142-------------------------------
143
144The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100145to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100146from the client to the server, a request is sent by the client through the
147connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200148will involve a new connection :
149
150 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
151
152In this mode, called the "HTTP close" mode, there are as many connection
153establishments as there are HTTP transactions. Since the connection is closed
154by the server after the response, the client does not need to know the content
155length.
156
157Due to the transactional nature of the protocol, it was possible to improve it
158to avoid closing a connection between two subsequent transactions. In this mode
159however, it is mandatory that the server indicates the content length for each
160response so that the client does not wait indefinitely. For this, a special
161header is used: "Content-length". This mode is called the "keep-alive" mode :
162
163 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
164
165Its advantages are a reduced latency between transactions, and less processing
166power required on the server side. It is generally better than the close mode,
167but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200168a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169
Willy Tarreau95c4e142017-11-26 12:18:55 +0100170Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171keep-alive, but the client does not wait for the first response to send the
172second request. This is useful for fetching large number of images composing a
173page :
174
175 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
176
177This can obviously have a tremendous benefit on performance because the network
178latency is eliminated between subsequent requests. Many HTTP agents do not
179correctly support pipelining since there is no way to associate a response with
180the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100181server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200182
Willy Tarreau95c4e142017-11-26 12:18:55 +0100183The next improvement is the multiplexed mode, as implemented in HTTP/2. This
184time, each transaction is assigned a single stream identifier, and all streams
185are multiplexed over an existing connection. Many requests can be sent in
186parallel by the client, and responses can arrive in any order since they also
187carry the stream identifier.
188
Willy Tarreau70dffda2014-01-30 03:07:23 +0100189By default HAProxy operates in keep-alive mode with regards to persistent
190connections: for each connection it processes each request and response, and
191leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100192start of a new request. When it receives HTTP/2 connections from a client, it
193processes all the requests in parallel and leaves the connection idling,
194waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200195
Christopher Faulet315b39c2018-09-21 16:26:19 +0200196HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +0100197 - keep alive : all requests and responses are processed (default)
198 - tunnel : only the first request and response are processed,
Christopher Faulet6c9bbb22019-03-26 21:37:23 +0100199 everything else is forwarded with no analysis (deprecated).
Willy Tarreau70dffda2014-01-30 03:07:23 +0100200 - server close : the server-facing connection is closed after the response.
Christopher Faulet315b39c2018-09-21 16:26:19 +0200201 - close : the connection is actively closed after end of response.
Willy Tarreau70dffda2014-01-30 03:07:23 +0100202
Davor Ocelice9ed2812017-12-25 17:49:28 +0100203For HTTP/2, the connection mode resembles more the "server close" mode : given
204the independence of all streams, there is currently no place to hook the idle
Willy Tarreau95c4e142017-11-26 12:18:55 +0100205server connection after a response, so it is closed after the response. HTTP/2
206is only supported for incoming connections, not on connections going to
207servers.
208
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200209
2101.2. HTTP request
211-----------------
212
213First, let's consider this HTTP request :
214
215 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100216 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200217 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
218 2 Host: www.mydomain.com
219 3 User-agent: my small browser
220 4 Accept: image/jpeg, image/gif
221 5 Accept: image/png
222
223
2241.2.1. The Request line
225-----------------------
226
227Line 1 is the "request line". It is always composed of 3 fields :
228
229 - a METHOD : GET
230 - a URI : /serv/login.php?lang=en&profile=2
231 - a version tag : HTTP/1.1
232
233All of them are delimited by what the standard calls LWS (linear white spaces),
234which are commonly spaces, but can also be tabs or line feeds/carriage returns
235followed by spaces/tabs. The method itself cannot contain any colon (':') and
236is limited to alphabetic letters. All those various combinations make it
237desirable that HAProxy performs the splitting itself rather than leaving it to
238the user to write a complex or inaccurate regular expression.
239
240The URI itself can have several forms :
241
242 - A "relative URI" :
243
244 /serv/login.php?lang=en&profile=2
245
246 It is a complete URL without the host part. This is generally what is
247 received by servers, reverse proxies and transparent proxies.
248
249 - An "absolute URI", also called a "URL" :
250
251 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
252
253 It is composed of a "scheme" (the protocol name followed by '://'), a host
254 name or address, optionally a colon (':') followed by a port number, then
255 a relative URI beginning at the first slash ('/') after the address part.
256 This is generally what proxies receive, but a server supporting HTTP/1.1
257 must accept this form too.
258
259 - a star ('*') : this form is only accepted in association with the OPTIONS
260 method and is not relayable. It is used to inquiry a next hop's
261 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100262
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200263 - an address:port combination : 192.168.0.12:80
264 This is used with the CONNECT method, which is used to establish TCP
265 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
266 other protocols too.
267
268In a relative URI, two sub-parts are identified. The part before the question
269mark is called the "path". It is typically the relative path to static objects
270on the server. The part after the question mark is called the "query string".
271It is mostly used with GET requests sent to dynamic scripts and is very
272specific to the language, framework or application in use.
273
Willy Tarreau95c4e142017-11-26 12:18:55 +0100274HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100275assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100276However, haproxy natively processes HTTP/1.x requests and headers, so requests
277received over an HTTP/2 connection are transcoded to HTTP/1.1 before being
278processed. This explains why they still appear as "HTTP/1.1" in haproxy's logs
279as well as in server logs.
280
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200281
2821.2.2. The request headers
283--------------------------
284
285The headers start at the second line. They are composed of a name at the
286beginning of the line, immediately followed by a colon (':'). Traditionally,
287an LWS is added after the colon but that's not required. Then come the values.
288Multiple identical headers may be folded into one single line, delimiting the
289values with commas, provided that their order is respected. This is commonly
290encountered in the "Cookie:" field. A header may span over multiple lines if
291the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
292define a total of 3 values for the "Accept:" header.
293
Davor Ocelice9ed2812017-12-25 17:49:28 +0100294Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200295their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100296"Connection:" header). In HTTP/2, header names are always sent in lower case,
297as can be seen when running in debug mode.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200298
299The end of the headers is indicated by the first empty line. People often say
300that it's a double line feed, which is not exact, even if a double line feed
301is one valid form of empty line.
302
303Fortunately, HAProxy takes care of all these complex combinations when indexing
304headers, checking values and counting them, so there is no reason to worry
305about the way they could be written, but it is important not to accuse an
306application of being buggy if it does unusual, valid things.
307
308Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000309 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200310 in the middle of headers by LWS in order to join multi-line headers. This
311 is necessary for proper analysis and helps less capable HTTP parsers to work
312 correctly and not to be fooled by such complex constructs.
313
314
3151.3. HTTP response
316------------------
317
318An HTTP response looks very much like an HTTP request. Both are called HTTP
319messages. Let's consider this HTTP response :
320
321 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100322 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200323 1 HTTP/1.1 200 OK
324 2 Content-length: 350
325 3 Content-Type: text/html
326
Willy Tarreau816b9792009-09-15 21:25:21 +0200327As a special case, HTTP supports so called "Informational responses" as status
328codes 1xx. These messages are special in that they don't convey any part of the
329response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100330continue to post its request for instance. In the case of a status 100 response
331the requested information will be carried by the next non-100 response message
332following the informational one. This implies that multiple responses may be
333sent to a single request, and that this only works when keep-alive is enabled
334(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
335correctly forward and skip them, and only process the next non-100 response. As
336such, these messages are neither logged nor transformed, unless explicitly
337state otherwise. Status 101 messages indicate that the protocol is changing
338over the same connection and that haproxy must switch to tunnel mode, just as
339if a CONNECT had occurred. Then the Upgrade header would contain additional
340information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200341
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200342
Davor Ocelice9ed2812017-12-25 17:49:28 +01003431.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200344------------------------
345
346Line 1 is the "response line". It is always composed of 3 fields :
347
348 - a version tag : HTTP/1.1
349 - a status code : 200
350 - a reason : OK
351
352The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100353 - 1xx = informational message to be skipped (e.g. 100, 101)
354 - 2xx = OK, content is following (e.g. 200, 206)
355 - 3xx = OK, no content following (e.g. 302, 304)
356 - 4xx = error caused by the client (e.g. 401, 403, 404)
357 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200358
Lukas Tribus23953682017-04-28 13:24:30 +0000359Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100360"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200361found there, but it's a common practice to respect the well-established
362messages. It can be composed of one or multiple words, such as "OK", "Found",
363or "Authentication Required".
364
Davor Ocelice9ed2812017-12-25 17:49:28 +0100365HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200366
367 Code When / reason
368 200 access to stats page, and when replying to monitoring requests
369 301 when performing a redirection, depending on the configured code
370 302 when performing a redirection, depending on the configured code
371 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100372 307 when performing a redirection, depending on the configured code
373 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200374 400 for an invalid or too large request
375 401 when an authentication is required to perform the action (when
376 accessing the stats page)
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200377 403 when a request is forbidden by a "http-request deny" rule
Florian Tham9205fea2020-01-08 13:35:30 +0100378 404 when the requested resource could not be found
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200379 408 when the request timeout strikes before the request is complete
Florian Tham272e29b2020-01-08 10:19:05 +0100380 410 when the requested resource is no longer available and will not
381 be available again
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200382 500 when haproxy encounters an unrecoverable internal error, such as a
383 memory allocation failure, which should never happen
384 502 when the server returns an empty, invalid or incomplete response, or
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200385 when an "http-response deny" rule blocks the response.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200386 503 when no server was available to handle the request, or in response to
387 monitoring requests which match the "monitor fail" condition
388 504 when the response timeout strikes before the server responds
389
390The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3914.2).
392
393
3941.3.2. The response headers
395---------------------------
396
397Response headers work exactly like request headers, and as such, HAProxy uses
398the same parsing function for both. Please refer to paragraph 1.2.2 for more
399details.
400
401
4022. Configuring HAProxy
403----------------------
404
4052.1. Configuration file format
406------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200407
408HAProxy's configuration process involves 3 major sources of parameters :
409
410 - the arguments from the command-line, which always take precedence
411 - the "global" section, which sets process-wide parameters
412 - the proxies sections which can take form of "defaults", "listen",
413 "frontend" and "backend".
414
Willy Tarreau0ba27502007-12-24 16:55:16 +0100415The configuration file syntax consists in lines beginning with a keyword
416referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200417delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100418
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200419
William Lallemandf9873ba2015-05-05 17:37:14 +02004202.2. Quoting and escaping
421-------------------------
422
423HAProxy's configuration introduces a quoting and escaping system similar to
424many programming languages. The configuration file supports 3 types: escaping
425with a backslash, weak quoting with double quotes, and strong quoting with
426single quotes.
427
428If spaces have to be entered in strings, then they must be escaped by preceding
429them by a backslash ('\') or by quoting them. Backslashes also have to be
430escaped by doubling or strong quoting them.
431
432Escaping is achieved by preceding a special character by a backslash ('\'):
433
434 \ to mark a space and differentiate it from a delimiter
435 \# to mark a hash and differentiate it from a comment
436 \\ to use a backslash
437 \' to use a single quote and differentiate it from strong quoting
438 \" to use a double quote and differentiate it from weak quoting
439
440Weak quoting is achieved by using double quotes (""). Weak quoting prevents
441the interpretation of:
442
443 space as a parameter separator
444 ' single quote as a strong quoting delimiter
445 # hash as a comment start
446
William Lallemandb2f07452015-05-12 14:27:13 +0200447Weak quoting permits the interpretation of variables, if you want to use a non
448-interpreted dollar within a double quoted string, you should escape it with a
449backslash ("\$"), it does not work outside weak quoting.
450
451Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200452quoting.
453
454Strong quoting is achieved by using single quotes (''). Inside single quotes,
455nothing is interpreted, it's the efficient way to quote regexes.
456
457Quoted and escaped strings are replaced in memory by their interpreted
458equivalent, it allows you to perform concatenation.
459
460 Example:
461 # those are equivalents:
462 log-format %{+Q}o\ %t\ %s\ %{-Q}r
463 log-format "%{+Q}o %t %s %{-Q}r"
464 log-format '%{+Q}o %t %s %{-Q}r'
465 log-format "%{+Q}o %t"' %s %{-Q}r'
466 log-format "%{+Q}o %t"' %s'\ %{-Q}r
467
468 # those are equivalents:
469 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
470 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
471 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
472 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
473
474
William Lallemandb2f07452015-05-12 14:27:13 +02004752.3. Environment variables
476--------------------------
477
478HAProxy's configuration supports environment variables. Those variables are
479interpreted only within double quotes. Variables are expanded during the
480configuration parsing. Variable names must be preceded by a dollar ("$") and
481optionally enclosed with braces ("{}") similarly to what is done in Bourne
482shell. Variable names can contain alphanumerical characters or the character
483underscore ("_") but should not start with a digit.
484
485 Example:
486
487 bind "fd@${FD_APP1}"
488
489 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
490
491 user "$HAPROXY_USER"
492
William Lallemand4d03e432019-06-14 15:35:37 +0200493Some variables are defined by HAProxy, they can be used in the configuration
494file, or could be inherited by a program (See 3.7. Programs):
William Lallemanddaf4cd22018-04-17 16:46:13 +0200495
William Lallemand4d03e432019-06-14 15:35:37 +0200496* HAPROXY_LOCALPEER: defined at the startup of the process which contains the
497 name of the local peer. (See "-L" in the management guide.)
498
499* HAPROXY_CFGFILES: list of the configuration files loaded by HAProxy,
500 separated by semicolons. Can be useful in the case you specified a
501 directory.
502
503* HAPROXY_MWORKER: In master-worker mode, this variable is set to 1.
504
John Roeslerfb2fce12019-07-10 15:45:51 -0500505* HAPROXY_CLI: configured listeners addresses of the stats socket for every
William Lallemand4d03e432019-06-14 15:35:37 +0200506 processes, separated by semicolons.
507
John Roeslerfb2fce12019-07-10 15:45:51 -0500508* HAPROXY_MASTER_CLI: In master-worker mode, listeners addresses of the master
William Lallemand4d03e432019-06-14 15:35:37 +0200509 CLI, separated by semicolons.
510
511See also "external-check command" for other variables.
William Lallemandb2f07452015-05-12 14:27:13 +0200512
5132.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200514----------------
515
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100516Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100517values are generally expressed in milliseconds (unless explicitly stated
518otherwise) but may be expressed in any other unit by suffixing the unit to the
519numeric value. It is important to consider this because it will not be repeated
520for every keyword. Supported units are :
521
522 - us : microseconds. 1 microsecond = 1/1000000 second
523 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
524 - s : seconds. 1s = 1000ms
525 - m : minutes. 1m = 60s = 60000ms
526 - h : hours. 1h = 60m = 3600s = 3600000ms
527 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
528
529
Lukas Tribusaa83a312017-03-21 09:25:09 +00005302.5. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200531-------------
532
533 # Simple configuration for an HTTP proxy listening on port 80 on all
534 # interfaces and forwarding requests to a single backend "servers" with a
535 # single server "server1" listening on 127.0.0.1:8000
536 global
537 daemon
538 maxconn 256
539
540 defaults
541 mode http
542 timeout connect 5000ms
543 timeout client 50000ms
544 timeout server 50000ms
545
546 frontend http-in
547 bind *:80
548 default_backend servers
549
550 backend servers
551 server server1 127.0.0.1:8000 maxconn 32
552
553
554 # The same configuration defined with a single listen block. Shorter but
555 # less expressive, especially in HTTP mode.
556 global
557 daemon
558 maxconn 256
559
560 defaults
561 mode http
562 timeout connect 5000ms
563 timeout client 50000ms
564 timeout server 50000ms
565
566 listen http-in
567 bind *:80
568 server server1 127.0.0.1:8000 maxconn 32
569
570
571Assuming haproxy is in $PATH, test these configurations in a shell with:
572
Willy Tarreauccb289d2010-12-11 20:19:38 +0100573 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200574
575
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005763. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200577--------------------
578
579Parameters in the "global" section are process-wide and often OS-specific. They
580are generally set once for all and do not need being changed once correct. Some
581of them have command-line equivalents.
582
583The following keywords are supported in the "global" section :
584
585 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200586 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200587 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200588 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200589 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200590 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200591 - description
592 - deviceatlas-json-file
593 - deviceatlas-log-level
594 - deviceatlas-separator
595 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900596 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200597 - gid
598 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100599 - hard-stop-after
Christopher Faulet98fbe952019-07-22 16:18:24 +0200600 - h1-case-adjust
601 - h1-case-adjust-file
Willy Tarreaud96f1122019-12-03 07:07:36 +0100602 - insecure-fork-wanted
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100603 - insecure-setuid-wanted
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +0100604 - issuers-chain-path
Willy Tarreau6a06a402007-07-15 20:15:28 +0200605 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200606 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100607 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200608 - lua-load
Tim Duesterhusdd74b5f2020-01-12 13:55:40 +0100609 - lua-prepend-path
William Lallemand27edc4b2019-05-07 17:49:33 +0200610 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200611 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200612 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200613 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200614 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100615 - presetenv
616 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200617 - uid
618 - ulimit-n
619 - user
Willy Tarreau636848a2019-04-15 19:38:50 +0200620 - set-dumpable
Willy Tarreau1d549722016-02-16 12:41:57 +0100621 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200622 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200623 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200624 - ssl-default-bind-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200625 - ssl-default-bind-options
626 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200627 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200628 - ssl-default-server-options
629 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100630 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100631 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100632 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100633 - 51degrees-data-file
634 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200635 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200636 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200637 - wurfl-data-file
638 - wurfl-information-list
639 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200640 - wurfl-cache-size
William Dauchy0fec3ab2019-10-27 20:08:11 +0100641 - strict-limits
Willy Tarreaud72758d2010-01-12 10:42:19 +0100642
Willy Tarreau6a06a402007-07-15 20:15:28 +0200643 * Performance tuning
William Dauchy0a8824f2019-10-27 20:08:09 +0100644 - busy-polling
Willy Tarreau1746eec2014-04-25 10:46:47 +0200645 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200646 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200647 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100648 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100649 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100650 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200651 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200652 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200653 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200654 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200655 - noepoll
656 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +0000657 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +0200658 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100659 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300660 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000661 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +0100662 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +0200663 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200664 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200665 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000666 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000667 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200668 - tune.buffers.limit
669 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200670 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200671 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100672 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200673 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200674 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200675 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100676 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200677 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200678 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100679 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100680 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100681 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100682 - tune.lua.session-timeout
683 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200684 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100685 - tune.maxaccept
686 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200687 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200688 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200689 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100690 - tune.rcvbuf.client
691 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100692 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200693 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100694 - tune.sndbuf.client
695 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100696 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100697 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200698 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100699 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200700 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200701 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100702 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200703 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100704 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200705 - tune.vars.reqres-max-size
706 - tune.vars.sess-max-size
707 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100708 - tune.zlib.memlevel
709 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100710
Willy Tarreau6a06a402007-07-15 20:15:28 +0200711 * Debugging
712 - debug
713 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200714
715
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007163.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200717------------------------------------
718
Emeric Brunc8e8d122012-10-02 18:42:10 +0200719ca-base <dir>
720 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200721 relative path is used with "ca-file" or "crl-file" directives. Absolute
722 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200723
Willy Tarreau6a06a402007-07-15 20:15:28 +0200724chroot <jail dir>
725 Changes current directory to <jail dir> and performs a chroot() there before
726 dropping privileges. This increases the security level in case an unknown
727 vulnerability would be exploited, since it would make it very hard for the
728 attacker to exploit the system. This only works when the process is started
729 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100730 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100731
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100732cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
733 On Linux 2.6 and above, it is possible to bind a process or a thread to a
734 specific CPU set. This means that the process or the thread will never run on
735 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
736 sets. The first argument is a process set, eventually followed by a thread
737 set. These sets have the format
738
739 all | odd | even | number[-[number]]
740
741 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100742 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100743 ignored. It is possible to specify a range with two such number delimited by
744 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100745 "all", only odd numbers using "odd" or even numbers using "even", just like
746 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100747 CPU sets. Each CPU set is either a unique number between 0 and 31 or 63 or a
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100748 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100749 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100750 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100751 specified. Each "cpu-map" directive will replace the previous ones when they
752 overlap. A thread will be bound on the intersection of its mapping and the
753 one of the process on which it is attached. If the intersection is null, no
754 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100755
Christopher Fauletff4121f2017-11-22 16:38:49 +0100756 Ranges can be partially defined. The higher bound can be omitted. In such
757 case, it is replaced by the corresponding maximum value, 32 or 64 depending
758 on the machine's word size.
759
Christopher Faulet26028f62017-11-22 15:01:51 +0100760 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100761 automatically bind a process or a thread to a CPU by incrementing
762 process/thread and CPU sets. To be valid, both sets must have the same
763 size. No matter the declaration order of the CPU sets, it will be bound from
764 the lowest to the highest bound. Having a process and a thread range with the
765 "auto:" prefix is not supported. Only one range is supported, the other one
766 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100767
768 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100769 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
770
771 cpu-map 1/all 0-3 # bind all threads of the first process on the
772 # first 4 CPUs
773
774 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
775 # or "cpu-map 1-32 0-31" depending on the machine's
776 # word size.
777
Christopher Faulet26028f62017-11-22 15:01:51 +0100778 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100779 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100780 cpu-map auto:1-4 0-3
781 cpu-map auto:1-4 0-1 2-3
782 cpu-map auto:1-4 3 2 1 0
783
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100784 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
785 # and so on.
786 cpu-map auto:1/1-4 0-3
787 cpu-map auto:1/1-4 0-1 2-3
788 cpu-map auto:1/1-4 3 2 1 0
789
Davor Ocelice9ed2812017-12-25 17:49:28 +0100790 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100791 cpu-map auto:all 0-63
792 cpu-map auto:even 0-31
793 cpu-map auto:odd 32-63
794
795 # invalid cpu-map because process and CPU sets have different sizes.
796 cpu-map auto:1-4 0 # invalid
797 cpu-map auto:1 0-3 # invalid
798
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100799 # invalid cpu-map because automatic binding is used with a process range
800 # and a thread range.
801 cpu-map auto:all/all 0 # invalid
802 cpu-map auto:all/1-4 0 # invalid
803 cpu-map auto:1-4/all 0 # invalid
804
Emeric Brunc8e8d122012-10-02 18:42:10 +0200805crt-base <dir>
806 Assigns a default directory to fetch SSL certificates from when a relative
William Dauchy238ea3b2020-01-11 13:09:12 +0100807 path is used with "crtfile" or "crt" directives. Absolute locations specified
808 prevail and ignore "crt-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200809
Willy Tarreau6a06a402007-07-15 20:15:28 +0200810daemon
811 Makes the process fork into background. This is the recommended mode of
812 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100813 disabled by the command line "-db" argument. This option is ignored in
814 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200815
David Carlier8167f302015-06-01 13:50:06 +0200816deviceatlas-json-file <path>
817 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100818 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200819
820deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100821 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200822 optional and set to 0 by default if not set.
823
824deviceatlas-separator <char>
825 Sets the character separator for the API properties results. This directive
826 is optional and set to | by default if not set.
827
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100828deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200829 Sets the client cookie's name used for the detection if the DeviceAtlas
830 Client-side component was used during the request. This directive is optional
831 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100832
Simon Horman98637e52014-06-20 12:30:16 +0900833external-check
Willy Tarreaud96f1122019-12-03 07:07:36 +0100834 Allows the use of an external agent to perform health checks. This is
835 disabled by default as a security precaution, and even when enabled, checks
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100836 may still fail unless "insecure-fork-wanted" is enabled as well. If the
837 program launched makes use of a setuid executable (it should really not),
838 you may also need to set "insecure-setuid-wanted" in the global section.
839 See "option external-check", and "insecure-fork-wanted", and
840 "insecure-setuid-wanted".
Simon Horman98637e52014-06-20 12:30:16 +0900841
Willy Tarreau6a06a402007-07-15 20:15:28 +0200842gid <number>
843 Changes the process' group ID to <number>. It is recommended that the group
844 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
845 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100846 Note that if haproxy is started from a user having supplementary groups, it
847 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200848 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100849
Willy Tarreau11770ce2019-12-03 08:29:22 +0100850group <group name>
851 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
852 See also "gid" and "user".
853
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100854hard-stop-after <time>
855 Defines the maximum time allowed to perform a clean soft-stop.
856
857 Arguments :
858 <time> is the maximum time (by default in milliseconds) for which the
859 instance will remain alive when a soft-stop is received via the
860 SIGUSR1 signal.
861
862 This may be used to ensure that the instance will quit even if connections
863 remain opened during a soft-stop (for example with long timeouts for a proxy
864 in tcp mode). It applies both in TCP and HTTP mode.
865
866 Example:
867 global
868 hard-stop-after 30s
869
Christopher Faulet98fbe952019-07-22 16:18:24 +0200870h1-case-adjust <from> <to>
871 Defines the case adjustment to apply, when enabled, to the header name
872 <from>, to change it to <to> before sending it to HTTP/1 clients or
873 servers. <from> must be in lower case, and <from> and <to> must not differ
874 except for their case. It may be repeated if several header names need to be
875 ajusted. Duplicate entries are not allowed. If a lot of header names have to
876 be adjusted, it might be more convenient to use "h1-case-adjust-file".
877 Please note that no transformation will be applied unless "option
878 h1-case-adjust-bogus-client" or "option h1-case-adjust-bogus-server" is
879 specified in a proxy.
880
881 There is no standard case for header names because, as stated in RFC7230,
882 they are case-insensitive. So applications must handle them in a case-
883 insensitive manner. But some bogus applications violate the standards and
884 erroneously rely on the cases most commonly used by browsers. This problem
885 becomes critical with HTTP/2 because all header names must be exchanged in
886 lower case, and HAProxy follows the same convention. All header names are
887 sent in lower case to clients and servers, regardless of the HTTP version.
888
889 Applications which fail to properly process requests or responses may require
890 to temporarily use such workarounds to adjust header names sent to them for
891 the time it takes the application to be fixed. Please note that an
892 application which requires such workarounds might be vulnerable to content
893 smuggling attacks and must absolutely be fixed.
894
895 Example:
896 global
897 h1-case-adjust content-length Content-Length
898
899 See "h1-case-adjust-file", "option h1-case-adjust-bogus-client" and
900 "option h1-case-adjust-bogus-server".
901
902h1-case-adjust-file <hdrs-file>
903 Defines a file containing a list of key/value pairs used to adjust the case
904 of some header names before sending them to HTTP/1 clients or servers. The
905 file <hdrs-file> must contain 2 header names per line. The first one must be
906 in lower case and both must not differ except for their case. Lines which
907 start with '#' are ignored, just like empty lines. Leading and trailing tabs
908 and spaces are stripped. Duplicate entries are not allowed. Please note that
909 no transformation will be applied unless "option h1-case-adjust-bogus-client"
910 or "option h1-case-adjust-bogus-server" is specified in a proxy.
911
912 If this directive is repeated, only the last one will be processed. It is an
913 alternative to the directive "h1-case-adjust" if a lot of header names need
914 to be adjusted. Please read the risks associated with using this.
915
916 See "h1-case-adjust", "option h1-case-adjust-bogus-client" and
917 "option h1-case-adjust-bogus-server".
918
Willy Tarreaud96f1122019-12-03 07:07:36 +0100919insecure-fork-wanted
920 By default haproxy tries hard to prevent any thread and process creation
921 after it starts. Doing so is particularly important when using Lua files of
922 uncertain origin, and when experimenting with development versions which may
923 still contain bugs whose exploitability is uncertain. And generally speaking
924 it's good hygiene to make sure that no unexpected background activity can be
925 triggered by traffic. But this prevents external checks from working, and may
926 break some very specific Lua scripts which actively rely on the ability to
927 fork. This option is there to disable this protection. Note that it is a bad
928 idea to disable it, as a vulnerability in a library or within haproxy itself
929 will be easier to exploit once disabled. In addition, forking from Lua or
930 anywhere else is not reliable as the forked process may randomly embed a lock
931 set by another thread and never manage to finish an operation. As such it is
932 highly recommended that this option is never used and that any workload
933 requiring such a fork be reconsidered and moved to a safer solution (such as
934 agents instead of external checks). This option supports the "no" prefix to
935 disable it.
936
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100937insecure-setuid-wanted
938 HAProxy doesn't need to call executables at run time (except when using
939 external checks which are strongly recommended against), and is even expected
940 to isolate itself into an empty chroot. As such, there basically is no valid
941 reason to allow a setuid executable to be called without the user being fully
942 aware of the risks. In a situation where haproxy would need to call external
943 checks and/or disable chroot, exploiting a vulnerability in a library or in
944 haproxy itself could lead to the execution of an external program. On Linux
945 it is possible to lock the process so that any setuid bit present on such an
946 executable is ignored. This significantly reduces the risk of privilege
947 escalation in such a situation. This is what haproxy does by default. In case
948 this causes a problem to an external check (for example one which would need
949 the "ping" command), then it is possible to disable this protection by
950 explicitly adding this directive in the global section. If enabled, it is
951 possible to turn it back off by prefixing it with the "no" keyword.
952
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +0100953issuers-chain-path <dir>
954 Assigns a directory to load certificate chain for issuer completion. All
955 files must be in PEM format. For certificates loaded with "crt" or "crt-list",
956 if certificate chain is not included in PEM (also commonly known as
957 intermediate certificate), haproxy will complete chain if the issuer of the
958 certificate corresponds to the first certificate of the chain loaded with
959 "issuers-chain-path".
960 A "crt" file with PrivateKey+Certificate+IntermediateCA2+IntermediateCA1
961 could be replaced with PrivateKey+Certificate. HAProxy will complete the
962 chain if a file with IntermediateCA2+IntermediateCA1 is present in
963 "issuers-chain-path" directory. All other certificates with the same issuer
964 will share the chain in memory.
965
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200966log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
967 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100968 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100969 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100970 configured with "log global".
971
972 <address> can be one of:
973
Willy Tarreau2769aa02007-12-27 18:26:09 +0100974 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100975 no port is specified, 514 is used by default (the standard syslog
976 port).
977
David du Colombier24bb5f52011-03-17 10:40:23 +0100978 - An IPv6 address followed by a colon and optionally a UDP port. If
979 no port is specified, 514 is used by default (the standard syslog
980 port).
981
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100982 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +0100983 considerations for chroot (be sure the path is accessible inside
984 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100985 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100986
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100987 - A file descriptor number in the form "fd@<number>", which may point
988 to a pipe, terminal, or socket. In this case unbuffered logs are used
989 and one writev() call per log is performed. This is a bit expensive
990 but acceptable for most workloads. Messages sent this way will not be
991 truncated but may be dropped, in which case the DroppedLogs counter
992 will be incremented. The writev() call is atomic even on pipes for
993 messages up to PIPE_BUF size, which POSIX recommends to be at least
994 512 and which is 4096 bytes on most modern operating systems. Any
995 larger message may be interleaved with messages from other processes.
996 Exceptionally for debugging purposes the file descriptor may also be
997 directed to a file, but doing so will significantly slow haproxy down
998 as non-blocking calls will be ignored. Also there will be no way to
999 purge nor rotate this file without restarting the process. Note that
1000 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +01001001 for use with a TCP syslog server. See also the "short" and "raw"
1002 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01001003
1004 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
1005 "fd@2", see above.
1006
Willy Tarreauc046d162019-08-30 15:24:59 +02001007 - A ring buffer in the form "ring@<name>", which will correspond to an
1008 in-memory ring buffer accessible over the CLI using the "show events"
1009 command, which will also list existing rings and their sizes. Such
1010 buffers are lost on reload or restart but when used as a complement
1011 this can help troubleshooting by having the logs instantly available.
1012
William Lallemandb2f07452015-05-12 14:27:13 +02001013 You may want to reference some environment variables in the address
1014 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001015
Willy Tarreau18324f52014-06-27 18:10:07 +02001016 <length> is an optional maximum line length. Log lines larger than this value
1017 will be truncated before being sent. The reason is that syslog
1018 servers act differently on log line length. All servers support the
1019 default value of 1024, but some servers simply drop larger lines
1020 while others do log them. If a server supports long lines, it may
1021 make sense to set this value here in order to avoid truncating long
1022 lines. Similarly, if a server drops long lines, it is preferable to
1023 truncate them before sending them. Accepted values are 80 to 65535
1024 inclusive. The default value of 1024 is generally fine for all
1025 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +01001026 JSON-formatted logs may require larger values. You may also need to
1027 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +02001028
Dragan Dosen7ad31542015-09-28 17:16:47 +02001029 <format> is the log format used when generating syslog messages. It may be
1030 one of the following :
1031
1032 rfc3164 The RFC3164 syslog message format. This is the default.
1033 (https://tools.ietf.org/html/rfc3164)
1034
1035 rfc5424 The RFC5424 syslog message format.
1036 (https://tools.ietf.org/html/rfc5424)
1037
Willy Tarreaue8746a02018-11-12 08:45:00 +01001038 short A message containing only a level between angle brackets such as
1039 '<3>', followed by the text. The PID, date, time, process name
1040 and system name are omitted. This is designed to be used with a
1041 local log server. This format is compatible with what the systemd
1042 logger consumes.
1043
Willy Tarreauc1b06452018-11-12 11:57:56 +01001044 raw A message containing only the text. The level, PID, date, time,
1045 process name and system name are omitted. This is designed to be
1046 used in containers or during development, where the severity only
1047 depends on the file descriptor used (stdout/stderr).
1048
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02001049 <ranges> A list of comma-separated ranges to identify the logs to sample.
1050 This is used to balance the load of the logs to send to the log
1051 server. The limits of the ranges cannot be null. They are numbered
1052 from 1. The size or period (in number of logs) of the sample must be
1053 set with <sample_size> parameter.
1054
1055 <sample_size>
1056 The size of the sample in number of logs to consider when balancing
1057 their logging loads. It is used to balance the load of the logs to
1058 send to the syslog server. This size must be greater or equal to the
1059 maximum of the high limits of the ranges.
1060 (see also <ranges> parameter).
1061
Robert Tsai81ae1952007-12-05 10:47:29 +01001062 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001063
Willy Tarreaue8746a02018-11-12 08:45:00 +01001064 kern user mail daemon auth syslog lpr news
1065 uucp cron auth2 ftp ntp audit alert cron2
1066 local0 local1 local2 local3 local4 local5 local6 local7
1067
Willy Tarreauc1b06452018-11-12 11:57:56 +01001068 Note that the facility is ignored for the "short" and "raw"
1069 formats, but still required as a positional field. It is
1070 recommended to use "daemon" in this case to make it clear that
1071 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001072
1073 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +02001074 all messages are sent. If a maximum level is specified, only messages with a
1075 severity at least as important as this level will be sent. An optional minimum
1076 level can be specified. If it is set, logs emitted with a more severe level
1077 than this one will be capped to this level. This is used to avoid sending
1078 "emerg" messages on all terminals on some default syslog configurations.
1079 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001080
Cyril Bontédc4d9032012-04-08 21:57:39 +02001081 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +02001082
Joe Williamsdf5b38f2010-12-29 17:05:48 +01001083log-send-hostname [<string>]
1084 Sets the hostname field in the syslog header. If optional "string" parameter
1085 is set the header is set to the string contents, otherwise uses the hostname
1086 of the system. Generally used if one is not relaying logs through an
1087 intermediate syslog server or for simply customizing the hostname printed in
1088 the logs.
1089
Kevinm48936af2010-12-22 16:08:21 +00001090log-tag <string>
1091 Sets the tag field in the syslog header to this string. It defaults to the
1092 program name as launched from the command line, which usually is "haproxy".
1093 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +01001094 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +00001095
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001096lua-load <file>
1097 This global directive loads and executes a Lua file. This directive can be
1098 used multiple times.
1099
Tim Duesterhusdd74b5f2020-01-12 13:55:40 +01001100lua-prepend-path <string> [<type>]
1101 Prepends the given string followed by a semicolon to Lua's package.<type>
1102 variable.
1103 <type> must either be "path" or "cpath". If <type> is not given it defaults
1104 to "path".
1105
1106 Lua's paths are semicolon delimited lists of patterns that specify how the
1107 `require` function attempts to find the source file of a library. Question
1108 marks (?) within a pattern will be replaced by module name. The path is
1109 evaluated left to right. This implies that paths that are prepended later
1110 will be checked earlier.
1111
1112 As an example by specifying the following path:
1113
1114 lua-prepend-path /usr/share/haproxy-lua/?/init.lua
1115 lua-prepend-path /usr/share/haproxy-lua/?.lua
1116
1117 When `require "example"` is being called Lua will first attempt to load the
1118 /usr/share/haproxy-lua/example.lua script, if that does not exist the
1119 /usr/share/haproxy-lua/example/init.lua will be attempted and the default
1120 paths if that does not exist either.
1121
1122 See https://www.lua.org/pil/8.1.html for the details within the Lua
1123 documentation.
1124
William Lallemand4cfede82017-11-24 22:02:34 +01001125master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +02001126 Master-worker mode. It is equivalent to the command line "-W" argument.
1127 This mode will launch a "master" which will monitor the "workers". Using
1128 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001129 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +02001130 or daemon mode. It is recommended to use this mode with multiprocess and
1131 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +01001132 By default, if a worker exits with a bad return code, in the case of a
1133 segfault for example, all workers will be killed, and the master will leave.
1134 It is convenient to combine this behavior with Restart=on-failure in a
1135 systemd unit file in order to relaunch the whole process. If you don't want
1136 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +02001137
William Lallemand4cfede82017-11-24 22:02:34 +01001138 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +02001139
William Lallemand27edc4b2019-05-07 17:49:33 +02001140mworker-max-reloads <number>
1141 In master-worker mode, this option limits the number of time a worker can
John Roeslerfb2fce12019-07-10 15:45:51 -05001142 survive to a reload. If the worker did not leave after a reload, once its
William Lallemand27edc4b2019-05-07 17:49:33 +02001143 number of reloads is greater than this number, the worker will receive a
1144 SIGTERM. This option helps to keep under control the number of workers.
1145 See also "show proc" in the Management Guide.
1146
Willy Tarreau6a06a402007-07-15 20:15:28 +02001147nbproc <number>
1148 Creates <number> processes when going daemon. This requires the "daemon"
1149 mode. By default, only one process is created, which is the recommended mode
1150 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +01001151 process, it may be needed to fork multiple daemons. When set to a value
1152 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreau1f672a82019-01-26 14:20:55 +01001153 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon" and
1154 "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001155
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001156nbthread <number>
1157 This setting is only available when support for threads was built in. It
Willy Tarreau26f6ae12019-02-02 12:56:15 +01001158 makes haproxy run on <number> threads. This is exclusive with "nbproc". While
1159 "nbproc" historically used to be the only way to use multiple processors, it
1160 also involved a number of shortcomings related to the lack of synchronization
1161 between processes (health-checks, peers, stick-tables, stats, ...) which do
1162 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +01001163 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
1164 HAProxy is started in foreground. On some platforms supporting CPU affinity,
1165 when nbproc is not used, the default "nbthread" value is automatically set to
1166 the number of CPUs the process is bound to upon startup. This means that the
1167 thread count can easily be adjusted from the calling process using commands
1168 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1169 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001170
Willy Tarreau6a06a402007-07-15 20:15:28 +02001171pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001172 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +02001173 the "-p" command line argument. The file must be accessible to the user
1174 starting the process. See also "daemon".
1175
Willy Tarreau1d549722016-02-16 12:41:57 +01001176presetenv <name> <value>
1177 Sets environment variable <name> to value <value>. If the variable exists, it
1178 is NOT overwritten. The changes immediately take effect so that the next line
1179 in the configuration file sees the new value. See also "setenv", "resetenv",
1180 and "unsetenv".
1181
1182resetenv [<name> ...]
1183 Removes all environment variables except the ones specified in argument. It
1184 allows to use a clean controlled environment before setting new values with
1185 setenv or unsetenv. Please note that some internal functions may make use of
1186 some environment variables, such as time manipulation functions, but also
1187 OpenSSL or even external checks. This must be used with extreme care and only
1188 after complete validation. The changes immediately take effect so that the
1189 next line in the configuration file sees the new environment. See also
1190 "setenv", "presetenv", and "unsetenv".
1191
Christopher Fauletff4121f2017-11-22 16:38:49 +01001192stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001193 Limits the stats socket to a certain set of processes numbers. By default the
1194 stats socket is bound to all processes, causing a warning to be emitted when
1195 nbproc is greater than 1 because there is no way to select the target process
1196 when connecting. However, by using this setting, it becomes possible to pin
1197 the stats socket to a specific set of processes, typically the first one. The
1198 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001199 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001200 word size (32 or 64). Ranges can be partially defined. The higher bound can
1201 be omitted. In such case, it is replaced by the corresponding maximum
1202 value. A better option consists in using the "process" setting of the "stats
1203 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001204
Baptiste Assmann5626f482015-08-23 10:00:10 +02001205server-state-base <directory>
1206 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001207 file names which do not start with a '/'. See also "server-state-file",
1208 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001209
1210server-state-file <file>
1211 Specifies the path to the file containing state of servers. If the path starts
1212 with a slash ('/'), it is considered absolute, otherwise it is considered
1213 relative to the directory specified using "server-state-base" (if set) or to
1214 the current directory. Before reloading HAProxy, it is possible to save the
1215 servers' current state using the stats command "show servers state". The
1216 output of this command must be written in the file pointed by <file>. When
1217 starting up, before handling traffic, HAProxy will read, load and apply state
1218 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001219 configuration. See also "server-state-base" and "show servers state",
1220 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001221
Willy Tarreau1d549722016-02-16 12:41:57 +01001222setenv <name> <value>
1223 Sets environment variable <name> to value <value>. If the variable exists, it
1224 is overwritten. The changes immediately take effect so that the next line in
1225 the configuration file sees the new value. See also "presetenv", "resetenv",
1226 and "unsetenv".
1227
Willy Tarreau636848a2019-04-15 19:38:50 +02001228set-dumpable
1229 This option is better left disabled by default and enabled only upon a
William Dauchyec730982019-10-27 20:08:10 +01001230 developer's request. If it has been enabled, it may still be forcibly
1231 disabled by prefixing it with the "no" keyword. It has no impact on
1232 performance nor stability but will try hard to re-enable core dumps that were
1233 possibly disabled by file size limitations (ulimit -f), core size limitations
1234 (ulimit -c), or "dumpability" of a process after changing its UID/GID (such
1235 as /proc/sys/fs/suid_dumpable on Linux). Core dumps might still be limited by
1236 the current directory's permissions (check what directory the file is started
1237 from), the chroot directory's permission (it may be needed to temporarily
1238 disable the chroot directive or to move it to a dedicated writable location),
1239 or any other system-specific constraint. For example, some Linux flavours are
1240 notorious for replacing the default core file with a path to an executable
1241 not even installed on the system (check /proc/sys/kernel/core_pattern). Often,
1242 simply writing "core", "core.%p" or "/var/log/core/core.%p" addresses the
1243 issue. When trying to enable this option waiting for a rare issue to
1244 re-appear, it's often a good idea to first try to obtain such a dump by
1245 issuing, for example, "kill -11" to the haproxy process and verify that it
1246 leaves a core where expected when dying.
Willy Tarreau636848a2019-04-15 19:38:50 +02001247
Willy Tarreau610f04b2014-02-13 11:36:41 +01001248ssl-default-bind-ciphers <ciphers>
1249 This setting is only available when support for OpenSSL was built in. It sets
1250 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001251 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001252 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001253 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1254 information and recommendations see e.g.
1255 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1256 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1257 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1258 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001259
1260ssl-default-bind-ciphersuites <ciphersuites>
1261 This setting is only available when support for OpenSSL was built in and
1262 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1263 describing the list of cipher algorithms ("cipher suite") that are negotiated
1264 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1265 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001266 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1267 cipher configuration for TLSv1.2 and earlier, please check the
1268 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001269 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001270
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001271ssl-default-bind-options [<option>]...
1272 This setting is only available when support for OpenSSL was built in. It sets
1273 default ssl-options to force on all "bind" lines. Please check the "bind"
1274 keyword to see available options.
1275
1276 Example:
1277 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001278 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001279
Willy Tarreau610f04b2014-02-13 11:36:41 +01001280ssl-default-server-ciphers <ciphers>
1281 This setting is only available when support for OpenSSL was built in. It
1282 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001283 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001284 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001285 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1286 information and recommendations see e.g.
1287 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1288 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1289 For TLSv1.3 cipher configuration, please check the
1290 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1291 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001292
1293ssl-default-server-ciphersuites <ciphersuites>
1294 This setting is only available when support for OpenSSL was built in and
1295 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1296 string describing the list of cipher algorithms that are negotiated during
1297 the TLSv1.3 handshake with the server, for all "server" lines which do not
1298 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001299 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1300 cipher configuration for TLSv1.2 and earlier, please check the
1301 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1302 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001303
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001304ssl-default-server-options [<option>]...
1305 This setting is only available when support for OpenSSL was built in. It sets
1306 default ssl-options to force on all "server" lines. Please check the "server"
1307 keyword to see available options.
1308
Remi Gacogne47783ef2015-05-29 15:53:22 +02001309ssl-dh-param-file <file>
1310 This setting is only available when support for OpenSSL was built in. It sets
1311 the default DH parameters that are used during the SSL/TLS handshake when
1312 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001313 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001314 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001315 are not specified either by using ssl-dh-param-file or by setting them
1316 directly in the certificate file, pre-generated DH parameters of the size
1317 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1318 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001319 Custom DH parameters may be generated by using the OpenSSL command
1320 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1321 parameters should not be considered secure anymore.
1322
William Lallemand4c5adbf2020-02-24 14:23:22 +01001323ssl-load-extra-files <none|all|bundle|sctl|ocsp|issuer|key>*
William Lallemand3af48e72020-02-03 17:15:52 +01001324 This setting alters the way HAProxy will look for unspecified files during
1325 the loading of the SSL certificates.
1326
1327 By default, HAProxy discovers automatically a lot of files not specified in
1328 the configuration, and you may want to disable this behavior if you want to
1329 optimize the startup time.
1330
1331 "none": Only load the files specified in the configuration. Don't try to load
1332 a certificate bundle if the file does not exist. In the case of a directory,
1333 it won't try to bundle the certificates if they have the same basename.
1334
1335 "all": This is the default behavior, it will try to load everything,
William Lallemand4c5adbf2020-02-24 14:23:22 +01001336 bundles, sctl, ocsp, issuer, key.
William Lallemand3af48e72020-02-03 17:15:52 +01001337
1338 "bundle": When a file specified in the configuration does not exist, HAProxy
1339 will try to load a certificate bundle. This is done by looking for
1340 <basename>.rsa, .ecdsa and .dsa. In the case of directories, HAProxy will
1341 try to gather the files with the same basename in a multi-certificate bundle.
1342 The bundles were introduced with OpenSSL 1.0.2 and were the only way back
1343 then to load an ECDSA certificate and a RSA one, with the same SNI. Since
1344 OpenSSL 1.1.1 it is not recommended anymore, you can specifiy both the ECDSA
1345 and the RSA file on the bind line.
1346
1347 "sctl": Try to load "<basename>.sctl" for each crt keyword.
1348
1349 "ocsp": Try to load "<basename>.ocsp" for each crt keyword.
1350
1351 "issuer": Try to load "<basename>.issuer" if the issuer of the OCSP file is
1352 not provided in the PEM file.
1353
William Lallemand4c5adbf2020-02-24 14:23:22 +01001354 "key": If the private key was not provided by the PEM file, try to load a
1355 file "<basename>.key" containing a private key.
1356
William Lallemand3af48e72020-02-03 17:15:52 +01001357 The default behavior is "all".
1358
1359 Example:
1360 ssl-load-extra-files bundle sctl
1361 ssl-load-extra-files sctl ocsp issuer
1362 ssl-load-extra-files none
1363
1364 See also: "crt", section 5.1 about bind options.
1365
Emeric Brun850efd52014-01-29 12:24:34 +01001366ssl-server-verify [none|required]
1367 The default behavior for SSL verify on servers side. If specified to 'none',
1368 servers certificates are not verified. The default is 'required' except if
1369 forced using cmdline option '-dV'.
1370
Willy Tarreauabb175f2012-09-24 12:43:26 +02001371stats socket [<address:port>|<path>] [param*]
1372 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1373 Connections to this socket will return various statistics outputs and even
1374 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001375 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001376 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001377
Willy Tarreauabb175f2012-09-24 12:43:26 +02001378 All parameters supported by "bind" lines are supported, for instance to
1379 restrict access to some users or their access rights. Please consult
1380 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001381
1382stats timeout <timeout, in milliseconds>
1383 The default timeout on the stats socket is set to 10 seconds. It is possible
1384 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001385 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001386
1387stats maxconn <connections>
1388 By default, the stats socket is limited to 10 concurrent connections. It is
1389 possible to change this value with "stats maxconn".
1390
Willy Tarreau6a06a402007-07-15 20:15:28 +02001391uid <number>
1392 Changes the process' user ID to <number>. It is recommended that the user ID
1393 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1394 be started with superuser privileges in order to be able to switch to another
1395 one. See also "gid" and "user".
1396
1397ulimit-n <number>
1398 Sets the maximum number of per-process file-descriptors to <number>. By
1399 default, it is automatically computed, so it is recommended not to use this
1400 option.
1401
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001402unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1403 [ group <group> ] [ gid <gid> ]
1404
1405 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1406 This is mainly used to simplify declaration of those UNIX sockets and reduce
1407 the risk of errors, since those settings are most commonly required but are
1408 also process-specific. The <prefix> setting can be used to force all socket
1409 path to be relative to that directory. This might be needed to access another
1410 component's chroot. Note that those paths are resolved before haproxy chroots
1411 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1412 all have the same meaning as their homonyms used by the "bind" statement. If
1413 both are specified, the "bind" statement has priority, meaning that the
1414 "unix-bind" settings may be seen as process-wide default settings.
1415
Willy Tarreau1d549722016-02-16 12:41:57 +01001416unsetenv [<name> ...]
1417 Removes environment variables specified in arguments. This can be useful to
1418 hide some sensitive information that are occasionally inherited from the
1419 user's environment during some operations. Variables which did not exist are
1420 silently ignored so that after the operation, it is certain that none of
1421 these variables remain. The changes immediately take effect so that the next
1422 line in the configuration file will not see these variables. See also
1423 "setenv", "presetenv", and "resetenv".
1424
Willy Tarreau6a06a402007-07-15 20:15:28 +02001425user <user name>
1426 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1427 See also "uid" and "group".
1428
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001429node <name>
1430 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1431
1432 This statement is useful in HA configurations where two or more processes or
1433 servers share the same IP address. By setting a different node-name on all
1434 nodes, it becomes easy to immediately spot what server is handling the
1435 traffic.
1436
1437description <text>
1438 Add a text that describes the instance.
1439
1440 Please note that it is required to escape certain characters (# for example)
1441 and this text is inserted into a html page so you should avoid using
1442 "<" and ">" characters.
1443
Thomas Holmesdb04f192015-05-18 13:21:39 +0100144451degrees-data-file <file path>
1445 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001446 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001447
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001448 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001449 compiled with USE_51DEGREES.
1450
Ben Shillitof25e8e52016-12-02 14:25:37 +0000145151degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001452 A list of 51Degrees property names to be load from the dataset. A full list
1453 of names is available on the 51Degrees website:
1454 https://51degrees.com/resources/property-dictionary
1455
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001456 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001457 compiled with USE_51DEGREES.
1458
Dragan Dosen93b38d92015-06-29 16:43:25 +0200145951degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001460 A char that will be appended to every property value in a response header
1461 containing 51Degrees results. If not set that will be set as ','.
1462
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001463 Please note that this option is only available when haproxy has been
1464 compiled with USE_51DEGREES.
1465
146651degrees-cache-size <number>
1467 Sets the size of the 51Degrees converter cache to <number> entries. This
1468 is an LRU cache which reminds previous device detections and their results.
1469 By default, this cache is disabled.
1470
1471 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001472 compiled with USE_51DEGREES.
1473
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001474wurfl-data-file <file path>
1475 The path of the WURFL data file to provide device detection services. The
1476 file should be accessible by HAProxy with relevant permissions.
1477
1478 Please note that this option is only available when haproxy has been compiled
1479 with USE_WURFL=1.
1480
1481wurfl-information-list [<capability>]*
1482 A space-delimited list of WURFL capabilities, virtual capabilities, property
1483 names we plan to use in injected headers. A full list of capability and
1484 virtual capability names is available on the Scientiamobile website :
1485
1486 https://www.scientiamobile.com/wurflCapability
1487
1488 Valid WURFL properties are:
1489 - wurfl_id Contains the device ID of the matched device.
1490
1491 - wurfl_root_id Contains the device root ID of the matched
1492 device.
1493
1494 - wurfl_isdevroot Tells if the matched device is a root device.
1495 Possible values are "TRUE" or "FALSE".
1496
1497 - wurfl_useragent The original useragent coming with this
1498 particular web request.
1499
1500 - wurfl_api_version Contains a string representing the currently
1501 used Libwurfl API version.
1502
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001503 - wurfl_info A string containing information on the parsed
1504 wurfl.xml and its full path.
1505
1506 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1507 WURFL has been loaded successfully.
1508
1509 - wurfl_normalized_useragent The normalized useragent.
1510
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001511 Please note that this option is only available when haproxy has been compiled
1512 with USE_WURFL=1.
1513
1514wurfl-information-list-separator <char>
1515 A char that will be used to separate values in a response header containing
1516 WURFL results. If not set that a comma (',') will be used by default.
1517
1518 Please note that this option is only available when haproxy has been compiled
1519 with USE_WURFL=1.
1520
1521wurfl-patch-file [<file path>]
1522 A list of WURFL patch file paths. Note that patches are loaded during startup
1523 thus before the chroot.
1524
1525 Please note that this option is only available when haproxy has been compiled
1526 with USE_WURFL=1.
1527
paulborilebad132c2019-04-18 11:57:04 +02001528wurfl-cache-size <size>
1529 Sets the WURFL Useragent cache size. For faster lookups, already processed user
1530 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001531 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02001532 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001533
1534 Please note that this option is only available when haproxy has been compiled
1535 with USE_WURFL=1.
1536
William Dauchy0fec3ab2019-10-27 20:08:11 +01001537strict-limits
1538 Makes process fail at startup when a setrlimit fails. Haproxy is tries to set
1539 the best setrlimit according to what has been calculated. If it fails, it
1540 will emit a warning. Use this option if you want an explicit failure of
1541 haproxy when those limits fail. This option is disabled by default. If it has
1542 been enabled, it may still be forcibly disabled by prefixing it with the "no"
1543 keyword.
1544
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015453.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001546-----------------------
1547
Willy Tarreaubeb859a2018-11-22 18:07:59 +01001548busy-polling
1549 In some situations, especially when dealing with low latency on processors
1550 supporting a variable frequency or when running inside virtual machines, each
1551 time the process waits for an I/O using the poller, the processor goes back
1552 to sleep or is offered to another VM for a long time, and it causes
1553 excessively high latencies. This option provides a solution preventing the
1554 processor from sleeping by always using a null timeout on the pollers. This
1555 results in a significant latency reduction (30 to 100 microseconds observed)
1556 at the expense of a risk to overheat the processor. It may even be used with
1557 threads, in which case improperly bound threads may heavily conflict,
1558 resulting in a worse performance and high values for the CPU stolen fields
1559 in "show info" output, indicating which threads are misconfigured. It is
1560 important not to let the process run on the same processor as the network
1561 interrupts when this option is used. It is also better to avoid using it on
1562 multiple CPU threads sharing the same core. This option is disabled by
1563 default. If it has been enabled, it may still be forcibly disabled by
1564 prefixing it with the "no" keyword. It is ignored by the "select" and
1565 "poll" pollers.
1566
William Dauchy3894d972019-12-28 15:36:02 +01001567 This option is automatically disabled on old processes in the context of
1568 seamless reload; it avoids too much cpu conflicts when multiple processes
1569 stay around for some time waiting for the end of their current connections.
1570
Willy Tarreau1746eec2014-04-25 10:46:47 +02001571max-spread-checks <delay in milliseconds>
1572 By default, haproxy tries to spread the start of health checks across the
1573 smallest health check interval of all the servers in a farm. The principle is
1574 to avoid hammering services running on the same server. But when using large
1575 check intervals (10 seconds or more), the last servers in the farm take some
1576 time before starting to be tested, which can be a problem. This parameter is
1577 used to enforce an upper bound on delay between the first and the last check,
1578 even if the servers' check intervals are larger. When servers run with
1579 shorter intervals, their intervals will be respected though.
1580
Willy Tarreau6a06a402007-07-15 20:15:28 +02001581maxconn <number>
1582 Sets the maximum per-process number of concurrent connections to <number>. It
1583 is equivalent to the command-line argument "-n". Proxies will stop accepting
1584 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001585 automatically adjusted according to this value. See also "ulimit-n". Note:
1586 the "select" poller cannot reliably use more than 1024 file descriptors on
1587 some platforms. If your platform only supports select and reports "select
1588 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01001589 below 500 in general). If this value is not set, it will automatically be
1590 calculated based on the current file descriptors limit reported by the
1591 "ulimit -n" command, possibly reduced to a lower value if a memory limit
1592 is enforced, based on the buffer size, memory allocated to compression, SSL
1593 cache size, and use or not of SSL and the associated maxsslconn (which can
1594 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001595
Willy Tarreau81c25d02011-09-07 15:17:21 +02001596maxconnrate <number>
1597 Sets the maximum per-process number of connections per second to <number>.
1598 Proxies will stop accepting connections when this limit is reached. It can be
1599 used to limit the global capacity regardless of each frontend capacity. It is
1600 important to note that this can only be used as a service protection measure,
1601 as there will not necessarily be a fair share between frontends when the
1602 limit is reached, so it's a good idea to also limit each frontend to some
1603 value close to its expected share. Also, lowering tune.maxaccept can improve
1604 fairness.
1605
William Lallemandd85f9172012-11-09 17:05:39 +01001606maxcomprate <number>
1607 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001608 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001609 level will be decreased during the session. If the maximum is reached at the
1610 beginning of a session, the session will not compress at all. If the maximum
1611 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001612 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001613 default value.
1614
William Lallemand072a2bf2012-11-20 17:01:01 +01001615maxcompcpuusage <number>
1616 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1617 for new requests or decreasing the compression level of current requests.
1618 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1619 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1620 case of multiple processes (nbproc > 1), each process manages its individual
1621 usage. A value of 100 disable the limit. The default value is 100. Setting
1622 a lower value will prevent the compression work from slowing the whole
1623 process down and from introducing high latencies.
1624
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001625maxpipes <number>
1626 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1627 are only used by kernel-based tcp splicing. Since a pipe contains two file
1628 descriptors, the "ulimit-n" value will be increased accordingly. The default
1629 value is maxconn/4, which seems to be more than enough for most heavy usages.
1630 The splice code dynamically allocates and releases pipes, and can fall back
1631 to standard copy, so setting this value too low may only impact performance.
1632
Willy Tarreau93e7c002013-10-07 18:51:07 +02001633maxsessrate <number>
1634 Sets the maximum per-process number of sessions per second to <number>.
1635 Proxies will stop accepting connections when this limit is reached. It can be
1636 used to limit the global capacity regardless of each frontend capacity. It is
1637 important to note that this can only be used as a service protection measure,
1638 as there will not necessarily be a fair share between frontends when the
1639 limit is reached, so it's a good idea to also limit each frontend to some
1640 value close to its expected share. Also, lowering tune.maxaccept can improve
1641 fairness.
1642
Willy Tarreau403edff2012-09-06 11:58:37 +02001643maxsslconn <number>
1644 Sets the maximum per-process number of concurrent SSL connections to
1645 <number>. By default there is no SSL-specific limit, which means that the
1646 global maxconn setting will apply to all connections. Setting this limit
1647 avoids having openssl use too much memory and crash when malloc returns NULL
1648 (since it unfortunately does not reliably check for such conditions). Note
1649 that the limit applies both to incoming and outgoing connections, so one
1650 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001651 If this value is not set, but a memory limit is enforced, this value will be
1652 automatically computed based on the memory limit, maxconn, the buffer size,
1653 memory allocated to compression, SSL cache size, and use of SSL in either
1654 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1655 when there is a memory limit, haproxy will automatically adjust these values
1656 so that 100% of the connections can be made over SSL with no risk, and will
1657 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001658
Willy Tarreaue43d5322013-10-07 20:01:52 +02001659maxsslrate <number>
1660 Sets the maximum per-process number of SSL sessions per second to <number>.
1661 SSL listeners will stop accepting connections when this limit is reached. It
1662 can be used to limit the global SSL CPU usage regardless of each frontend
1663 capacity. It is important to note that this can only be used as a service
1664 protection measure, as there will not necessarily be a fair share between
1665 frontends when the limit is reached, so it's a good idea to also limit each
1666 frontend to some value close to its expected share. It is also important to
1667 note that the sessions are accounted before they enter the SSL stack and not
1668 after, which also protects the stack against bad handshakes. Also, lowering
1669 tune.maxaccept can improve fairness.
1670
William Lallemand9d5f5482012-11-07 16:12:57 +01001671maxzlibmem <number>
1672 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1673 When the maximum amount is reached, future sessions will not compress as long
1674 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001675 The default value is 0. The value is available in bytes on the UNIX socket
1676 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1677 "ZlibMemUsage" in bytes.
1678
Willy Tarreau6a06a402007-07-15 20:15:28 +02001679noepoll
1680 Disables the use of the "epoll" event polling system on Linux. It is
1681 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001682 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001683
1684nokqueue
1685 Disables the use of the "kqueue" event polling system on BSD. It is
1686 equivalent to the command-line argument "-dk". The next polling system
1687 used will generally be "poll". See also "nopoll".
1688
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001689noevports
1690 Disables the use of the event ports event polling system on SunOS systems
1691 derived from Solaris 10 and later. It is equivalent to the command-line
1692 argument "-dv". The next polling system used will generally be "poll". See
1693 also "nopoll".
1694
Willy Tarreau6a06a402007-07-15 20:15:28 +02001695nopoll
1696 Disables the use of the "poll" event polling system. It is equivalent to the
1697 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001698 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001699 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
1700 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001701
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001702nosplice
1703 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001704 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001705 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001706 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001707 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1708 be used. This option makes it easier to globally disable kernel splicing in
1709 case of doubt. See also "option splice-auto", "option splice-request" and
1710 "option splice-response".
1711
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001712nogetaddrinfo
1713 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1714 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1715
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001716noreuseport
1717 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1718 command line argument "-dR".
1719
Willy Tarreaud2d33482019-04-25 17:09:07 +02001720profiling.tasks { auto | on | off }
1721 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
1722 the profiling automatically turns on a thread when it starts to suffer from
1723 an average latency of 1000 microseconds or higher as reported in the
1724 "avg_loop_us" activity field, and automatically turns off when the latency
John Roeslerfb2fce12019-07-10 15:45:51 -05001725 returns below 990 microseconds (this value is an average over the last 1024
Willy Tarreaud2d33482019-04-25 17:09:07 +02001726 loops so it does not vary quickly and tends to significantly smooth short
1727 spikes). It may also spontaneously trigger from time to time on overloaded
1728 systems, containers, or virtual machines, or when the system swaps (which
1729 must absolutely never happen on a load balancer).
1730
1731 CPU profiling per task can be very convenient to report where the time is
1732 spent and which requests have what effect on which other request. Enabling
1733 it will typically affect the overall's performance by less than 1%, thus it
1734 is recommended to leave it to the default 'auto' value so that it only
1735 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01001736 supporting the clock_gettime(2) syscall with clock identifiers
1737 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
1738 be zero. This option may be changed at run time using "set profiling" on the
1739 CLI.
1740
Willy Tarreaufe255b72007-10-14 23:09:26 +02001741spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001742 Sometimes it is desirable to avoid sending agent and health checks to
1743 servers at exact intervals, for instance when many logical servers are
1744 located on the same physical server. With the help of this parameter, it
1745 becomes possible to add some randomness in the check interval between 0
1746 and +/- 50%. A value between 2 and 5 seems to show good results. The
1747 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001748
Davor Ocelice9ed2812017-12-25 17:49:28 +01001749ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001750 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001751 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001752 multiple times, it will simply enable multiple crypto engines. Referencing an
1753 unsupported engine will prevent haproxy from starting. Note that many engines
1754 will lead to lower HTTPS performance than pure software with recent
1755 processors. The optional command "algo" sets the default algorithms an ENGINE
1756 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001757 of "ALL" uses the engine for all cryptographic operations. If no list of
1758 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001759 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1760 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1761 openssl configuration file uses:
1762 https://www.openssl.org/docs/man1.0.2/apps/config.html
1763
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001764ssl-mode-async
1765 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001766 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001767 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1768 doesn't support moving read/write buffers and is not compliant with
1769 haproxy's buffer management. So the asynchronous mode is disabled on
John Roeslerfb2fce12019-07-10 15:45:51 -05001770 read/write operations (it is only enabled during initial and renegotiation
Emeric Brunb5e42a82017-06-06 12:35:14 +00001771 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001772
Willy Tarreau33cb0652014-12-23 22:52:37 +01001773tune.buffers.limit <number>
1774 Sets a hard limit on the number of buffers which may be allocated per process.
1775 The default value is zero which means unlimited. The minimum non-zero value
1776 will always be greater than "tune.buffers.reserve" and should ideally always
1777 be about twice as large. Forcing this value can be particularly useful to
1778 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001779 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001780 another one to be released by another session. Since buffers are dynamically
1781 allocated and released, the waiting time is very short and not perceptible
1782 provided that limits remain reasonable. In fact sometimes reducing the limit
1783 may even increase performance by increasing the CPU cache's efficiency. Tests
1784 have shown good results on average HTTP traffic with a limit to 1/10 of the
1785 expected global maxconn setting, which also significantly reduces memory
1786 usage. The memory savings come from the fact that a number of connections
1787 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1788 advised to do so by an haproxy core developer.
1789
Willy Tarreau1058ae72014-12-23 22:40:40 +01001790tune.buffers.reserve <number>
1791 Sets the number of buffers which are pre-allocated and reserved for use only
1792 during memory shortage conditions resulting in failed memory allocations. The
1793 minimum value is 2 and is also the default. There is no reason a user would
1794 want to change this value, it's mostly aimed at haproxy core developers.
1795
Willy Tarreau27a674e2009-08-17 07:23:33 +02001796tune.bufsize <number>
1797 Sets the buffer size to this size (in bytes). Lower values allow more
1798 sessions to coexist in the same amount of RAM, and higher values allow some
1799 applications with very large cookies to work. The default value is 16384 and
1800 can be changed at build time. It is strongly recommended not to change this
1801 from the default value, as very low values will break some services such as
1802 statistics, and values larger than default size will increase memory usage,
1803 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001804 parameter should be decreased by the same factor as this one is increased. In
1805 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1806 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001807 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Willy Tarreauc77d3642018-12-12 06:19:42 +01001808 than this size, haproxy will return HTTP 502 (Bad Gateway). Note that the
1809 value set using this parameter will automatically be rounded up to the next
1810 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02001811
Willy Tarreau43961d52010-10-04 20:39:20 +02001812tune.chksize <number>
1813 Sets the check buffer size to this size (in bytes). Higher values may help
1814 find string or regex patterns in very large pages, though doing so may imply
1815 more memory and CPU usage. The default value is 16384 and can be changed at
1816 build time. It is not recommended to change this value, but to use better
1817 checks whenever possible.
1818
William Lallemandf3747832012-11-09 12:33:10 +01001819tune.comp.maxlevel <number>
1820 Sets the maximum compression level. The compression level affects CPU
1821 usage during compression. This value affects CPU usage during compression.
1822 Each session using compression initializes the compression algorithm with
1823 this value. The default value is 1.
1824
Willy Tarreauc299e1e2019-02-27 11:35:12 +01001825tune.fail-alloc
1826 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
1827 allocation attempt fails. Must be between 0 (no failure) and 100 (no
1828 success). This is useful to debug and make sure memory failures are handled
1829 gracefully.
1830
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001831tune.h2.header-table-size <number>
1832 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1833 cannot be larger than 65536 bytes. A larger value may help certain clients
1834 send more compact requests, depending on their capabilities. This amount of
1835 memory is consumed for each HTTP/2 connection. It is recommended not to
1836 change it.
1837
Willy Tarreaue6baec02017-07-27 11:45:11 +02001838tune.h2.initial-window-size <number>
1839 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001840 can upload before waiting for an acknowledgment from haproxy. This setting
1841 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001842 The default value is 65535, which roughly allows up to 5 Mbps of upload
1843 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1844 over a 1-ms local network. It can make sense to increase this value to allow
1845 faster uploads, or to reduce it to increase fairness when dealing with many
1846 clients. It doesn't affect resource usage.
1847
Willy Tarreau5242ef82017-07-27 11:47:28 +02001848tune.h2.max-concurrent-streams <number>
1849 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1850 number of outstanding requests on a single connection). The default value is
1851 100. A larger one may slightly improve page load time for complex sites when
1852 visited over high latency networks, but increases the amount of resources a
1853 single client may allocate. A value of zero disables the limit so a single
1854 client may create as many streams as allocatable by haproxy. It is highly
1855 recommended not to change this value.
1856
Willy Tarreaua24b35c2019-02-21 13:24:36 +01001857tune.h2.max-frame-size <number>
1858 Sets the HTTP/2 maximum frame size that haproxy announces it is willing to
1859 receive to its peers. The default value is the largest between 16384 and the
1860 buffer size (tune.bufsize). In any case, haproxy will not announce support
1861 for frame sizes larger than buffers. The main purpose of this setting is to
1862 allow to limit the maximum frame size setting when using large buffers. Too
1863 large frame sizes might have performance impact or cause some peers to
1864 misbehave. It is highly recommended not to change this value.
1865
Willy Tarreau193b8c62012-11-22 00:17:38 +01001866tune.http.cookielen <number>
1867 Sets the maximum length of captured cookies. This is the maximum value that
1868 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1869 will automatically be truncated to this one. It is important not to set too
1870 high a value because all cookie captures still allocate this size whatever
1871 their configured value (they share a same pool). This value is per request
1872 per response, so the memory allocated is twice this value per connection.
1873 When not specified, the limit is set to 63 characters. It is recommended not
1874 to change this value.
1875
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001876tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001877 Sets the maximum length of request URI in logs. This prevents truncating long
1878 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001879 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001880 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001881 configuration directives too.
1882 The default value is 1024.
1883
Willy Tarreauac1932d2011-10-24 19:14:41 +02001884tune.http.maxhdr <number>
1885 Sets the maximum number of headers in a request. When a request comes with a
1886 number of headers greater than this value (including the first line), it is
1887 rejected with a "400 Bad Request" status code. Similarly, too large responses
1888 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1889 for all usages, considering that the widely deployed Apache server uses the
1890 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001891 a buggy application to work by the time it gets fixed. The accepted range is
1892 1..32767. Keep in mind that each new header consumes 32bits of memory for
1893 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001894
Willy Tarreau7e312732014-02-12 16:35:14 +01001895tune.idletimer <timeout>
1896 Sets the duration after which haproxy will consider that an empty buffer is
1897 probably associated with an idle stream. This is used to optimally adjust
1898 some packet sizes while forwarding large and small data alternatively. The
1899 decision to use splice() or to send large buffers in SSL is modulated by this
1900 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1901 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001902 which seems to correctly detect end user pauses (e.g. read a page before
John Roeslerfb2fce12019-07-10 15:45:51 -05001903 clicking). There should be no reason for changing this value. Please check
Willy Tarreau7e312732014-02-12 16:35:14 +01001904 tune.ssl.maxrecord below.
1905
Willy Tarreau7ac908b2019-02-27 12:02:18 +01001906tune.listener.multi-queue { on | off }
1907 Enables ('on') or disables ('off') the listener's multi-queue accept which
1908 spreads the incoming traffic to all threads a "bind" line is allowed to run
1909 on instead of taking them for itself. This provides a smoother traffic
1910 distribution and scales much better, especially in environments where threads
1911 may be unevenly loaded due to external activity (network interrupts colliding
1912 with one thread for example). This option is enabled by default, but it may
1913 be forcefully disabled for troubleshooting or for situations where it is
1914 estimated that the operating system already provides a good enough
1915 distribution and connections are extremely short-lived.
1916
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001917tune.lua.forced-yield <number>
1918 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001919 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001920 HAProxy scheduler to process other tasks like accepting connections or
1921 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001922 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001923 lowered. If the Lua code is quite long and its result is absolutely required
1924 to process the data, the <number> can be increased.
1925
Willy Tarreau32f61e22015-03-18 17:54:59 +01001926tune.lua.maxmem
1927 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1928 default it is zero which means unlimited. It is important to set a limit to
1929 ensure that a bug in a script will not result in the system running out of
1930 memory.
1931
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001932tune.lua.session-timeout <timeout>
1933 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001934 preventing infinite loops or spending too much time in Lua. This timeout
1935 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001936 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001937
1938tune.lua.task-timeout <timeout>
1939 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1940 dedicated to the tasks. By default, this timeout isn't set because a task may
1941 remain alive during of the lifetime of HAProxy. For example, a task used to
1942 check servers.
1943
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001944tune.lua.service-timeout <timeout>
1945 This is the execution timeout for the Lua services. This is useful for
1946 preventing infinite loops or spending too much time in Lua. This timeout
1947 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001948 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001949
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001950tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001951 Sets the maximum number of consecutive connections a process may accept in a
1952 row before switching to other work. In single process mode, higher numbers
1953 give better performance at high connection rates. However in multi-process
1954 modes, keeping a bit of fairness between processes generally is better to
1955 increase performance. This value applies individually to each listener, so
1956 that the number of processes a listener is bound to is taken into account.
1957 This value defaults to 64. In multi-process mode, it is divided by twice
1958 the number of processes the listener is bound to. Setting this value to -1
1959 completely disables the limitation. It should normally not be needed to tweak
1960 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001961
1962tune.maxpollevents <number>
1963 Sets the maximum amount of events that can be processed at once in a call to
1964 the polling system. The default value is adapted to the operating system. It
1965 has been noticed that reducing it below 200 tends to slightly decrease
1966 latency at the expense of network bandwidth, and increasing it above 200
1967 tends to trade latency for slightly increased bandwidth.
1968
Willy Tarreau27a674e2009-08-17 07:23:33 +02001969tune.maxrewrite <number>
1970 Sets the reserved buffer space to this size in bytes. The reserved space is
1971 used for header rewriting or appending. The first reads on sockets will never
1972 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1973 bufsize, though that does not make much sense since there are rarely large
1974 numbers of headers to add. Setting it too high prevents processing of large
1975 requests or responses. Setting it too low prevents addition of new headers
1976 to already large requests or to POST requests. It is generally wise to set it
1977 to about 1024. It is automatically readjusted to half of bufsize if it is
1978 larger than that. This means you don't have to worry about it when changing
1979 bufsize.
1980
Willy Tarreauf3045d22015-04-29 16:24:50 +02001981tune.pattern.cache-size <number>
1982 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1983 cache which reminds previous lookups and their results. It is used by ACLs
1984 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1985 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1986 strings. It applies to pattern expressions which means that it will be able
1987 to memorize the result of a lookup among all the patterns specified on a
1988 configuration line (including all those loaded from files). It automatically
1989 invalidates entries which are updated using HTTP actions or on the CLI. The
1990 default cache size is set to 10000 entries, which limits its footprint to
Willy Tarreau403bfbb2019-10-23 06:59:31 +02001991 about 5 MB per process/thread on 32-bit systems and 8 MB per process/thread
1992 on 64-bit systems, as caches are thread/process local. There is a very low
Willy Tarreauf3045d22015-04-29 16:24:50 +02001993 risk of collision in this cache, which is in the order of the size of the
1994 cache divided by 2^64. Typically, at 10000 requests per second with the
1995 default cache size of 10000 entries, there's 1% chance that a brute force
1996 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1997 This is considered much lower than the risk of a memory corruption caused by
1998 aging components. If this is not acceptable, the cache can be disabled by
1999 setting this parameter to 0.
2000
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02002001tune.pipesize <number>
2002 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
2003 are the default size for the system. But sometimes when using TCP splicing,
2004 it can improve performance to increase pipe sizes, especially if it is
2005 suspected that pipes are not filled and that many calls to splice() are
2006 performed. This has an impact on the kernel's memory footprint, so this must
2007 not be changed if impacts are not understood.
2008
Olivier Houchard88698d92019-04-16 19:07:22 +02002009tune.pool-low-fd-ratio <number>
2010 This setting sets the max number of file descriptors (in percentage) used by
2011 haproxy globally against the maximum number of file descriptors haproxy can
2012 use before we stop putting connection into the idle pool for reuse. The
2013 default is 20.
2014
2015tune.pool-high-fd-ratio <number>
2016 This setting sets the max number of file descriptors (in percentage) used by
2017 haproxy globally against the maximum number of file descriptors haproxy can
2018 use before we start killing idle connections when we can't reuse a connection
2019 and we have to create a new one. The default is 25 (one quarter of the file
2020 descriptor will mean that roughly half of the maximum front connections can
2021 keep an idle connection behind, anything beyond this probably doesn't make
John Roeslerfb2fce12019-07-10 15:45:51 -05002022 much sense in the general case when targeting connection reuse).
Olivier Houchard88698d92019-04-16 19:07:22 +02002023
Willy Tarreaue803de22010-01-21 17:43:04 +01002024tune.rcvbuf.client <number>
2025tune.rcvbuf.server <number>
2026 Forces the kernel socket receive buffer size on the client or the server side
2027 to the specified value in bytes. This value applies to all TCP/HTTP frontends
2028 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05002029 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01002030 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01002031 order to save kernel memory by preventing it from buffering too large amounts
2032 of received data. Lower values will significantly increase CPU usage though.
2033
Willy Tarreaub22fc302015-12-14 12:04:35 +01002034tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01002035 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01002036 socket buffers. One of them is that a read returns more than <recv_enough>
2037 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
2038 may be changed by this setting to better deal with workloads involving lots
2039 of short messages such as telnet or SSH sessions.
2040
Olivier Houchard1599b802018-05-24 18:59:04 +02002041tune.runqueue-depth <number>
John Roeslerfb2fce12019-07-10 15:45:51 -05002042 Sets the maximum amount of task that can be processed at once when running
Olivier Houchard1599b802018-05-24 18:59:04 +02002043 tasks. The default value is 200. Increasing it may incur latency when
2044 dealing with I/Os, making it too small can incur extra overhead.
2045
Willy Tarreaue803de22010-01-21 17:43:04 +01002046tune.sndbuf.client <number>
2047tune.sndbuf.server <number>
2048 Forces the kernel socket send buffer size on the client or the server side to
2049 the specified value in bytes. This value applies to all TCP/HTTP frontends
2050 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05002051 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01002052 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01002053 order to save kernel memory by preventing it from buffering too large amounts
2054 of received data. Lower values will significantly increase CPU usage though.
2055 Another use case is to prevent write timeouts with extremely slow clients due
2056 to the kernel waiting for a large part of the buffer to be read before
2057 notifying haproxy again.
2058
Willy Tarreau6ec58db2012-11-16 16:32:15 +01002059tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01002060 Sets the size of the global SSL session cache, in a number of blocks. A block
2061 is large enough to contain an encoded session without peer certificate.
2062 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002063 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01002064 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01002065 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01002066 and reassigned. Higher values reduce the occurrence of such a purge, hence
2067 the number of CPU-intensive SSL handshakes by ensuring that all users keep
2068 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01002069 and are shared between all processes if "nbproc" is greater than 1. Setting
2070 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01002071
Emeric Brun8dc60392014-05-09 13:52:00 +02002072tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02002073 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02002074 should normally not be used since it will force many renegotiations due to
2075 clients hitting a random process. But it may be required on some operating
2076 systems where none of the SSL cache synchronization method may be used. In
2077 this case, adding a first layer of hash-based load balancing before the SSL
2078 layer might limit the impact of the lack of session sharing.
2079
Emeric Brun4f65bff2012-11-16 15:11:00 +01002080tune.ssl.lifetime <timeout>
2081 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002082 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01002083 does not guarantee that sessions will last that long, because if the cache is
2084 full, the longest idle sessions will be purged despite their configured
2085 lifetime. The real usefulness of this setting is to prevent sessions from
2086 being used for too long.
2087
Willy Tarreaubfd59462013-02-21 07:46:09 +01002088tune.ssl.maxrecord <number>
2089 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
2090 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
2091 data only once it has received a full record. With large records, it means
2092 that clients might have to download up to 16kB of data before starting to
2093 process them. Limiting the value can improve page load times on browsers
2094 located over high latency or low bandwidth networks. It is suggested to find
2095 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
2096 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
2097 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
2098 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01002099 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01002100 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01002101
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002102tune.ssl.default-dh-param <number>
2103 Sets the maximum size of the Diffie-Hellman parameters used for generating
2104 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
2105 final size will try to match the size of the server's RSA (or DSA) key (e.g,
2106 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
2107 this maximum value. Default value if 1024. Only 1024 or higher values are
2108 allowed. Higher values will increase the CPU load, and values greater than
2109 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02002110 used if static Diffie-Hellman parameters are supplied either directly
2111 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002112
Christopher Faulet31af49d2015-06-09 17:29:50 +02002113tune.ssl.ssl-ctx-cache-size <number>
2114 Sets the size of the cache used to store generated certificates to <number>
2115 entries. This is a LRU cache. Because generating a SSL certificate
2116 dynamically is expensive, they are cached. The default cache size is set to
2117 1000 entries.
2118
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01002119tune.ssl.capture-cipherlist-size <number>
2120 Sets the maximum size of the buffer used for capturing client-hello cipher
2121 list. If the value is 0 (default value) the capture is disabled, otherwise
2122 a buffer is allocated for each SSL/TLS connection.
2123
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002124tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002125tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002126tune.vars.reqres-max-size <size>
2127tune.vars.sess-max-size <size>
2128tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002129 These five tunes help to manage the maximum amount of memory used by the
2130 variables system. "global" limits the overall amount of memory available for
2131 all scopes. "proc" limits the memory for the process scope, "sess" limits the
2132 memory for the session scope, "txn" for the transaction scope, and "reqres"
2133 limits the memory for each request or response processing.
2134 Memory accounting is hierarchical, meaning more coarse grained limits include
2135 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
2136 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002137
Daniel Schneller0b547052016-03-21 20:46:57 +01002138 For example, when "tune.vars.sess-max-size" is limited to 100,
2139 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
2140 100 either. If we create a variable "txn.var" that contains 100 bytes,
2141 all available space is consumed.
2142 Notice that exceeding the limits at runtime will not result in an error
2143 message, but values might be cut off or corrupted. So make sure to accurately
2144 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002145
William Lallemanda509e4c2012-11-07 16:54:34 +01002146tune.zlib.memlevel <number>
2147 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002148 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01002149 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01002150 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01002151 between 1 and 9. The default value is 8.
2152
2153tune.zlib.windowsize <number>
2154 Sets the window size (the size of the history buffer) as a parameter of the
2155 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01002156 in better compression at the expense of memory usage. Can be a value between
2157 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002158
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021593.3. Debugging
2160--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02002161
Willy Tarreau1b857852020-02-25 11:27:22 +01002162debug (deprecated)
Willy Tarreau6a06a402007-07-15 20:15:28 +02002163 Enables debug mode which dumps to stdout all exchanges, and disables forking
2164 into background. It is the equivalent of the command-line argument "-d". It
2165 should never be used in a production configuration since it may prevent full
2166 system startup.
2167
2168quiet
2169 Do not display any message during startup. It is equivalent to the command-
2170 line argument "-q".
2171
Emeric Brunf099e792010-09-27 12:05:28 +02002172
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010021733.4. Userlists
2174--------------
2175It is possible to control access to frontend/backend/listen sections or to
2176http stats by allowing only authenticated and authorized users. To do this,
2177it is required to create at least one userlist and to define users.
2178
2179userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01002180 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002181 used to store authentication & authorization data for independent customers.
2182
2183group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01002184 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002185 attach users to this group by using a comma separated list of names
2186 proceeded by "users" keyword.
2187
Cyril Bontéf0c60612010-02-06 14:44:47 +01002188user <username> [password|insecure-password <password>]
2189 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002190 Adds user <username> to the current userlist. Both secure (encrypted) and
2191 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002192 evaluated using the crypt(3) function, so depending on the system's
2193 capabilities, different algorithms are supported. For example, modern Glibc
2194 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
2195 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002196
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002197 Attention: Be aware that using encrypted passwords might cause significantly
2198 increased CPU usage, depending on the number of requests, and the algorithm
2199 used. For any of the hashed variants, the password for each request must
2200 be processed through the chosen algorithm, before it can be compared to the
2201 value specified in the config file. Most current algorithms are deliberately
2202 designed to be expensive to compute to achieve resistance against brute
2203 force attacks. They do not simply salt/hash the clear text password once,
2204 but thousands of times. This can quickly become a major factor in haproxy's
2205 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002206
2207 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01002208 userlist L1
2209 group G1 users tiger,scott
2210 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002211
Cyril Bontéf0c60612010-02-06 14:44:47 +01002212 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
2213 user scott insecure-password elgato
2214 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002215
Cyril Bontéf0c60612010-02-06 14:44:47 +01002216 userlist L2
2217 group G1
2218 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002219
Cyril Bontéf0c60612010-02-06 14:44:47 +01002220 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
2221 user scott insecure-password elgato groups G1,G2
2222 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002223
2224 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002225
Emeric Brunf099e792010-09-27 12:05:28 +02002226
22273.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002228----------
Emeric Brun94900952015-06-11 18:25:54 +02002229It is possible to propagate entries of any data-types in stick-tables between
2230several haproxy instances over TCP connections in a multi-master fashion. Each
2231instance pushes its local updates and insertions to remote peers. The pushed
2232values overwrite remote ones without aggregation. Interrupted exchanges are
2233automatically detected and recovered from the last known point.
2234In addition, during a soft restart, the old process connects to the new one
2235using such a TCP connection to push all its entries before the new process
2236tries to connect to other peers. That ensures very fast replication during a
2237reload, it typically takes a fraction of a second even for large tables.
2238Note that Server IDs are used to identify servers remotely, so it is important
2239that configurations look similar or at least that the same IDs are forced on
2240each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002241
2242peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002243 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002244 which is referenced by one or more stick-tables.
2245
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002246bind [<address>]:<port_range> [, ...] [param*]
2247 Defines the binding parameters of the local peer of this "peers" section.
2248 Such lines are not supported with "peer" line in the same "peers" section.
2249
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002250disabled
2251 Disables a peers section. It disables both listening and any synchronization
2252 related to this section. This is provided to disable synchronization of stick
2253 tables without having to comment out all "peers" references.
2254
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002255default-bind [param*]
2256 Defines the binding parameters for the local peer, excepted its address.
2257
2258default-server [param*]
2259 Change default options for a server in a "peers" section.
2260
2261 Arguments:
2262 <param*> is a list of parameters for this server. The "default-server"
2263 keyword accepts an important number of options and has a complete
2264 section dedicated to it. Please refer to section 5 for more
2265 details.
2266
2267
2268 See also: "server" and section 5 about server options
2269
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002270enable
2271 This re-enables a disabled peers section which was previously disabled.
2272
Frédéric Lécailleb6f759b2019-11-05 09:57:45 +01002273log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
2274 <facility> [<level> [<minlevel>]]
2275 "peers" sections support the same "log" keyword as for the proxies to
2276 log information about the "peers" listener. See "log" option for proxies for
2277 more details.
2278
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002279peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002280 Defines a peer inside a peers section.
2281 If <peername> is set to the local peer name (by default hostname, or forced
2282 using "-L" command line option), haproxy will listen for incoming remote peer
2283 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
2284 to join the remote peer, and <peername> is used at the protocol level to
2285 identify and validate the remote peer on the server side.
2286
2287 During a soft restart, local peer <ip>:<port> is used by the old instance to
2288 connect the new one and initiate a complete replication (teaching process).
2289
2290 It is strongly recommended to have the exact same peers declaration on all
2291 peers and to only rely on the "-L" command line argument to change the local
2292 peer name. This makes it easier to maintain coherent configuration files
2293 across all peers.
2294
William Lallemandb2f07452015-05-12 14:27:13 +02002295 You may want to reference some environment variables in the address
2296 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01002297
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002298 Note: "peer" keyword may transparently be replaced by "server" keyword (see
2299 "server" keyword explanation below).
2300
2301server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02002302 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002303 with a support for all "server" parameters found in 5.2 paragraph.
2304 If the underlying peer is local, <ip>:<port> parameters must not be present.
2305 These parameters must be provided on a "bind" line (see "bind" keyword
2306 of this "peers" section).
2307 Some of these parameters are irrelevant for "peers" sections.
2308
2309
Cyril Bontédc4d9032012-04-08 21:57:39 +02002310 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002311 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02002312 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002313 peer haproxy1 192.168.0.1:1024
2314 peer haproxy2 192.168.0.2:1024
2315 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02002316
2317 backend mybackend
2318 mode tcp
2319 balance roundrobin
2320 stick-table type ip size 20k peers mypeers
2321 stick on src
2322
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002323 server srv1 192.168.0.30:80
2324 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02002325
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002326 Example:
2327 peers mypeers
2328 bind 127.0.0.11:10001 ssl crt mycerts/pem
2329 default-server ssl verify none
2330 server hostA 127.0.0.10:10000
2331 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02002332
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002333
2334table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
2335 size <size> [expire <expire>] [nopurge] [store <data_type>]*
2336
2337 Configure a stickiness table for the current section. This line is parsed
2338 exactly the same way as the "stick-table" keyword in others section, except
John Roeslerfb2fce12019-07-10 15:45:51 -05002339 for the "peers" argument which is not required here and with an additional
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002340 mandatory first parameter to designate the stick-table. Contrary to others
2341 sections, there may be several "table" lines in "peers" sections (see also
2342 "stick-table" keyword).
2343
2344 Also be aware of the fact that "peers" sections have their own stick-table
2345 namespaces to avoid collisions between stick-table names identical in
2346 different "peers" section. This is internally handled prepending the "peers"
2347 sections names to the name of the stick-tables followed by a '/' character.
2348 If somewhere else in the configuration file you have to refer to such
2349 stick-tables declared in "peers" sections you must use the prefixed version
2350 of the stick-table name as follows:
2351
2352 peers mypeers
2353 peer A ...
2354 peer B ...
2355 table t1 ...
2356
2357 frontend fe1
2358 tcp-request content track-sc0 src table mypeers/t1
2359
2360 This is also this prefixed version of the stick-table names which must be
2361 used to refer to stick-tables through the CLI.
2362
2363 About "peers" protocol, as only "peers" belonging to the same section may
2364 communicate with each others, there is no need to do such a distinction.
2365 Several "peers" sections may declare stick-tables with the same name.
2366 This is shorter version of the stick-table name which is sent over the network.
2367 There is only a '/' character as prefix to avoid stick-table name collisions between
2368 stick-tables declared as backends and stick-table declared in "peers" sections
2369 as follows in this weird but supported configuration:
2370
2371 peers mypeers
2372 peer A ...
2373 peer B ...
2374 table t1 type string size 10m store gpc0
2375
2376 backend t1
2377 stick-table type string size 10m store gpc0 peers mypeers
2378
2379 Here "t1" table declared in "mypeeers" section has "mypeers/t1" as global name.
2380 "t1" table declared as a backend as "t1" as global name. But at peer protocol
2381 level the former table is named "/t1", the latter is again named "t1".
2382
Simon Horman51a1cf62015-02-03 13:00:44 +090023833.6. Mailers
2384------------
2385It is possible to send email alerts when the state of servers changes.
2386If configured email alerts are sent to each mailer that is configured
2387in a mailers section. Email is sent to mailers using SMTP.
2388
Pieter Baauw386a1272015-08-16 15:26:24 +02002389mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09002390 Creates a new mailer list with the name <mailersect>. It is an
2391 independent section which is referenced by one or more proxies.
2392
2393mailer <mailername> <ip>:<port>
2394 Defines a mailer inside a mailers section.
2395
2396 Example:
2397 mailers mymailers
2398 mailer smtp1 192.168.0.1:587
2399 mailer smtp2 192.168.0.2:587
2400
2401 backend mybackend
2402 mode tcp
2403 balance roundrobin
2404
2405 email-alert mailers mymailers
2406 email-alert from test1@horms.org
2407 email-alert to test2@horms.org
2408
2409 server srv1 192.168.0.30:80
2410 server srv2 192.168.0.31:80
2411
Pieter Baauw235fcfc2016-02-13 15:33:40 +01002412timeout mail <time>
2413 Defines the time available for a mail/connection to be made and send to
2414 the mail-server. If not defined the default value is 10 seconds. To allow
2415 for at least two SYN-ACK packets to be send during initial TCP handshake it
2416 is advised to keep this value above 4 seconds.
2417
2418 Example:
2419 mailers mymailers
2420 timeout mail 20s
2421 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09002422
William Lallemandc9515522019-06-12 16:32:11 +020024233.7. Programs
2424-------------
2425In master-worker mode, it is possible to launch external binaries with the
2426master, these processes are called programs. These programs are launched and
2427managed the same way as the workers.
2428
2429During a reload of HAProxy, those processes are dealing with the same
2430sequence as a worker:
2431
2432 - the master is re-executed
2433 - the master sends a SIGUSR1 signal to the program
2434 - if "option start-on-reload" is not disabled, the master launches a new
2435 instance of the program
2436
2437During a stop, or restart, a SIGTERM is sent to the programs.
2438
2439program <name>
2440 This is a new program section, this section will create an instance <name>
2441 which is visible in "show proc" on the master CLI. (See "9.4. Master CLI" in
2442 the management guide).
2443
2444command <command> [arguments*]
2445 Define the command to start with optional arguments. The command is looked
2446 up in the current PATH if it does not include an absolute path. This is a
2447 mandatory option of the program section. Arguments containing spaces must
2448 be enclosed in quotes or double quotes or be prefixed by a backslash.
2449
Andrew Heberle97236962019-07-12 11:50:26 +08002450user <user name>
2451 Changes the executed command user ID to the <user name> from /etc/passwd.
2452 See also "group".
2453
2454group <group name>
2455 Changes the executed command group ID to the <group name> from /etc/group.
2456 See also "user".
2457
William Lallemandc9515522019-06-12 16:32:11 +02002458option start-on-reload
2459no option start-on-reload
2460 Start (or not) a new instance of the program upon a reload of the master.
2461 The default is to start a new instance. This option may only be used in a
2462 program section.
2463
2464
Christopher Faulet76edc0f2020-01-13 15:52:01 +010024653.8. HTTP-errors
2466----------------
2467
2468It is possible to globally declare several groups of HTTP errors, to be
2469imported afterwards in any proxy section. Same group may be referenced at
2470several places and can be fully or partially imported.
2471
2472http-errors <name>
2473 Create a new http-errors group with the name <name>. It is an independent
2474 section that may be referenced by one or more proxies using its name.
2475
2476errorfile <code> <file>
2477 Associate a file contents to an HTTP error code
2478
2479 Arguments :
2480 <code> is the HTTP status code. Currently, HAProxy is capable of
2481 generating codes 200, 400, 403, 404, 405, 408, 410, 425, 429,
2482 500, 502, 503, and 504.
2483
2484 <file> designates a file containing the full HTTP response. It is
2485 recommended to follow the common practice of appending ".http" to
2486 the filename so that people do not confuse the response with HTML
2487 error pages, and to use absolute paths, since files are read
2488 before any chroot is performed.
2489
2490 Please referrers to "errorfile" keyword in section 4 for details.
2491
2492 Example:
2493 http-errors website-1
2494 errorfile 400 /etc/haproxy/errorfiles/site1/400.http
2495 errorfile 404 /etc/haproxy/errorfiles/site1/404.http
2496 errorfile 408 /dev/null # work around Chrome pre-connect bug
2497
2498 http-errors website-2
2499 errorfile 400 /etc/haproxy/errorfiles/site2/400.http
2500 errorfile 404 /etc/haproxy/errorfiles/site2/404.http
2501 errorfile 408 /dev/null # work around Chrome pre-connect bug
2502
2503
Willy Tarreauc57f0e22009-05-10 13:12:33 +020025044. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02002505----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002506
Willy Tarreau6a06a402007-07-15 20:15:28 +02002507Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02002508 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02002509 - frontend <name>
2510 - backend <name>
2511 - listen <name>
2512
2513A "defaults" section sets default parameters for all other sections following
2514its declaration. Those default parameters are reset by the next "defaults"
2515section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01002516section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002517
2518A "frontend" section describes a set of listening sockets accepting client
2519connections.
2520
2521A "backend" section describes a set of servers to which the proxy will connect
2522to forward incoming connections.
2523
2524A "listen" section defines a complete proxy with its frontend and backend
2525parts combined in one section. It is generally useful for TCP-only traffic.
2526
Willy Tarreau0ba27502007-12-24 16:55:16 +01002527All proxy names must be formed from upper and lower case letters, digits,
2528'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
2529case-sensitive, which means that "www" and "WWW" are two different proxies.
2530
2531Historically, all proxy names could overlap, it just caused troubles in the
2532logs. Since the introduction of content switching, it is mandatory that two
2533proxies with overlapping capabilities (frontend/backend) have different names.
2534However, it is still permitted that a frontend and a backend share the same
2535name, as this configuration seems to be commonly encountered.
2536
2537Right now, two major proxy modes are supported : "tcp", also known as layer 4,
2538and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002539bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002540protocol, and can interact with it by allowing, blocking, switching, adding,
2541modifying, or removing arbitrary contents in requests or responses, based on
2542arbitrary criteria.
2543
Willy Tarreau70dffda2014-01-30 03:07:23 +01002544In HTTP mode, the processing applied to requests and responses flowing over
2545a connection depends in the combination of the frontend's HTTP options and
Julien Pivotto21ad3152019-12-10 13:11:17 +01002546the backend's. HAProxy supports 3 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01002547
2548 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
2549 requests and responses are processed, and connections remain open but idle
2550 between responses and new requests.
2551
Willy Tarreau70dffda2014-01-30 03:07:23 +01002552 - SCL: server close ("option http-server-close") : the server-facing
2553 connection is closed after the end of the response is received, but the
2554 client-facing connection remains open.
2555
Christopher Faulet315b39c2018-09-21 16:26:19 +02002556 - CLO: close ("option httpclose"): the connection is closed after the end of
2557 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002558
2559The effective mode that will be applied to a connection passing through a
2560frontend and a backend can be determined by both proxy modes according to the
2561following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02002562weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002563
Christopher Faulet315b39c2018-09-21 16:26:19 +02002564 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01002565
Christopher Faulet315b39c2018-09-21 16:26:19 +02002566 | KAL | SCL | CLO
2567 ----+-----+-----+----
2568 KAL | KAL | SCL | CLO
2569 ----+-----+-----+----
Christopher Faulet315b39c2018-09-21 16:26:19 +02002570 mode SCL | SCL | SCL | CLO
2571 ----+-----+-----+----
2572 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01002573
Willy Tarreau0ba27502007-12-24 16:55:16 +01002574
Willy Tarreau70dffda2014-01-30 03:07:23 +01002575
Willy Tarreauc57f0e22009-05-10 13:12:33 +020025764.1. Proxy keywords matrix
2577--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002578
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002579The following list of keywords is supported. Most of them may only be used in a
2580limited set of section types. Some of them are marked as "deprecated" because
2581they are inherited from an old syntax which may be confusing or functionally
2582limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002583marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002584option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002585and must be disabled for a specific instance. Such options may also be prefixed
2586with "default" in order to restore default settings regardless of what has been
2587specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002588
Willy Tarreau6a06a402007-07-15 20:15:28 +02002589
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002590 keyword defaults frontend listen backend
2591------------------------------------+----------+----------+---------+---------
2592acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002593backlog X X X -
2594balance X - X X
2595bind - X X -
2596bind-process X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002597capture cookie - X X -
2598capture request header - X X -
2599capture response header - X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002600compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002601cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002602declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002603default-server X - X X
2604default_backend X X X -
2605description - X X X
2606disabled X X X X
2607dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002608email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002609email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002610email-alert mailers X X X X
2611email-alert myhostname X X X X
2612email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002613enabled X X X X
2614errorfile X X X X
Christopher Faulet76edc0f2020-01-13 15:52:01 +01002615errorfiles X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002616errorloc X X X X
2617errorloc302 X X X X
2618-- keyword -------------------------- defaults - frontend - listen -- backend -
2619errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002620force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002621filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002622fullconn X - X X
2623grace X X X X
2624hash-type X - X X
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01002625http-after-response - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002626http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002627http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002628http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002629http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002630http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002631http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002632http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002633id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002634ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002635load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002636log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002637log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002638log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002639log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002640max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002641maxconn X X X -
2642mode X X X X
2643monitor fail - X X -
2644monitor-net X X X -
2645monitor-uri X X X -
2646option abortonclose (*) X - X X
2647option accept-invalid-http-request (*) X X X -
2648option accept-invalid-http-response (*) X - X X
2649option allbackups (*) X - X X
2650option checkcache (*) X - X X
2651option clitcpka (*) X X X -
2652option contstats (*) X X X -
2653option dontlog-normal (*) X X X -
2654option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002655-- keyword -------------------------- defaults - frontend - listen -- backend -
2656option forwardfor X X X X
Christopher Faulet98fbe952019-07-22 16:18:24 +02002657option h1-case-adjust-bogus-client (*) X X X -
2658option h1-case-adjust-bogus-server (*) X - X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002659option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002660option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002661option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002662option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02002663option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002664option http-server-close (*) X X X X
2665option http-use-proxy-header (*) X X X -
2666option httpchk X - X X
2667option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01002668option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002669option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002670option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002671option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002672option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002673option log-health-checks (*) X - X X
2674option log-separate-errors (*) X X X -
2675option logasap (*) X X X -
2676option mysql-check X - X X
2677option nolinger (*) X X X X
2678option originalto X X X X
2679option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002680option pgsql-check X - X X
2681option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002682option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002683option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002684option smtpchk X - X X
2685option socket-stats (*) X X X -
2686option splice-auto (*) X X X X
2687option splice-request (*) X X X X
2688option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002689option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002690option srvtcpka (*) X - X X
2691option ssl-hello-chk X - X X
2692-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002693option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002694option tcp-smart-accept (*) X X X -
2695option tcp-smart-connect (*) X - X X
2696option tcpka X X X X
2697option tcplog X X X X
2698option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002699external-check command X - X X
2700external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002701persist rdp-cookie X - X X
2702rate-limit sessions X X X -
2703redirect - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002704-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002705retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02002706retry-on X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002707server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002708server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002709server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002710source X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002711stats admin - X X X
2712stats auth X X X X
2713stats enable X X X X
2714stats hide-version X X X X
2715stats http-request - X X X
2716stats realm X X X X
2717stats refresh X X X X
2718stats scope X X X X
2719stats show-desc X X X X
2720stats show-legends X X X X
2721stats show-node X X X X
2722stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002723-- keyword -------------------------- defaults - frontend - listen -- backend -
2724stick match - - X X
2725stick on - - X X
2726stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002727stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002728stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002729tcp-check connect - - X X
2730tcp-check expect - - X X
2731tcp-check send - - X X
2732tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002733tcp-request connection - X X -
2734tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002735tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002736tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002737tcp-response content - - X X
2738tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002739timeout check X - X X
2740timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002741timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002742timeout connect X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002743timeout http-keep-alive X X X X
2744timeout http-request X X X X
2745timeout queue X - X X
2746timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002747timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002748timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002749timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002750transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002751unique-id-format X X X -
2752unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002753use_backend - X X -
Christopher Fauletb30b3102019-09-12 23:03:09 +02002754use-fcgi-app - - X X
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002755use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002756------------------------------------+----------+----------+---------+---------
2757 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002758
Willy Tarreau0ba27502007-12-24 16:55:16 +01002759
Willy Tarreauc57f0e22009-05-10 13:12:33 +020027604.2. Alphabetically sorted keywords reference
2761---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002762
2763This section provides a description of each keyword and its usage.
2764
2765
2766acl <aclname> <criterion> [flags] [operator] <value> ...
2767 Declare or complete an access list.
2768 May be used in sections : defaults | frontend | listen | backend
2769 no | yes | yes | yes
2770 Example:
2771 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2772 acl invalid_src src_port 0:1023
2773 acl local_dst hdr(host) -i localhost
2774
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002775 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002776
2777
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002778backlog <conns>
2779 Give hints to the system about the approximate listen backlog desired size
2780 May be used in sections : defaults | frontend | listen | backend
2781 yes | yes | yes | no
2782 Arguments :
2783 <conns> is the number of pending connections. Depending on the operating
2784 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002785 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002786
2787 In order to protect against SYN flood attacks, one solution is to increase
2788 the system's SYN backlog size. Depending on the system, sometimes it is just
2789 tunable via a system parameter, sometimes it is not adjustable at all, and
2790 sometimes the system relies on hints given by the application at the time of
2791 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2792 to the listen() syscall. On systems which can make use of this value, it can
2793 sometimes be useful to be able to specify a different value, hence this
2794 backlog parameter.
2795
2796 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2797 used as a hint and the system accepts up to the smallest greater power of
2798 two, and never more than some limits (usually 32768).
2799
2800 See also : "maxconn" and the target operating system's tuning guide.
2801
2802
Willy Tarreau0ba27502007-12-24 16:55:16 +01002803balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002804balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002805 Define the load balancing algorithm to be used in a backend.
2806 May be used in sections : defaults | frontend | listen | backend
2807 yes | no | yes | yes
2808 Arguments :
2809 <algorithm> is the algorithm used to select a server when doing load
2810 balancing. This only applies when no persistence information
2811 is available, or when a connection is redispatched to another
2812 server. <algorithm> may be one of the following :
2813
2814 roundrobin Each server is used in turns, according to their weights.
2815 This is the smoothest and fairest algorithm when the server's
2816 processing time remains equally distributed. This algorithm
2817 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002818 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002819 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002820 large farms, when a server becomes up after having been down
2821 for a very short time, it may sometimes take a few hundreds
2822 requests for it to be re-integrated into the farm and start
2823 receiving traffic. This is normal, though very rare. It is
2824 indicated here in case you would have the chance to observe
2825 it, so that you don't worry.
2826
2827 static-rr Each server is used in turns, according to their weights.
2828 This algorithm is as similar to roundrobin except that it is
2829 static, which means that changing a server's weight on the
2830 fly will have no effect. On the other hand, it has no design
2831 limitation on the number of servers, and when a server goes
2832 up, it is always immediately reintroduced into the farm, once
2833 the full map is recomputed. It also uses slightly less CPU to
2834 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002835
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002836 leastconn The server with the lowest number of connections receives the
2837 connection. Round-robin is performed within groups of servers
2838 of the same load to ensure that all servers will be used. Use
2839 of this algorithm is recommended where very long sessions are
2840 expected, such as LDAP, SQL, TSE, etc... but is not very well
2841 suited for protocols using short sessions such as HTTP. This
2842 algorithm is dynamic, which means that server weights may be
2843 adjusted on the fly for slow starts for instance.
2844
Willy Tarreauf09c6602012-02-13 17:12:08 +01002845 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002846 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002847 identifier to the highest (see server parameter "id"), which
2848 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002849 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002850 not make sense to use this algorithm without setting maxconn.
2851 The purpose of this algorithm is to always use the smallest
2852 number of servers so that extra servers can be powered off
2853 during non-intensive hours. This algorithm ignores the server
2854 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002855 or IMAP than HTTP, though it can be useful there too. In
2856 order to use this algorithm efficiently, it is recommended
2857 that a cloud controller regularly checks server usage to turn
2858 them off when unused, and regularly checks backend queue to
2859 turn new servers on when the queue inflates. Alternatively,
2860 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002861
Willy Tarreau0ba27502007-12-24 16:55:16 +01002862 source The source IP address is hashed and divided by the total
2863 weight of the running servers to designate which server will
2864 receive the request. This ensures that the same client IP
2865 address will always reach the same server as long as no
2866 server goes down or up. If the hash result changes due to the
2867 number of running servers changing, many clients will be
2868 directed to a different server. This algorithm is generally
2869 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002870 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002871 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002872 static by default, which means that changing a server's
2873 weight on the fly will have no effect, but this can be
2874 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002875
Oskar Stolc8dc41842012-05-19 10:19:54 +01002876 uri This algorithm hashes either the left part of the URI (before
2877 the question mark) or the whole URI (if the "whole" parameter
2878 is present) and divides the hash value by the total weight of
2879 the running servers. The result designates which server will
2880 receive the request. This ensures that the same URI will
2881 always be directed to the same server as long as no server
2882 goes up or down. This is used with proxy caches and
2883 anti-virus proxies in order to maximize the cache hit rate.
2884 Note that this algorithm may only be used in an HTTP backend.
2885 This algorithm is static by default, which means that
2886 changing a server's weight on the fly will have no effect,
2887 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002888
Oskar Stolc8dc41842012-05-19 10:19:54 +01002889 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002890 "depth", both followed by a positive integer number. These
2891 options may be helpful when it is needed to balance servers
2892 based on the beginning of the URI only. The "len" parameter
2893 indicates that the algorithm should only consider that many
2894 characters at the beginning of the URI to compute the hash.
2895 Note that having "len" set to 1 rarely makes sense since most
2896 URIs start with a leading "/".
2897
2898 The "depth" parameter indicates the maximum directory depth
2899 to be used to compute the hash. One level is counted for each
2900 slash in the request. If both parameters are specified, the
2901 evaluation stops when either is reached.
2902
Willy Tarreau0ba27502007-12-24 16:55:16 +01002903 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002904 the query string of each HTTP GET request.
2905
2906 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002907 request entity will be searched for the parameter argument,
2908 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002909 ('?') in the URL. The message body will only start to be
2910 analyzed once either the advertised amount of data has been
2911 received or the request buffer is full. In the unlikely event
2912 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002913 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002914 be randomly balanced if at all. This keyword used to support
2915 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002916
2917 If the parameter is found followed by an equal sign ('=') and
2918 a value, then the value is hashed and divided by the total
2919 weight of the running servers. The result designates which
2920 server will receive the request.
2921
2922 This is used to track user identifiers in requests and ensure
2923 that a same user ID will always be sent to the same server as
2924 long as no server goes up or down. If no value is found or if
2925 the parameter is not found, then a round robin algorithm is
2926 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002927 backend. This algorithm is static by default, which means
2928 that changing a server's weight on the fly will have no
2929 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002930
Cyril Bontédc4d9032012-04-08 21:57:39 +02002931 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2932 request. Just as with the equivalent ACL 'hdr()' function,
2933 the header name in parenthesis is not case sensitive. If the
2934 header is absent or if it does not contain any value, the
2935 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002936
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002937 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002938 reducing the hash algorithm to the main domain part with some
2939 specific headers such as 'Host'. For instance, in the Host
2940 value "haproxy.1wt.eu", only "1wt" will be considered.
2941
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002942 This algorithm is static by default, which means that
2943 changing a server's weight on the fly will have no effect,
2944 but this can be changed using "hash-type".
2945
Willy Tarreau21c741a2019-01-14 18:14:27 +01002946 random
2947 random(<draws>)
2948 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02002949 hashing function. This means that the servers' weights are
2950 respected, dynamic weight changes immediately take effect, as
2951 well as new server additions. Random load balancing can be
2952 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01002953 or removed as it may avoid the hammering effect that could
2954 result from roundrobin or leastconn in this situation. The
2955 hash-balance-factor directive can be used to further improve
2956 fairness of the load balancing, especially in situations
2957 where servers show highly variable response times. When an
2958 argument <draws> is present, it must be an integer value one
2959 or greater, indicating the number of draws before selecting
2960 the least loaded of these servers. It was indeed demonstrated
2961 that picking the least loaded of two servers is enough to
2962 significantly improve the fairness of the algorithm, by
2963 always avoiding to pick the most loaded server within a farm
2964 and getting rid of any bias that could be induced by the
2965 unfair distribution of the consistent list. Higher values N
2966 will take away N-1 of the highest loaded servers at the
2967 expense of performance. With very high values, the algorithm
2968 will converge towards the leastconn's result but much slower.
2969 The default value is 2, which generally shows very good
2970 distribution and performance. This algorithm is also known as
2971 the Power of Two Random Choices and is described here :
2972 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02002973
Emeric Brun736aa232009-06-30 17:56:00 +02002974 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002975 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002976 The RDP cookie <name> (or "mstshash" if omitted) will be
2977 looked up and hashed for each incoming TCP request. Just as
2978 with the equivalent ACL 'req_rdp_cookie()' function, the name
2979 is not case-sensitive. This mechanism is useful as a degraded
2980 persistence mode, as it makes it possible to always send the
2981 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002982 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002983 used instead.
2984
2985 Note that for this to work, the frontend must ensure that an
2986 RDP cookie is already present in the request buffer. For this
2987 you must use 'tcp-request content accept' rule combined with
2988 a 'req_rdp_cookie_cnt' ACL.
2989
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002990 This algorithm is static by default, which means that
2991 changing a server's weight on the fly will have no effect,
2992 but this can be changed using "hash-type".
2993
Cyril Bontédc4d9032012-04-08 21:57:39 +02002994 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002995
Willy Tarreau0ba27502007-12-24 16:55:16 +01002996 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002997 algorithms. Right now, only "url_param" and "uri" support an
2998 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002999
Willy Tarreau3cd9af22009-03-15 14:06:41 +01003000 The load balancing algorithm of a backend is set to roundrobin when no other
3001 algorithm, mode nor option have been set. The algorithm may only be set once
3002 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003003
Lukas Tribus80512b12018-10-27 20:07:40 +02003004 With authentication schemes that require the same connection like NTLM, URI
John Roeslerfb2fce12019-07-10 15:45:51 -05003005 based algorithms must not be used, as they would cause subsequent requests
Lukas Tribus80512b12018-10-27 20:07:40 +02003006 to be routed to different backend servers, breaking the invalid assumptions
3007 NTLM relies on.
3008
Willy Tarreau0ba27502007-12-24 16:55:16 +01003009 Examples :
3010 balance roundrobin
3011 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003012 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01003013 balance hdr(User-Agent)
3014 balance hdr(host)
3015 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003016
3017 Note: the following caveats and limitations on using the "check_post"
3018 extension with "url_param" must be considered :
3019
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003020 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003021 to determine if the parameters will be found in the body or entity which
3022 may contain binary data. Therefore another method may be required to
3023 restrict consideration of POST requests that have no URL parameters in
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02003024 the body. (see acl http_end)
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003025
3026 - using a <max_wait> value larger than the request buffer size does not
3027 make sense and is useless. The buffer size is set at build time, and
3028 defaults to 16 kB.
3029
3030 - Content-Encoding is not supported, the parameter search will probably
3031 fail; and load balancing will fall back to Round Robin.
3032
3033 - Expect: 100-continue is not supported, load balancing will fall back to
3034 Round Robin.
3035
Lukas Tribus23953682017-04-28 13:24:30 +00003036 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003037 If the entire parameter value is not present in the first chunk, the
3038 selection of server is undefined (actually, defined by how little
3039 actually appeared in the first chunk).
3040
3041 - This feature does not support generation of a 100, 411 or 501 response.
3042
3043 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003044 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02003045 white space or control characters are found, indicating the end of what
3046 might be a URL parameter list. This is probably not a concern with SGML
3047 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003048
Willy Tarreau294d0f02015-08-10 19:40:12 +02003049 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003050
3051
Willy Tarreaub6205fd2012-09-24 12:27:33 +02003052bind [<address>]:<port_range> [, ...] [param*]
3053bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003054 Define one or several listening addresses and/or ports in a frontend.
3055 May be used in sections : defaults | frontend | listen | backend
3056 no | yes | yes | no
3057 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01003058 <address> is optional and can be a host name, an IPv4 address, an IPv6
3059 address, or '*'. It designates the address the frontend will
3060 listen on. If unset, all IPv4 addresses of the system will be
3061 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01003062 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01003063 Optionally, an address family prefix may be used before the
3064 address to force the family regardless of the address format,
3065 which can be useful to specify a path to a unix socket with
3066 no slash ('/'). Currently supported prefixes are :
3067 - 'ipv4@' -> address is always IPv4
3068 - 'ipv6@' -> address is always IPv6
3069 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02003070 - 'abns@' -> address is in abstract namespace (Linux only).
3071 Note: since abstract sockets are not "rebindable", they
3072 do not cope well with multi-process mode during
3073 soft-restart, so it is better to avoid them if
3074 nbproc is greater than 1. The effect is that if the
3075 new process fails to start, only one of the old ones
3076 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01003077 - 'fd@<n>' -> use file descriptor <n> inherited from the
3078 parent. The fd must be bound and may or may not already
3079 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02003080 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
3081 connected unix socket or of a socketpair. The bind waits
3082 to receive a FD over the unix socket and uses it as if it
3083 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02003084 You may want to reference some environment variables in the
3085 address parameter, see section 2.3 about environment
3086 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01003087
Willy Tarreauc5011ca2010-03-22 11:53:56 +01003088 <port_range> is either a unique TCP port, or a port range for which the
3089 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003090 above. The port is mandatory for TCP listeners. Note that in
3091 the case of an IPv6 address, the port is always the number
3092 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01003093 - a numerical port (ex: '80')
3094 - a dash-delimited ports range explicitly stating the lower
3095 and upper bounds (ex: '2000-2100') which are included in
3096 the range.
3097
3098 Particular care must be taken against port ranges, because
3099 every <address:port> couple consumes one socket (= a file
3100 descriptor), so it's easy to consume lots of descriptors
3101 with a simple range, and to run out of sockets. Also, each
3102 <address:port> couple must be used only once among all
3103 instances running on a same system. Please note that binding
3104 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04003105 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01003106 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003107
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003108 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01003109 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003110 receive UNIX connections on the socket located at this place.
3111 The path must begin with a slash and by default is absolute.
3112 It can be relative to the prefix defined by "unix-bind" in
3113 the global section. Note that the total length of the prefix
3114 followed by the socket path cannot exceed some system limits
3115 for UNIX sockets, which commonly are set to 107 characters.
3116
Willy Tarreaub6205fd2012-09-24 12:27:33 +02003117 <param*> is a list of parameters common to all sockets declared on the
3118 same line. These numerous parameters depend on OS and build
3119 options and have a complete section dedicated to them. Please
3120 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02003121
Willy Tarreau0ba27502007-12-24 16:55:16 +01003122 It is possible to specify a list of address:port combinations delimited by
3123 commas. The frontend will then listen on all of these addresses. There is no
3124 fixed limit to the number of addresses and ports which can be listened on in
3125 a frontend, as well as there is no limit to the number of "bind" statements
3126 in a frontend.
3127
3128 Example :
3129 listen http_proxy
3130 bind :80,:443
3131 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003132 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01003133
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02003134 listen http_https_proxy
3135 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02003136 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02003137
Willy Tarreau24709282013-03-10 21:32:12 +01003138 listen http_https_proxy_explicit
3139 bind ipv6@:80
3140 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
3141 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
3142
Willy Tarreaudad36a32013-03-11 01:20:04 +01003143 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02003144 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01003145
Willy Tarreau55dcaf62015-09-27 15:03:15 +02003146 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
3147 sun_path length is used for the address length. Some other programs
3148 such as socat use the string length only by default. Pass the option
3149 ",unix-tightsocklen=0" to any abstract socket definition in socat to
3150 make it compatible with HAProxy's.
3151
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003152 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02003153 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003154
3155
Christopher Fauletff4121f2017-11-22 16:38:49 +01003156bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003157 Limit visibility of an instance to a certain set of processes numbers.
3158 May be used in sections : defaults | frontend | listen | backend
3159 yes | yes | yes | yes
3160 Arguments :
3161 all All process will see this instance. This is the default. It
3162 may be used to override a default value.
3163
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003164 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003165 option may be combined with other numbers.
3166
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003167 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003168 option may be combined with other numbers. Do not use it
3169 with less than 2 processes otherwise some instances might be
3170 missing from all processes.
3171
Christopher Fauletff4121f2017-11-22 16:38:49 +01003172 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003173 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01003174 the machine's word size. Ranges can be partially defined. The
3175 higher bound can be omitted. In such case, it is replaced by
3176 the corresponding maximum value. If a proxy is bound to
3177 process numbers greater than the configured global.nbproc, it
3178 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02003179 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003180
3181 This keyword limits binding of certain instances to certain processes. This
3182 is useful in order not to have too many processes listening to the same
3183 ports. For instance, on a dual-core machine, it might make sense to set
3184 'nbproc 2' in the global section, then distributes the listeners among 'odd'
3185 and 'even' instances.
3186
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003187 At the moment, it is not possible to reference more than 32 or 64 processes
3188 using this keyword, but this should be more than enough for most setups.
3189 Please note that 'all' really means all processes regardless of the machine's
3190 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003191
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003192 Each "bind" line may further be limited to a subset of the proxy's processes,
3193 please consult the "process" bind keyword in section 5.1.
3194
Willy Tarreaub369a042014-09-16 13:21:03 +02003195 When a frontend has no explicit "bind-process" line, it tries to bind to all
3196 the processes referenced by its "bind" lines. That means that frontends can
3197 easily adapt to their listeners' processes.
3198
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003199 If some backends are referenced by frontends bound to other processes, the
3200 backend automatically inherits the frontend's processes.
3201
3202 Example :
3203 listen app_ip1
3204 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003205 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003206
3207 listen app_ip2
3208 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003209 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003210
3211 listen management
3212 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003213 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003214
Willy Tarreau110ecc12012-11-15 17:50:01 +01003215 listen management
3216 bind 10.0.0.4:80
3217 bind-process 1-4
3218
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003219 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003220
3221
Willy Tarreau0ba27502007-12-24 16:55:16 +01003222capture cookie <name> len <length>
3223 Capture and log a cookie in the request and in the response.
3224 May be used in sections : defaults | frontend | listen | backend
3225 no | yes | yes | no
3226 Arguments :
3227 <name> is the beginning of the name of the cookie to capture. In order
3228 to match the exact name, simply suffix the name with an equal
3229 sign ('='). The full name will appear in the logs, which is
3230 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01003231 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003232
3233 <length> is the maximum number of characters to report in the logs, which
3234 include the cookie name, the equal sign and the value, all in the
3235 standard "name=value" form. The string will be truncated on the
3236 right if it exceeds <length>.
3237
3238 Only the first cookie is captured. Both the "cookie" request headers and the
3239 "set-cookie" response headers are monitored. This is particularly useful to
3240 check for application bugs causing session crossing or stealing between
3241 users, because generally the user's cookies can only change on a login page.
3242
3243 When the cookie was not presented by the client, the associated log column
3244 will report "-". When a request does not cause a cookie to be assigned by the
3245 server, a "-" is reported in the response column.
3246
3247 The capture is performed in the frontend only because it is necessary that
3248 the log format does not change for a given frontend depending on the
3249 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01003250 "capture cookie" statement in a frontend. The maximum capture length is set
3251 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
3252 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003253
3254 Example:
3255 capture cookie ASPSESSION len 32
3256
3257 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003258 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003259
3260
3261capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003262 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003263 May be used in sections : defaults | frontend | listen | backend
3264 no | yes | yes | no
3265 Arguments :
3266 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003267 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003268 appear in the requests, with the first letter of each word in
3269 upper case. The header name will not appear in the logs, only the
3270 value is reported, but the position in the logs is respected.
3271
3272 <length> is the maximum number of characters to extract from the value and
3273 report in the logs. The string will be truncated on the right if
3274 it exceeds <length>.
3275
Willy Tarreau4460d032012-11-21 23:37:37 +01003276 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003277 value will be added to the logs between braces ('{}'). If multiple headers
3278 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003279 in the same order they were declared in the configuration. Non-existent
3280 headers will be logged just as an empty string. Common uses for request
3281 header captures include the "Host" field in virtual hosting environments, the
3282 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003283 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003284 environments to find where the request came from.
3285
3286 Note that when capturing headers such as "User-agent", some spaces may be
3287 logged, making the log analysis more difficult. Thus be careful about what
3288 you log if you know your log parser is not smart enough to rely on the
3289 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003290
Willy Tarreau0900abb2012-11-22 00:21:46 +01003291 There is no limit to the number of captured request headers nor to their
3292 length, though it is wise to keep them low to limit memory usage per session.
3293 In order to keep log format consistent for a same frontend, header captures
3294 can only be declared in a frontend. It is not possible to specify a capture
3295 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003296
3297 Example:
3298 capture request header Host len 15
3299 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01003300 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01003301
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003302 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003303 about logging.
3304
3305
3306capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003307 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003308 May be used in sections : defaults | frontend | listen | backend
3309 no | yes | yes | no
3310 Arguments :
3311 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003312 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003313 appear in the response, with the first letter of each word in
3314 upper case. The header name will not appear in the logs, only the
3315 value is reported, but the position in the logs is respected.
3316
3317 <length> is the maximum number of characters to extract from the value and
3318 report in the logs. The string will be truncated on the right if
3319 it exceeds <length>.
3320
Willy Tarreau4460d032012-11-21 23:37:37 +01003321 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003322 result will be added to the logs between braces ('{}') after the captured
3323 request headers. If multiple headers are captured, they will be delimited by
3324 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003325 the configuration. Non-existent headers will be logged just as an empty
3326 string. Common uses for response header captures include the "Content-length"
3327 header which indicates how many bytes are expected to be returned, the
3328 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003329
Willy Tarreau0900abb2012-11-22 00:21:46 +01003330 There is no limit to the number of captured response headers nor to their
3331 length, though it is wise to keep them low to limit memory usage per session.
3332 In order to keep log format consistent for a same frontend, header captures
3333 can only be declared in a frontend. It is not possible to specify a capture
3334 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003335
3336 Example:
3337 capture response header Content-length len 9
3338 capture response header Location len 15
3339
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003340 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003341 about logging.
3342
3343
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003344compression algo <algorithm> ...
3345compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02003346compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02003347 Enable HTTP compression.
3348 May be used in sections : defaults | frontend | listen | backend
3349 yes | yes | yes | yes
3350 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003351 algo is followed by the list of supported compression algorithms.
3352 type is followed by the list of MIME types that will be compressed.
3353 offload makes haproxy work as a compression offloader only (see notes).
3354
3355 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01003356 identity this is mostly for debugging, and it was useful for developing
3357 the compression feature. Identity does not apply any change on
3358 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003359
Willy Tarreauc91840a2015-03-28 17:00:39 +01003360 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01003361 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01003362
3363 deflate same as "gzip", but with deflate algorithm and zlib format.
3364 Note that this algorithm has ambiguous support on many
3365 browsers and no support at all from recent ones. It is
3366 strongly recommended not to use it for anything else than
3367 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01003368 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003369
Willy Tarreauc91840a2015-03-28 17:00:39 +01003370 raw-deflate same as "deflate" without the zlib wrapper, and used as an
3371 alternative when the browser wants "deflate". All major
3372 browsers understand it and despite violating the standards,
3373 it is known to work better than "deflate", at least on MSIE
3374 and some versions of Safari. Do not use it in conjunction
3375 with "deflate", use either one or the other since both react
3376 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01003377 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003378
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04003379 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003380 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003381 If backend servers support HTTP compression, these directives
3382 will be no-op: haproxy will see the compressed response and will not
3383 compress again. If backend servers do not support HTTP compression and
3384 there is Accept-Encoding header in request, haproxy will compress the
3385 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02003386
3387 The "offload" setting makes haproxy remove the Accept-Encoding header to
3388 prevent backend servers from compressing responses. It is strongly
3389 recommended not to do this because this means that all the compression work
3390 will be done on the single point where haproxy is located. However in some
3391 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003392 with broken HTTP compression implementation which can't be turned off.
3393 In that case haproxy can be used to prevent that gateway from emitting
3394 invalid payloads. In this case, simply removing the header in the
3395 configuration does not work because it applies before the header is parsed,
3396 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02003397 then be used for such scenarios. Note: for now, the "offload" setting is
3398 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02003399
William Lallemand05097442012-11-20 12:14:28 +01003400 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003401 * the request does not advertise a supported compression algorithm in the
3402 "Accept-Encoding" header
3403 * the response message is not HTTP/1.1
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003404 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003405 * response contain neither a "Content-Length" header nor a
3406 "Transfer-Encoding" whose last value is "chunked"
3407 * response contains a "Content-Type" header whose first value starts with
3408 "multipart"
3409 * the response contains the "no-transform" value in the "Cache-control"
3410 header
3411 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
3412 and later
3413 * The response contains a "Content-Encoding" header, indicating that the
3414 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003415 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01003416
Tim Duesterhusb229f012019-01-29 16:38:56 +01003417 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01003418
William Lallemand82fe75c2012-10-23 10:25:10 +02003419 Examples :
3420 compression algo gzip
3421 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01003422
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003423
Willy Tarreau55165fe2009-05-10 12:02:55 +02003424cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003425 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3426 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Christopher Faulet2f533902020-01-21 11:06:48 +01003427 [ dynamic ] [ attr <value> ]*
Willy Tarreau0ba27502007-12-24 16:55:16 +01003428 Enable cookie-based persistence in a backend.
3429 May be used in sections : defaults | frontend | listen | backend
3430 yes | no | yes | yes
3431 Arguments :
3432 <name> is the name of the cookie which will be monitored, modified or
3433 inserted in order to bring persistence. This cookie is sent to
3434 the client via a "Set-Cookie" header in the response, and is
3435 brought back by the client in a "Cookie" header in all requests.
3436 Special care should be taken to choose a name which does not
3437 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003438 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003439 HTTP/HTTPS), care should be taken to use different cookie names
3440 between all backends if persistence between them is not desired.
3441
3442 rewrite This keyword indicates that the cookie will be provided by the
3443 server and that haproxy will have to modify its value to set the
3444 server's identifier in it. This mode is handy when the management
3445 of complex combinations of "Set-cookie" and "Cache-control"
3446 headers is left to the application. The application can then
3447 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003448 cookie. Since all responses should be monitored, this mode
3449 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003450 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003451 start with this mode for new deployments. This keyword is
3452 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003453
3454 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003455 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003456
Willy Tarreaua79094d2010-08-31 22:54:15 +02003457 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003458 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02003459 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003460 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003461 existing configurations running in the "rewrite" mode. The cookie
3462 will only be a session cookie and will not be stored on the
3463 client's disk. By default, unless the "indirect" option is added,
3464 the server will see the cookies emitted by the client. Due to
3465 caching effects, it is generally wise to add the "nocache" or
3466 "postonly" keywords (see below). The "insert" keyword is not
3467 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003468
3469 prefix This keyword indicates that instead of relying on a dedicated
3470 cookie for the persistence, an existing one will be completed.
3471 This may be needed in some specific environments where the client
3472 does not support more than one single cookie and the application
3473 already needs it. In this case, whenever the server sets a cookie
3474 named <name>, it will be prefixed with the server's identifier
3475 and a delimiter. The prefix will be removed from all client
3476 requests so that the server still finds the cookie it emitted.
3477 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003478 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003479 not compatible with "rewrite" and "insert". Note: it is highly
3480 recommended not to use "indirect" with "prefix", otherwise server
3481 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003482
Willy Tarreaua79094d2010-08-31 22:54:15 +02003483 indirect When this option is specified, no cookie will be emitted to a
3484 client which already has a valid one for the server which has
3485 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003486 it will be removed, unless the "preserve" option is also set. In
3487 "insert" mode, this will additionally remove cookies from the
3488 requests transmitted to the server, making the persistence
3489 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003490 Note: it is highly recommended not to use "indirect" with
3491 "prefix", otherwise server cookie updates would not be sent to
3492 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003493
3494 nocache This option is recommended in conjunction with the insert mode
3495 when there is a cache between the client and HAProxy, as it
3496 ensures that a cacheable response will be tagged non-cacheable if
3497 a cookie needs to be inserted. This is important because if all
3498 persistence cookies are added on a cacheable home page for
3499 instance, then all customers will then fetch the page from an
3500 outer cache and will all share the same persistence cookie,
3501 leading to one server receiving much more traffic than others.
3502 See also the "insert" and "postonly" options.
3503
3504 postonly This option ensures that cookie insertion will only be performed
3505 on responses to POST requests. It is an alternative to the
3506 "nocache" option, because POST responses are not cacheable, so
3507 this ensures that the persistence cookie will never get cached.
3508 Since most sites do not need any sort of persistence before the
3509 first POST which generally is a login request, this is a very
3510 efficient method to optimize caching without risking to find a
3511 persistence cookie in the cache.
3512 See also the "insert" and "nocache" options.
3513
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003514 preserve This option may only be used with "insert" and/or "indirect". It
3515 allows the server to emit the persistence cookie itself. In this
3516 case, if a cookie is found in the response, haproxy will leave it
3517 untouched. This is useful in order to end persistence after a
3518 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003519 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003520 the past. By combining this mechanism with the "disable-on-404"
3521 check option, it is possible to perform a completely graceful
3522 shutdown because users will definitely leave the server after
3523 they logout.
3524
Willy Tarreau4992dd22012-05-31 21:02:17 +02003525 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3526 when a cookie is inserted. This attribute is used so that a
3527 user agent doesn't share the cookie with non-HTTP components.
3528 Please check RFC6265 for more information on this attribute.
3529
3530 secure This option tells haproxy to add a "Secure" cookie attribute when
3531 a cookie is inserted. This attribute is used so that a user agent
3532 never emits this cookie over non-secure channels, which means
3533 that a cookie learned with this flag will be presented only over
3534 SSL/TLS connections. Please check RFC6265 for more information on
3535 this attribute.
3536
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003537 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003538 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003539 name. If the domain begins with a dot, the browser is allowed to
3540 use it for any host ending with that name. It is also possible to
3541 specify several domain names by invoking this option multiple
3542 times. Some browsers might have small limits on the number of
3543 domains, so be careful when doing that. For the record, sending
3544 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003545
Willy Tarreau996a92c2010-10-13 19:30:47 +02003546 maxidle This option allows inserted cookies to be ignored after some idle
3547 time. It only works with insert-mode cookies. When a cookie is
3548 sent to the client, the date this cookie was emitted is sent too.
3549 Upon further presentations of this cookie, if the date is older
3550 than the delay indicated by the parameter (in seconds), it will
3551 be ignored. Otherwise, it will be refreshed if needed when the
3552 response is sent to the client. This is particularly useful to
3553 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003554 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003555 this option is set and a cookie has no date, it is always
3556 accepted, but gets refreshed in the response. This maintains the
3557 ability for admins to access their sites. Cookies that have a
3558 date in the future further than 24 hours are ignored. Doing so
3559 lets admins fix timezone issues without risking kicking users off
3560 the site.
3561
3562 maxlife This option allows inserted cookies to be ignored after some life
3563 time, whether they're in use or not. It only works with insert
3564 mode cookies. When a cookie is first sent to the client, the date
3565 this cookie was emitted is sent too. Upon further presentations
3566 of this cookie, if the date is older than the delay indicated by
3567 the parameter (in seconds), it will be ignored. If the cookie in
3568 the request has no date, it is accepted and a date will be set.
3569 Cookies that have a date in the future further than 24 hours are
3570 ignored. Doing so lets admins fix timezone issues without risking
3571 kicking users off the site. Contrary to maxidle, this value is
3572 not refreshed, only the first visit date counts. Both maxidle and
3573 maxlife may be used at the time. This is particularly useful to
3574 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003575 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003576 is stronger than the maxidle method in that it forces a
3577 redispatch after some absolute delay.
3578
Olivier Houchard4e694042017-03-14 20:01:29 +01003579 dynamic Activate dynamic cookies. When used, a session cookie is
3580 dynamically created for each server, based on the IP and port
3581 of the server, and a secret key, specified in the
3582 "dynamic-cookie-key" backend directive.
3583 The cookie will be regenerated each time the IP address change,
3584 and is only generated for IPv4/IPv6.
3585
Christopher Faulet2f533902020-01-21 11:06:48 +01003586 attr This option tells haproxy to add an extra attribute when a
3587 cookie is inserted. The attribute value can contain any
3588 characters except control ones or ";". This option may be
3589 repeated.
3590
Willy Tarreau0ba27502007-12-24 16:55:16 +01003591 There can be only one persistence cookie per HTTP backend, and it can be
3592 declared in a defaults section. The value of the cookie will be the value
3593 indicated after the "cookie" keyword in a "server" statement. If no cookie
3594 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003595
Willy Tarreau0ba27502007-12-24 16:55:16 +01003596 Examples :
3597 cookie JSESSIONID prefix
3598 cookie SRV insert indirect nocache
3599 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003600 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003601
Willy Tarreau294d0f02015-08-10 19:40:12 +02003602 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003603
Willy Tarreau983e01e2010-01-11 18:42:06 +01003604
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003605declare capture [ request | response ] len <length>
3606 Declares a capture slot.
3607 May be used in sections : defaults | frontend | listen | backend
3608 no | yes | yes | no
3609 Arguments:
3610 <length> is the length allowed for the capture.
3611
3612 This declaration is only available in the frontend or listen section, but the
3613 reserved slot can be used in the backends. The "request" keyword allocates a
3614 capture slot for use in the request, and "response" allocates a capture slot
3615 for use in the response.
3616
3617 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003618 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003619 "http-request capture" and "http-response capture".
3620
3621
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003622default-server [param*]
3623 Change default options for a server in a backend
3624 May be used in sections : defaults | frontend | listen | backend
3625 yes | no | yes | yes
3626 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003627 <param*> is a list of parameters for this server. The "default-server"
3628 keyword accepts an important number of options and has a complete
3629 section dedicated to it. Please refer to section 5 for more
3630 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003631
Willy Tarreau983e01e2010-01-11 18:42:06 +01003632 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003633 default-server inter 1000 weight 13
3634
3635 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003636
Willy Tarreau983e01e2010-01-11 18:42:06 +01003637
Willy Tarreau0ba27502007-12-24 16:55:16 +01003638default_backend <backend>
3639 Specify the backend to use when no "use_backend" rule has been matched.
3640 May be used in sections : defaults | frontend | listen | backend
3641 yes | yes | yes | no
3642 Arguments :
3643 <backend> is the name of the backend to use.
3644
3645 When doing content-switching between frontend and backends using the
3646 "use_backend" keyword, it is often useful to indicate which backend will be
3647 used when no rule has matched. It generally is the dynamic backend which
3648 will catch all undetermined requests.
3649
Willy Tarreau0ba27502007-12-24 16:55:16 +01003650 Example :
3651
3652 use_backend dynamic if url_dyn
3653 use_backend static if url_css url_img extension_img
3654 default_backend dynamic
3655
Willy Tarreau98d04852015-05-26 12:18:29 +02003656 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003657
Willy Tarreau0ba27502007-12-24 16:55:16 +01003658
Baptiste Assmann27f51342013-10-09 06:51:49 +02003659description <string>
3660 Describe a listen, frontend or backend.
3661 May be used in sections : defaults | frontend | listen | backend
3662 no | yes | yes | yes
3663 Arguments : string
3664
3665 Allows to add a sentence to describe the related object in the HAProxy HTML
3666 stats page. The description will be printed on the right of the object name
3667 it describes.
3668 No need to backslash spaces in the <string> arguments.
3669
3670
Willy Tarreau0ba27502007-12-24 16:55:16 +01003671disabled
3672 Disable a proxy, frontend or backend.
3673 May be used in sections : defaults | frontend | listen | backend
3674 yes | yes | yes | yes
3675 Arguments : none
3676
3677 The "disabled" keyword is used to disable an instance, mainly in order to
3678 liberate a listening port or to temporarily disable a service. The instance
3679 will still be created and its configuration will be checked, but it will be
3680 created in the "stopped" state and will appear as such in the statistics. It
3681 will not receive any traffic nor will it send any health-checks or logs. It
3682 is possible to disable many instances at once by adding the "disabled"
3683 keyword in a "defaults" section.
3684
3685 See also : "enabled"
3686
3687
Willy Tarreau5ce94572010-06-07 14:35:41 +02003688dispatch <address>:<port>
3689 Set a default server address
3690 May be used in sections : defaults | frontend | listen | backend
3691 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003692 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003693
3694 <address> is the IPv4 address of the default server. Alternatively, a
3695 resolvable hostname is supported, but this name will be resolved
3696 during start-up.
3697
3698 <ports> is a mandatory port specification. All connections will be sent
3699 to this port, and it is not permitted to use port offsets as is
3700 possible with normal servers.
3701
Willy Tarreau787aed52011-04-15 06:45:37 +02003702 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003703 server can take the connection. In the past it was used to forward non
3704 persistent connections to an auxiliary load balancer. Due to its simple
3705 syntax, it has also been used for simple TCP relays. It is recommended not to
3706 use it for more clarity, and to use the "server" directive instead.
3707
3708 See also : "server"
3709
Olivier Houchard4e694042017-03-14 20:01:29 +01003710
3711dynamic-cookie-key <string>
3712 Set the dynamic cookie secret key for a backend.
3713 May be used in sections : defaults | frontend | listen | backend
3714 yes | no | yes | yes
3715 Arguments : The secret key to be used.
3716
3717 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003718 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003719 specified on the "server" line), using a hash of the IP address of the
3720 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003721 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003722 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003723
Willy Tarreau0ba27502007-12-24 16:55:16 +01003724enabled
3725 Enable a proxy, frontend or backend.
3726 May be used in sections : defaults | frontend | listen | backend
3727 yes | yes | yes | yes
3728 Arguments : none
3729
3730 The "enabled" keyword is used to explicitly enable an instance, when the
3731 defaults has been set to "disabled". This is very rarely used.
3732
3733 See also : "disabled"
3734
3735
3736errorfile <code> <file>
3737 Return a file contents instead of errors generated by HAProxy
3738 May be used in sections : defaults | frontend | listen | backend
3739 yes | yes | yes | yes
3740 Arguments :
3741 <code> is the HTTP status code. Currently, HAProxy is capable of
Florian Tham9205fea2020-01-08 13:35:30 +01003742 generating codes 200, 400, 403, 404, 405, 408, 410, 425, 429, 500,
3743 502, 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003744
3745 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003746 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003747 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003748 error pages, and to use absolute paths, since files are read
3749 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003750
3751 It is important to understand that this keyword is not meant to rewrite
3752 errors returned by the server, but errors detected and returned by HAProxy.
3753 This is why the list of supported errors is limited to a small set.
3754
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003755 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3756
Willy Tarreau0ba27502007-12-24 16:55:16 +01003757 The files are returned verbatim on the TCP socket. This allows any trick such
3758 as redirections to another URL or site, as well as tricks to clean cookies,
3759 force enable or disable caching, etc... The package provides default error
3760 files returning the same contents as default errors.
3761
Willy Tarreau59140a22009-02-22 12:02:30 +01003762 The files should not exceed the configured buffer size (BUFSIZE), which
3763 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003764 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003765 loops between the client and HAProxy when all servers are down, causing an
3766 error to be returned instead of an image. For better HTTP compliance, it is
3767 recommended that all header lines end with CR-LF and not LF alone.
3768
Willy Tarreau0ba27502007-12-24 16:55:16 +01003769 The files are read at the same time as the configuration and kept in memory.
3770 For this reason, the errors continue to be returned even when the process is
3771 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003772 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003773 403 status code and interrogating a blocked URL.
3774
3775 See also : "errorloc", "errorloc302", "errorloc303"
3776
Willy Tarreau59140a22009-02-22 12:02:30 +01003777 Example :
3778 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003779 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003780 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3781 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3782
Willy Tarreau2769aa02007-12-27 18:26:09 +01003783
Christopher Faulet76edc0f2020-01-13 15:52:01 +01003784errorfiles <name> [<code> ...]
3785 Import, fully or partially, the error files defined in the <name> http-errors
3786 section.
3787 May be used in sections : defaults | frontend | listen | backend
3788 yes | yes | yes | yes
3789 Arguments :
3790 <name> is the name of an existing http-errors section.
3791
3792 <code> is a HTTP status code. Several status code may be listed.
3793 Currently, HAProxy is capable of generating codes 200, 400, 403,
3794 404, 405, 408, 410, 425, 429, 500, 502, 503, and 504.
3795
3796 Errors defined in the http-errors section with the name <name> are imported
3797 in the current proxy. If no status code is specified, all error files of the
3798 http-errors section are imported. Otherwise, only error files associated to
3799 the listed status code are imported. Those error files override the already
3800 defined custom errors for the proxy. And they may be overridden by following
3801 ones. Fonctionnly, it is exactly the same than declaring all error files by
3802 hand using "errorfile" directives.
3803
3804 See also : "errorfile", "errorloc", "errorloc302" , "errorloc303" and section
3805 3.8 about http-errors.
3806
3807 Example :
3808 errorfiles generic
3809 errorfiles site-1 403 404
3810
3811
Willy Tarreau2769aa02007-12-27 18:26:09 +01003812errorloc <code> <url>
3813errorloc302 <code> <url>
3814 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3815 May be used in sections : defaults | frontend | listen | backend
3816 yes | yes | yes | yes
3817 Arguments :
3818 <code> is the HTTP status code. Currently, HAProxy is capable of
Florian Tham9205fea2020-01-08 13:35:30 +01003819 generating codes 200, 400, 403, 404, 405, 408, 410, 425, 429, 500,
3820 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003821
3822 <url> it is the exact contents of the "Location" header. It may contain
3823 either a relative URI to an error page hosted on the same site,
3824 or an absolute URI designating an error page on another site.
3825 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003826 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003827
3828 It is important to understand that this keyword is not meant to rewrite
3829 errors returned by the server, but errors detected and returned by HAProxy.
3830 This is why the list of supported errors is limited to a small set.
3831
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003832 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3833
Willy Tarreau2769aa02007-12-27 18:26:09 +01003834 Note that both keyword return the HTTP 302 status code, which tells the
3835 client to fetch the designated URL using the same HTTP method. This can be
3836 quite problematic in case of non-GET methods such as POST, because the URL
3837 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003838 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003839 status code, indicating to the client that the URL must be fetched with a GET
3840 request.
3841
3842 See also : "errorfile", "errorloc303"
3843
3844
3845errorloc303 <code> <url>
3846 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3847 May be used in sections : defaults | frontend | listen | backend
3848 yes | yes | yes | yes
3849 Arguments :
3850 <code> is the HTTP status code. Currently, HAProxy is capable of
Florian Tham9205fea2020-01-08 13:35:30 +01003851 generating codes 200, 400, 403, 404, 405, 408, 410, 425, 429, 500,
3852 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003853
3854 <url> it is the exact contents of the "Location" header. It may contain
3855 either a relative URI to an error page hosted on the same site,
3856 or an absolute URI designating an error page on another site.
3857 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003858 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003859
3860 It is important to understand that this keyword is not meant to rewrite
3861 errors returned by the server, but errors detected and returned by HAProxy.
3862 This is why the list of supported errors is limited to a small set.
3863
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003864 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3865
Willy Tarreau2769aa02007-12-27 18:26:09 +01003866 Note that both keyword return the HTTP 303 status code, which tells the
3867 client to fetch the designated URL using the same HTTP GET method. This
3868 solves the usual problems associated with "errorloc" and the 302 code. It is
3869 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003870 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003871
3872 See also : "errorfile", "errorloc", "errorloc302"
3873
3874
Simon Horman51a1cf62015-02-03 13:00:44 +09003875email-alert from <emailaddr>
3876 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003877 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003878 May be used in sections: defaults | frontend | listen | backend
3879 yes | yes | yes | yes
3880
3881 Arguments :
3882
3883 <emailaddr> is the from email address to use when sending email alerts
3884
3885 Also requires "email-alert mailers" and "email-alert to" to be set
3886 and if so sending email alerts is enabled for the proxy.
3887
Simon Horman64e34162015-02-06 11:11:57 +09003888 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003889 "email-alert myhostname", "email-alert to", section 3.6 about
3890 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003891
3892
3893email-alert level <level>
3894 Declare the maximum log level of messages for which email alerts will be
3895 sent. This acts as a filter on the sending of email alerts.
3896 May be used in sections: defaults | frontend | listen | backend
3897 yes | yes | yes | yes
3898
3899 Arguments :
3900
3901 <level> One of the 8 syslog levels:
3902 emerg alert crit err warning notice info debug
3903 The above syslog levels are ordered from lowest to highest.
3904
3905 By default level is alert
3906
3907 Also requires "email-alert from", "email-alert mailers" and
3908 "email-alert to" to be set and if so sending email alerts is enabled
3909 for the proxy.
3910
Simon Horman1421e212015-04-30 13:10:35 +09003911 Alerts are sent when :
3912
3913 * An un-paused server is marked as down and <level> is alert or lower
3914 * A paused server is marked as down and <level> is notice or lower
3915 * A server is marked as up or enters the drain state and <level>
3916 is notice or lower
3917 * "option log-health-checks" is enabled, <level> is info or lower,
3918 and a health check status update occurs
3919
Simon Horman64e34162015-02-06 11:11:57 +09003920 See also : "email-alert from", "email-alert mailers",
3921 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003922 section 3.6 about mailers.
3923
3924
3925email-alert mailers <mailersect>
3926 Declare the mailers to be used when sending email alerts
3927 May be used in sections: defaults | frontend | listen | backend
3928 yes | yes | yes | yes
3929
3930 Arguments :
3931
3932 <mailersect> is the name of the mailers section to send email alerts.
3933
3934 Also requires "email-alert from" and "email-alert to" to be set
3935 and if so sending email alerts is enabled for the proxy.
3936
Simon Horman64e34162015-02-06 11:11:57 +09003937 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3938 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003939
3940
3941email-alert myhostname <hostname>
3942 Declare the to hostname address to be used when communicating with
3943 mailers.
3944 May be used in sections: defaults | frontend | listen | backend
3945 yes | yes | yes | yes
3946
3947 Arguments :
3948
Baptiste Assmann738bad92015-12-21 15:27:53 +01003949 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003950
3951 By default the systems hostname is used.
3952
3953 Also requires "email-alert from", "email-alert mailers" and
3954 "email-alert to" to be set and if so sending email alerts is enabled
3955 for the proxy.
3956
Simon Horman64e34162015-02-06 11:11:57 +09003957 See also : "email-alert from", "email-alert level", "email-alert mailers",
3958 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003959
3960
3961email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003962 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003963 header of email alerts. This is the address that email alerts are sent to.
3964 May be used in sections: defaults | frontend | listen | backend
3965 yes | yes | yes | yes
3966
3967 Arguments :
3968
3969 <emailaddr> is the to email address to use when sending email alerts
3970
3971 Also requires "email-alert mailers" and "email-alert to" to be set
3972 and if so sending email alerts is enabled for the proxy.
3973
Simon Horman64e34162015-02-06 11:11:57 +09003974 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003975 "email-alert myhostname", section 3.6 about mailers.
3976
3977
Willy Tarreau4de91492010-01-22 19:10:05 +01003978force-persist { if | unless } <condition>
3979 Declare a condition to force persistence on down servers
3980 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003981 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003982
3983 By default, requests are not dispatched to down servers. It is possible to
3984 force this using "option persist", but it is unconditional and redispatches
3985 to a valid server if "option redispatch" is set. That leaves with very little
3986 possibilities to force some requests to reach a server which is artificially
3987 marked down for maintenance operations.
3988
3989 The "force-persist" statement allows one to declare various ACL-based
3990 conditions which, when met, will cause a request to ignore the down status of
3991 a server and still try to connect to it. That makes it possible to start a
3992 server, still replying an error to the health checks, and run a specially
3993 configured browser to test the service. Among the handy methods, one could
3994 use a specific source IP address, or a specific cookie. The cookie also has
3995 the advantage that it can easily be added/removed on the browser from a test
3996 page. Once the service is validated, it is then possible to open the service
3997 to the world by returning a valid response to health checks.
3998
3999 The forced persistence is enabled when an "if" condition is met, or unless an
4000 "unless" condition is met. The final redispatch is always disabled when this
4001 is used.
4002
Cyril Bonté0d4bf012010-04-25 23:21:46 +02004003 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02004004 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01004005
Christopher Fauletc3fe5332016-04-07 15:30:10 +02004006
4007filter <name> [param*]
4008 Add the filter <name> in the filter list attached to the proxy.
4009 May be used in sections : defaults | frontend | listen | backend
4010 no | yes | yes | yes
4011 Arguments :
4012 <name> is the name of the filter. Officially supported filters are
4013 referenced in section 9.
4014
Tim Düsterhus4896c442016-11-29 02:15:19 +01004015 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02004016 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01004017 filter. Please refer to the documentation of the corresponding
4018 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02004019
4020 Multiple occurrences of the filter line can be used for the same proxy. The
4021 same filter can be referenced many times if needed.
4022
4023 Example:
4024 listen
4025 bind *:80
4026
4027 filter trace name BEFORE-HTTP-COMP
4028 filter compression
4029 filter trace name AFTER-HTTP-COMP
4030
4031 compression algo gzip
4032 compression offload
4033
4034 server srv1 192.168.0.1:80
4035
4036 See also : section 9.
4037
Willy Tarreau4de91492010-01-22 19:10:05 +01004038
Willy Tarreau2769aa02007-12-27 18:26:09 +01004039fullconn <conns>
4040 Specify at what backend load the servers will reach their maxconn
4041 May be used in sections : defaults | frontend | listen | backend
4042 yes | no | yes | yes
4043 Arguments :
4044 <conns> is the number of connections on the backend which will make the
4045 servers use the maximal number of connections.
4046
Willy Tarreau198a7442008-01-17 12:05:32 +01004047 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01004048 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01004049 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01004050 load. The server will then always accept at least <minconn> connections,
4051 never more than <maxconn>, and the limit will be on the ramp between both
4052 values when the backend has less than <conns> concurrent connections. This
4053 makes it possible to limit the load on the servers during normal loads, but
4054 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004055 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01004056
Willy Tarreaufbb78422011-06-05 15:38:35 +02004057 Since it's hard to get this value right, haproxy automatically sets it to
4058 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01004059 backend (based on "use_backend" and "default_backend" rules). That way it's
4060 safe to leave it unset. However, "use_backend" involving dynamic names are
4061 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02004062
Willy Tarreau2769aa02007-12-27 18:26:09 +01004063 Example :
4064 # The servers will accept between 100 and 1000 concurrent connections each
4065 # and the maximum of 1000 will be reached when the backend reaches 10000
4066 # connections.
4067 backend dynamic
4068 fullconn 10000
4069 server srv1 dyn1:80 minconn 100 maxconn 1000
4070 server srv2 dyn2:80 minconn 100 maxconn 1000
4071
4072 See also : "maxconn", "server"
4073
4074
4075grace <time>
4076 Maintain a proxy operational for some time after a soft stop
4077 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01004078 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01004079 Arguments :
4080 <time> is the time (by default in milliseconds) for which the instance
4081 will remain operational with the frontend sockets still listening
4082 when a soft-stop is received via the SIGUSR1 signal.
4083
4084 This may be used to ensure that the services disappear in a certain order.
4085 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004086 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01004087 needed by the equipment to detect the failure.
4088
4089 Note that currently, there is very little benefit in using this parameter,
4090 and it may in fact complicate the soft-reconfiguration process more than
4091 simplify it.
4092
Willy Tarreau0ba27502007-12-24 16:55:16 +01004093
Andrew Rodland17be45e2016-10-25 17:04:12 -04004094hash-balance-factor <factor>
4095 Specify the balancing factor for bounded-load consistent hashing
4096 May be used in sections : defaults | frontend | listen | backend
4097 yes | no | no | yes
4098 Arguments :
4099 <factor> is the control for the maximum number of concurrent requests to
4100 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01004101 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04004102
4103 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
4104 enables an algorithm that prevents any one server from getting too many
4105 requests at once, even if some hash buckets receive many more requests than
4106 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
4107 <factor> is a percentage greater than 100. For example, if <factor> is 150,
4108 then no server will be allowed to have a load more than 1.5 times the average.
4109 If server weights are used, they will be respected.
4110
4111 If the first-choice server is disqualified, the algorithm will choose another
4112 server based on the request hash, until a server with additional capacity is
4113 found. A higher <factor> allows more imbalance between the servers, while a
4114 lower <factor> means that more servers will be checked on average, affecting
4115 performance. Reasonable values are from 125 to 200.
4116
Willy Tarreau760e81d2018-05-03 07:20:40 +02004117 This setting is also used by "balance random" which internally relies on the
4118 consistent hashing mechanism.
4119
Andrew Rodland17be45e2016-10-25 17:04:12 -04004120 See also : "balance" and "hash-type".
4121
4122
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004123hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004124 Specify a method to use for mapping hashes to servers
4125 May be used in sections : defaults | frontend | listen | backend
4126 yes | no | yes | yes
4127 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04004128 <method> is the method used to select a server from the hash computed by
4129 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004130
Bhaskar98634f02013-10-29 23:30:51 -04004131 map-based the hash table is a static array containing all alive servers.
4132 The hashes will be very smooth, will consider weights, but
4133 will be static in that weight changes while a server is up
4134 will be ignored. This means that there will be no slow start.
4135 Also, since a server is selected by its position in the array,
4136 most mappings are changed when the server count changes. This
4137 means that when a server goes up or down, or when a server is
4138 added to a farm, most connections will be redistributed to
4139 different servers. This can be inconvenient with caches for
4140 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01004141
Bhaskar98634f02013-10-29 23:30:51 -04004142 consistent the hash table is a tree filled with many occurrences of each
4143 server. The hash key is looked up in the tree and the closest
4144 server is chosen. This hash is dynamic, it supports changing
4145 weights while the servers are up, so it is compatible with the
4146 slow start feature. It has the advantage that when a server
4147 goes up or down, only its associations are moved. When a
4148 server is added to the farm, only a few part of the mappings
4149 are redistributed, making it an ideal method for caches.
4150 However, due to its principle, the distribution will never be
4151 very smooth and it may sometimes be necessary to adjust a
4152 server's weight or its ID to get a more balanced distribution.
4153 In order to get the same distribution on multiple load
4154 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004155 same IDs. Note: consistent hash uses sdbm and avalanche if no
4156 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04004157
4158 <function> is the hash function to be used :
4159
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004160 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04004161 reimplementation of ndbm) database library. It was found to do
4162 well in scrambling bits, causing better distribution of the keys
4163 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004164 function with good distribution, unless the total server weight
4165 is a multiple of 64, in which case applying the avalanche
4166 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04004167
4168 djb2 this function was first proposed by Dan Bernstein many years ago
4169 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004170 function provides a better distribution than sdbm. It generally
4171 works well with text-based inputs though it can perform extremely
4172 poorly with numeric-only input or when the total server weight is
4173 a multiple of 33, unless the avalanche modifier is also used.
4174
Willy Tarreaua0f42712013-11-14 14:30:35 +01004175 wt6 this function was designed for haproxy while testing other
4176 functions in the past. It is not as smooth as the other ones, but
4177 is much less sensible to the input data set or to the number of
4178 servers. It can make sense as an alternative to sdbm+avalanche or
4179 djb2+avalanche for consistent hashing or when hashing on numeric
4180 data such as a source IP address or a visitor identifier in a URL
4181 parameter.
4182
Willy Tarreau324f07f2015-01-20 19:44:50 +01004183 crc32 this is the most common CRC32 implementation as used in Ethernet,
4184 gzip, PNG, etc. It is slower than the other ones but may provide
4185 a better distribution or less predictable results especially when
4186 used on strings.
4187
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004188 <modifier> indicates an optional method applied after hashing the key :
4189
4190 avalanche This directive indicates that the result from the hash
4191 function above should not be used in its raw form but that
4192 a 4-byte full avalanche hash must be applied first. The
4193 purpose of this step is to mix the resulting bits from the
4194 previous hash in order to avoid any undesired effect when
4195 the input contains some limited values or when the number of
4196 servers is a multiple of one of the hash's components (64
4197 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
4198 result less predictable, but it's also not as smooth as when
4199 using the original function. Some testing might be needed
4200 with some workloads. This hash is one of the many proposed
4201 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004202
Bhaskar98634f02013-10-29 23:30:51 -04004203 The default hash type is "map-based" and is recommended for most usages. The
4204 default function is "sdbm", the selection of a function should be based on
4205 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004206
Andrew Rodland17be45e2016-10-25 17:04:12 -04004207 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004208
4209
Christopher Faulet6d0c3df2020-01-22 09:26:35 +01004210http-after-response <action> <options...> [ { if | unless } <condition> ]
4211 Access control for all Layer 7 responses (server, applet/service and internal
4212 ones).
4213
4214 May be used in sections: defaults | frontend | listen | backend
4215 no | yes | yes | yes
4216
4217 The http-after-response statement defines a set of rules which apply to layer
4218 7 processing. The rules are evaluated in their declaration order when they
4219 are met in a frontend, listen or backend section. Any rule may optionally be
4220 followed by an ACL-based condition, in which case it will only be evaluated
4221 if the condition is true. Since these rules apply on responses, the backend
4222 rules are applied first, followed by the frontend's rules.
4223
4224 Unlike http-response rules, these ones are applied on all responses, the
4225 server ones but also to all responses generated by HAProxy. These rules are
4226 evaluated at the end of the responses analysis, before the data forwarding.
4227
4228 The first keyword is the rule's action. The supported actions are described
4229 below.
4230
4231 There is no limit to the number of http-after-response statements per
4232 instance.
4233
4234 Example:
4235 http-after-response set-header Strict-Transport-Security "max-age=31536000"
4236 http-after-response set-header Cache-Control "no-store,no-cache,private"
4237 http-after-response set-header Pragma "no-cache"
4238
4239http-after-response add-header <name> <fmt> [ { if | unless } <condition> ]
4240
4241 This appends an HTTP header field whose name is specified in <name> and whose
4242 value is defined by <fmt> which follows the log-format rules (see Custom Log
4243 Format in section 8.2.4). This may be used to send a cookie to a client for
4244 example, or to pass some internal information.
4245 This rule is not final, so it is possible to add other similar rules.
4246 Note that header addition is performed immediately, so one rule might reuse
4247 the resulting header from a previous rule.
4248
4249http-after-response allow [ { if | unless } <condition> ]
4250
4251 This stops the evaluation of the rules and lets the response pass the check.
4252 No further "http-after-response" rules are evaluated.
4253
4254http-after-response del-header <name> [ { if | unless } <condition> ]
4255
4256 This removes all HTTP header fields whose name is specified in <name>.
4257
4258http-after-response replace-header <name> <regex-match> <replace-fmt>
4259 [ { if | unless } <condition> ]
4260
4261 This works like "http-response replace-header".
4262
4263 Example:
4264 http-after-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
4265
4266 # applied to:
4267 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
4268
4269 # outputs:
4270 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
4271
4272 # assuming the backend IP is 192.168.1.20.
4273
4274http-after-response replace-value <name> <regex-match> <replace-fmt>
4275 [ { if | unless } <condition> ]
4276
4277 This works like "http-response replace-value".
4278
4279 Example:
4280 http-after-response replace-value Cache-control ^public$ private
4281
4282 # applied to:
4283 Cache-Control: max-age=3600, public
4284
4285 # outputs:
4286 Cache-Control: max-age=3600, private
4287
4288http-after-response set-header <name> <fmt> [ { if | unless } <condition> ]
4289
4290 This does the same as "add-header" except that the header name is first
4291 removed if it existed. This is useful when passing security information to
4292 the server, where the header must not be manipulated by external users.
4293
4294http-after-response set-status <status> [reason <str>]
4295 [ { if | unless } <condition> ]
4296
4297 This replaces the response status code with <status> which must be an integer
4298 between 100 and 999. Optionally, a custom reason text can be provided defined
4299 by <str>, or the default reason for the specified code will be used as a
4300 fallback.
4301
4302 Example:
4303 # return "431 Request Header Fields Too Large"
4304 http-response set-status 431
4305 # return "503 Slow Down", custom reason
4306 http-response set-status 503 reason "Slow Down"
4307
4308http-after-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
4309
4310 This is used to set the contents of a variable. The variable is declared
4311 inline.
4312
4313 Arguments:
4314 <var-name> The name of the variable starts with an indication about its
4315 scope. The scopes allowed are:
4316 "proc" : the variable is shared with the whole process
4317 "sess" : the variable is shared with the whole session
4318 "txn" : the variable is shared with the transaction
4319 (request and response)
4320 "req" : the variable is shared only during request
4321 processing
4322 "res" : the variable is shared only during response
4323 processing
4324 This prefix is followed by a name. The separator is a '.'.
4325 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
4326 and '_'.
4327
4328 <expr> Is a standard HAProxy expression formed by a sample-fetch
4329 followed by some converters.
4330
4331 Example:
4332 http-after-response set-var(sess.last_redir) res.hdr(location)
4333
4334http-after-response strict-mode { on | off }
4335
4336 This enables or disables the strict rewriting mode for following rules. It
4337 does not affect rules declared before it and it is only applicable on rules
4338 performing a rewrite on the responses. When the strict mode is enabled, any
4339 rewrite failure triggers an internal error. Otherwise, such errors are
4340 silently ignored. The purpose of the strict rewriting mode is to make some
4341 rewrites optionnal while others must be performed to continue the response
4342 processing.
4343
4344 By default, the strict rewriting mode is enabled. Its value is also reset
4345 when a ruleset evaluation ends. So, for instance, if you change the mode on
4346 the bacnkend, the default mode is restored when HAProxy starts the frontend
4347 rules evaluation.
4348
4349http-after-response unset-var(<var-name>) [ { if | unless } <condition> ]
4350
4351 This is used to unset a variable. See "http-after-response set-var" for
4352 details about <var-name>.
4353
4354 Example:
4355 http-after-response unset-var(sess.last_redir)
4356
4357
Willy Tarreau0ba27502007-12-24 16:55:16 +01004358http-check disable-on-404
4359 Enable a maintenance mode upon HTTP/404 response to health-checks
4360 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01004361 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01004362 Arguments : none
4363
4364 When this option is set, a server which returns an HTTP code 404 will be
4365 excluded from further load-balancing, but will still receive persistent
4366 connections. This provides a very convenient method for Web administrators
4367 to perform a graceful shutdown of their servers. It is also important to note
4368 that a server which is detected as failed while it was in this mode will not
4369 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
4370 will immediately be reinserted into the farm. The status on the stats page
4371 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01004372 option only works in conjunction with the "httpchk" option. If this option
4373 is used with "http-check expect", then it has precedence over it so that 404
4374 responses will still be considered as soft-stop.
4375
4376 See also : "option httpchk", "http-check expect"
4377
4378
4379http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004380 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01004381 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02004382 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01004383 Arguments :
4384 <match> is a keyword indicating how to look for a specific pattern in the
4385 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004386 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01004387 exclamation mark ("!") to negate the match. Spaces are allowed
4388 between the exclamation mark and the keyword. See below for more
4389 details on the supported keywords.
4390
4391 <pattern> is the pattern to look for. It may be a string or a regular
4392 expression. If the pattern contains spaces, they must be escaped
4393 with the usual backslash ('\').
4394
4395 By default, "option httpchk" considers that response statuses 2xx and 3xx
4396 are valid, and that others are invalid. When "http-check expect" is used,
4397 it defines what is considered valid or invalid. Only one "http-check"
4398 statement is supported in a backend. If a server fails to respond or times
4399 out, the check obviously fails. The available matches are :
4400
4401 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004402 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004403 response's status code is exactly this string. If the
4404 "status" keyword is prefixed with "!", then the response
4405 will be considered invalid if the status code matches.
4406
4407 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004408 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004409 response's status code matches the expression. If the
4410 "rstatus" keyword is prefixed with "!", then the response
4411 will be considered invalid if the status code matches.
4412 This is mostly used to check for multiple codes.
4413
4414 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004415 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004416 response's body contains this exact string. If the
4417 "string" keyword is prefixed with "!", then the response
4418 will be considered invalid if the body contains this
4419 string. This can be used to look for a mandatory word at
4420 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004421 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01004422 trace).
4423
4424 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004425 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004426 response's body matches this expression. If the "rstring"
4427 keyword is prefixed with "!", then the response will be
4428 considered invalid if the body matches the expression.
4429 This can be used to look for a mandatory word at the end
4430 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01004431 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01004432
4433 It is important to note that the responses will be limited to a certain size
4434 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
4435 Thus, too large responses may not contain the mandatory pattern when using
4436 "string" or "rstring". If a large response is absolutely required, it is
4437 possible to change the default max size by setting the global variable.
4438 However, it is worth keeping in mind that parsing very large responses can
4439 waste some CPU cycles, especially when regular expressions are used, and that
4440 it is always better to focus the checks on smaller resources.
4441
Cyril Bonté32602d22015-01-30 00:07:07 +01004442 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
4443 will automatically append a "Connection: close" header, meaning that this
4444 header should not be present in the request provided by "option httpchk".
4445
Willy Tarreaubd741542010-03-16 18:46:54 +01004446 Last, if "http-check expect" is combined with "http-check disable-on-404",
4447 then this last one has precedence when the server responds with 404.
4448
4449 Examples :
4450 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004451 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01004452
4453 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004454 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01004455
4456 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004457 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01004458
4459 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004460 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01004461
Willy Tarreaubd741542010-03-16 18:46:54 +01004462 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004463
4464
Willy Tarreauef781042010-01-27 11:53:01 +01004465http-check send-state
4466 Enable emission of a state header with HTTP health checks
4467 May be used in sections : defaults | frontend | listen | backend
4468 yes | no | yes | yes
4469 Arguments : none
4470
4471 When this option is set, haproxy will systematically send a special header
4472 "X-Haproxy-Server-State" with a list of parameters indicating to each server
4473 how they are seen by haproxy. This can be used for instance when a server is
4474 manipulated without access to haproxy and the operator needs to know whether
4475 haproxy still sees it up or not, or if the server is the last one in a farm.
4476
4477 The header is composed of fields delimited by semi-colons, the first of which
4478 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
4479 checks on the total number before transition, just as appears in the stats
4480 interface. Next headers are in the form "<variable>=<value>", indicating in
4481 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08004482 - a variable "address", containing the address of the backend server.
4483 This corresponds to the <address> field in the server declaration. For
4484 unix domain sockets, it will read "unix".
4485
4486 - a variable "port", containing the port of the backend server. This
4487 corresponds to the <port> field in the server declaration. For unix
4488 domain sockets, it will read "unix".
4489
Willy Tarreauef781042010-01-27 11:53:01 +01004490 - a variable "name", containing the name of the backend followed by a slash
4491 ("/") then the name of the server. This can be used when a server is
4492 checked in multiple backends.
4493
4494 - a variable "node" containing the name of the haproxy node, as set in the
4495 global "node" variable, otherwise the system's hostname if unspecified.
4496
4497 - a variable "weight" indicating the weight of the server, a slash ("/")
4498 and the total weight of the farm (just counting usable servers). This
4499 helps to know if other servers are available to handle the load when this
4500 one fails.
4501
4502 - a variable "scur" indicating the current number of concurrent connections
4503 on the server, followed by a slash ("/") then the total number of
4504 connections on all servers of the same backend.
4505
4506 - a variable "qcur" indicating the current number of requests in the
4507 server's queue.
4508
4509 Example of a header received by the application server :
4510 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
4511 scur=13/22; qcur=0
4512
4513 See also : "option httpchk", "http-check disable-on-404"
4514
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004515
4516http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004517 Access control for Layer 7 requests
4518
4519 May be used in sections: defaults | frontend | listen | backend
4520 no | yes | yes | yes
4521
Willy Tarreau20b0de52012-12-24 15:45:22 +01004522 The http-request statement defines a set of rules which apply to layer 7
4523 processing. The rules are evaluated in their declaration order when they are
4524 met in a frontend, listen or backend section. Any rule may optionally be
4525 followed by an ACL-based condition, in which case it will only be evaluated
4526 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004527
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004528 The first keyword is the rule's action. The supported actions are described
4529 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004530
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004531 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004532
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004533 Example:
4534 acl nagios src 192.168.129.3
4535 acl local_net src 192.168.0.0/16
4536 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01004537
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004538 http-request allow if nagios
4539 http-request allow if local_net auth_ok
4540 http-request auth realm Gimme if local_net auth_ok
4541 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01004542
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004543 Example:
4544 acl key req.hdr(X-Add-Acl-Key) -m found
4545 acl add path /addacl
4546 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01004547
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004548 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01004549
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004550 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4551 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004552
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004553 Example:
4554 acl value req.hdr(X-Value) -m found
4555 acl setmap path /setmap
4556 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004557
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004558 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004559
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004560 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4561 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004562
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004563 See also : "stats http-request", section 3.4 about userlists and section 7
4564 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004565
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004566http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004567
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004568 This is used to add a new entry into an ACL. The ACL must be loaded from a
4569 file (even a dummy empty file). The file name of the ACL to be updated is
4570 passed between parentheses. It takes one argument: <key fmt>, which follows
4571 log-format rules, to collect content of the new entry. It performs a lookup
4572 in the ACL before insertion, to avoid duplicated (or more) values. This
4573 lookup is done by a linear search and can be expensive with large lists!
4574 It is the equivalent of the "add acl" command from the stats socket, but can
4575 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06004576
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004577http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004578
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004579 This appends an HTTP header field whose name is specified in <name> and
4580 whose value is defined by <fmt> which follows the log-format rules (see
4581 Custom Log Format in section 8.2.4). This is particularly useful to pass
4582 connection-specific information to the server (e.g. the client's SSL
4583 certificate), or to combine several headers into one. This rule is not
4584 final, so it is possible to add other similar rules. Note that header
4585 addition is performed immediately, so one rule might reuse the resulting
4586 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004587
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004588http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004589
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004590 This stops the evaluation of the rules and lets the request pass the check.
4591 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06004592
Sasha Pachev218f0642014-06-16 12:05:59 -06004593
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004594http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004595
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004596 This stops the evaluation of the rules and immediately responds with an
4597 HTTP 401 or 407 error code to invite the user to present a valid user name
4598 and password. No further "http-request" rules are evaluated. An optional
4599 "realm" parameter is supported, it sets the authentication realm that is
4600 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06004601
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004602 Example:
4603 acl auth_ok http_auth_group(L1) G1
4604 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06004605
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004606http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004607
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02004608 See section 6.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004609
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004610http-request capture <sample> [ len <length> | id <id> ]
4611 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004612
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004613 This captures sample expression <sample> from the request buffer, and
4614 converts it to a string of at most <len> characters. The resulting string is
4615 stored into the next request "capture" slot, so it will possibly appear next
4616 to some captured HTTP headers. It will then automatically appear in the logs,
4617 and it will be possible to extract it using sample fetch rules to feed it
4618 into headers or anything. The length should be limited given that this size
4619 will be allocated for each capture during the whole session life.
4620 Please check section 7.3 (Fetching samples) and "capture request header" for
4621 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004622
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004623 If the keyword "id" is used instead of "len", the action tries to store the
4624 captured string in a previously declared capture slot. This is useful to run
4625 captures in backends. The slot id can be declared by a previous directive
Baptiste Assmann19a69b32020-01-16 14:34:22 +01004626 "http-request capture" or with the "declare capture" keyword.
4627
4628 When using this action in a backend, double check that the relevant
4629 frontend(s) have the required capture slots otherwise, this rule will be
4630 ignored at run time. This can't be detected at configuration parsing time
4631 due to HAProxy's ability to dynamically resolve backend name at runtime.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004632
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004633http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004634
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004635 This is used to delete an entry from an ACL. The ACL must be loaded from a
4636 file (even a dummy empty file). The file name of the ACL to be updated is
4637 passed between parentheses. It takes one argument: <key fmt>, which follows
4638 log-format rules, to collect content of the entry to delete.
4639 It is the equivalent of the "del acl" command from the stats socket, but can
4640 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004641
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004642http-request del-header <name> [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004643
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004644 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004645
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004646http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004647
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004648 This is used to delete an entry from a MAP. The MAP must be loaded from a
4649 file (even a dummy empty file). The file name of the MAP to be updated is
4650 passed between parentheses. It takes one argument: <key fmt>, which follows
4651 log-format rules, to collect content of the entry to delete.
4652 It takes one argument: "file name" It is the equivalent of the "del map"
4653 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02004654
Christopher Faulet554c0eb2020-01-14 12:00:28 +01004655http-request deny [deny_status <status>] [ { errorfile | errorfiles } <err> ]
4656 [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04004657
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004658 This stops the evaluation of the rules and immediately rejects the request
4659 and emits an HTTP 403 error, or optionally the status code specified as an
4660 argument to "deny_status". The list of permitted status codes is limited to
Christopher Faulet554c0eb2020-01-14 12:00:28 +01004661 those that can be overridden by the "errorfile" directive. A specific error
4662 message may be specified. It may be an error file, using the "errorfile"
4663 keyword followed by the file containing the full HTTP response. It may also
4664 be an error from an http-errors section, using the "errorfiles" keyword
4665 followed by the section name.
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004666 No further "http-request" rules are evaluated.
Patrick Hemmer268a7072018-05-11 12:52:31 -04004667
Olivier Houchard602bf7d2019-05-10 13:59:15 +02004668http-request disable-l7-retry [ { if | unless } <condition> ]
4669 This disables any attempt to retry the request if it fails for any other
4670 reason than a connection failure. This can be useful for example to make
4671 sure POST requests aren't retried on failure.
4672
Baptiste Assmann333939c2019-01-21 08:34:50 +01004673http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
4674
4675 This action performs a DNS resolution of the output of <expr> and stores
4676 the result in the variable <var>. It uses the DNS resolvers section
4677 pointed by <resolvers>.
4678 It is possible to choose a resolution preference using the optional
4679 arguments 'ipv4' or 'ipv6'.
4680 When performing the DNS resolution, the client side connection is on
4681 pause waiting till the end of the resolution.
4682 If an IP address can be found, it is stored into <var>. If any kind of
4683 error occurs, then <var> is not set.
4684 One can use this action to discover a server IP address at run time and
4685 based on information found in the request (IE a Host header).
4686 If this action is used to find the server's IP address (using the
4687 "set-dst" action), then the server IP address in the backend must be set
4688 to 0.0.0.0.
4689
4690 Example:
4691 resolvers mydns
4692 nameserver local 127.0.0.53:53
4693 nameserver google 8.8.8.8:53
4694 timeout retry 1s
4695 hold valid 10s
4696 hold nx 3s
4697 hold other 3s
4698 hold obsolete 0s
4699 accepted_payload_size 8192
4700
4701 frontend fe
4702 bind 10.42.0.1:80
4703 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
4704 http-request capture var(txn.myip) len 40
4705
4706 # return 503 when the variable is not set,
4707 # which mean DNS resolution error
4708 use_backend b_503 unless { var(txn.myip) -m found }
4709
4710 default_backend be
4711
4712 backend b_503
4713 # dummy backend used to return 503.
4714 # one can use the errorfile directive to send a nice
4715 # 503 error page to end users
4716
4717 backend be
4718 # rule to prevent HAProxy from reconnecting to services
4719 # on the local network (forged DNS name used to scan the network)
4720 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
4721 http-request set-dst var(txn.myip)
4722 server clear 0.0.0.0:0
4723
4724 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
4725 be used to scan the network or worst won't loop over itself...
4726
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004727http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
4728
4729 This is used to build an HTTP 103 Early Hints response prior to any other one.
4730 This appends an HTTP header field to this response whose name is specified in
4731 <name> and whose value is defined by <fmt> which follows the log-format rules
4732 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01004733 to the client some Link headers to preload resources required to render the
4734 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004735
4736 See RFC 8297 for more information.
4737
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004738http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004739
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004740 This performs an HTTP redirection based on a redirect rule. This is exactly
4741 the same as the "redirect" statement except that it inserts a redirect rule
4742 which can be processed in the middle of other "http-request" rules and that
4743 these rules use the "log-format" strings. See the "redirect" keyword for the
4744 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004745
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004746http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004747
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004748 This stops the evaluation of the rules and immediately closes the connection
4749 without sending any response. It acts similarly to the
4750 "tcp-request content reject" rules. It can be useful to force an immediate
4751 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004752
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004753http-request replace-header <name> <match-regex> <replace-fmt>
4754 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02004755
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004756 This matches the value of all occurences of header field <name> against
4757 <match-regex>. Matching is performed case-sensitively. Matching values are
4758 completely replaced by <replace-fmt>. Format characters are allowed in
4759 <replace-fmt> and work like <fmt> arguments in "http-request add-header".
4760 Standard back-references using the backslash ('\') followed by a number are
4761 supported.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004762
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004763 This action acts on whole header lines, regardless of the number of values
4764 they may contain. Thus it is well-suited to process headers naturally
4765 containing commas in their value, such as If-Modified-Since. Headers that
4766 contain a comma-separated list of values, such as Accept, should be processed
4767 using "http-request replace-value".
William Lallemand86d0df02017-11-24 21:36:45 +01004768
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004769 Example:
4770 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
4771
4772 # applied to:
4773 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4774
4775 # outputs:
4776 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4777
4778 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02004779
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004780 http-request replace-header User-Agent curl foo
4781
4782 # applied to:
4783 User-Agent: curl/7.47.0
Willy Tarreau09448f72014-06-25 18:12:15 +02004784
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004785 # outputs:
4786 User-Agent: foo
Willy Tarreau09448f72014-06-25 18:12:15 +02004787
Willy Tarreau262c3f12019-12-17 06:52:51 +01004788http-request replace-path <match-regex> <replace-fmt>
4789 [ { if | unless } <condition> ]
4790
4791 This works like "replace-header" except that it works on the request's path
4792 component instead of a header. The path component starts at the first '/'
4793 after an optional scheme+authority. It does contain the query string if any
4794 is present. The replacement does not modify the scheme nor authority.
4795
4796 It is worth noting that regular expressions may be more expensive to evaluate
4797 than certain ACLs, so rare replacements may benefit from a condition to avoid
4798 performing the evaluation at all if it does not match.
4799
4800 Example:
4801 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4802 http-request replace-path (.*) /foo\1
4803
4804 # suffix /foo : turn /bar?q=1 into /bar/foo?q=1 :
4805 http-request replace-path ([^?]*)(\?(.*))? \1/foo\2
4806
4807 # strip /foo : turn /foo/bar?q=1 into /bar?q=1
4808 http-request replace-path /foo/(.*) /\1
4809 # or more efficient if only some requests match :
4810 http-request replace-path /foo/(.*) /\1 if { url_beg /foo/ }
4811
Willy Tarreau33810222019-06-12 17:44:02 +02004812http-request replace-uri <match-regex> <replace-fmt>
4813 [ { if | unless } <condition> ]
4814
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004815 This works like "replace-header" except that it works on the request's URI part
4816 instead of a header. The URI part may contain an optional scheme, authority or
4817 query string. These are considered to be part of the value that is matched
4818 against.
4819
4820 It is worth noting that regular expressions may be more expensive to evaluate
4821 than certain ACLs, so rare replacements may benefit from a condition to avoid
4822 performing the evaluation at all if it does not match.
Willy Tarreau33810222019-06-12 17:44:02 +02004823
Willy Tarreau62b59132019-12-17 06:51:20 +01004824 IMPORTANT NOTE: historically in HTTP/1.x, the vast majority of requests sent
4825 by browsers use the "origin form", which differs from the "absolute form" in
4826 that they do not contain a scheme nor authority in the URI portion. Mostly
4827 only requests sent to proxies, those forged by hand and some emitted by
4828 certain applications use the absolute form. As such, "replace-uri" usually
4829 works fine most of the time in HTTP/1.x with rules starting with a "/". But
4830 with HTTP/2, clients are encouraged to send absolute URIs only, which look
4831 like the ones HTTP/1 clients use to talk to proxies. Such partial replace-uri
4832 rules may then fail in HTTP/2 when they work in HTTP/1. Either the rules need
Willy Tarreau262c3f12019-12-17 06:52:51 +01004833 to be adapted to optionally match a scheme and authority, or replace-path
4834 should be used.
Willy Tarreau33810222019-06-12 17:44:02 +02004835
Willy Tarreau62b59132019-12-17 06:51:20 +01004836 Example:
4837 # rewrite all "http" absolute requests to "https":
4838 http-request replace-uri ^http://(.*) https://\1
Willy Tarreau33810222019-06-12 17:44:02 +02004839
Willy Tarreau62b59132019-12-17 06:51:20 +01004840 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4841 http-request replace-uri ([^/:]*://[^/]*)?(.*) \1/foo\2
Willy Tarreau33810222019-06-12 17:44:02 +02004842
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004843http-request replace-value <name> <match-regex> <replace-fmt>
4844 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02004845
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004846 This works like "replace-header" except that it matches the regex against
4847 every comma-delimited value of the header field <name> instead of the
4848 entire header. This is suited for all headers which are allowed to carry
4849 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02004850
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004851 Example:
4852 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004853
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004854 # applied to:
4855 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004856
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004857 # outputs:
4858 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004859
Christopher Faulet24231ab2020-01-24 17:44:23 +01004860http-request return [status <code>] [content-type <type>]
4861 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
4862 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
Christopher Faulet4a2c1422020-01-31 17:36:01 +01004863 [ hdr <name> <fmt> ]*
Christopher Faulet24231ab2020-01-24 17:44:23 +01004864 [ { if | unless } <condition> ]
4865
4866 This stops the evaluation of the rules and immediatly returns a response. The
4867 default status code used for the response is 200. It can be optionally
4868 specified as an arguments to "status". The response content-type may also be
4869 specified as an argument to "content-type". Finally the response itselft may
4870 be defined. If can be a full HTTP response specifying the errorfile to use,
4871 or the response payload specifing the file or the string to use. These rules
4872 are followed to create the response :
4873
4874 * If neither the errorfile nor the payload to use is defined, a dummy
4875 response is returned. Only the "status" argument is considered. It can be
4876 any code in the range [200, 599]. The "content-type" argument, if any, is
4877 ignored.
4878
4879 * If "default-errorfiles" argument is set, the proxy's errorfiles are
4880 considered. If the "status" argument is defined, it must be one of the
4881 status code handled by hparoxy (200, 400, 403, 404, 405, 408, 410, 425,
4882 429, 500, 502, 503, and 504). The "content-type" argument, if any, is
4883 ignored.
4884
4885 * If a specific errorfile is defined, with an "errorfile" argument, the
4886 corresponding file, containing a full HTTP response, is returned. Only the
4887 "status" argument is considered. It must be one of the status code handled
4888 by hparoxy (200, 400, 403, 404, 405, 408, 410, 425, 429, 500, 502, 503, and
4889 504). The "content-type" argument, if any, is ignored.
4890
4891 * If an http-errors section is defined, with an "errorfiles" argument, the
4892 corresponding file in the specified http-errors section, containing a full
4893 HTTP response, is returned. Only the "status" argument is considered. It
4894 must be one of the status code handled by hparoxy (200, 400, 403, 404, 405,
4895 408, 410, 425, 429, 500, 502, 503, and 504). The "content-type" argument,
4896 if any, is ignored.
4897
4898 * If a "file" or a "lf-file" argument is specified, the file's content is
4899 used as the response payload. If the file is not empty, its content-type
4900 must be set as argument to "content-type". Otherwise, any "content-type"
4901 argument is ignored. With a "lf-file" argument, the file's content is
4902 evaluated as a log-format string. With a "file" argument, it is considered
4903 as a raw content.
4904
4905 * If a "string" or "lf-string" argument is specified, the defined string is
4906 used as the response payload. The content-type must always be set as
4907 argument to "content-type". With a "lf-string" argument, the string is
4908 evaluated as a log-format string. With a "string" argument, it is
4909 considered as a raw string.
4910
Christopher Faulet4a2c1422020-01-31 17:36:01 +01004911 When the response is not based an errorfile, it is possible to appends HTTP
4912 header fields to the response using "hdr" arguments. Otherwise, all "hdr"
4913 arguments are ignored. For each one, the header name is specified in <name>
4914 and its value is defined by <fmt> which follows the log-format rules.
4915
Christopher Faulet24231ab2020-01-24 17:44:23 +01004916 Note that the generated response must be smaller than a buffer. And to avoid
4917 any warning, when an errorfile or a raw file is loaded, the buffer space
4918 reserved to the headers rewritting should also be free.
4919
4920 No further "http-request" rules are evaluated.
4921
4922 Example:
4923 http-request return errorfile /etc/haproy/errorfiles/200.http \
4924 if { path /ping }
4925
4926 http-request return content-type image/x-icon file /var/www/favicon.ico \
4927 if { path /favicon.ico }
4928
4929 http-request return status 403 content-type text/plain \
4930 lf-string "Access denied. IP %[src] is blacklisted." \
4931 if { src -f /etc/haproxy/blacklist.lst }
4932
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004933http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4934http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004935
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004936 This actions increments the GPC0 or GPC1 counter according with the sticky
4937 counter designated by <sc-id>. If an error occurs, this action silently fails
4938 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004939
Cédric Dufour0d7712d2019-11-06 18:38:53 +01004940http-request sc-set-gpt0(<sc-id>) { <int> | <expr> }
4941 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004942
Cédric Dufour0d7712d2019-11-06 18:38:53 +01004943 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
4944 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
4945 boolean. If an error occurs, this action silently fails and the actions
4946 evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004947
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004948http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004949
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004950 This is used to set the destination IP address to the value of specified
4951 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
4952 but provides the correct IP in a HTTP header; or you want to mask the IP for
4953 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
4954 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004955
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004956 Arguments:
4957 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4958 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004959
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004960 Example:
4961 http-request set-dst hdr(x-dst)
4962 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01004963
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004964 When possible, set-dst preserves the original destination port as long as the
4965 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004966
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004967http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004968
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004969 This is used to set the destination port address to the value of specified
4970 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
4971 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004972
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004973 Arguments:
4974 <expr> Is a standard HAProxy expression formed by a sample-fetch
4975 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004976
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004977 Example:
4978 http-request set-dst-port hdr(x-port)
4979 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004980
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004981 When possible, set-dst-port preserves the original destination address as
4982 long as the address family supports a port, otherwise it forces the
4983 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004984
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004985http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004986
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004987 This does the same as "http-request add-header" except that the header name
4988 is first removed if it existed. This is useful when passing security
4989 information to the server, where the header must not be manipulated by
4990 external users. Note that the new value is computed before the removal so it
4991 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02004992
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004993 Example:
4994 http-request set-header X-Haproxy-Current-Date %T
4995 http-request set-header X-SSL %[ssl_fc]
4996 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
4997 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4998 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4999 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
5000 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
5001 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
5002 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02005003
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005004http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02005005
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005006 This is used to change the log level of the current request when a certain
5007 condition is met. Valid levels are the 8 syslog levels (see the "log"
5008 keyword) plus the special level "silent" which disables logging for this
5009 request. This rule is not final so the last matching rule wins. This rule
5010 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02005011
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005012http-request set-map(<file-name>) <key fmt> <value fmt>
5013 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02005014
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005015 This is used to add a new entry into a MAP. The MAP must be loaded from a
5016 file (even a dummy empty file). The file name of the MAP to be updated is
5017 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5018 log-format rules, used to collect MAP key, and <value fmt>, which follows
5019 log-format rules, used to collect content for the new entry.
5020 It performs a lookup in the MAP before insertion, to avoid duplicated (or
5021 more) values. This lookup is done by a linear search and can be expensive
5022 with large lists! It is the equivalent of the "set map" command from the
5023 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02005024
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005025http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02005026
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005027 This is used to set the Netfilter MARK on all packets sent to the client to
5028 the value passed in <mark> on platforms which support it. This value is an
5029 unsigned 32 bit value which can be matched by netfilter and by the routing
5030 table. It can be expressed both in decimal or hexadecimal format (prefixed by
5031 "0x"). This can be useful to force certain packets to take a different route
5032 (for example a cheaper network path for bulk downloads). This works on Linux
5033 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02005034
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005035http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02005036
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005037 This rewrites the request method with the result of the evaluation of format
5038 string <fmt>. There should be very few valid reasons for having to do so as
5039 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02005040
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005041http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02005042
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005043 This sets the "nice" factor of the current request being processed. It only
5044 has effect against the other requests being processed at the same time.
5045 The default value is 0, unless altered by the "nice" setting on the "bind"
5046 line. The accepted range is -1024..1024. The higher the value, the nicest
5047 the request will be. Lower values will make the request more important than
5048 other ones. This can be useful to improve the speed of some requests, or
5049 lower the priority of non-important requests. Using this setting without
5050 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02005051
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005052http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02005053
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005054 This rewrites the request path with the result of the evaluation of format
5055 string <fmt>. The query string, if any, is left intact. If a scheme and
5056 authority is found before the path, they are left intact as well. If the
5057 request doesn't have a path ("*"), this one is replaced with the format.
5058 This can be used to prepend a directory component in front of a path for
5059 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02005060
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005061 Example :
5062 # prepend the host name before the path
5063 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005064
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005065http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02005066
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005067 This is used to set the queue priority class of the current request.
5068 The value must be a sample expression which converts to an integer in the
5069 range -2047..2047. Results outside this range will be truncated.
5070 The priority class determines the order in which queued requests are
5071 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005072
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005073http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005074
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005075 This is used to set the queue priority timestamp offset of the current
5076 request. The value must be a sample expression which converts to an integer
5077 in the range -524287..524287. Results outside this range will be truncated.
5078 When a request is queued, it is ordered first by the priority class, then by
5079 the current timestamp adjusted by the given offset in milliseconds. Lower
5080 values have higher priority.
5081 Note that the resulting timestamp is is only tracked with enough precision
5082 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
5083 adjusted timestamp exceeds this value, it will be misidentified as highest
5084 priority. Thus it is important to set "timeout queue" to a value, where when
5085 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005086
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005087http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01005088
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005089 This rewrites the request's query string which appears after the first
5090 question mark ("?") with the result of the evaluation of format string <fmt>.
5091 The part prior to the question mark is left intact. If the request doesn't
5092 contain a question mark and the new value is not empty, then one is added at
5093 the end of the URI, followed by the new value. If a question mark was
5094 present, it will never be removed even if the value is empty. This can be
5095 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005096
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005097 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005098
5099 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005100 # replace "%3D" with "=" in the query string
5101 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005102
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005103http-request set-src <expr> [ { if | unless } <condition> ]
5104 This is used to set the source IP address to the value of specified
5105 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
5106 provides the correct IP in a HTTP header; or you want to mask source IP for
5107 privacy.
5108
5109 Arguments :
5110 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
5111 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005112
Cyril Bonté78caf842010-03-10 22:41:43 +01005113 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005114 http-request set-src hdr(x-forwarded-for)
5115 http-request set-src src,ipmask(24)
5116
5117 When possible, set-src preserves the original source port as long as the
5118 address family allows it, otherwise the source port is set to 0.
5119
5120http-request set-src-port <expr> [ { if | unless } <condition> ]
5121
5122 This is used to set the source port address to the value of specified
5123 expression.
5124
5125 Arguments:
5126 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
5127 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01005128
Willy Tarreau20b0de52012-12-24 15:45:22 +01005129 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005130 http-request set-src-port hdr(x-port)
5131 http-request set-src-port int(4000)
5132
5133 When possible, set-src-port preserves the original source address as long as
5134 the address family supports a port, otherwise it forces the source address to
5135 IPv4 "0.0.0.0" before rewriting the port.
5136
5137http-request set-tos <tos> [ { if | unless } <condition> ]
5138
5139 This is used to set the TOS or DSCP field value of packets sent to the client
5140 to the value passed in <tos> on platforms which support this. This value
5141 represents the whole 8 bits of the IP TOS field, and can be expressed both in
5142 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
5143 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
5144 be used to adjust some routing behavior on border routers based on some
5145 information from the request.
5146
5147 See RFC 2474, 2597, 3260 and 4594 for more information.
5148
5149http-request set-uri <fmt> [ { if | unless } <condition> ]
5150
5151 This rewrites the request URI with the result of the evaluation of format
5152 string <fmt>. The scheme, authority, path and query string are all replaced
5153 at once. This can be used to rewrite hosts in front of proxies, or to
5154 perform complex modifications to the URI such as moving parts between the
5155 path and the query string.
5156 See also "http-request set-path" and "http-request set-query".
5157
5158http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5159
5160 This is used to set the contents of a variable. The variable is declared
5161 inline.
5162
5163 Arguments:
5164 <var-name> The name of the variable starts with an indication about its
5165 scope. The scopes allowed are:
5166 "proc" : the variable is shared with the whole process
5167 "sess" : the variable is shared with the whole session
5168 "txn" : the variable is shared with the transaction
5169 (request and response)
5170 "req" : the variable is shared only during request
5171 processing
5172 "res" : the variable is shared only during response
5173 processing
5174 This prefix is followed by a name. The separator is a '.'.
5175 The name may only contain characters 'a-z', 'A-Z', '0-9'
5176 and '_'.
5177
5178 <expr> Is a standard HAProxy expression formed by a sample-fetch
5179 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01005180
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005181 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005182 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005183
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005184http-request send-spoe-group <engine-name> <group-name>
5185 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005186
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005187 This action is used to trigger sending of a group of SPOE messages. To do so,
5188 the SPOE engine used to send messages must be defined, as well as the SPOE
5189 group to send. Of course, the SPOE engine must refer to an existing SPOE
5190 filter. If not engine name is provided on the SPOE filter line, the SPOE
5191 agent name must be used.
5192
5193 Arguments:
5194 <engine-name> The SPOE engine name.
5195
5196 <group-name> The SPOE group name as specified in the engine
5197 configuration.
5198
5199http-request silent-drop [ { if | unless } <condition> ]
5200
5201 This stops the evaluation of the rules and makes the client-facing connection
5202 suddenly disappear using a system-dependent way that tries to prevent the
5203 client from being notified. The effect it then that the client still sees an
5204 established connection while there's none on HAProxy. The purpose is to
5205 achieve a comparable effect to "tarpit" except that it doesn't use any local
5206 resource at all on the machine running HAProxy. It can resist much higher
5207 loads than "tarpit", and slow down stronger attackers. It is important to
5208 understand the impact of using this mechanism. All stateful equipment placed
5209 between the client and HAProxy (firewalls, proxies, load balancers) will also
5210 keep the established connection for a long time and may suffer from this
5211 action.
5212 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5213 option is used to block the emission of a TCP reset. On other systems, the
5214 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5215 router, though it's still delivered to local networks. Do not use it unless
5216 you fully understand how it works.
5217
Christopher Faulet46f95542019-12-20 10:07:22 +01005218http-request strict-mode { on | off }
5219
5220 This enables or disables the strict rewriting mode for following rules. It
5221 does not affect rules declared before it and it is only applicable on rules
5222 performing a rewrite on the requests. When the strict mode is enabled, any
5223 rewrite failure triggers an internal error. Otherwise, such errors are
5224 silently ignored. The purpose of the strict rewriting mode is to make some
5225 rewrites optionnal while others must be performed to continue the request
5226 processing.
5227
Christopher Faulet1aea50e2020-01-17 16:03:53 +01005228 By default, the strict rewriting mode is enabled. Its value is also reset
Christopher Faulet46f95542019-12-20 10:07:22 +01005229 when a ruleset evaluation ends. So, for instance, if you change the mode on
5230 the frontend, the default mode is restored when HAProxy starts the backend
5231 rules evaluation.
5232
Christopher Faulet554c0eb2020-01-14 12:00:28 +01005233http-request tarpit [deny_status <status>] [ { errorfile | errorfiles } <err> ]
5234 [ { if | unless } <condition> ]
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005235
5236 This stops the evaluation of the rules and immediately blocks the request
5237 without responding for a delay specified by "timeout tarpit" or
5238 "timeout connect" if the former is not set. After that delay, if the client
5239 is still connected, an HTTP error 500 (or optionally the status code
5240 specified as an argument to "deny_status") is returned so that the client
5241 does not suspect it has been tarpitted. Logs will report the flags "PT".
5242 The goal of the tarpit rule is to slow down robots during an attack when
5243 they're limited on the number of concurrent requests. It can be very
5244 efficient against very dumb robots, and will significantly reduce the load
5245 on firewalls compared to a "deny" rule. But when facing "correctly"
5246 developed robots, it can make things worse by forcing haproxy and the front
Christopher Faulet554c0eb2020-01-14 12:00:28 +01005247 firewall to support insane number of concurrent connections. A specific error
5248 message may be specified. It may be an error file, using the "errorfile"
5249 keyword followed by the file containing the full HTTP response. It may also
5250 be an error from an http-errors section, using the "errorfiles" keyword
5251 followed by the section name.
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005252 See also the "silent-drop" action.
5253
5254http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5255http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5256http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
5257
5258 This enables tracking of sticky counters from current request. These rules do
5259 not stop evaluation and do not change default action. The number of counters
5260 that may be simultaneously tracked by the same connection is set in
5261 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
5262 so the track-sc number is between 0 and (MAX_SESS_STCKTR-1). The first
5263 "track-sc0" rule executed enables tracking of the counters of the specified
5264 table as the first set. The first "track-sc1" rule executed enables tracking
5265 of the counters of the specified table as the second set. The first
5266 "track-sc2" rule executed enables tracking of the counters of the specified
5267 table as the third set. It is a recommended practice to use the first set of
5268 counters for the per-frontend counters and the second set for the per-backend
5269 ones. But this is just a guideline, all may be used everywhere.
5270
5271 Arguments :
5272 <key> is mandatory, and is a sample expression rule as described in
5273 section 7.3. It describes what elements of the incoming request or
5274 connection will be analyzed, extracted, combined, and used to
5275 select which table entry to update the counters.
5276
5277 <table> is an optional table to be used instead of the default one, which
5278 is the stick-table declared in the current proxy. All the counters
5279 for the matches and updates for the key will then be performed in
5280 that table until the session ends.
5281
5282 Once a "track-sc*" rule is executed, the key is looked up in the table and if
5283 it is not found, an entry is allocated for it. Then a pointer to that entry
5284 is kept during all the session's life, and this entry's counters are updated
5285 as often as possible, every time the session's counters are updated, and also
5286 systematically when the session ends. Counters are only updated for events
5287 that happen after the tracking has been started. As an exception, connection
5288 counters and request counters are systematically updated so that they reflect
5289 useful information.
5290
5291 If the entry tracks concurrent connection counters, one connection is counted
5292 for as long as the entry is tracked, and the entry will not expire during
5293 that time. Tracking counters also provides a performance advantage over just
5294 checking the keys, because only one table lookup is performed for all ACL
5295 checks that make use of it.
5296
5297http-request unset-var(<var-name>) [ { if | unless } <condition> ]
5298
5299 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005300
5301 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005302 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005303
Christopher Faulet579d83b2019-11-22 15:34:17 +01005304http-request use-service <service-name> [ { if | unless } <condition> ]
5305
5306 This directive executes the configured HTTP service to reply to the request
5307 and stops the evaluation of the rules. An HTTP service may choose to reply by
5308 sending any valid HTTP response or it may immediately close the connection
5309 without sending any response. Outside natives services, for instance the
5310 Prometheus exporter, it is possible to write your own services in Lua. No
5311 further "http-request" rules are evaluated.
5312
5313 Arguments :
5314 <service-name> is mandatory. It is the service to call
5315
5316 Example:
5317 http-request use-service prometheus-exporter if { path /metrics }
5318
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005319http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005320
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02005321 This will delay the processing of the request until the SSL handshake
5322 happened. This is mostly useful to delay processing early data until we're
5323 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005324
Willy Tarreauef781042010-01-27 11:53:01 +01005325
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005326http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005327 Access control for Layer 7 responses
5328
5329 May be used in sections: defaults | frontend | listen | backend
5330 no | yes | yes | yes
5331
5332 The http-response statement defines a set of rules which apply to layer 7
5333 processing. The rules are evaluated in their declaration order when they are
5334 met in a frontend, listen or backend section. Any rule may optionally be
5335 followed by an ACL-based condition, in which case it will only be evaluated
5336 if the condition is true. Since these rules apply on responses, the backend
5337 rules are applied first, followed by the frontend's rules.
5338
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005339 The first keyword is the rule's action. The supported actions are described
5340 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005341
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005342 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005343
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005344 Example:
5345 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02005346
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005347 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06005348
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005349 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
5350 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06005351
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005352 Example:
5353 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06005354
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005355 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06005356
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005357 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
5358 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06005359
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005360 See also : "http-request", section 3.4 about userlists and section 7 about
5361 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06005362
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005363http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005364
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005365 This is used to add a new entry into an ACL. The ACL must be loaded from a
5366 file (even a dummy empty file). The file name of the ACL to be updated is
5367 passed between parentheses. It takes one argument: <key fmt>, which follows
5368 log-format rules, to collect content of the new entry. It performs a lookup
5369 in the ACL before insertion, to avoid duplicated (or more) values.
5370 This lookup is done by a linear search and can be expensive with large lists!
5371 It is the equivalent of the "add acl" command from the stats socket, but can
5372 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06005373
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005374http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005375
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005376 This appends an HTTP header field whose name is specified in <name> and whose
5377 value is defined by <fmt> which follows the log-format rules (see Custom Log
5378 Format in section 8.2.4). This may be used to send a cookie to a client for
5379 example, or to pass some internal information.
5380 This rule is not final, so it is possible to add other similar rules.
5381 Note that header addition is performed immediately, so one rule might reuse
5382 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06005383
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005384http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005385
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005386 This stops the evaluation of the rules and lets the response pass the check.
5387 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06005388
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02005389http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005390
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02005391 See section 6.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06005392
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005393http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005394
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005395 This captures sample expression <sample> from the response buffer, and
5396 converts it to a string. The resulting string is stored into the next request
5397 "capture" slot, so it will possibly appear next to some captured HTTP
5398 headers. It will then automatically appear in the logs, and it will be
5399 possible to extract it using sample fetch rules to feed it into headers or
5400 anything. Please check section 7.3 (Fetching samples) and
5401 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005402
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005403 The keyword "id" is the id of the capture slot which is used for storing the
5404 string. The capture slot must be defined in an associated frontend.
5405 This is useful to run captures in backends. The slot id can be declared by a
5406 previous directive "http-response capture" or with the "declare capture"
5407 keyword.
Baptiste Assmann19a69b32020-01-16 14:34:22 +01005408
5409 When using this action in a backend, double check that the relevant
5410 frontend(s) have the required capture slots otherwise, this rule will be
5411 ignored at run time. This can't be detected at configuration parsing time
5412 due to HAProxy's ability to dynamically resolve backend name at runtime.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005413
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005414http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005415
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005416 This is used to delete an entry from an ACL. The ACL must be loaded from a
5417 file (even a dummy empty file). The file name of the ACL to be updated is
5418 passed between parentheses. It takes one argument: <key fmt>, which follows
5419 log-format rules, to collect content of the entry to delete.
5420 It is the equivalent of the "del acl" command from the stats socket, but can
5421 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02005422
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005423http-response del-header <name> [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02005424
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005425 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02005426
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005427http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02005428
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005429 This is used to delete an entry from a MAP. The MAP must be loaded from a
5430 file (even a dummy empty file). The file name of the MAP to be updated is
5431 passed between parentheses. It takes one argument: <key fmt>, which follows
5432 log-format rules, to collect content of the entry to delete.
5433 It takes one argument: "file name" It is the equivalent of the "del map"
5434 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005435
Christopher Faulet554c0eb2020-01-14 12:00:28 +01005436http-response deny [deny_status <status>] [ { errorfile | errorfiles } <err> ]
5437 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005438
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005439 This stops the evaluation of the rules and immediately rejects the response
Christopher Faulet040c8cd2020-01-13 16:43:45 +01005440 and emits an HTTP 502 error, or optionally the status code specified as an
5441 argument to "deny_status". The list of permitted status codes is limited to
Christopher Faulet554c0eb2020-01-14 12:00:28 +01005442 those that can be overridden by the "errorfile" directive. A specific error
5443 message may be specified. It may be an error file, using the "errorfile"
5444 keyword followed by the file containing the full HTTP response. It may also
5445 be an error from an http-errors section, using the "errorfiles" keyword
5446 followed by the section name.
Christopher Faulet040c8cd2020-01-13 16:43:45 +01005447 No further "http-response" rules are evaluated.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005448
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005449http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005450
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005451 This performs an HTTP redirection based on a redirect rule.
5452 This supports a format string similarly to "http-request redirect" rules,
5453 with the exception that only the "location" type of redirect is possible on
5454 the response. See the "redirect" keyword for the rule's syntax. When a
5455 redirect rule is applied during a response, connections to the server are
5456 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005457
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005458http-response replace-header <name> <regex-match> <replace-fmt>
5459 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005460
Tim Duesterhus2252beb2019-10-29 00:05:13 +01005461 This works like "http-request replace-header" except that it works on the
5462 server's response instead of the client's request.
William Lallemand86d0df02017-11-24 21:36:45 +01005463
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005464 Example:
5465 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02005466
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005467 # applied to:
5468 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005469
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005470 # outputs:
5471 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 +02005472
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005473 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005474
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005475http-response replace-value <name> <regex-match> <replace-fmt>
5476 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005477
Tim Duesterhus6bd909b2020-01-17 15:53:18 +01005478 This works like "http-request replace-value" except that it works on the
Tim Duesterhus2252beb2019-10-29 00:05:13 +01005479 server's response instead of the client's request.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005480
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005481 Example:
5482 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005483
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005484 # applied to:
5485 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01005486
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005487 # outputs:
5488 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005489
Christopher Faulet24231ab2020-01-24 17:44:23 +01005490http-response return [status <code>] [content-type <type>]
5491 [ { default-errorfiles | errorfile <file> | errorfiles <name> |
5492 file <file> | lf-file <file> | string <str> | lf-string <fmt> } ]
Christopher Faulet4a2c1422020-01-31 17:36:01 +01005493 [ hdr <name> <value> ]*
Christopher Faulet24231ab2020-01-24 17:44:23 +01005494 [ { if | unless } <condition> ]
5495
5496 This stops the evaluation of the rules and immediatly returns a response. The
5497 default status code used for the response is 200. It can be optionally
5498 specified as an arguments to "status". The response content-type may also be
5499 specified as an argument to "content-type". Finally the response itselft may
5500 be defined. If can be a full HTTP response specifying the errorfile to use,
5501 or the response payload specifing the file or the string to use. These rules
5502 are followed to create the response :
5503
5504 * If neither the errorfile nor the payload to use is defined, a dummy
5505 response is returned. Only the "status" argument is considered. It can be
5506 any code in the range [200, 599]. The "content-type" argument, if any, is
5507 ignored.
5508
5509 * If "default-errorfiles" argument is set, the proxy's errorfiles are
5510 considered. If the "status" argument is defined, it must be one of the
5511 status code handled by hparoxy (200, 400, 403, 404, 405, 408, 410, 425,
5512 429, 500, 502, 503, and 504). The "content-type" argument, if any, is
5513 ignored.
5514
5515 * If a specific errorfile is defined, with an "errorfile" argument, the
5516 corresponding file, containing a full HTTP response, is returned. Only the
5517 "status" argument is considered. It must be one of the status code handled
5518 by hparoxy (200, 400, 403, 404, 405, 408, 410, 425, 429, 500, 502, 503, and
5519 504). The "content-type" argument, if any, is ignored.
5520
5521 * If an http-errors section is defined, with an "errorfiles" argument, the
5522 corresponding file in the specified http-errors section, containing a full
5523 HTTP response, is returned. Only the "status" argument is considered. It
5524 must be one of the status code handled by hparoxy (200, 400, 403, 404, 405,
5525 408, 410, 425, 429, 500, 502, 503, and 504). The "content-type" argument,
5526 if any, is ignored.
5527
5528 * If a "file" or a "lf-file" argument is specified, the file's content is
5529 used as the response payload. If the file is not empty, its content-type
5530 must be set as argument to "content-type". Otherwise, any "content-type"
5531 argument is ignored. With a "lf-file" argument, the file's content is
5532 evaluated as a log-format string. With a "file" argument, it is considered
5533 as a raw content.
5534
5535 * If a "string" or "lf-string" argument is specified, the defined string is
5536 used as the response payload. The content-type must always be set as
5537 argument to "content-type". With a "lf-string" argument, the string is
5538 evaluated as a log-format string. With a "string" argument, it is
5539 considered as a raw string.
5540
Christopher Faulet4a2c1422020-01-31 17:36:01 +01005541 When the response is not based an errorfile, it is possible to appends HTTP
5542 header fields to the response using "hdr" arguments. Otherwise, all "hdr"
5543 arguments are ignored. For each one, the header name is specified in <name>
5544 and its value is defined by <fmt> which follows the log-format rules.
5545
Christopher Faulet24231ab2020-01-24 17:44:23 +01005546 Note that the generated response must be smaller than a buffer. And to avoid
5547 any warning, when an errorfile or a raw file is loaded, the buffer space
5548 reserved to the headers rewritting should also be free.
5549
5550 No further "http-response" rules are evaluated.
5551
5552 Example:
5553 http-response return errorfile /etc/haproy/errorfiles/200.http \
5554 if { status eq 404 }
5555
5556 http-response return content-type text/plain \
5557 string "This is the end !" \
5558 if { status eq 500 }
5559
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005560http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
5561http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08005562
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005563 This action increments the GPC0 or GPC1 counter according with the sticky
5564 counter designated by <sc-id>. If an error occurs, this action silently fails
5565 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02005566
Cédric Dufour0d7712d2019-11-06 18:38:53 +01005567http-response sc-set-gpt0(<sc-id>) { <int> | <expr> }
5568 [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02005569
Cédric Dufour0d7712d2019-11-06 18:38:53 +01005570 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
5571 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
5572 boolean. If an error occurs, this action silently fails and the actions
5573 evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01005574
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005575http-response send-spoe-group [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02005576
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005577 This action is used to trigger sending of a group of SPOE messages. To do so,
5578 the SPOE engine used to send messages must be defined, as well as the SPOE
5579 group to send. Of course, the SPOE engine must refer to an existing SPOE
5580 filter. If not engine name is provided on the SPOE filter line, the SPOE
5581 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005582
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005583 Arguments:
5584 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005585
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005586 <group-name> The SPOE group name as specified in the engine
5587 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005588
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005589http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005590
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005591 This does the same as "add-header" except that the header name is first
5592 removed if it existed. This is useful when passing security information to
5593 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005594
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005595http-response set-log-level <level> [ { if | unless } <condition> ]
5596
5597 This is used to change the log level of the current request when a certain
5598 condition is met. Valid levels are the 8 syslog levels (see the "log"
5599 keyword) plus the special level "silent" which disables logging for this
5600 request. This rule is not final so the last matching rule wins. This rule can
5601 be useful to disable health checks coming from another equipment.
5602
5603http-response set-map(<file-name>) <key fmt> <value fmt>
5604
5605 This is used to add a new entry into a MAP. The MAP must be loaded from a
5606 file (even a dummy empty file). The file name of the MAP to be updated is
5607 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5608 log-format rules, used to collect MAP key, and <value fmt>, which follows
5609 log-format rules, used to collect content for the new entry. It performs a
5610 lookup in the MAP before insertion, to avoid duplicated (or more) values.
5611 This lookup is done by a linear search and can be expensive with large lists!
5612 It is the equivalent of the "set map" command from the stats socket, but can
5613 be triggered by an HTTP response.
5614
5615http-response set-mark <mark> [ { if | unless } <condition> ]
5616
5617 This is used to set the Netfilter MARK on all packets sent to the client to
5618 the value passed in <mark> on platforms which support it. This value is an
5619 unsigned 32 bit value which can be matched by netfilter and by the routing
5620 table. It can be expressed both in decimal or hexadecimal format (prefixed
5621 by "0x"). This can be useful to force certain packets to take a different
5622 route (for example a cheaper network path for bulk downloads). This works on
5623 Linux kernels 2.6.32 and above and requires admin privileges.
5624
5625http-response set-nice <nice> [ { if | unless } <condition> ]
5626
5627 This sets the "nice" factor of the current request being processed.
5628 It only has effect against the other requests being processed at the same
5629 time. The default value is 0, unless altered by the "nice" setting on the
5630 "bind" line. The accepted range is -1024..1024. The higher the value, the
5631 nicest the request will be. Lower values will make the request more important
5632 than other ones. This can be useful to improve the speed of some requests, or
5633 lower the priority of non-important requests. Using this setting without
5634 prior experimentation can cause some major slowdown.
5635
5636http-response set-status <status> [reason <str>]
5637 [ { if | unless } <condition> ]
5638
5639 This replaces the response status code with <status> which must be an integer
5640 between 100 and 999. Optionally, a custom reason text can be provided defined
5641 by <str>, or the default reason for the specified code will be used as a
5642 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005643
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005644 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005645 # return "431 Request Header Fields Too Large"
5646 http-response set-status 431
5647 # return "503 Slow Down", custom reason
5648 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005649
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005650http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005651
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005652 This is used to set the TOS or DSCP field value of packets sent to the client
5653 to the value passed in <tos> on platforms which support this.
5654 This value represents the whole 8 bits of the IP TOS field, and can be
5655 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
5656 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
5657 always 0. This can be used to adjust some routing behavior on border routers
5658 based on some information from the request.
5659
5660 See RFC 2474, 2597, 3260 and 4594 for more information.
5661
5662http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5663
5664 This is used to set the contents of a variable. The variable is declared
5665 inline.
5666
5667 Arguments:
5668 <var-name> The name of the variable starts with an indication about its
5669 scope. The scopes allowed are:
5670 "proc" : the variable is shared with the whole process
5671 "sess" : the variable is shared with the whole session
5672 "txn" : the variable is shared with the transaction
5673 (request and response)
5674 "req" : the variable is shared only during request
5675 processing
5676 "res" : the variable is shared only during response
5677 processing
5678 This prefix is followed by a name. The separator is a '.'.
5679 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5680 and '_'.
5681
5682 <expr> Is a standard HAProxy expression formed by a sample-fetch
5683 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005684
5685 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005686 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005687
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005688http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005689
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005690 This stops the evaluation of the rules and makes the client-facing connection
5691 suddenly disappear using a system-dependent way that tries to prevent the
5692 client from being notified. The effect it then that the client still sees an
5693 established connection while there's none on HAProxy. The purpose is to
5694 achieve a comparable effect to "tarpit" except that it doesn't use any local
5695 resource at all on the machine running HAProxy. It can resist much higher
5696 loads than "tarpit", and slow down stronger attackers. It is important to
5697 understand the impact of using this mechanism. All stateful equipment placed
5698 between the client and HAProxy (firewalls, proxies, load balancers) will also
5699 keep the established connection for a long time and may suffer from this
5700 action.
5701 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5702 option is used to block the emission of a TCP reset. On other systems, the
5703 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5704 router, though it's still delivered to local networks. Do not use it unless
5705 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005706
Christopher Faulet46f95542019-12-20 10:07:22 +01005707http-response strict-mode { on | off }
5708
5709 This enables or disables the strict rewriting mode for following rules. It
5710 does not affect rules declared before it and it is only applicable on rules
5711 performing a rewrite on the responses. When the strict mode is enabled, any
5712 rewrite failure triggers an internal error. Otherwise, such errors are
5713 silently ignored. The purpose of the strict rewriting mode is to make some
5714 rewrites optionnal while others must be performed to continue the response
5715 processing.
5716
Christopher Faulet1aea50e2020-01-17 16:03:53 +01005717 By default, the strict rewriting mode is enabled. Its value is also reset
Christopher Faulet46f95542019-12-20 10:07:22 +01005718 when a ruleset evaluation ends. So, for instance, if you change the mode on
5719 the bacnkend, the default mode is restored when HAProxy starts the frontend
5720 rules evaluation.
5721
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005722http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5723http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5724http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005725
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005726 This enables tracking of sticky counters from current response. Please refer
5727 to "http-request track-sc" for a complete description. The only difference
5728 from "http-request track-sc" is the <key> sample expression can only make use
5729 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
5730 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
5731 supported, haproxy will fail and warn while parsing the config.
5732
5733http-response unset-var(<var-name>) [ { if | unless } <condition> ]
5734
5735 This is used to unset a variable. See "http-response set-var" for details
5736 about <var-name>.
5737
5738 Example:
5739 http-response unset-var(sess.last_redir)
5740
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02005741
Willy Tarreau30631952015-08-06 15:05:24 +02005742http-reuse { never | safe | aggressive | always }
5743 Declare how idle HTTP connections may be shared between requests
5744
5745 May be used in sections: defaults | frontend | listen | backend
5746 yes | no | yes | yes
5747
5748 By default, a connection established between haproxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01005749 which is considered safe for reuse is moved back to the server's idle
5750 connections pool so that any other request can make use of it. This is the
5751 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02005752
5753 The argument indicates the desired connection reuse strategy :
5754
Olivier Houchard86006a52018-12-14 19:37:49 +01005755 - "never" : idle connections are never shared between sessions. This mode
5756 may be enforced to cancel a different strategy inherited from
5757 a defaults section or for troubleshooting. For example, if an
5758 old bogus application considers that multiple requests over
5759 the same connection come from the same client and it is not
5760 possible to fix the application, it may be desirable to
5761 disable connection sharing in a single backend. An example of
5762 such an application could be an old haproxy using cookie
5763 insertion in tunnel mode and not checking any request past the
5764 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02005765
Olivier Houchard86006a52018-12-14 19:37:49 +01005766 - "safe" : this is the default and the recommended strategy. The first
5767 request of a session is always sent over its own connection,
5768 and only subsequent requests may be dispatched over other
5769 existing connections. This ensures that in case the server
5770 closes the connection when the request is being sent, the
5771 browser can decide to silently retry it. Since it is exactly
5772 equivalent to regular keep-alive, there should be no side
5773 effects.
Willy Tarreau30631952015-08-06 15:05:24 +02005774
5775 - "aggressive" : this mode may be useful in webservices environments where
5776 all servers are not necessarily known and where it would be
5777 appreciable to deliver most first requests over existing
5778 connections. In this case, first requests are only delivered
5779 over existing connections that have been reused at least once,
5780 proving that the server correctly supports connection reuse.
5781 It should only be used when it's sure that the client can
5782 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02005783 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02005784 downsides of rare connection failures.
5785
5786 - "always" : this mode is only recommended when the path to the server is
5787 known for never breaking existing connections quickly after
5788 releasing them. It allows the first request of a session to be
5789 sent to an existing connection. This can provide a significant
5790 performance increase over the "safe" strategy when the backend
5791 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005792 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02005793 sharing. It is recommended that the "http-keep-alive" timeout
5794 remains low in this mode so that no dead connections remain
5795 usable. In most cases, this will lead to the same performance
5796 gains as "aggressive" but with more risks. It should only be
5797 used when it improves the situation over "aggressive".
5798
5799 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005800 connection properties and compatibility. Specifically :
5801 - connections made with "usesrc" followed by a client-dependent value
5802 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005803
5804 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01005805 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005806
Lukas Tribusfd9b68c2018-10-27 20:06:59 +02005807 - connections with certain bogus authentication schemes (relying on the
5808 connection) like NTLM are detected, marked private and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005809
Lukas Tribuse8adfeb2019-11-06 11:50:25 +01005810 A connection pool is involved and configurable with "pool-max-conn".
Willy Tarreau30631952015-08-06 15:05:24 +02005811
5812 Note: connection reuse improves the accuracy of the "server maxconn" setting,
5813 because almost no new connection will be established while idle connections
5814 remain available. This is particularly true with the "always" strategy.
5815
5816 See also : "option http-keep-alive", "server maxconn"
5817
5818
Mark Lamourinec2247f02012-01-04 13:02:01 -05005819http-send-name-header [<header>]
5820 Add the server name to a request. Use the header string given by <header>
Mark Lamourinec2247f02012-01-04 13:02:01 -05005821 May be used in sections: defaults | frontend | listen | backend
5822 yes | no | yes | yes
Mark Lamourinec2247f02012-01-04 13:02:01 -05005823 Arguments :
Mark Lamourinec2247f02012-01-04 13:02:01 -05005824 <header> The header string to use to send the server name
5825
Willy Tarreau81bef7e2019-10-07 14:58:02 +02005826 The "http-send-name-header" statement causes the header field named <header>
5827 to be set to the name of the target server at the moment the request is about
5828 to be sent on the wire. Any existing occurrences of this header are removed.
5829 Upon retries and redispatches, the header field is updated to always reflect
5830 the server being attempted to connect to. Given that this header is modified
5831 very late in the connection setup, it may have unexpected effects on already
5832 modified headers. For example using it with transport-level header such as
5833 connection, content-length, transfer-encoding and so on will likely result in
5834 invalid requests being sent to the server. Additionally it has been reported
5835 that this directive is currently being used as a way to overwrite the Host
5836 header field in outgoing requests; while this trick has been known to work
5837 as a side effect of the feature for some time, it is not officially supported
5838 and might possibly not work anymore in a future version depending on the
5839 technical difficulties this feature induces. A long-term solution instead
5840 consists in fixing the application which required this trick so that it binds
5841 to the correct host name.
Mark Lamourinec2247f02012-01-04 13:02:01 -05005842
5843 See also : "server"
5844
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005845id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02005846 Set a persistent ID to a proxy.
5847 May be used in sections : defaults | frontend | listen | backend
5848 no | yes | yes | yes
5849 Arguments : none
5850
5851 Set a persistent ID for the proxy. This ID must be unique and positive.
5852 An unused ID will automatically be assigned if unset. The first assigned
5853 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005854
5855
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005856ignore-persist { if | unless } <condition>
5857 Declare a condition to ignore persistence
5858 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005859 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005860
5861 By default, when cookie persistence is enabled, every requests containing
5862 the cookie are unconditionally persistent (assuming the target server is up
5863 and running).
5864
5865 The "ignore-persist" statement allows one to declare various ACL-based
5866 conditions which, when met, will cause a request to ignore persistence.
5867 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005868 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005869 persistence for a specific User-Agent (for example, some web crawler bots).
5870
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005871 The persistence is ignored when an "if" condition is met, or unless an
5872 "unless" condition is met.
5873
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005874 Example:
5875 acl url_static path_beg /static /images /img /css
5876 acl url_static path_end .gif .png .jpg .css .js
5877 ignore-persist if url_static
5878
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005879 See also : "force-persist", "cookie", and section 7 about ACL usage.
5880
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005881load-server-state-from-file { global | local | none }
5882 Allow seamless reload of HAProxy
5883 May be used in sections: defaults | frontend | listen | backend
5884 yes | no | yes | yes
5885
5886 This directive points HAProxy to a file where server state from previous
5887 running process has been saved. That way, when starting up, before handling
5888 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01005889 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005890 to tell haproxy which file to use. For now, only 2 arguments to either prevent
5891 loading state or load states from a file containing all backends and servers.
5892 The state file can be generated by running the command "show servers state"
5893 over the stats socket and redirect output.
5894
Davor Ocelice9ed2812017-12-25 17:49:28 +01005895 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005896 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02005897 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005898
5899 Arguments:
5900 global load the content of the file pointed by the global directive
5901 named "server-state-file".
5902
5903 local load the content of the file pointed by the directive
5904 "server-state-file-name" if set. If not set, then the backend
5905 name is used as a file name.
5906
5907 none don't load any stat for this backend
5908
5909 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01005910 - server's IP address is preserved across reloads by default, but the
5911 order can be changed thanks to the server's "init-addr" setting. This
5912 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01005913 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01005914 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005915
5916 - server's weight is applied from previous running process unless it has
5917 has changed between previous and new configuration files.
5918
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005919 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005920
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005921 global
5922 stats socket /tmp/socket
5923 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005924
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005925 defaults
5926 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005927
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005928 backend bk
5929 server s1 127.0.0.1:22 check weight 11
5930 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005931
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005932
5933 Then one can run :
5934
5935 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
5936
5937 Content of the file /tmp/server_state would be like this:
5938
5939 1
5940 # <field names skipped for the doc example>
5941 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5942 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5943
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005944 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005945
5946 global
5947 stats socket /tmp/socket
5948 server-state-base /etc/haproxy/states
5949
5950 defaults
5951 load-server-state-from-file local
5952
5953 backend bk
5954 server s1 127.0.0.1:22 check weight 11
5955 server s2 127.0.0.1:22 check weight 12
5956
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005957
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005958 Then one can run :
5959
5960 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
5961
5962 Content of the file /etc/haproxy/states/bk would be like this:
5963
5964 1
5965 # <field names skipped for the doc example>
5966 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5967 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5968
5969 See also: "server-state-file", "server-state-file-name", and
5970 "show servers state"
5971
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005972
Willy Tarreau2769aa02007-12-27 18:26:09 +01005973log global
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005974log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
5975 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02005976no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01005977 Enable per-instance logging of events and traffic.
5978 May be used in sections : defaults | frontend | listen | backend
5979 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02005980
5981 Prefix :
5982 no should be used when the logger list must be flushed. For example,
5983 if you don't want to inherit from the default logger list. This
5984 prefix does not allow arguments.
5985
Willy Tarreau2769aa02007-12-27 18:26:09 +01005986 Arguments :
5987 global should be used when the instance's logging parameters are the
5988 same as the global ones. This is the most common usage. "global"
5989 replaces <address>, <facility> and <level> with those of the log
5990 entries found in the "global" section. Only one "log global"
5991 statement may be used per instance, and this form takes no other
5992 parameter.
5993
5994 <address> indicates where to send the logs. It takes the same format as
5995 for the "global" section's logs, and can be one of :
5996
5997 - An IPv4 address optionally followed by a colon (':') and a UDP
5998 port. If no port is specified, 514 is used by default (the
5999 standard syslog port).
6000
David du Colombier24bb5f52011-03-17 10:40:23 +01006001 - An IPv6 address followed by a colon (':') and optionally a UDP
6002 port. If no port is specified, 514 is used by default (the
6003 standard syslog port).
6004
Willy Tarreau2769aa02007-12-27 18:26:09 +01006005 - A filesystem path to a UNIX domain socket, keeping in mind
6006 considerations for chroot (be sure the path is accessible
6007 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01006008 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01006009
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01006010 - A file descriptor number in the form "fd@<number>", which may
6011 point to a pipe, terminal, or socket. In this case unbuffered
6012 logs are used and one writev() call per log is performed. This
6013 is a bit expensive but acceptable for most workloads. Messages
6014 sent this way will not be truncated but may be dropped, in
6015 which case the DroppedLogs counter will be incremented. The
6016 writev() call is atomic even on pipes for messages up to
6017 PIPE_BUF size, which POSIX recommends to be at least 512 and
6018 which is 4096 bytes on most modern operating systems. Any
6019 larger message may be interleaved with messages from other
6020 processes. Exceptionally for debugging purposes the file
6021 descriptor may also be directed to a file, but doing so will
6022 significantly slow haproxy down as non-blocking calls will be
6023 ignored. Also there will be no way to purge nor rotate this
6024 file without restarting the process. Note that the configured
6025 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01006026 with a TCP syslog server. See also the "short" and "raw"
6027 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01006028
6029 - "stdout" / "stderr", which are respectively aliases for "fd@1"
6030 and "fd@2", see above.
6031
Willy Tarreauc046d162019-08-30 15:24:59 +02006032 - A ring buffer in the form "ring@<name>", which will correspond
6033 to an in-memory ring buffer accessible over the CLI using the
6034 "show events" command, which will also list existing rings and
6035 their sizes. Such buffers are lost on reload or restart but
6036 when used as a complement this can help troubleshooting by
6037 having the logs instantly available.
6038
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01006039 You may want to reference some environment variables in the
6040 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01006041
Willy Tarreau18324f52014-06-27 18:10:07 +02006042 <length> is an optional maximum line length. Log lines larger than this
6043 value will be truncated before being sent. The reason is that
6044 syslog servers act differently on log line length. All servers
6045 support the default value of 1024, but some servers simply drop
6046 larger lines while others do log them. If a server supports long
6047 lines, it may make sense to set this value here in order to avoid
6048 truncating long lines. Similarly, if a server drops long lines,
6049 it is preferable to truncate them before sending them. Accepted
6050 values are 80 to 65535 inclusive. The default value of 1024 is
6051 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01006052 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02006053
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02006054 <ranges> A list of comma-separated ranges to identify the logs to sample.
6055 This is used to balance the load of the logs to send to the log
6056 server. The limits of the ranges cannot be null. They are numbered
6057 from 1. The size or period (in number of logs) of the sample must
6058 be set with <sample_size> parameter.
6059
6060 <sample_size>
6061 The size of the sample in number of logs to consider when balancing
6062 their logging loads. It is used to balance the load of the logs to
6063 send to the syslog server. This size must be greater or equal to the
6064 maximum of the high limits of the ranges.
6065 (see also <ranges> parameter).
6066
Willy Tarreauadb345d2018-11-12 07:56:13 +01006067 <format> is the log format used when generating syslog messages. It may be
6068 one of the following :
6069
6070 rfc3164 The RFC3164 syslog message format. This is the default.
6071 (https://tools.ietf.org/html/rfc3164)
6072
6073 rfc5424 The RFC5424 syslog message format.
6074 (https://tools.ietf.org/html/rfc5424)
6075
Willy Tarreaue8746a02018-11-12 08:45:00 +01006076 short A message containing only a level between angle brackets such as
6077 '<3>', followed by the text. The PID, date, time, process name
6078 and system name are omitted. This is designed to be used with a
6079 local log server. This format is compatible with what the
6080 systemd logger consumes.
6081
Willy Tarreauc1b06452018-11-12 11:57:56 +01006082 raw A message containing only the text. The level, PID, date, time,
6083 process name and system name are omitted. This is designed to
6084 be used in containers or during development, where the severity
6085 only depends on the file descriptor used (stdout/stderr).
6086
Willy Tarreau2769aa02007-12-27 18:26:09 +01006087 <facility> must be one of the 24 standard syslog facilities :
6088
Willy Tarreaue8746a02018-11-12 08:45:00 +01006089 kern user mail daemon auth syslog lpr news
6090 uucp cron auth2 ftp ntp audit alert cron2
6091 local0 local1 local2 local3 local4 local5 local6 local7
6092
Willy Tarreauc1b06452018-11-12 11:57:56 +01006093 Note that the facility is ignored for the "short" and "raw"
6094 formats, but still required as a positional field. It is
6095 recommended to use "daemon" in this case to make it clear that
6096 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006097
6098 <level> is optional and can be specified to filter outgoing messages. By
6099 default, all messages are sent. If a level is specified, only
6100 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02006101 will be sent. An optional minimum level can be specified. If it
6102 is set, logs emitted with a more severe level than this one will
6103 be capped to this level. This is used to avoid sending "emerg"
6104 messages on all terminals on some default syslog configurations.
6105 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01006106
6107 emerg alert crit err warning notice info debug
6108
William Lallemand0f99e342011-10-12 17:50:54 +02006109 It is important to keep in mind that it is the frontend which decides what to
6110 log from a connection, and that in case of content switching, the log entries
6111 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01006112
6113 However, backend log declaration define how and where servers status changes
6114 will be logged. Level "notice" will be used to indicate a server going up,
6115 "warning" will be used for termination signals and definitive service
6116 termination, and "alert" will be used for when a server goes down.
6117
6118 Note : According to RFC3164, messages are truncated to 1024 bytes before
6119 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006120
6121 Example :
6122 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01006123 log stdout format short daemon # send log to systemd
6124 log stdout format raw daemon # send everything to stdout
6125 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02006126 log 127.0.0.1:514 local0 notice # only send important events
6127 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02006128 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01006129
Willy Tarreau2769aa02007-12-27 18:26:09 +01006130
William Lallemand48940402012-01-30 16:47:22 +01006131log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01006132 Specifies the log format string to use for traffic logs
6133 May be used in sections: defaults | frontend | listen | backend
6134 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01006135
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01006136 This directive specifies the log format string that will be used for all logs
6137 resulting from traffic passing through the frontend using this line. If the
6138 directive is used in a defaults section, all subsequent frontends will use
6139 the same log format. Please see section 8.2.4 which covers the log format
6140 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01006141
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006142 "log-format" directive overrides previous "option tcplog", "log-format" and
6143 "option httplog" directives.
6144
Dragan Dosen7ad31542015-09-28 17:16:47 +02006145log-format-sd <string>
6146 Specifies the RFC5424 structured-data log format string
6147 May be used in sections: defaults | frontend | listen | backend
6148 yes | yes | yes | no
6149
6150 This directive specifies the RFC5424 structured-data log format string that
6151 will be used for all logs resulting from traffic passing through the frontend
6152 using this line. If the directive is used in a defaults section, all
6153 subsequent frontends will use the same log format. Please see section 8.2.4
6154 which covers the log format string in depth.
6155
6156 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
6157 about the RFC5424 structured-data part.
6158
6159 Note : This log format string will be used only for loggers that have set
6160 log format to "rfc5424".
6161
6162 Example :
6163 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
6164
6165
Willy Tarreau094af4e2015-01-07 15:03:42 +01006166log-tag <string>
6167 Specifies the log tag to use for all outgoing logs
6168 May be used in sections: defaults | frontend | listen | backend
6169 yes | yes | yes | yes
6170
6171 Sets the tag field in the syslog header to this string. It defaults to the
6172 log-tag set in the global section, otherwise the program name as launched
6173 from the command line, which usually is "haproxy". Sometimes it can be useful
6174 to differentiate between multiple processes running on the same host, or to
6175 differentiate customer instances running in the same process. In the backend,
6176 logs about servers up/down will use this tag. As a hint, it can be convenient
6177 to set a log-tag related to a hosted customer in a defaults section then put
6178 all the frontends and backends for that customer, then start another customer
6179 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006180
Willy Tarreauc35362a2014-04-25 13:58:37 +02006181max-keep-alive-queue <value>
6182 Set the maximum server queue size for maintaining keep-alive connections
6183 May be used in sections: defaults | frontend | listen | backend
6184 yes | no | yes | yes
6185
6186 HTTP keep-alive tries to reuse the same server connection whenever possible,
6187 but sometimes it can be counter-productive, for example if a server has a lot
6188 of connections while other ones are idle. This is especially true for static
6189 servers.
6190
6191 The purpose of this setting is to set a threshold on the number of queued
6192 connections at which haproxy stops trying to reuse the same server and prefers
6193 to find another one. The default value, -1, means there is no limit. A value
6194 of zero means that keep-alive requests will never be queued. For very close
6195 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01006196 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02006197 use a value of 10 or less). For remote servers suffering from a high latency,
6198 higher values might be needed to cover for the latency and/or the cost of
6199 picking a different server.
6200
6201 Note that this has no impact on responses which are maintained to the same
6202 server consecutively to a 401 response. They will still go to the same server
6203 even if they have to be queued.
6204
6205 See also : "option http-server-close", "option prefer-last-server", server
6206 "maxconn" and cookie persistence.
6207
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01006208max-session-srv-conns <nb>
6209 Set the maximum number of outgoing connections we can keep idling for a given
6210 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
6211 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02006212
Willy Tarreau2769aa02007-12-27 18:26:09 +01006213maxconn <conns>
6214 Fix the maximum number of concurrent connections on a frontend
6215 May be used in sections : defaults | frontend | listen | backend
6216 yes | yes | yes | no
6217 Arguments :
6218 <conns> is the maximum number of concurrent connections the frontend will
6219 accept to serve. Excess connections will be queued by the system
6220 in the socket's listen queue and will be served once a connection
6221 closes.
6222
6223 If the system supports it, it can be useful on big sites to raise this limit
6224 very high so that haproxy manages connection queues, instead of leaving the
6225 clients with unanswered connection attempts. This value should not exceed the
6226 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01006227 of tune.bufsize (16kB by default) each, as well as some other data resulting
6228 in about 33 kB of RAM being consumed per established connection. That means
6229 that a medium system equipped with 1GB of RAM can withstand around
6230 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006231
6232 Also, when <conns> is set to large values, it is possible that the servers
6233 are not sized to accept such loads, and for this reason it is generally wise
6234 to assign them some reasonable connection limits.
6235
Willy Tarreauc8d5b952019-02-27 17:25:52 +01006236 When this value is set to zero, which is the default, the global "maxconn"
6237 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02006238
Willy Tarreau2769aa02007-12-27 18:26:09 +01006239 See also : "server", global section's "maxconn", "fullconn"
6240
6241
6242mode { tcp|http|health }
6243 Set the running mode or protocol of the instance
6244 May be used in sections : defaults | frontend | listen | backend
6245 yes | yes | yes | yes
6246 Arguments :
6247 tcp The instance will work in pure TCP mode. A full-duplex connection
6248 will be established between clients and servers, and no layer 7
6249 examination will be performed. This is the default mode. It
6250 should be used for SSL, SSH, SMTP, ...
6251
6252 http The instance will work in HTTP mode. The client request will be
6253 analyzed in depth before connecting to any server. Any request
6254 which is not RFC-compliant will be rejected. Layer 7 filtering,
6255 processing and switching will be possible. This is the mode which
6256 brings HAProxy most of its value.
6257
6258 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02006259 to incoming connections and close the connection. Alternatively,
6260 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
6261 instead. Nothing will be logged in either case. This mode is used
6262 to reply to external components health checks. This mode is
6263 deprecated and should not be used anymore as it is possible to do
6264 the same and even better by combining TCP or HTTP modes with the
6265 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006266
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006267 When doing content switching, it is mandatory that the frontend and the
6268 backend are in the same mode (generally HTTP), otherwise the configuration
6269 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006270
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006271 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01006272 defaults http_instances
6273 mode http
6274
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006275 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01006276
Willy Tarreau0ba27502007-12-24 16:55:16 +01006277
Cyril Bontéf0c60612010-02-06 14:44:47 +01006278monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01006279 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006280 May be used in sections : defaults | frontend | listen | backend
6281 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01006282 Arguments :
6283 if <cond> the monitor request will fail if the condition is satisfied,
6284 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006285 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01006286 are met, for instance a low number of servers both in a
6287 backend and its backup.
6288
6289 unless <cond> the monitor request will succeed only if the condition is
6290 satisfied, and will fail otherwise. Such a condition may be
6291 based on a test on the presence of a minimum number of active
6292 servers in a list of backends.
6293
6294 This statement adds a condition which can force the response to a monitor
6295 request to report a failure. By default, when an external component queries
6296 the URI dedicated to monitoring, a 200 response is returned. When one of the
6297 conditions above is met, haproxy will return 503 instead of 200. This is
6298 very useful to report a site failure to an external component which may base
6299 routing advertisements between multiple sites on the availability reported by
6300 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02006301 criterion. Note that "monitor fail" only works in HTTP mode. Both status
6302 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006303
6304 Example:
6305 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01006306 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01006307 acl site_dead nbsrv(dynamic) lt 2
6308 acl site_dead nbsrv(static) lt 2
6309 monitor-uri /site_alive
6310 monitor fail if site_dead
6311
Willy Tarreauae94d4d2011-05-11 16:28:49 +02006312 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01006313
6314
6315monitor-net <source>
6316 Declare a source network which is limited to monitor requests
6317 May be used in sections : defaults | frontend | listen | backend
6318 yes | yes | yes | no
6319 Arguments :
6320 <source> is the source IPv4 address or network which will only be able to
6321 get monitor responses to any request. It can be either an IPv4
6322 address, a host name, or an address followed by a slash ('/')
6323 followed by a mask.
6324
6325 In TCP mode, any connection coming from a source matching <source> will cause
6326 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006327 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01006328 forwarding the connection to a remote server.
6329
6330 In HTTP mode, a connection coming from a source matching <source> will be
6331 accepted, the following response will be sent without waiting for a request,
6332 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
6333 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02006334 running without forwarding the request to a backend server. Note that this
6335 response is sent in raw format, without any transformation. This is important
6336 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01006337
Willy Tarreau82569f92012-09-27 23:48:56 +02006338 Monitor requests are processed very early, just after tcp-request connection
6339 ACLs which are the only ones able to block them. These connections are short
6340 lived and never wait for any data from the client. They cannot be logged, and
6341 it is the intended purpose. They are only used to report HAProxy's health to
6342 an upper component, nothing more. Please note that "monitor fail" rules do
6343 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01006344
Willy Tarreau95cd2832010-03-04 23:36:33 +01006345 Last, please note that only one "monitor-net" statement can be specified in
6346 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006347
Willy Tarreau2769aa02007-12-27 18:26:09 +01006348 Example :
6349 # addresses .252 and .253 are just probing us.
6350 frontend www
6351 monitor-net 192.168.0.252/31
6352
6353 See also : "monitor fail", "monitor-uri"
6354
6355
6356monitor-uri <uri>
6357 Intercept a URI used by external components' monitor requests
6358 May be used in sections : defaults | frontend | listen | backend
6359 yes | yes | yes | no
6360 Arguments :
6361 <uri> is the exact URI which we want to intercept to return HAProxy's
6362 health status instead of forwarding the request.
6363
6364 When an HTTP request referencing <uri> will be received on a frontend,
6365 HAProxy will not forward it nor log it, but instead will return either
6366 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
6367 conditions defined with "monitor fail". This is normally enough for any
6368 front-end HTTP probe to detect that the service is UP and running without
6369 forwarding the request to a backend server. Note that the HTTP method, the
6370 version and all headers are ignored, but the request must at least be valid
6371 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
6372
Willy Tarreau721d8e02017-12-01 18:25:08 +01006373 Monitor requests are processed very early, just after the request is parsed
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02006374 and even before any "http-request". The only rulesets applied before are the
6375 tcp-request ones. They cannot be logged either, and it is the intended
6376 purpose. They are only used to report HAProxy's health to an upper component,
6377 nothing more. However, it is possible to add any number of conditions using
6378 "monitor fail" and ACLs so that the result can be adjusted to whatever check
6379 can be imagined (most often the number of available servers in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01006380
Christopher Faulet6072beb2020-02-18 15:34:58 +01006381 Note: if <uri> starts by a slash ('/'), the matching is performed against the
6382 request's path instead of the request's uri. It is a workaround to let
6383 the HTTP/2 requests match the monitor-uri. Indeed, in HTTP/2, clients
6384 are encouraged to send absolute URIs only.
6385
Willy Tarreau2769aa02007-12-27 18:26:09 +01006386 Example :
6387 # Use /haproxy_test to report haproxy's status
6388 frontend www
6389 mode http
6390 monitor-uri /haproxy_test
6391
6392 See also : "monitor fail", "monitor-net"
6393
Willy Tarreau0ba27502007-12-24 16:55:16 +01006394
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006395option abortonclose
6396no option abortonclose
6397 Enable or disable early dropping of aborted requests pending in queues.
6398 May be used in sections : defaults | frontend | listen | backend
6399 yes | no | yes | yes
6400 Arguments : none
6401
6402 In presence of very high loads, the servers will take some time to respond.
6403 The per-instance connection queue will inflate, and the response time will
6404 increase respective to the size of the queue times the average per-session
6405 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01006406 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006407 the queue, and slowing down other users, and the servers as well, because the
6408 request will eventually be served, then aborted at the first error
6409 encountered while delivering the response.
6410
6411 As there is no way to distinguish between a full STOP and a simple output
6412 close on the client side, HTTP agents should be conservative and consider
6413 that the client might only have closed its output channel while waiting for
6414 the response. However, this introduces risks of congestion when lots of users
6415 do the same, and is completely useless nowadays because probably no client at
6416 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01006417 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006418 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01006419 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006420 of being the single component to break rare but valid traffic is extremely
6421 low, which adds to the temptation to be able to abort a session early while
6422 still not served and not pollute the servers.
6423
Davor Ocelice9ed2812017-12-25 17:49:28 +01006424 In HAProxy, the user can choose the desired behavior using the option
6425 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006426 compliant and aborted requests will be served. But when the option is
6427 specified, a session with an incoming channel closed will be aborted while
6428 it is still possible, either pending in the queue for a connection slot, or
6429 during the connection establishment if the server has not yet acknowledged
6430 the connection request. This considerably reduces the queue size and the load
6431 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01006432 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006433
6434 If this option has been enabled in a "defaults" section, it can be disabled
6435 in a specific instance by prepending the "no" keyword before it.
6436
6437 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
6438
6439
Willy Tarreau4076a152009-04-02 15:18:36 +02006440option accept-invalid-http-request
6441no option accept-invalid-http-request
6442 Enable or disable relaxing of HTTP request parsing
6443 May be used in sections : defaults | frontend | listen | backend
6444 yes | yes | yes | no
6445 Arguments : none
6446
Willy Tarreau91852eb2015-05-01 13:26:00 +02006447 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02006448 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01006449 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02006450 forbidden characters are essentially used to build attacks exploiting server
6451 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
6452 server will emit invalid header names for whatever reason (configuration,
6453 implementation) and the issue will not be immediately fixed. In such a case,
6454 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01006455 even if that does not make sense, by specifying this option. Similarly, the
6456 list of characters allowed to appear in a URI is well defined by RFC3986, and
6457 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
6458 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01006459 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02006460 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02006461 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
6462 to pass through (no version specified) and multiple digits for both the major
6463 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02006464
6465 This option should never be enabled by default as it hides application bugs
6466 and open security breaches. It should only be deployed after a problem has
6467 been confirmed.
6468
6469 When this option is enabled, erroneous header names will still be accepted in
6470 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01006471 analysis using the "show errors" request on the UNIX stats socket. Similarly,
6472 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02006473 also helps confirming that the issue has been solved.
6474
6475 If this option has been enabled in a "defaults" section, it can be disabled
6476 in a specific instance by prepending the "no" keyword before it.
6477
6478 See also : "option accept-invalid-http-response" and "show errors" on the
6479 stats socket.
6480
6481
6482option accept-invalid-http-response
6483no option accept-invalid-http-response
6484 Enable or disable relaxing of HTTP response parsing
6485 May be used in sections : defaults | frontend | listen | backend
6486 yes | no | yes | yes
6487 Arguments : none
6488
Willy Tarreau91852eb2015-05-01 13:26:00 +02006489 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02006490 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01006491 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02006492 forbidden characters are essentially used to build attacks exploiting server
6493 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
6494 server will emit invalid header names for whatever reason (configuration,
6495 implementation) and the issue will not be immediately fixed. In such a case,
6496 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02006497 even if that does not make sense, by specifying this option. This option also
6498 relaxes the test on the HTTP version format, it allows multiple digits for
6499 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02006500
6501 This option should never be enabled by default as it hides application bugs
6502 and open security breaches. It should only be deployed after a problem has
6503 been confirmed.
6504
6505 When this option is enabled, erroneous header names will still be accepted in
6506 responses, but the complete response will be captured in order to permit
6507 later analysis using the "show errors" request on the UNIX stats socket.
6508 Doing this also helps confirming that the issue has been solved.
6509
6510 If this option has been enabled in a "defaults" section, it can be disabled
6511 in a specific instance by prepending the "no" keyword before it.
6512
6513 See also : "option accept-invalid-http-request" and "show errors" on the
6514 stats socket.
6515
6516
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006517option allbackups
6518no option allbackups
6519 Use either all backup servers at a time or only the first one
6520 May be used in sections : defaults | frontend | listen | backend
6521 yes | no | yes | yes
6522 Arguments : none
6523
6524 By default, the first operational backup server gets all traffic when normal
6525 servers are all down. Sometimes, it may be preferred to use multiple backups
6526 at once, because one will not be enough. When "option allbackups" is enabled,
6527 the load balancing will be performed among all backup servers when all normal
6528 ones are unavailable. The same load balancing algorithm will be used and the
6529 servers' weights will be respected. Thus, there will not be any priority
6530 order between the backup servers anymore.
6531
6532 This option is mostly used with static server farms dedicated to return a
6533 "sorry" page when an application is completely offline.
6534
6535 If this option has been enabled in a "defaults" section, it can be disabled
6536 in a specific instance by prepending the "no" keyword before it.
6537
6538
6539option checkcache
6540no option checkcache
Godbach7056a352013-12-11 20:01:07 +08006541 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006542 May be used in sections : defaults | frontend | listen | backend
6543 yes | no | yes | yes
6544 Arguments : none
6545
6546 Some high-level frameworks set application cookies everywhere and do not
6547 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006548 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006549 high risk of session crossing or stealing between users traversing the same
6550 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006551 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006552
6553 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006554 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01006555 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006556 response to check if there's a risk of caching a cookie on a client-side
6557 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01006558 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01006559 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01006560 - all those with a return code other than 200, 203, 204, 206, 300, 301,
6561 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01006562 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01006563 - all those that result from a request using a method other than GET, HEAD,
6564 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01006565 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006566 - those with a 'Pragma: no-cache' header
6567 - those with a 'Cache-control: private' header
6568 - those with a 'Cache-control: no-store' header
6569 - those with a 'Cache-control: max-age=0' header
6570 - those with a 'Cache-control: s-maxage=0' header
6571 - those with a 'Cache-control: no-cache' header
6572 - those with a 'Cache-control: no-cache="set-cookie"' header
6573 - those with a 'Cache-control: no-cache="set-cookie,' header
6574 (allowing other fields after set-cookie)
6575
6576 If a response doesn't respect these requirements, then it will be blocked
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02006577 just as if it was from an "http-response deny" rule, with an "HTTP 502 bad
6578 gateway". The session state shows "PH--" meaning that the proxy blocked the
6579 response during headers processing. Additionally, an alert will be sent in
6580 the logs so that admins are informed that there's something to be fixed.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006581
6582 Due to the high impact on the application, the application should be tested
6583 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006584 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01006585 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006586
6587 If this option has been enabled in a "defaults" section, it can be disabled
6588 in a specific instance by prepending the "no" keyword before it.
6589
6590
6591option clitcpka
6592no option clitcpka
6593 Enable or disable the sending of TCP keepalive packets on the client side
6594 May be used in sections : defaults | frontend | listen | backend
6595 yes | yes | yes | no
6596 Arguments : none
6597
6598 When there is a firewall or any session-aware component between a client and
6599 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006600 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006601 components decides to expire a session which has remained idle for too long.
6602
6603 Enabling socket-level TCP keep-alives makes the system regularly send packets
6604 to the other end of the connection, leaving it active. The delay between
6605 keep-alive probes is controlled by the system only and depends both on the
6606 operating system and its tuning parameters.
6607
6608 It is important to understand that keep-alive packets are neither emitted nor
6609 received at the application level. It is only the network stacks which sees
6610 them. For this reason, even if one side of the proxy already uses keep-alives
6611 to maintain its connection alive, those keep-alive packets will not be
6612 forwarded to the other side of the proxy.
6613
6614 Please note that this has nothing to do with HTTP keep-alive.
6615
6616 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
6617 client side of a connection, which should help when session expirations are
6618 noticed between HAProxy and a client.
6619
6620 If this option has been enabled in a "defaults" section, it can be disabled
6621 in a specific instance by prepending the "no" keyword before it.
6622
6623 See also : "option srvtcpka", "option tcpka"
6624
6625
Willy Tarreau0ba27502007-12-24 16:55:16 +01006626option contstats
6627 Enable continuous traffic statistics updates
6628 May be used in sections : defaults | frontend | listen | backend
6629 yes | yes | yes | no
6630 Arguments : none
6631
6632 By default, counters used for statistics calculation are incremented
6633 only when a session finishes. It works quite well when serving small
6634 objects, but with big ones (for example large images or archives) or
6635 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01006636 a hedgehog. With this option enabled counters get incremented frequently
6637 along the session, typically every 5 seconds, which is often enough to
6638 produce clean graphs. Recounting touches a hotpath directly so it is not
6639 not enabled by default, as it can cause a lot of wakeups for very large
6640 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006641
6642
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006643option dontlog-normal
6644no option dontlog-normal
6645 Enable or disable logging of normal, successful connections
6646 May be used in sections : defaults | frontend | listen | backend
6647 yes | yes | yes | no
6648 Arguments : none
6649
6650 There are large sites dealing with several thousand connections per second
6651 and for which logging is a major pain. Some of them are even forced to turn
6652 logs off and cannot debug production issues. Setting this option ensures that
6653 normal connections, those which experience no error, no timeout, no retry nor
6654 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
6655 mode, the response status code is checked and return codes 5xx will still be
6656 logged.
6657
6658 It is strongly discouraged to use this option as most of the time, the key to
6659 complex issues is in the normal logs which will not be logged here. If you
6660 need to separate logs, see the "log-separate-errors" option instead.
6661
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006662 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006663 logging.
6664
6665
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006666option dontlognull
6667no option dontlognull
6668 Enable or disable logging of null connections
6669 May be used in sections : defaults | frontend | listen | backend
6670 yes | yes | yes | no
6671 Arguments : none
6672
6673 In certain environments, there are components which will regularly connect to
6674 various systems to ensure that they are still alive. It can be the case from
6675 another load balancer as well as from monitoring systems. By default, even a
6676 simple port probe or scan will produce a log. If those connections pollute
6677 the logs too much, it is possible to enable option "dontlognull" to indicate
6678 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02006679 which typically corresponds to those probes. Note that errors will still be
6680 returned to the client and accounted for in the stats. If this is not what is
6681 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006682
6683 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01006684 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006685 would not be logged.
6686
6687 If this option has been enabled in a "defaults" section, it can be disabled
6688 in a specific instance by prepending the "no" keyword before it.
6689
Willy Tarreau0f228a02015-05-01 15:37:53 +02006690 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
6691 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006692
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006693
Willy Tarreau87cf5142011-08-19 22:57:24 +02006694option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006695 Enable insertion of the X-Forwarded-For header to requests sent to servers
6696 May be used in sections : defaults | frontend | listen | backend
6697 yes | yes | yes | yes
6698 Arguments :
6699 <network> is an optional argument used to disable this option for sources
6700 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02006701 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01006702 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006703
6704 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
6705 their client address. This is sometimes annoying when the client's IP address
6706 is expected in server logs. To solve this problem, the well-known HTTP header
6707 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
6708 This header contains a value representing the client's IP address. Since this
6709 header is always appended at the end of the existing header list, the server
6710 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02006711 the server's manual to find how to enable use of this standard header. Note
6712 that only the last occurrence of the header must be used, since it is really
6713 possible that the client has already brought one.
6714
Willy Tarreaud72758d2010-01-12 10:42:19 +01006715 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02006716 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01006717 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01006718 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006719 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02006720 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01006721
6722 Sometimes, a same HAProxy instance may be shared between a direct client
6723 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6724 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6725 header for a known source address or network by adding the "except" keyword
6726 followed by the network address. In this case, any source IP matching the
6727 network will not cause an addition of this header. Most common uses are with
6728 private networks or 127.0.0.1.
6729
Willy Tarreau87cf5142011-08-19 22:57:24 +02006730 Alternatively, the keyword "if-none" states that the header will only be
6731 added if it is not present. This should only be used in perfectly trusted
6732 environment, as this might cause a security issue if headers reaching haproxy
6733 are under the control of the end-user.
6734
Willy Tarreauc27debf2008-01-06 08:57:02 +01006735 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02006736 least one of them uses it, the header will be added. Note that the backend's
6737 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02006738 both are defined. In the case of the "if-none" argument, if at least one of
6739 the frontend or the backend does not specify it, it wants the addition to be
6740 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006741
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02006742 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01006743 # Public HTTP address also used by stunnel on the same machine
6744 frontend www
6745 mode http
6746 option forwardfor except 127.0.0.1 # stunnel already adds the header
6747
Ross Westaf72a1d2008-08-03 10:51:45 +02006748 # Those servers want the IP Address in X-Client
6749 backend www
6750 mode http
6751 option forwardfor header X-Client
6752
Willy Tarreau87cf5142011-08-19 22:57:24 +02006753 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006754 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01006755
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006756
Christopher Faulet98fbe952019-07-22 16:18:24 +02006757option h1-case-adjust-bogus-client
6758no option h1-case-adjust-bogus-client
6759 Enable or disable the case adjustment of HTTP/1 headers sent to bogus clients
6760 May be used in sections : defaults | frontend | listen | backend
6761 yes | yes | yes | no
6762 Arguments : none
6763
6764 There is no standard case for header names because, as stated in RFC7230,
6765 they are case-insensitive. So applications must handle them in a case-
6766 insensitive manner. But some bogus applications violate the standards and
6767 erroneously rely on the cases most commonly used by browsers. This problem
6768 becomes critical with HTTP/2 because all header names must be exchanged in
6769 lower case, and HAProxy follows the same convention. All header names are
6770 sent in lower case to clients and servers, regardless of the HTTP version.
6771
6772 When HAProxy receives an HTTP/1 response, its header names are converted to
6773 lower case and manipulated and sent this way to the clients. If a client is
6774 known to violate the HTTP standards and to fail to process a response coming
6775 from HAProxy, it is possible to transform the lower case header names to a
6776 different format when the response is formatted and sent to the client, by
6777 enabling this option and specifying the list of headers to be reformatted
6778 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6779 must only be a temporary workaround for the time it takes the client to be
6780 fixed, because clients which require such workarounds might be vulnerable to
6781 content smuggling attacks and must absolutely be fixed.
6782
6783 Please note that this option will not affect standards-compliant clients.
6784
6785 If this option has been enabled in a "defaults" section, it can be disabled
6786 in a specific instance by prepending the "no" keyword before it.
6787
6788 See also: "option h1-case-adjust-bogus-server", "h1-case-adjust",
6789 "h1-case-adjust-file".
6790
6791
6792option h1-case-adjust-bogus-server
6793no option h1-case-adjust-bogus-server
6794 Enable or disable the case adjustment of HTTP/1 headers sent to bogus servers
6795 May be used in sections : defaults | frontend | listen | backend
6796 yes | no | yes | yes
6797 Arguments : none
6798
6799 There is no standard case for header names because, as stated in RFC7230,
6800 they are case-insensitive. So applications must handle them in a case-
6801 insensitive manner. But some bogus applications violate the standards and
6802 erroneously rely on the cases most commonly used by browsers. This problem
6803 becomes critical with HTTP/2 because all header names must be exchanged in
6804 lower case, and HAProxy follows the same convention. All header names are
6805 sent in lower case to clients and servers, regardless of the HTTP version.
6806
6807 When HAProxy receives an HTTP/1 request, its header names are converted to
6808 lower case and manipulated and sent this way to the servers. If a server is
6809 known to violate the HTTP standards and to fail to process a request coming
6810 from HAProxy, it is possible to transform the lower case header names to a
6811 different format when the request is formatted and sent to the server, by
6812 enabling this option and specifying the list of headers to be reformatted
6813 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6814 must only be a temporary workaround for the time it takes the server to be
6815 fixed, because servers which require such workarounds might be vulnerable to
6816 content smuggling attacks and must absolutely be fixed.
6817
6818 Please note that this option will not affect standards-compliant servers.
6819
6820 If this option has been enabled in a "defaults" section, it can be disabled
6821 in a specific instance by prepending the "no" keyword before it.
6822
6823 See also: "option h1-case-adjust-bogus-client", "h1-case-adjust",
6824 "h1-case-adjust-file".
6825
6826
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006827option http-buffer-request
6828no option http-buffer-request
6829 Enable or disable waiting for whole HTTP request body before proceeding
6830 May be used in sections : defaults | frontend | listen | backend
6831 yes | yes | yes | yes
6832 Arguments : none
6833
6834 It is sometimes desirable to wait for the body of an HTTP request before
6835 taking a decision. This is what is being done by "balance url_param" for
6836 example. The first use case is to buffer requests from slow clients before
6837 connecting to the server. Another use case consists in taking the routing
6838 decision based on the request body's contents. This option placed in a
6839 frontend or backend forces the HTTP processing to wait until either the whole
Christopher Faulet6db8a2e2019-11-19 16:27:25 +01006840 body is received or the request buffer is full. It can have undesired side
6841 effects with some applications abusing HTTP by expecting unbuffered
6842 transmissions between the frontend and the backend, so this should definitely
6843 not be used by default.
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006844
Baptiste Assmanneccdf432015-10-28 13:49:01 +01006845 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006846
6847
Willy Tarreau0f228a02015-05-01 15:37:53 +02006848option http-ignore-probes
6849no option http-ignore-probes
6850 Enable or disable logging of null connections and request timeouts
6851 May be used in sections : defaults | frontend | listen | backend
6852 yes | yes | yes | no
6853 Arguments : none
6854
6855 Recently some browsers started to implement a "pre-connect" feature
6856 consisting in speculatively connecting to some recently visited web sites
6857 just in case the user would like to visit them. This results in many
6858 connections being established to web sites, which end up in 408 Request
6859 Timeout if the timeout strikes first, or 400 Bad Request when the browser
6860 decides to close them first. These ones pollute the log and feed the error
6861 counters. There was already "option dontlognull" but it's insufficient in
6862 this case. Instead, this option does the following things :
6863 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01006864 was received over a connection before it was closed;
6865 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02006866 - prevent any error counter from being incremented
6867
6868 That way the empty connection is silently ignored. Note that it is better
6869 not to use this unless it is clear that it is needed, because it will hide
6870 real problems. The most common reason for not receiving a request and seeing
6871 a 408 is due to an MTU inconsistency between the client and an intermediary
6872 element such as a VPN, which blocks too large packets. These issues are
6873 generally seen with POST requests as well as GET with large cookies. The logs
6874 are often the only way to detect them.
6875
6876 If this option has been enabled in a "defaults" section, it can be disabled
6877 in a specific instance by prepending the "no" keyword before it.
6878
6879 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
6880
6881
Willy Tarreau16bfb022010-01-16 19:48:41 +01006882option http-keep-alive
6883no option http-keep-alive
6884 Enable or disable HTTP keep-alive from client to server
6885 May be used in sections : defaults | frontend | listen | backend
6886 yes | yes | yes | yes
6887 Arguments : none
6888
Willy Tarreau70dffda2014-01-30 03:07:23 +01006889 By default HAProxy operates in keep-alive mode with regards to persistent
6890 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02006891 leaves the connection idle on both sides between the end of a response and
6892 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02006893 as "option http-server-close" or "option httpclose". This option allows to
6894 set back the keep-alive mode, which can be useful when another mode was used
6895 in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01006896
6897 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
6898 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01006899 network) and the fastest session reuse on the server side at the expense
6900 of maintaining idle connections to the servers. In general, it is possible
6901 with this option to achieve approximately twice the request rate that the
6902 "http-server-close" option achieves on small objects. There are mainly two
6903 situations where this option may be useful :
6904
6905 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01006906 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01006907
6908 - when the cost of establishing the connection to the server is significant
6909 compared to the cost of retrieving the associated object from the server.
6910
6911 This last case can happen when the server is a fast static server of cache.
6912 In this case, the server will need to be properly tuned to support high enough
6913 connection counts because connections will last until the client sends another
6914 request.
6915
6916 If the client request has to go to another backend or another server due to
6917 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01006918 immediately be closed and a new one re-opened. Option "prefer-last-server" is
6919 available to try optimize server selection so that if the server currently
6920 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01006921
Willy Tarreau16bfb022010-01-16 19:48:41 +01006922 At the moment, logs will not indicate whether requests came from the same
6923 session or not. The accept date reported in the logs corresponds to the end
6924 of the previous request, and the request time corresponds to the time spent
6925 waiting for a new request. The keep-alive request time is still bound to the
6926 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6927 not set.
6928
Christopher Faulet159e6672019-07-16 15:09:52 +02006929 This option disables and replaces any previous "option httpclose" or "option
6930 http-server-close". When backend and frontend options differ, all of these 4
6931 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006932
Christopher Faulet315b39c2018-09-21 16:26:19 +02006933 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01006934 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01006935 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006936
6937
Willy Tarreau96e31212011-05-30 18:10:30 +02006938option http-no-delay
6939no option http-no-delay
6940 Instruct the system to favor low interactive delays over performance in HTTP
6941 May be used in sections : defaults | frontend | listen | backend
6942 yes | yes | yes | yes
6943 Arguments : none
6944
6945 In HTTP, each payload is unidirectional and has no notion of interactivity.
6946 Any agent is expected to queue data somewhat for a reasonably low delay.
6947 There are some very rare server-to-server applications that abuse the HTTP
6948 protocol and expect the payload phase to be highly interactive, with many
6949 interleaved data chunks in both directions within a single request. This is
6950 absolutely not supported by the HTTP specification and will not work across
6951 most proxies or servers. When such applications attempt to do this through
6952 haproxy, it works but they will experience high delays due to the network
6953 optimizations which favor performance by instructing the system to wait for
6954 enough data to be available in order to only send full packets. Typical
6955 delays are around 200 ms per round trip. Note that this only happens with
6956 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
6957 affected.
6958
6959 When "option http-no-delay" is present in either the frontend or the backend
6960 used by a connection, all such optimizations will be disabled in order to
6961 make the exchanges as fast as possible. Of course this offers no guarantee on
6962 the functionality, as it may break at any other place. But if it works via
6963 HAProxy, it will work as fast as possible. This option should never be used
6964 by default, and should never be used at all unless such a buggy application
6965 is discovered. The impact of using this option is an increase of bandwidth
6966 usage and CPU usage, which may significantly lower performance in high
6967 latency environments.
6968
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006969 See also : "option http-buffer-request"
6970
Willy Tarreau96e31212011-05-30 18:10:30 +02006971
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006972option http-pretend-keepalive
6973no option http-pretend-keepalive
6974 Define whether haproxy will announce keepalive to the server or not
6975 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02006976 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006977 Arguments : none
6978
Christopher Faulet315b39c2018-09-21 16:26:19 +02006979 When running with "option http-server-close" or "option httpclose", haproxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006980 adds a "Connection: close" header to the request forwarded to the server.
6981 Unfortunately, when some servers see this header, they automatically refrain
6982 from using the chunked encoding for responses of unknown length, while this
6983 is totally unrelated. The immediate effect is that this prevents haproxy from
6984 maintaining the client connection alive. A second effect is that a client or
6985 a cache could receive an incomplete response without being aware of it, and
6986 consider the response complete.
6987
6988 By setting "option http-pretend-keepalive", haproxy will make the server
6989 believe it will keep the connection alive. The server will then not fall back
6990 to the abnormal undesired above. When haproxy gets the whole response, it
6991 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02006992 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006993 connection is correctly closed on the server side.
6994
6995 It is recommended not to enable this option by default, because most servers
6996 will more efficiently close the connection themselves after the last packet,
6997 and release its buffers slightly earlier. Also, the added packet on the
6998 network could slightly reduce the overall peak performance. However it is
6999 worth noting that when this option is enabled, haproxy will have slightly
7000 less work to do. So if haproxy is the bottleneck on the whole architecture,
7001 enabling this option might save a few CPU cycles.
7002
Christopher Faulet98db9762018-09-21 10:25:19 +02007003 This option may be set in backend and listen sections. Using it in a frontend
7004 section will be ignored and a warning will be reported during startup. It is
7005 a backend related option, so there is no real reason to set it on a
7006 frontend. This option may be combined with "option httpclose", which will
7007 cause keepalive to be announced to the server and close to be announced to
7008 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02007009
7010 If this option has been enabled in a "defaults" section, it can be disabled
7011 in a specific instance by prepending the "no" keyword before it.
7012
Christopher Faulet315b39c2018-09-21 16:26:19 +02007013 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01007014 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02007015
Willy Tarreauc27debf2008-01-06 08:57:02 +01007016
Willy Tarreaub608feb2010-01-02 22:47:18 +01007017option http-server-close
7018no option http-server-close
7019 Enable or disable HTTP connection closing on the server side
7020 May be used in sections : defaults | frontend | listen | backend
7021 yes | yes | yes | yes
7022 Arguments : none
7023
Willy Tarreau70dffda2014-01-30 03:07:23 +01007024 By default HAProxy operates in keep-alive mode with regards to persistent
7025 connections: for each connection it processes each request and response, and
7026 leaves the connection idle on both sides between the end of a response and
7027 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02007028 as "option http-server-close" or "option httpclose". Setting "option
7029 http-server-close" enables HTTP connection-close mode on the server side
7030 while keeping the ability to support HTTP keep-alive and pipelining on the
7031 client side. This provides the lowest latency on the client side (slow
7032 network) and the fastest session reuse on the server side to save server
7033 resources, similarly to "option httpclose". It also permits non-keepalive
7034 capable servers to be served in keep-alive mode to the clients if they
7035 conform to the requirements of RFC7230. Please note that some servers do not
7036 always conform to those requirements when they see "Connection: close" in the
7037 request. The effect will be that keep-alive will never be used. A workaround
7038 consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01007039
7040 At the moment, logs will not indicate whether requests came from the same
7041 session or not. The accept date reported in the logs corresponds to the end
7042 of the previous request, and the request time corresponds to the time spent
7043 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01007044 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
7045 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01007046
7047 This option may be set both in a frontend and in a backend. It is enabled if
7048 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02007049 It disables and replaces any previous "option httpclose" or "option
7050 http-keep-alive". Please check section 4 ("Proxies") to see how this option
7051 combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01007052
7053 If this option has been enabled in a "defaults" section, it can be disabled
7054 in a specific instance by prepending the "no" keyword before it.
7055
Christopher Faulet315b39c2018-09-21 16:26:19 +02007056 See also : "option httpclose", "option http-pretend-keepalive",
7057 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01007058
Willy Tarreau88d349d2010-01-25 12:15:43 +01007059option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01007060no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01007061 Make use of non-standard Proxy-Connection header instead of Connection
7062 May be used in sections : defaults | frontend | listen | backend
7063 yes | yes | yes | no
7064 Arguments : none
7065
Lukas Tribus23953682017-04-28 13:24:30 +00007066 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01007067 Connection header to indicate their wish of persistent or non-persistent
7068 connections, both browsers and proxies ignore this header for proxied
7069 connections and make use of the undocumented, non-standard Proxy-Connection
7070 header instead. The issue begins when trying to put a load balancer between
7071 browsers and such proxies, because there will be a difference between what
7072 haproxy understands and what the client and the proxy agree on.
7073
7074 By setting this option in a frontend, haproxy can automatically switch to use
7075 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01007076 defined here as one where the URI begins with neither a '/' nor a '*'. This
7077 is incompatible with the HTTP tunnel mode. Note that this option can only be
7078 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01007079
Willy Tarreau844a7e72010-01-31 21:46:18 +01007080 Also, when this option is set, a request which requires authentication will
7081 automatically switch to use proxy authentication headers if it is itself a
7082 proxied request. That makes it possible to check or enforce authentication in
7083 front of an existing proxy.
7084
Willy Tarreau88d349d2010-01-25 12:15:43 +01007085 This option should normally never be used, except in front of a proxy.
7086
Christopher Faulet315b39c2018-09-21 16:26:19 +02007087 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01007088
Willy Tarreaud63335a2010-02-26 12:56:52 +01007089option httpchk
7090option httpchk <uri>
7091option httpchk <method> <uri>
7092option httpchk <method> <uri> <version>
7093 Enable HTTP protocol to check on the servers health
7094 May be used in sections : defaults | frontend | listen | backend
7095 yes | no | yes | yes
7096 Arguments :
7097 <method> is the optional HTTP method used with the requests. When not set,
7098 the "OPTIONS" method is used, as it generally requires low server
7099 processing and is easy to filter out from the logs. Any method
7100 may be used, though it is not recommended to invent non-standard
7101 ones.
7102
7103 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
7104 which is accessible by default on almost any server, but may be
7105 changed to any other URI. Query strings are permitted.
7106
7107 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
7108 but some servers might behave incorrectly in HTTP 1.0, so turning
7109 it to HTTP/1.1 may sometimes help. Note that the Host field is
7110 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
7111 after "\r\n" following the version string.
7112
7113 By default, server health checks only consist in trying to establish a TCP
7114 connection. When "option httpchk" is specified, a complete HTTP request is
7115 sent once the TCP connection is established, and responses 2xx and 3xx are
7116 considered valid, while all other ones indicate a server failure, including
7117 the lack of any response.
7118
7119 The port and interval are specified in the server configuration.
7120
7121 This option does not necessarily require an HTTP backend, it also works with
7122 plain TCP backends. This is particularly useful to check simple scripts bound
7123 to some dedicated ports using the inetd daemon.
7124
7125 Examples :
7126 # Relay HTTPS traffic to Apache instance and check service availability
7127 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
7128 backend https_relay
7129 mode tcp
7130 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
7131 server apache1 192.168.1.1:443 check port 80
7132
Simon Hormanafc47ee2013-11-25 10:46:35 +09007133 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
7134 "option pgsql-check", "http-check" and the "check", "port" and
7135 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01007136
7137
Willy Tarreauc27debf2008-01-06 08:57:02 +01007138option httpclose
7139no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02007140 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01007141 May be used in sections : defaults | frontend | listen | backend
7142 yes | yes | yes | yes
7143 Arguments : none
7144
Willy Tarreau70dffda2014-01-30 03:07:23 +01007145 By default HAProxy operates in keep-alive mode with regards to persistent
7146 connections: for each connection it processes each request and response, and
7147 leaves the connection idle on both sides between the end of a response and
7148 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02007149 as "option http-server-close" or "option httpclose".
Willy Tarreau70dffda2014-01-30 03:07:23 +01007150
Christopher Faulet315b39c2018-09-21 16:26:19 +02007151 If "option httpclose" is set, HAProxy will close connections with the server
7152 and the client as soon as the request and the response are received. It will
John Roeslerfb2fce12019-07-10 15:45:51 -05007153 also check if a "Connection: close" header is already set in each direction,
Christopher Faulet315b39c2018-09-21 16:26:19 +02007154 and will add one if missing. Any "Connection" header different from "close"
7155 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01007156
Christopher Faulet315b39c2018-09-21 16:26:19 +02007157 This option may also be combined with "option http-pretend-keepalive", which
7158 will disable sending of the "Connection: close" header, but will still cause
7159 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01007160
7161 This option may be set both in a frontend and in a backend. It is enabled if
7162 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02007163 It disables and replaces any previous "option http-server-close" or "option
7164 http-keep-alive". Please check section 4 ("Proxies") to see how this option
7165 combines with others when frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01007166
7167 If this option has been enabled in a "defaults" section, it can be disabled
7168 in a specific instance by prepending the "no" keyword before it.
7169
Christopher Faulet315b39c2018-09-21 16:26:19 +02007170 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01007171
7172
Emeric Brun3a058f32009-06-30 18:26:00 +02007173option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01007174 Enable logging of HTTP request, session state and timers
7175 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01007176 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02007177 Arguments :
7178 clf if the "clf" argument is added, then the output format will be
7179 the CLF format instead of HAProxy's default HTTP format. You can
7180 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01007181 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02007182 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01007183
7184 By default, the log output format is very poor, as it only contains the
7185 source and destination addresses, and the instance name. By specifying
7186 "option httplog", each log line turns into a much richer format including,
7187 but not limited to, the HTTP request, the connection timers, the session
7188 status, the connections numbers, the captured headers and cookies, the
7189 frontend, backend and server name, and of course the source address and
7190 ports.
7191
PiBa-NLbd556bf2014-12-11 21:31:54 +01007192 Specifying only "option httplog" will automatically clear the 'clf' mode
7193 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02007194
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007195 "option httplog" overrides any previous "log-format" directive.
7196
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007197 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01007198
Willy Tarreau55165fe2009-05-10 12:02:55 +02007199
7200option http_proxy
7201no option http_proxy
7202 Enable or disable plain HTTP proxy mode
7203 May be used in sections : defaults | frontend | listen | backend
7204 yes | yes | yes | yes
7205 Arguments : none
7206
7207 It sometimes happens that people need a pure HTTP proxy which understands
7208 basic proxy requests without caching nor any fancy feature. In this case,
7209 it may be worth setting up an HAProxy instance with the "option http_proxy"
7210 set. In this mode, no server is declared, and the connection is forwarded to
7211 the IP address and port found in the URL after the "http://" scheme.
7212
7213 No host address resolution is performed, so this only works when pure IP
7214 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01007215 it will probably be used only by experts who know they need exactly it. This
7216 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02007217
7218 If this option has been enabled in a "defaults" section, it can be disabled
7219 in a specific instance by prepending the "no" keyword before it.
7220
7221 Example :
7222 # this backend understands HTTP proxy requests and forwards them directly.
7223 backend direct_forward
7224 option httpclose
7225 option http_proxy
7226
7227 See also : "option httpclose"
7228
Willy Tarreau211ad242009-10-03 21:45:07 +02007229
Jamie Gloudon801a0a32012-08-25 00:18:33 -04007230option independent-streams
7231no option independent-streams
7232 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02007233 May be used in sections : defaults | frontend | listen | backend
7234 yes | yes | yes | yes
7235 Arguments : none
7236
7237 By default, when data is sent over a socket, both the write timeout and the
7238 read timeout for that socket are refreshed, because we consider that there is
7239 activity on that socket, and we have no other means of guessing if we should
7240 receive data or not.
7241
Davor Ocelice9ed2812017-12-25 17:49:28 +01007242 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02007243 exists a situation where it is desirable to disable it, and only refresh the
7244 read timeout if there are incoming data. This happens on sessions with large
7245 timeouts and low amounts of exchanged data such as telnet session. If the
7246 server suddenly disappears, the output data accumulates in the system's
7247 socket buffers, both timeouts are correctly refreshed, and there is no way
7248 to know the server does not receive them, so we don't timeout. However, when
7249 the underlying protocol always echoes sent data, it would be enough by itself
7250 to detect the issue using the read timeout. Note that this problem does not
7251 happen with more verbose protocols because data won't accumulate long in the
7252 socket buffers.
7253
7254 When this option is set on the frontend, it will disable read timeout updates
7255 on data sent to the client. There probably is little use of this case. When
7256 the option is set on the backend, it will disable read timeout updates on
7257 data sent to the server. Doing so will typically break large HTTP posts from
7258 slow lines, so use it with caution.
7259
Willy Tarreauce887fd2012-05-12 12:50:00 +02007260 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02007261
7262
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02007263option ldap-check
7264 Use LDAPv3 health checks for server testing
7265 May be used in sections : defaults | frontend | listen | backend
7266 yes | no | yes | yes
7267 Arguments : none
7268
7269 It is possible to test that the server correctly talks LDAPv3 instead of just
7270 testing that it accepts the TCP connection. When this option is set, an
7271 LDAPv3 anonymous simple bind message is sent to the server, and the response
7272 is analyzed to find an LDAPv3 bind response message.
7273
7274 The server is considered valid only when the LDAP response contains success
7275 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
7276
7277 Logging of bind requests is server dependent see your documentation how to
7278 configure it.
7279
7280 Example :
7281 option ldap-check
7282
7283 See also : "option httpchk"
7284
7285
Simon Horman98637e52014-06-20 12:30:16 +09007286option external-check
7287 Use external processes for server health checks
7288 May be used in sections : defaults | frontend | listen | backend
7289 yes | no | yes | yes
7290
7291 It is possible to test the health of a server using an external command.
7292 This is achieved by running the executable set using "external-check
7293 command".
7294
7295 Requires the "external-check" global to be set.
7296
7297 See also : "external-check", "external-check command", "external-check path"
7298
7299
Willy Tarreau211ad242009-10-03 21:45:07 +02007300option log-health-checks
7301no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02007302 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02007303 May be used in sections : defaults | frontend | listen | backend
7304 yes | no | yes | yes
7305 Arguments : none
7306
Willy Tarreaubef1b322014-05-13 21:01:39 +02007307 By default, failed health check are logged if server is UP and successful
7308 health checks are logged if server is DOWN, so the amount of additional
7309 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02007310
Willy Tarreaubef1b322014-05-13 21:01:39 +02007311 When this option is enabled, any change of the health check status or to
7312 the server's health will be logged, so that it becomes possible to know
7313 that a server was failing occasional checks before crashing, or exactly when
7314 it failed to respond a valid HTTP status, then when the port started to
7315 reject connections, then when the server stopped responding at all.
7316
Davor Ocelice9ed2812017-12-25 17:49:28 +01007317 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02007318 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02007319
Willy Tarreaubef1b322014-05-13 21:01:39 +02007320 See also: "option httpchk", "option ldap-check", "option mysql-check",
7321 "option pgsql-check", "option redis-check", "option smtpchk",
7322 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02007323
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02007324
7325option log-separate-errors
7326no option log-separate-errors
7327 Change log level for non-completely successful connections
7328 May be used in sections : defaults | frontend | listen | backend
7329 yes | yes | yes | no
7330 Arguments : none
7331
7332 Sometimes looking for errors in logs is not easy. This option makes haproxy
7333 raise the level of logs containing potentially interesting information such
7334 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
7335 level changes from "info" to "err". This makes it possible to log them
7336 separately to a different file with most syslog daemons. Be careful not to
7337 remove them from the original file, otherwise you would lose ordering which
7338 provides very important information.
7339
7340 Using this option, large sites dealing with several thousand connections per
7341 second may log normal traffic to a rotating buffer and only archive smaller
7342 error logs.
7343
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007344 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02007345 logging.
7346
Willy Tarreauc27debf2008-01-06 08:57:02 +01007347
7348option logasap
7349no option logasap
7350 Enable or disable early logging of HTTP requests
7351 May be used in sections : defaults | frontend | listen | backend
7352 yes | yes | yes | no
7353 Arguments : none
7354
7355 By default, HTTP requests are logged upon termination so that the total
7356 transfer time and the number of bytes appear in the logs. When large objects
7357 are being transferred, it may take a while before the request appears in the
7358 logs. Using "option logasap", the request gets logged as soon as the server
7359 sends the complete headers. The only missing information in the logs will be
7360 the total number of bytes which will indicate everything except the amount
7361 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007362 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01007363 "Content-Length" response header so that the logs at least indicate how many
7364 bytes are expected to be transferred.
7365
Willy Tarreaucc6c8912009-02-22 10:53:55 +01007366 Examples :
7367 listen http_proxy 0.0.0.0:80
7368 mode http
7369 option httplog
7370 option logasap
7371 log 192.168.2.200 local3
7372
7373 >>> Feb 6 12:14:14 localhost \
7374 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
7375 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
7376 "GET /image.iso HTTP/1.0"
7377
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007378 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01007379 logging.
7380
7381
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02007382option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007383 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007384 May be used in sections : defaults | frontend | listen | backend
7385 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007386 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007387 <username> This is the username which will be used when connecting to MySQL
7388 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02007389 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007390
7391 If you specify a username, the check consists of sending two MySQL packet,
7392 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01007393 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007394 Error packet. It is a basic but useful test which does not produce error nor
7395 aborted connect on the server. However, it requires adding an authorization
7396 in the MySQL table, like this :
7397
7398 USE mysql;
7399 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
7400 FLUSH PRIVILEGES;
7401
7402 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007403 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007404 Error packet, we don't send anything in this mode. It was reported that it
7405 can generate lockout if check is too frequent and/or if there is not enough
7406 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
7407 value as if a connection is established successfully within fewer than MySQL
7408 "max_connect_errors" attempts after a previous connection was interrupted,
7409 the error count for the host is cleared to zero. If HAProxy's server get
7410 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
7411
7412 Remember that this does not check database presence nor database consistency.
7413 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007414
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02007415 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007416
7417 Most often, an incoming MySQL server needs to see the client's IP address for
7418 various purposes, including IP privilege matching and connection logging.
7419 When possible, it is often wise to masquerade the client's IP address when
7420 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007421 which requires the transparent proxy feature to be compiled in, and the MySQL
7422 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007423
7424 See also: "option httpchk"
7425
7426
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007427option nolinger
7428no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007429 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007430 May be used in sections: defaults | frontend | listen | backend
7431 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007432 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007433
Davor Ocelice9ed2812017-12-25 17:49:28 +01007434 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007435 physically disconnected), the session timeouts triggers and the session is
7436 closed. But it will remain in FIN_WAIT1 state for some time in the system,
7437 using some resources and possibly limiting the ability to establish newer
7438 connections.
7439
7440 When this happens, it is possible to activate "option nolinger" which forces
7441 the system to immediately remove any socket's pending data on close. Thus,
7442 the session is instantly purged from the system's tables. This usually has
7443 side effects such as increased number of TCP resets due to old retransmits
7444 getting immediately rejected. Some firewalls may sometimes complain about
7445 this too.
7446
7447 For this reason, it is not recommended to use this option when not absolutely
7448 needed. You know that you need it when you have thousands of FIN_WAIT1
7449 sessions on your system (TIME_WAIT ones do not count).
7450
7451 This option may be used both on frontends and backends, depending on the side
7452 where it is required. Use it on the frontend for clients, and on the backend
7453 for servers.
7454
7455 If this option has been enabled in a "defaults" section, it can be disabled
7456 in a specific instance by prepending the "no" keyword before it.
7457
7458
Willy Tarreau55165fe2009-05-10 12:02:55 +02007459option originalto [ except <network> ] [ header <name> ]
7460 Enable insertion of the X-Original-To header to requests sent to servers
7461 May be used in sections : defaults | frontend | listen | backend
7462 yes | yes | yes | yes
7463 Arguments :
7464 <network> is an optional argument used to disable this option for sources
7465 matching <network>
7466 <name> an optional argument to specify a different "X-Original-To"
7467 header name.
7468
7469 Since HAProxy can work in transparent mode, every request from a client can
7470 be redirected to the proxy and HAProxy itself can proxy every request to a
7471 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
7472 be lost. This is annoying when you want access rules based on destination ip
7473 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
7474 added by HAProxy to all requests sent to the server. This header contains a
7475 value representing the original destination IP address. Since this must be
7476 configured to always use the last occurrence of this header only. Note that
7477 only the last occurrence of the header must be used, since it is really
7478 possible that the client has already brought one.
7479
7480 The keyword "header" may be used to supply a different header name to replace
7481 the default "X-Original-To". This can be useful where you might already
7482 have a "X-Original-To" header from a different application, and you need
7483 preserve it. Also if your backend server doesn't use the "X-Original-To"
7484 header and requires different one.
7485
7486 Sometimes, a same HAProxy instance may be shared between a direct client
7487 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
7488 used to decrypt HTTPS traffic). It is possible to disable the addition of the
7489 header for a known source address or network by adding the "except" keyword
7490 followed by the network address. In this case, any source IP matching the
7491 network will not cause an addition of this header. Most common uses are with
7492 private networks or 127.0.0.1.
7493
7494 This option may be specified either in the frontend or in the backend. If at
7495 least one of them uses it, the header will be added. Note that the backend's
7496 setting of the header subargument takes precedence over the frontend's if
7497 both are defined.
7498
Willy Tarreau55165fe2009-05-10 12:02:55 +02007499 Examples :
7500 # Original Destination address
7501 frontend www
7502 mode http
7503 option originalto except 127.0.0.1
7504
7505 # Those servers want the IP Address in X-Client-Dst
7506 backend www
7507 mode http
7508 option originalto header X-Client-Dst
7509
Christopher Faulet315b39c2018-09-21 16:26:19 +02007510 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02007511
7512
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007513option persist
7514no option persist
7515 Enable or disable forced persistence on down servers
7516 May be used in sections: defaults | frontend | listen | backend
7517 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007518 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007519
7520 When an HTTP request reaches a backend with a cookie which references a dead
7521 server, by default it is redispatched to another server. It is possible to
7522 force the request to be sent to the dead server first using "option persist"
7523 if absolutely needed. A common use case is when servers are under extreme
7524 load and spend their time flapping. In this case, the users would still be
7525 directed to the server they opened the session on, in the hope they would be
7526 correctly served. It is recommended to use "option redispatch" in conjunction
7527 with this option so that in the event it would not be possible to connect to
7528 the server at all (server definitely dead), the client would finally be
7529 redirected to another valid server.
7530
7531 If this option has been enabled in a "defaults" section, it can be disabled
7532 in a specific instance by prepending the "no" keyword before it.
7533
Willy Tarreau4de91492010-01-22 19:10:05 +01007534 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007535
7536
Willy Tarreau0c122822013-12-15 18:49:01 +01007537option pgsql-check [ user <username> ]
7538 Use PostgreSQL health checks for server testing
7539 May be used in sections : defaults | frontend | listen | backend
7540 yes | no | yes | yes
7541 Arguments :
7542 <username> This is the username which will be used when connecting to
7543 PostgreSQL server.
7544
7545 The check sends a PostgreSQL StartupMessage and waits for either
7546 Authentication request or ErrorResponse message. It is a basic but useful
7547 test which does not produce error nor aborted connect on the server.
7548 This check is identical with the "mysql-check".
7549
7550 See also: "option httpchk"
7551
7552
Willy Tarreau9420b122013-12-15 18:58:25 +01007553option prefer-last-server
7554no option prefer-last-server
7555 Allow multiple load balanced requests to remain on the same server
7556 May be used in sections: defaults | frontend | listen | backend
7557 yes | no | yes | yes
7558 Arguments : none
7559
7560 When the load balancing algorithm in use is not deterministic, and a previous
7561 request was sent to a server to which haproxy still holds a connection, it is
7562 sometimes desirable that subsequent requests on a same session go to the same
7563 server as much as possible. Note that this is different from persistence, as
7564 we only indicate a preference which haproxy tries to apply without any form
7565 of warranty. The real use is for keep-alive connections sent to servers. When
7566 this option is used, haproxy will try to reuse the same connection that is
7567 attached to the server instead of rebalancing to another server, causing a
7568 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01007569 not make much sense to use this in combination with hashing algorithms. Note,
7570 haproxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02007571 to a proxy which sends a 407 (authentication required), when the load
7572 balancing algorithm is not deterministic. This is mandatory for use with the
7573 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01007574 troubleshooting some faulty applications. Option prefer-last-server might be
7575 desirable in these environments as well, to avoid redistributing the traffic
7576 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01007577
7578 If this option has been enabled in a "defaults" section, it can be disabled
7579 in a specific instance by prepending the "no" keyword before it.
7580
7581 See also: "option http-keep-alive"
7582
7583
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007584option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07007585option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007586no option redispatch
7587 Enable or disable session redistribution in case of connection failure
7588 May be used in sections: defaults | frontend | listen | backend
7589 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07007590 Arguments :
7591 <interval> The optional integer value that controls how often redispatches
7592 occur when retrying connections. Positive value P indicates a
7593 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01007594 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07007595 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007596 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07007597 positive value of 1 would indicate a redispatch on every retry,
7598 and a positive value of 3 would indicate a redispatch on every
7599 third retry. You can disable redispatches with a value of 0.
7600
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007601
7602 In HTTP mode, if a server designated by a cookie is down, clients may
7603 definitely stick to it because they cannot flush the cookie, so they will not
7604 be able to access the service anymore.
7605
Willy Tarreau59884a62019-01-02 14:48:31 +01007606 Specifying "option redispatch" will allow the proxy to break cookie or
7607 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007608
Joseph Lynch726ab712015-05-11 23:25:34 -07007609 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007610 connection failures. Of course, it requires having "retries" set to a nonzero
7611 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007612
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007613 If this option has been enabled in a "defaults" section, it can be disabled
7614 in a specific instance by prepending the "no" keyword before it.
7615
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02007616 See also : "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007617
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007618
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007619option redis-check
7620 Use redis health checks for server testing
7621 May be used in sections : defaults | frontend | listen | backend
7622 yes | no | yes | yes
7623 Arguments : none
7624
7625 It is possible to test that the server correctly talks REDIS protocol instead
7626 of just testing that it accepts the TCP connection. When this option is set,
7627 a PING redis command is sent to the server, and the response is analyzed to
7628 find the "+PONG" response message.
7629
7630 Example :
7631 option redis-check
7632
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007633 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007634
7635
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007636option smtpchk
7637option smtpchk <hello> <domain>
7638 Use SMTP health checks for server testing
7639 May be used in sections : defaults | frontend | listen | backend
7640 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01007641 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007642 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02007643 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007644 values will be turned into the default command ("HELO").
7645
7646 <domain> is the domain name to present to the server. It may only be
7647 specified (and is mandatory) if the hello command has been
7648 specified. By default, "localhost" is used.
7649
7650 When "option smtpchk" is set, the health checks will consist in TCP
7651 connections followed by an SMTP command. By default, this command is
7652 "HELO localhost". The server's return code is analyzed and only return codes
7653 starting with a "2" will be considered as valid. All other responses,
7654 including a lack of response will constitute an error and will indicate a
7655 dead server.
7656
7657 This test is meant to be used with SMTP servers or relays. Depending on the
7658 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01007659 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007660 25 is often easier than adjusting the configuration.
7661
7662 Most often, an incoming SMTP server needs to see the client's IP address for
7663 various purposes, including spam filtering, anti-spoofing and logging. When
7664 possible, it is often wise to masquerade the client's IP address when
7665 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007666 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007667
7668 Example :
7669 option smtpchk HELO mydomain.org
7670
7671 See also : "option httpchk", "source"
7672
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007673
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02007674option socket-stats
7675no option socket-stats
7676
7677 Enable or disable collecting & providing separate statistics for each socket.
7678 May be used in sections : defaults | frontend | listen | backend
7679 yes | yes | yes | no
7680
7681 Arguments : none
7682
7683
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007684option splice-auto
7685no option splice-auto
7686 Enable or disable automatic kernel acceleration on sockets in both directions
7687 May be used in sections : defaults | frontend | listen | backend
7688 yes | yes | yes | yes
7689 Arguments : none
7690
7691 When this option is enabled either on a frontend or on a backend, haproxy
7692 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007693 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007694 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007695 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007696 are not much aggressive in order to limit excessive use of splicing. This
7697 option requires splicing to be enabled at compile time, and may be globally
7698 disabled with the global option "nosplice". Since splice uses pipes, using it
7699 requires that there are enough spare pipes.
7700
7701 Important note: kernel-based TCP splicing is a Linux-specific feature which
7702 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
7703 transfer data between sockets without copying these data to user-space, thus
7704 providing noticeable performance gains and CPU cycles savings. Since many
7705 early implementations are buggy, corrupt data and/or are inefficient, this
7706 feature is not enabled by default, and it should be used with extreme care.
7707 While it is not possible to detect the correctness of an implementation,
7708 2.6.29 is the first version offering a properly working implementation. In
7709 case of doubt, splicing may be globally disabled using the global "nosplice"
7710 keyword.
7711
7712 Example :
7713 option splice-auto
7714
7715 If this option has been enabled in a "defaults" section, it can be disabled
7716 in a specific instance by prepending the "no" keyword before it.
7717
7718 See also : "option splice-request", "option splice-response", and global
7719 options "nosplice" and "maxpipes"
7720
7721
7722option splice-request
7723no option splice-request
7724 Enable or disable automatic kernel acceleration on sockets for requests
7725 May be used in sections : defaults | frontend | listen | backend
7726 yes | yes | yes | yes
7727 Arguments : none
7728
7729 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007730 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007731 the client to the server. It might still use the recv/send scheme if there
7732 are no spare pipes left. This option requires splicing to be enabled at
7733 compile time, and may be globally disabled with the global option "nosplice".
7734 Since splice uses pipes, using it requires that there are enough spare pipes.
7735
7736 Important note: see "option splice-auto" for usage limitations.
7737
7738 Example :
7739 option splice-request
7740
7741 If this option has been enabled in a "defaults" section, it can be disabled
7742 in a specific instance by prepending the "no" keyword before it.
7743
7744 See also : "option splice-auto", "option splice-response", and global options
7745 "nosplice" and "maxpipes"
7746
7747
7748option splice-response
7749no option splice-response
7750 Enable or disable automatic kernel acceleration on sockets for responses
7751 May be used in sections : defaults | frontend | listen | backend
7752 yes | yes | yes | yes
7753 Arguments : none
7754
7755 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007756 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007757 the server to the client. It might still use the recv/send scheme if there
7758 are no spare pipes left. This option requires splicing to be enabled at
7759 compile time, and may be globally disabled with the global option "nosplice".
7760 Since splice uses pipes, using it requires that there are enough spare pipes.
7761
7762 Important note: see "option splice-auto" for usage limitations.
7763
7764 Example :
7765 option splice-response
7766
7767 If this option has been enabled in a "defaults" section, it can be disabled
7768 in a specific instance by prepending the "no" keyword before it.
7769
7770 See also : "option splice-auto", "option splice-request", and global options
7771 "nosplice" and "maxpipes"
7772
7773
Christopher Fauletba7bc162016-11-07 21:07:38 +01007774option spop-check
7775 Use SPOP health checks for server testing
7776 May be used in sections : defaults | frontend | listen | backend
7777 no | no | no | yes
7778 Arguments : none
7779
7780 It is possible to test that the server correctly talks SPOP protocol instead
7781 of just testing that it accepts the TCP connection. When this option is set,
7782 a HELLO handshake is performed between HAProxy and the server, and the
7783 response is analyzed to check no error is reported.
7784
7785 Example :
7786 option spop-check
7787
7788 See also : "option httpchk"
7789
7790
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007791option srvtcpka
7792no option srvtcpka
7793 Enable or disable the sending of TCP keepalive packets on the server side
7794 May be used in sections : defaults | frontend | listen | backend
7795 yes | no | yes | yes
7796 Arguments : none
7797
7798 When there is a firewall or any session-aware component between a client and
7799 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007800 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007801 components decides to expire a session which has remained idle for too long.
7802
7803 Enabling socket-level TCP keep-alives makes the system regularly send packets
7804 to the other end of the connection, leaving it active. The delay between
7805 keep-alive probes is controlled by the system only and depends both on the
7806 operating system and its tuning parameters.
7807
7808 It is important to understand that keep-alive packets are neither emitted nor
7809 received at the application level. It is only the network stacks which sees
7810 them. For this reason, even if one side of the proxy already uses keep-alives
7811 to maintain its connection alive, those keep-alive packets will not be
7812 forwarded to the other side of the proxy.
7813
7814 Please note that this has nothing to do with HTTP keep-alive.
7815
7816 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
7817 server side of a connection, which should help when session expirations are
7818 noticed between HAProxy and a server.
7819
7820 If this option has been enabled in a "defaults" section, it can be disabled
7821 in a specific instance by prepending the "no" keyword before it.
7822
7823 See also : "option clitcpka", "option tcpka"
7824
7825
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007826option ssl-hello-chk
7827 Use SSLv3 client hello health checks for server testing
7828 May be used in sections : defaults | frontend | listen | backend
7829 yes | no | yes | yes
7830 Arguments : none
7831
7832 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
7833 possible to test that the server correctly talks SSL instead of just testing
7834 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
7835 SSLv3 client hello messages are sent once the connection is established to
7836 the server, and the response is analyzed to find an SSL server hello message.
7837 The server is considered valid only when the response contains this server
7838 hello message.
7839
7840 All servers tested till there correctly reply to SSLv3 client hello messages,
7841 and most servers tested do not even log the requests containing only hello
7842 messages, which is appreciable.
7843
Willy Tarreau763a95b2012-10-04 23:15:39 +02007844 Note that this check works even when SSL support was not built into haproxy
7845 because it forges the SSL message. When SSL support is available, it is best
7846 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007847
Willy Tarreau763a95b2012-10-04 23:15:39 +02007848 See also: "option httpchk", "check-ssl"
7849
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007850
Willy Tarreaued179852013-12-16 01:07:00 +01007851option tcp-check
7852 Perform health checks using tcp-check send/expect sequences
7853 May be used in sections: defaults | frontend | listen | backend
7854 yes | no | yes | yes
7855
7856 This health check method is intended to be combined with "tcp-check" command
7857 lists in order to support send/expect types of health check sequences.
7858
7859 TCP checks currently support 4 modes of operations :
7860 - no "tcp-check" directive : the health check only consists in a connection
7861 attempt, which remains the default mode.
7862
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007863 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01007864 used to send a string along with a connection opening. With some
7865 protocols, it helps sending a "QUIT" message for example that prevents
7866 the server from logging a connection error for each health check. The
7867 check result will still be based on the ability to open the connection
7868 only.
7869
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007870 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01007871 The connection is opened and haproxy waits for the server to present some
7872 contents which must validate some rules. The check result will be based
7873 on the matching between the contents and the rules. This is suited for
7874 POP, IMAP, SMTP, FTP, SSH, TELNET.
7875
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007876 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007877 used to test a hello-type protocol. HAProxy sends a message, the server
7878 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007879 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01007880 suited for protocols which require a binding or a request/response model.
7881 LDAP, MySQL, Redis and SSL are example of such protocols, though they
7882 already all have their dedicated checks with a deeper understanding of
7883 the respective protocols.
7884 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01007885 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01007886
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007887 A fifth mode can be used to insert comments in different steps of the
7888 script.
7889
7890 For each tcp-check rule you create, you can add a "comment" directive,
7891 followed by a string. This string will be reported in the log and stderr
7892 in debug mode. It is useful to make user-friendly error reporting.
7893 The "comment" is of course optional.
7894
7895
Willy Tarreaued179852013-12-16 01:07:00 +01007896 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007897 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007898 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007899 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007900
Davor Ocelice9ed2812017-12-25 17:49:28 +01007901 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007902 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007903 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007904
7905 # look for the redis master server after ensuring it speaks well
7906 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01007907 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01007908 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007909 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007910 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02007911 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007912 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01007913 tcp-check send info\ replication\r\n
7914 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007915 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007916 tcp-check send QUIT\r\n
7917 tcp-check expect string +OK
7918
Davor Ocelice9ed2812017-12-25 17:49:28 +01007919 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01007920 (send many headers before analyzing)
7921 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007922 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01007923 tcp-check send HEAD\ /\ HTTP/1.1\r\n
7924 tcp-check send Host:\ www.mydomain.com\r\n
7925 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
7926 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007927 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01007928
7929
7930 See also : "tcp-check expect", "tcp-check send"
7931
7932
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007933option tcp-smart-accept
7934no option tcp-smart-accept
7935 Enable or disable the saving of one ACK packet during the accept sequence
7936 May be used in sections : defaults | frontend | listen | backend
7937 yes | yes | yes | no
7938 Arguments : none
7939
7940 When an HTTP connection request comes in, the system acknowledges it on
7941 behalf of HAProxy, then the client immediately sends its request, and the
7942 system acknowledges it too while it is notifying HAProxy about the new
7943 connection. HAProxy then reads the request and responds. This means that we
7944 have one TCP ACK sent by the system for nothing, because the request could
7945 very well be acknowledged by HAProxy when it sends its response.
7946
7947 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
7948 sending this useless ACK on platforms which support it (currently at least
7949 Linux). It must not cause any problem, because the system will send it anyway
7950 after 40 ms if the response takes more time than expected to come.
7951
7952 During complex network debugging sessions, it may be desirable to disable
7953 this optimization because delayed ACKs can make troubleshooting more complex
7954 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007955 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007956
7957 It is also possible to force it for non-HTTP proxies by simply specifying
7958 "option tcp-smart-accept". For instance, it can make sense with some services
7959 such as SMTP where the server speaks first.
7960
7961 It is recommended to avoid forcing this option in a defaults section. In case
7962 of doubt, consider setting it back to automatic values by prepending the
7963 "default" keyword before it, or disabling it using the "no" keyword.
7964
Willy Tarreaud88edf22009-06-14 15:48:17 +02007965 See also : "option tcp-smart-connect"
7966
7967
7968option tcp-smart-connect
7969no option tcp-smart-connect
7970 Enable or disable the saving of one ACK packet during the connect sequence
7971 May be used in sections : defaults | frontend | listen | backend
7972 yes | no | yes | yes
7973 Arguments : none
7974
7975 On certain systems (at least Linux), HAProxy can ask the kernel not to
7976 immediately send an empty ACK upon a connection request, but to directly
7977 send the buffer request instead. This saves one packet on the network and
7978 thus boosts performance. It can also be useful for some servers, because they
7979 immediately get the request along with the incoming connection.
7980
7981 This feature is enabled when "option tcp-smart-connect" is set in a backend.
7982 It is not enabled by default because it makes network troubleshooting more
7983 complex.
7984
7985 It only makes sense to enable it with protocols where the client speaks first
7986 such as HTTP. In other situations, if there is no data to send in place of
7987 the ACK, a normal ACK is sent.
7988
7989 If this option has been enabled in a "defaults" section, it can be disabled
7990 in a specific instance by prepending the "no" keyword before it.
7991
7992 See also : "option tcp-smart-accept"
7993
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007994
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007995option tcpka
7996 Enable or disable the sending of TCP keepalive packets on both sides
7997 May be used in sections : defaults | frontend | listen | backend
7998 yes | yes | yes | yes
7999 Arguments : none
8000
8001 When there is a firewall or any session-aware component between a client and
8002 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01008003 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008004 components decides to expire a session which has remained idle for too long.
8005
8006 Enabling socket-level TCP keep-alives makes the system regularly send packets
8007 to the other end of the connection, leaving it active. The delay between
8008 keep-alive probes is controlled by the system only and depends both on the
8009 operating system and its tuning parameters.
8010
8011 It is important to understand that keep-alive packets are neither emitted nor
8012 received at the application level. It is only the network stacks which sees
8013 them. For this reason, even if one side of the proxy already uses keep-alives
8014 to maintain its connection alive, those keep-alive packets will not be
8015 forwarded to the other side of the proxy.
8016
8017 Please note that this has nothing to do with HTTP keep-alive.
8018
8019 Using option "tcpka" enables the emission of TCP keep-alive probes on both
8020 the client and server sides of a connection. Note that this is meaningful
8021 only in "defaults" or "listen" sections. If this option is used in a
8022 frontend, only the client side will get keep-alives, and if this option is
8023 used in a backend, only the server side will get keep-alives. For this
8024 reason, it is strongly recommended to explicitly use "option clitcpka" and
8025 "option srvtcpka" when the configuration is split between frontends and
8026 backends.
8027
8028 See also : "option clitcpka", "option srvtcpka"
8029
Willy Tarreau844e3c52008-01-11 16:28:18 +01008030
8031option tcplog
8032 Enable advanced logging of TCP connections with session state and timers
8033 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01008034 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01008035 Arguments : none
8036
8037 By default, the log output format is very poor, as it only contains the
8038 source and destination addresses, and the instance name. By specifying
8039 "option tcplog", each log line turns into a much richer format including, but
8040 not limited to, the connection timers, the session status, the connections
8041 numbers, the frontend, backend and server name, and of course the source
8042 address and ports. This option is useful for pure TCP proxies in order to
8043 find which of the client or server disconnects or times out. For normal HTTP
8044 proxies, it's better to use "option httplog" which is even more complete.
8045
Guillaume de Lafond29f45602017-03-31 19:52:15 +02008046 "option tcplog" overrides any previous "log-format" directive.
8047
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008048 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008049
8050
Willy Tarreau844e3c52008-01-11 16:28:18 +01008051option transparent
8052no option transparent
8053 Enable client-side transparent proxying
8054 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01008055 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01008056 Arguments : none
8057
8058 This option was introduced in order to provide layer 7 persistence to layer 3
8059 load balancers. The idea is to use the OS's ability to redirect an incoming
8060 connection for a remote address to a local process (here HAProxy), and let
8061 this process know what address was initially requested. When this option is
8062 used, sessions without cookies will be forwarded to the original destination
8063 IP address of the incoming request (which should match that of another
8064 equipment), while requests with cookies will still be forwarded to the
8065 appropriate server.
8066
8067 Note that contrary to a common belief, this option does NOT make HAProxy
8068 present the client's IP to the server when establishing the connection.
8069
Willy Tarreaua1146052011-03-01 09:51:54 +01008070 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008071 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008072
Willy Tarreaubf1f8162007-12-28 17:42:56 +01008073
Simon Horman98637e52014-06-20 12:30:16 +09008074external-check command <command>
8075 Executable to run when performing an external-check
8076 May be used in sections : defaults | frontend | listen | backend
8077 yes | no | yes | yes
8078
8079 Arguments :
8080 <command> is the external command to run
8081
Simon Horman98637e52014-06-20 12:30:16 +09008082 The arguments passed to the to the command are:
8083
Cyril Bonté777be862014-12-02 21:21:35 +01008084 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09008085
Cyril Bonté777be862014-12-02 21:21:35 +01008086 The <proxy_address> and <proxy_port> are derived from the first listener
8087 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
8088 listener the proxy_address will be the path of the socket and the
8089 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
8090 possible to determine a listener, and both <proxy_address> and <proxy_port>
8091 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09008092
Cyril Bonté72cda2a2014-12-27 22:28:39 +01008093 Some values are also provided through environment variables.
8094
8095 Environment variables :
8096 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
8097 applicable, for example in a "backend" section).
8098
8099 HAPROXY_PROXY_ID The backend id.
8100
8101 HAPROXY_PROXY_NAME The backend name.
8102
8103 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
8104 applicable, for example in a "backend" section or
8105 for a UNIX socket).
8106
8107 HAPROXY_SERVER_ADDR The server address.
8108
8109 HAPROXY_SERVER_CURCONN The current number of connections on the server.
8110
8111 HAPROXY_SERVER_ID The server id.
8112
8113 HAPROXY_SERVER_MAXCONN The server max connections.
8114
8115 HAPROXY_SERVER_NAME The server name.
8116
8117 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
8118 socket).
8119
8120 PATH The PATH environment variable used when executing
8121 the command may be set using "external-check path".
8122
William Lallemand4d03e432019-06-14 15:35:37 +02008123 See also "2.3. Environment variables" for other variables.
8124
Simon Horman98637e52014-06-20 12:30:16 +09008125 If the command executed and exits with a zero status then the check is
8126 considered to have passed, otherwise the check is considered to have
8127 failed.
8128
8129 Example :
8130 external-check command /bin/true
8131
8132 See also : "external-check", "option external-check", "external-check path"
8133
8134
8135external-check path <path>
8136 The value of the PATH environment variable used when running an external-check
8137 May be used in sections : defaults | frontend | listen | backend
8138 yes | no | yes | yes
8139
8140 Arguments :
8141 <path> is the path used when executing external command to run
8142
8143 The default path is "".
8144
8145 Example :
8146 external-check path "/usr/bin:/bin"
8147
8148 See also : "external-check", "option external-check",
8149 "external-check command"
8150
8151
Emeric Brun647caf12009-06-30 17:57:00 +02008152persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02008153persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02008154 Enable RDP cookie-based persistence
8155 May be used in sections : defaults | frontend | listen | backend
8156 yes | no | yes | yes
8157 Arguments :
8158 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02008159 default cookie name "msts" will be used. There currently is no
8160 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02008161
8162 This statement enables persistence based on an RDP cookie. The RDP cookie
8163 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01008164 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02008165 and if an RDP cookie is found, it is decoded. If it matches a known server
8166 which is still UP (or if "option persist" is set), then the connection is
8167 forwarded to this server.
8168
8169 Note that this only makes sense in a TCP backend, but for this to work, the
8170 frontend must have waited long enough to ensure that an RDP cookie is present
8171 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008172 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02008173 a single "listen" section.
8174
Willy Tarreau61e28f22010-05-16 22:31:05 +02008175 Also, it is important to understand that the terminal server will emit this
8176 RDP cookie only if it is configured for "token redirection mode", which means
8177 that the "IP address redirection" option is disabled.
8178
Emeric Brun647caf12009-06-30 17:57:00 +02008179 Example :
8180 listen tse-farm
8181 bind :3389
8182 # wait up to 5s for an RDP cookie in the request
8183 tcp-request inspect-delay 5s
8184 tcp-request content accept if RDP_COOKIE
8185 # apply RDP cookie persistence
8186 persist rdp-cookie
8187 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02008188 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02008189 balance rdp-cookie
8190 server srv1 1.1.1.1:3389
8191 server srv2 1.1.1.2:3389
8192
Simon Hormanab814e02011-06-24 14:50:20 +09008193 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
8194 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02008195
8196
Willy Tarreau3a7d2072009-03-05 23:48:25 +01008197rate-limit sessions <rate>
8198 Set a limit on the number of new sessions accepted per second on a frontend
8199 May be used in sections : defaults | frontend | listen | backend
8200 yes | yes | yes | no
8201 Arguments :
8202 <rate> The <rate> parameter is an integer designating the maximum number
8203 of new sessions per second to accept on the frontend.
8204
8205 When the frontend reaches the specified number of new sessions per second, it
8206 stops accepting new connections until the rate drops below the limit again.
8207 During this time, the pending sessions will be kept in the socket's backlog
8208 (in system buffers) and haproxy will not even be aware that sessions are
8209 pending. When applying very low limit on a highly loaded service, it may make
8210 sense to increase the socket's backlog using the "backlog" keyword.
8211
8212 This feature is particularly efficient at blocking connection-based attacks
8213 or service abuse on fragile servers. Since the session rate is measured every
8214 millisecond, it is extremely accurate. Also, the limit applies immediately,
8215 no delay is needed at all to detect the threshold.
8216
8217 Example : limit the connection rate on SMTP to 10 per second max
8218 listen smtp
8219 mode tcp
8220 bind :25
8221 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02008222 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01008223
Willy Tarreaua17c2d92011-07-25 08:16:20 +02008224 Note : when the maximum rate is reached, the frontend's status is not changed
8225 but its sockets appear as "WAITING" in the statistics if the
8226 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01008227
8228 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
8229
8230
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008231redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
8232redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
8233redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02008234 Return an HTTP redirection if/unless a condition is matched
8235 May be used in sections : defaults | frontend | listen | backend
8236 no | yes | yes | yes
8237
8238 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01008239 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02008240
Willy Tarreau0140f252008-11-19 21:07:09 +01008241 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008242 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01008243 the HTTP "Location" header. When used in an "http-request" rule,
8244 <loc> value follows the log-format rules and can include some
8245 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008246
8247 <pfx> With "redirect prefix", the "Location" header is built from the
8248 concatenation of <pfx> and the complete URI path, including the
8249 query string, unless the "drop-query" option is specified (see
8250 below). As a special case, if <pfx> equals exactly "/", then
8251 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01008252 redirect to the same URL (for instance, to insert a cookie). When
8253 used in an "http-request" rule, <pfx> value follows the log-format
8254 rules and can include some dynamic values (see Custom Log Format
8255 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008256
8257 <sch> With "redirect scheme", then the "Location" header is built by
8258 concatenating <sch> with "://" then the first occurrence of the
8259 "Host" header, and then the URI path, including the query string
8260 unless the "drop-query" option is specified (see below). If no
8261 path is found or if the path is "*", then "/" is used instead. If
8262 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008263 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008264 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01008265 HTTPS. When used in an "http-request" rule, <sch> value follows
8266 the log-format rules and can include some dynamic values (see
8267 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01008268
8269 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01008270 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
8271 with 302 used by default if no code is specified. 301 means
8272 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02008273 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01008274 cache the redirection. 303 is equivalent to 302 except that the
8275 browser will fetch the location with a GET method. 307 is just
8276 like 302 but makes it clear that the same method must be reused.
8277 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01008278
8279 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008280 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01008281
8282 - "drop-query"
8283 When this keyword is used in a prefix-based redirection, then the
8284 location will be set without any possible query-string, which is useful
8285 for directing users to a non-secure page for instance. It has no effect
8286 with a location-type redirect.
8287
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01008288 - "append-slash"
8289 This keyword may be used in conjunction with "drop-query" to redirect
8290 users who use a URL not ending with a '/' to the same one with the '/'.
8291 It can be useful to ensure that search engines will only see one URL.
8292 For this, a return code 301 is preferred.
8293
Willy Tarreau0140f252008-11-19 21:07:09 +01008294 - "set-cookie NAME[=value]"
8295 A "Set-Cookie" header will be added with NAME (and optionally "=value")
8296 to the response. This is sometimes used to indicate that a user has
8297 been seen, for instance to protect against some types of DoS. No other
8298 cookie option is added, so the cookie will be a session cookie. Note
8299 that for a browser, a sole cookie name without an equal sign is
8300 different from a cookie with an equal sign.
8301
8302 - "clear-cookie NAME[=]"
8303 A "Set-Cookie" header will be added with NAME (and optionally "="), but
8304 with the "Max-Age" attribute set to zero. This will tell the browser to
8305 delete this cookie. It is useful for instance on logout pages. It is
8306 important to note that clearing the cookie "NAME" will not remove a
8307 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
8308 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02008309
8310 Example: move the login URL only to HTTPS.
8311 acl clear dst_port 80
8312 acl secure dst_port 8080
8313 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01008314 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01008315 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01008316 acl cookie_set hdr_sub(cookie) SEEN=1
8317
8318 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01008319 redirect prefix https://mysite.com if login_page !secure
8320 redirect prefix http://mysite.com drop-query if login_page !uid_given
8321 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01008322 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02008323
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01008324 Example: send redirects for request for articles without a '/'.
8325 acl missing_slash path_reg ^/article/[^/]*$
8326 redirect code 301 prefix / drop-query append-slash if missing_slash
8327
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008328 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01008329 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02008330
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01008331 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02008332 http-request redirect code 301 location \
8333 http://www.%[hdr(host)]%[capture.req.uri] \
8334 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01008335
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008336 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02008337
Willy Tarreau303c0352008-01-17 19:01:39 +01008338
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008339retries <value>
8340 Set the number of retries to perform on a server after a connection failure
8341 May be used in sections: defaults | frontend | listen | backend
8342 yes | no | yes | yes
8343 Arguments :
8344 <value> is the number of times a connection attempt should be retried on
8345 a server when a connection either is refused or times out. The
8346 default value is 3.
8347
8348 It is important to understand that this value applies to the number of
8349 connection attempts, not full requests. When a connection has effectively
8350 been established to a server, there will be no more retry.
8351
8352 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07008353 a turn-around timer of min("timeout connect", one second) is applied before
8354 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008355
8356 When "option redispatch" is set, the last retry may be performed on another
8357 server even if a cookie references a different server.
8358
8359 See also : "option redispatch"
8360
8361
Olivier Houcharda254a372019-04-05 15:30:12 +02008362retry-on [list of keywords]
8363 Specify when to attempt to automatically retry a failed request
8364 May be used in sections: defaults | frontend | listen | backend
8365 yes | no | yes | yes
8366 Arguments :
8367 <keywords> is a list of keywords or HTTP status codes, each representing a
8368 type of failure event on which an attempt to retry the request
8369 is desired. Please read the notes at the bottom before changing
8370 this setting. The following keywords are supported :
8371
8372 none never retry
8373
8374 conn-failure retry when the connection or the SSL handshake failed
8375 and the request could not be sent. This is the default.
8376
8377 empty-response retry when the server connection was closed after part
8378 of the request was sent, and nothing was received from
8379 the server. This type of failure may be caused by the
8380 request timeout on the server side, poor network
8381 condition, or a server crash or restart while
8382 processing the request.
8383
Olivier Houcharde3249a92019-05-03 23:01:47 +02008384 junk-response retry when the server returned something not looking
8385 like a complete HTTP response. This includes partial
8386 responses headers as well as non-HTTP contents. It
8387 usually is a bad idea to retry on such events, which
8388 may be caused a configuration issue (wrong server port)
8389 or by the request being harmful to the server (buffer
8390 overflow attack for example).
8391
Olivier Houcharda254a372019-04-05 15:30:12 +02008392 response-timeout the server timeout stroke while waiting for the server
8393 to respond to the request. This may be caused by poor
8394 network condition, the reuse of an idle connection
8395 which has expired on the path, or by the request being
8396 extremely expensive to process. It generally is a bad
8397 idea to retry on such events on servers dealing with
8398 heavy database processing (full scans, etc) as it may
8399 amplify denial of service attacks.
8400
Olivier Houchard865d8392019-05-03 22:46:27 +02008401 0rtt-rejected retry requests which were sent over early data and were
8402 rejected by the server. These requests are generally
8403 considered to be safe to retry.
8404
Olivier Houcharda254a372019-04-05 15:30:12 +02008405 <status> any HTTP status code among "404" (Not Found), "408"
8406 (Request Timeout), "425" (Too Early), "500" (Server
8407 Error), "501" (Not Implemented), "502" (Bad Gateway),
8408 "503" (Service Unavailable), "504" (Gateway Timeout).
8409
Olivier Houchardddf0e032019-05-10 18:05:40 +02008410 all-retryable-errors
8411 retry request for any error that are considered
8412 retryable. This currently activates "conn-failure",
8413 "empty-response", "junk-response", "response-timeout",
8414 "0rtt-rejected", "500", "502", "503", and "504".
8415
Olivier Houcharda254a372019-04-05 15:30:12 +02008416 Using this directive replaces any previous settings with the new ones; it is
8417 not cumulative.
8418
8419 Please note that using anything other than "none" and "conn-failure" requires
8420 to allocate a buffer and copy the whole request into it, so it has memory and
8421 performance impacts. Requests not fitting in a single buffer will never be
8422 retried (see the global tune.bufsize setting).
8423
8424 You have to make sure the application has a replay protection mechanism built
8425 in such as a unique transaction IDs passed in requests, or that replaying the
8426 same request has no consequence, or it is very dangerous to use any retry-on
8427 value beside "conn-failure" and "none". Static file servers and caches are
8428 generally considered safe against any type of retry. Using a status code can
8429 be useful to quickly leave a server showing an abnormal behavior (out of
8430 memory, file system issues, etc), but in this case it may be a good idea to
8431 immediately redispatch the connection to another server (please see "option
8432 redispatch" for this). Last, it is important to understand that most causes
8433 of failures are the requests themselves and that retrying a request causing a
8434 server to misbehave will often make the situation even worse for this server,
8435 or for the whole service in case of redispatch.
8436
8437 Unless you know exactly how the application deals with replayed requests, you
8438 should not use this directive.
8439
8440 The default is "conn-failure".
8441
8442 See also: "retries", "option redispatch", "tune.bufsize"
8443
David du Colombier486df472011-03-17 10:40:26 +01008444server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008445 Declare a server in a backend
8446 May be used in sections : defaults | frontend | listen | backend
8447 no | no | yes | yes
8448 Arguments :
8449 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01008450 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05008451 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008452
David du Colombier486df472011-03-17 10:40:26 +01008453 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
8454 resolvable hostname is supported, but this name will be resolved
8455 during start-up. Address "0.0.0.0" or "*" has a special meaning.
8456 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02008457 address as the one from the client connection. This is useful in
8458 transparent proxy architectures where the client's connection is
8459 intercepted and haproxy must forward to the original destination
8460 address. This is more or less what the "transparent" keyword does
8461 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01008462 to report statistics. Optionally, an address family prefix may be
8463 used before the address to force the family regardless of the
8464 address format, which can be useful to specify a path to a unix
8465 socket with no slash ('/'). Currently supported prefixes are :
8466 - 'ipv4@' -> address is always IPv4
8467 - 'ipv6@' -> address is always IPv6
8468 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008469 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +02008470 - 'sockpair@' -> address is the FD of a connected unix
8471 socket or of a socketpair. During a connection, the
8472 backend creates a pair of connected sockets, and passes
8473 one of them over the FD. The bind part will use the
8474 received socket as the client FD. Should be used
8475 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02008476 You may want to reference some environment variables in the
8477 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01008478 variables. The "init-addr" setting can be used to modify the way
8479 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008480
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008481 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008482 be sent to this port. If unset, the same port the client
8483 connected to will be used. The port may also be prefixed by a "+"
8484 or a "-". In this case, the server's port will be determined by
8485 adding this value to the client's port.
8486
8487 <param*> is a list of parameters for this server. The "server" keywords
8488 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008489 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008490
8491 Examples :
8492 server first 10.1.1.1:1080 cookie first check inter 1000
8493 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01008494 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02008495 server backup "${SRV_BACKUP}:1080" backup
8496 server www1_dc1 "${LAN_DC1}.101:80"
8497 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008498
Willy Tarreau55dcaf62015-09-27 15:03:15 +02008499 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
8500 sun_path length is used for the address length. Some other programs
8501 such as socat use the string length only by default. Pass the option
8502 ",unix-tightsocklen=0" to any abstract socket definition in socat to
8503 make it compatible with HAProxy's.
8504
Mark Lamourinec2247f02012-01-04 13:02:01 -05008505 See also: "default-server", "http-send-name-header" and section 5 about
8506 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008507
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02008508server-state-file-name [<file>]
8509 Set the server state file to read, load and apply to servers available in
8510 this backend. It only applies when the directive "load-server-state-from-file"
8511 is set to "local". When <file> is not provided or if this directive is not
8512 set, then backend name is used. If <file> starts with a slash '/', then it is
8513 considered as an absolute path. Otherwise, <file> is concatenated to the
8514 global directive "server-state-file-base".
8515
8516 Example: the minimal configuration below would make HAProxy look for the
8517 state server file '/etc/haproxy/states/bk':
8518
8519 global
8520 server-state-file-base /etc/haproxy/states
8521
8522 backend bk
8523 load-server-state-from-file
8524
8525 See also: "server-state-file-base", "load-server-state-from-file", and
8526 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008527
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02008528server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
8529 Set a template to initialize servers with shared parameters.
8530 The names of these servers are built from <prefix> and <num | range> parameters.
8531 May be used in sections : defaults | frontend | listen | backend
8532 no | no | yes | yes
8533
8534 Arguments:
8535 <prefix> A prefix for the server names to be built.
8536
8537 <num | range>
8538 If <num> is provided, this template initializes <num> servers
8539 with 1 up to <num> as server name suffixes. A range of numbers
8540 <num_low>-<num_high> may also be used to use <num_low> up to
8541 <num_high> as server name suffixes.
8542
8543 <fqdn> A FQDN for all the servers this template initializes.
8544
8545 <port> Same meaning as "server" <port> argument (see "server" keyword).
8546
8547 <params*>
8548 Remaining server parameters among all those supported by "server"
8549 keyword.
8550
8551 Examples:
8552 # Initializes 3 servers with srv1, srv2 and srv3 as names,
8553 # google.com as FQDN, and health-check enabled.
8554 server-template srv 1-3 google.com:80 check
8555
8556 # or
8557 server-template srv 3 google.com:80 check
8558
8559 # would be equivalent to:
8560 server srv1 google.com:80 check
8561 server srv2 google.com:80 check
8562 server srv3 google.com:80 check
8563
8564
8565
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008566source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008567source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008568source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008569 Set the source address for outgoing connections
8570 May be used in sections : defaults | frontend | listen | backend
8571 yes | no | yes | yes
8572 Arguments :
8573 <addr> is the IPv4 address HAProxy will bind to before connecting to a
8574 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01008575
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008576 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01008577 the most appropriate address to reach its destination. Optionally
8578 an address family prefix may be used before the address to force
8579 the family regardless of the address format, which can be useful
8580 to specify a path to a unix socket with no slash ('/'). Currently
8581 supported prefixes are :
8582 - 'ipv4@' -> address is always IPv4
8583 - 'ipv6@' -> address is always IPv6
8584 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008585 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02008586 You may want to reference some environment variables in the
8587 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008588
8589 <port> is an optional port. It is normally not needed but may be useful
8590 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008591 the system will select a free port. Note that port ranges are not
8592 supported in the backend. If you want to force port ranges, you
8593 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008594
8595 <addr2> is the IP address to present to the server when connections are
8596 forwarded in full transparent proxy mode. This is currently only
8597 supported on some patched Linux kernels. When this address is
8598 specified, clients connecting to the server will be presented
8599 with this address, while health checks will still use the address
8600 <addr>.
8601
8602 <port2> is the optional port to present to the server when connections
8603 are forwarded in full transparent proxy mode (see <addr2> above).
8604 The default value of zero means the system will select a free
8605 port.
8606
Willy Tarreaubce70882009-09-07 11:51:47 +02008607 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
8608 This is the name of a comma-separated header list which can
8609 contain multiple IP addresses. By default, the last occurrence is
8610 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01008611 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02008612 by previous proxy, typically Stunnel. In order to use another
8613 occurrence from the last one, please see the <occ> parameter
8614 below. When the header (or occurrence) is not found, no binding
8615 is performed so that the proxy's default IP address is used. Also
8616 keep in mind that the header name is case insensitive, as for any
8617 HTTP header.
8618
8619 <occ> is the occurrence number of a value to be used in a multi-value
8620 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008621 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02008622 address. Positive values indicate a position from the first
8623 occurrence, 1 being the first one. Negative values indicate
8624 positions relative to the last one, -1 being the last one. This
8625 is helpful for situations where an X-Forwarded-For header is set
8626 at the entry point of an infrastructure and must be used several
8627 proxy layers away. When this value is not specified, -1 is
8628 assumed. Passing a zero here disables the feature.
8629
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008630 <name> is an optional interface name to which to bind to for outgoing
8631 traffic. On systems supporting this features (currently, only
8632 Linux), this allows one to bind all traffic to the server to
8633 this interface even if it is not the one the system would select
8634 based on routing tables. This should be used with extreme care.
8635 Note that using this option requires root privileges.
8636
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008637 The "source" keyword is useful in complex environments where a specific
8638 address only is allowed to connect to the servers. It may be needed when a
8639 private address must be used through a public gateway for instance, and it is
8640 known that the system cannot determine the adequate source address by itself.
8641
8642 An extension which is available on certain patched Linux kernels may be used
8643 through the "usesrc" optional keyword. It makes it possible to connect to the
8644 servers with an IP address which does not belong to the system itself. This
8645 is called "full transparent proxy mode". For this to work, the destination
8646 servers have to route their traffic back to this address through the machine
8647 running HAProxy, and IP forwarding must generally be enabled on this machine.
8648
8649 In this "full transparent proxy" mode, it is possible to force a specific IP
8650 address to be presented to the servers. This is not much used in fact. A more
8651 common use is to tell HAProxy to present the client's IP address. For this,
8652 there are two methods :
8653
8654 - present the client's IP and port addresses. This is the most transparent
8655 mode, but it can cause problems when IP connection tracking is enabled on
8656 the machine, because a same connection may be seen twice with different
8657 states. However, this solution presents the huge advantage of not
8658 limiting the system to the 64k outgoing address+port couples, because all
8659 of the client ranges may be used.
8660
8661 - present only the client's IP address and select a spare port. This
8662 solution is still quite elegant but slightly less transparent (downstream
8663 firewalls logs will not match upstream's). It also presents the downside
8664 of limiting the number of concurrent connections to the usual 64k ports.
8665 However, since the upstream and downstream ports are different, local IP
8666 connection tracking on the machine will not be upset by the reuse of the
8667 same session.
8668
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008669 This option sets the default source for all servers in the backend. It may
8670 also be specified in a "defaults" section. Finer source address specification
8671 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008672 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008673
Baptiste Assmann91bd3372015-07-17 21:59:42 +02008674 In order to work, "usesrc" requires root privileges.
8675
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008676 Examples :
8677 backend private
8678 # Connect to the servers using our 192.168.1.200 source address
8679 source 192.168.1.200
8680
8681 backend transparent_ssl1
8682 # Connect to the SSL farm from the client's source address
8683 source 192.168.1.200 usesrc clientip
8684
8685 backend transparent_ssl2
8686 # Connect to the SSL farm from the client's source address and port
8687 # not recommended if IP conntrack is present on the local machine.
8688 source 192.168.1.200 usesrc client
8689
8690 backend transparent_ssl3
8691 # Connect to the SSL farm from the client's source address. It
8692 # is more conntrack-friendly.
8693 source 192.168.1.200 usesrc clientip
8694
8695 backend transparent_smtp
8696 # Connect to the SMTP farm from the client's source address/port
8697 # with Tproxy version 4.
8698 source 0.0.0.0 usesrc clientip
8699
Willy Tarreaubce70882009-09-07 11:51:47 +02008700 backend transparent_http
8701 # Connect to the servers using the client's IP as seen by previous
8702 # proxy.
8703 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
8704
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008705 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008706 the Linux kernel on www.balabit.com, the "bind" keyword.
8707
Willy Tarreau844e3c52008-01-11 16:28:18 +01008708
Cyril Bonté66c327d2010-10-12 00:14:37 +02008709stats admin { if | unless } <cond>
8710 Enable statistics admin level if/unless a condition is matched
8711 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008712 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008713
8714 This statement enables the statistics admin level if/unless a condition is
8715 matched.
8716
8717 The admin level allows to enable/disable servers from the web interface. By
8718 default, statistics page is read-only for security reasons.
8719
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008720 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8721 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008722 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008723
Cyril Bonté23b39d92011-02-10 22:54:44 +01008724 Currently, the POST request is limited to the buffer size minus the reserved
8725 buffer space, which means that if the list of servers is too long, the
8726 request won't be processed. It is recommended to alter few servers at a
8727 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008728
8729 Example :
8730 # statistics admin level only for localhost
8731 backend stats_localhost
8732 stats enable
8733 stats admin if LOCALHOST
8734
8735 Example :
8736 # statistics admin level always enabled because of the authentication
8737 backend stats_auth
8738 stats enable
8739 stats auth admin:AdMiN123
8740 stats admin if TRUE
8741
8742 Example :
8743 # statistics admin level depends on the authenticated user
8744 userlist stats-auth
8745 group admin users admin
8746 user admin insecure-password AdMiN123
8747 group readonly users haproxy
8748 user haproxy insecure-password haproxy
8749
8750 backend stats_auth
8751 stats enable
8752 acl AUTH http_auth(stats-auth)
8753 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8754 stats http-request auth unless AUTH
8755 stats admin if AUTH_ADMIN
8756
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008757 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8758 "bind-process", section 3.4 about userlists and section 7 about
8759 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008760
8761
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008762stats auth <user>:<passwd>
8763 Enable statistics with authentication and grant access to an account
8764 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008765 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008766 Arguments :
8767 <user> is a user name to grant access to
8768
8769 <passwd> is the cleartext password associated to this user
8770
8771 This statement enables statistics with default settings, and restricts access
8772 to declared users only. It may be repeated as many times as necessary to
8773 allow as many users as desired. When a user tries to access the statistics
8774 without a valid account, a "401 Forbidden" response will be returned so that
8775 the browser asks the user to provide a valid user and password. The real
8776 which will be returned to the browser is configurable using "stats realm".
8777
8778 Since the authentication method is HTTP Basic Authentication, the passwords
8779 circulate in cleartext on the network. Thus, it was decided that the
8780 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008781 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008782
8783 It is also possible to reduce the scope of the proxies which appear in the
8784 report using "stats scope".
8785
8786 Though this statement alone is enough to enable statistics reporting, it is
8787 recommended to set all other settings in order to avoid relying on default
8788 unobvious parameters.
8789
8790 Example :
8791 # public access (limited to this backend only)
8792 backend public_www
8793 server srv1 192.168.0.1:80
8794 stats enable
8795 stats hide-version
8796 stats scope .
8797 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008798 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008799 stats auth admin1:AdMiN123
8800 stats auth admin2:AdMiN321
8801
8802 # internal monitoring access (unlimited)
8803 backend private_monitoring
8804 stats enable
8805 stats uri /admin?stats
8806 stats refresh 5s
8807
8808 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8809
8810
8811stats enable
8812 Enable statistics reporting with default settings
8813 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008814 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008815 Arguments : none
8816
8817 This statement enables statistics reporting with default settings defined
8818 at build time. Unless stated otherwise, these settings are used :
8819 - stats uri : /haproxy?stats
8820 - stats realm : "HAProxy Statistics"
8821 - stats auth : no authentication
8822 - stats scope : no restriction
8823
8824 Though this statement alone is enough to enable statistics reporting, it is
8825 recommended to set all other settings in order to avoid relying on default
8826 unobvious parameters.
8827
8828 Example :
8829 # public access (limited to this backend only)
8830 backend public_www
8831 server srv1 192.168.0.1:80
8832 stats enable
8833 stats hide-version
8834 stats scope .
8835 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008836 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008837 stats auth admin1:AdMiN123
8838 stats auth admin2:AdMiN321
8839
8840 # internal monitoring access (unlimited)
8841 backend private_monitoring
8842 stats enable
8843 stats uri /admin?stats
8844 stats refresh 5s
8845
8846 See also : "stats auth", "stats realm", "stats uri"
8847
8848
Willy Tarreaud63335a2010-02-26 12:56:52 +01008849stats hide-version
8850 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008851 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008852 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008853 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008854
Willy Tarreaud63335a2010-02-26 12:56:52 +01008855 By default, the stats page reports some useful status information along with
8856 the statistics. Among them is HAProxy's version. However, it is generally
8857 considered dangerous to report precise version to anyone, as it can help them
8858 target known weaknesses with specific attacks. The "stats hide-version"
8859 statement removes the version from the statistics report. This is recommended
8860 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008861
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008862 Though this statement alone is enough to enable statistics reporting, it is
8863 recommended to set all other settings in order to avoid relying on default
8864 unobvious parameters.
8865
Willy Tarreaud63335a2010-02-26 12:56:52 +01008866 Example :
8867 # public access (limited to this backend only)
8868 backend public_www
8869 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008870 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008871 stats hide-version
8872 stats scope .
8873 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008874 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008875 stats auth admin1:AdMiN123
8876 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008877
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008878 # internal monitoring access (unlimited)
8879 backend private_monitoring
8880 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008881 stats uri /admin?stats
8882 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008883
Willy Tarreaud63335a2010-02-26 12:56:52 +01008884 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008885
Willy Tarreau983e01e2010-01-11 18:42:06 +01008886
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008887stats http-request { allow | deny | auth [realm <realm>] }
8888 [ { if | unless } <condition> ]
8889 Access control for statistics
8890
8891 May be used in sections: defaults | frontend | listen | backend
8892 no | no | yes | yes
8893
8894 As "http-request", these set of options allow to fine control access to
8895 statistics. Each option may be followed by if/unless and acl.
8896 First option with matched condition (or option without condition) is final.
8897 For "deny" a 403 error will be returned, for "allow" normal processing is
8898 performed, for "auth" a 401/407 error code is returned so the client
8899 should be asked to enter a username and password.
8900
8901 There is no fixed limit to the number of http-request statements per
8902 instance.
8903
8904 See also : "http-request", section 3.4 about userlists and section 7
8905 about ACL usage.
8906
8907
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008908stats realm <realm>
8909 Enable statistics and set authentication realm
8910 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008911 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008912 Arguments :
8913 <realm> is the name of the HTTP Basic Authentication realm reported to
8914 the browser. The browser uses it to display it in the pop-up
8915 inviting the user to enter a valid username and password.
8916
8917 The realm is read as a single word, so any spaces in it should be escaped
8918 using a backslash ('\').
8919
8920 This statement is useful only in conjunction with "stats auth" since it is
8921 only related to authentication.
8922
8923 Though this statement alone is enough to enable statistics reporting, it is
8924 recommended to set all other settings in order to avoid relying on default
8925 unobvious parameters.
8926
8927 Example :
8928 # public access (limited to this backend only)
8929 backend public_www
8930 server srv1 192.168.0.1:80
8931 stats enable
8932 stats hide-version
8933 stats scope .
8934 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008935 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008936 stats auth admin1:AdMiN123
8937 stats auth admin2:AdMiN321
8938
8939 # internal monitoring access (unlimited)
8940 backend private_monitoring
8941 stats enable
8942 stats uri /admin?stats
8943 stats refresh 5s
8944
8945 See also : "stats auth", "stats enable", "stats uri"
8946
8947
8948stats refresh <delay>
8949 Enable statistics with automatic refresh
8950 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008951 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008952 Arguments :
8953 <delay> is the suggested refresh delay, specified in seconds, which will
8954 be returned to the browser consulting the report page. While the
8955 browser is free to apply any delay, it will generally respect it
8956 and refresh the page this every seconds. The refresh interval may
8957 be specified in any other non-default time unit, by suffixing the
8958 unit after the value, as explained at the top of this document.
8959
8960 This statement is useful on monitoring displays with a permanent page
8961 reporting the load balancer's activity. When set, the HTML report page will
8962 include a link "refresh"/"stop refresh" so that the user can select whether
8963 he wants automatic refresh of the page or not.
8964
8965 Though this statement alone is enough to enable statistics reporting, it is
8966 recommended to set all other settings in order to avoid relying on default
8967 unobvious parameters.
8968
8969 Example :
8970 # public access (limited to this backend only)
8971 backend public_www
8972 server srv1 192.168.0.1:80
8973 stats enable
8974 stats hide-version
8975 stats scope .
8976 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008977 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008978 stats auth admin1:AdMiN123
8979 stats auth admin2:AdMiN321
8980
8981 # internal monitoring access (unlimited)
8982 backend private_monitoring
8983 stats enable
8984 stats uri /admin?stats
8985 stats refresh 5s
8986
8987 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8988
8989
8990stats scope { <name> | "." }
8991 Enable statistics and limit access scope
8992 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008993 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008994 Arguments :
8995 <name> is the name of a listen, frontend or backend section to be
8996 reported. The special name "." (a single dot) designates the
8997 section in which the statement appears.
8998
8999 When this statement is specified, only the sections enumerated with this
9000 statement will appear in the report. All other ones will be hidden. This
9001 statement may appear as many times as needed if multiple sections need to be
9002 reported. Please note that the name checking is performed as simple string
9003 comparisons, and that it is never checked that a give section name really
9004 exists.
9005
9006 Though this statement alone is enough to enable statistics reporting, it is
9007 recommended to set all other settings in order to avoid relying on default
9008 unobvious parameters.
9009
9010 Example :
9011 # public access (limited to this backend only)
9012 backend public_www
9013 server srv1 192.168.0.1:80
9014 stats enable
9015 stats hide-version
9016 stats scope .
9017 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009018 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009019 stats auth admin1:AdMiN123
9020 stats auth admin2:AdMiN321
9021
9022 # internal monitoring access (unlimited)
9023 backend private_monitoring
9024 stats enable
9025 stats uri /admin?stats
9026 stats refresh 5s
9027
9028 See also : "stats auth", "stats enable", "stats realm", "stats uri"
9029
Willy Tarreaud63335a2010-02-26 12:56:52 +01009030
Willy Tarreauc9705a12010-07-27 20:05:50 +02009031stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01009032 Enable reporting of a description on the statistics page.
9033 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009034 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01009035
Willy Tarreauc9705a12010-07-27 20:05:50 +02009036 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01009037 description from global section is automatically used instead.
9038
9039 This statement is useful for users that offer shared services to their
9040 customers, where node or description should be different for each customer.
9041
9042 Though this statement alone is enough to enable statistics reporting, it is
9043 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01009044 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009045
9046 Example :
9047 # internal monitoring access (unlimited)
9048 backend private_monitoring
9049 stats enable
9050 stats show-desc Master node for Europe, Asia, Africa
9051 stats uri /admin?stats
9052 stats refresh 5s
9053
9054 See also: "show-node", "stats enable", "stats uri" and "description" in
9055 global section.
9056
9057
9058stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02009059 Enable reporting additional information on the statistics page
9060 May be used in sections : defaults | frontend | listen | backend
9061 yes | yes | yes | yes
9062 Arguments : none
9063
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009064 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01009065 - cap: capabilities (proxy)
9066 - mode: one of tcp, http or health (proxy)
9067 - id: SNMP ID (proxy, socket, server)
9068 - IP (socket, server)
9069 - cookie (backend, server)
9070
9071 Though this statement alone is enough to enable statistics reporting, it is
9072 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01009073 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009074
9075 See also: "stats enable", "stats uri".
9076
9077
9078stats show-node [ <name> ]
9079 Enable reporting of a host name on the statistics page.
9080 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009081 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01009082 Arguments:
9083 <name> is an optional name to be reported. If unspecified, the
9084 node name from global section is automatically used instead.
9085
9086 This statement is useful for users that offer shared services to their
9087 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01009088 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009089
9090 Though this statement alone is enough to enable statistics reporting, it is
9091 recommended to set all other settings in order to avoid relying on default
9092 unobvious parameters.
9093
9094 Example:
9095 # internal monitoring access (unlimited)
9096 backend private_monitoring
9097 stats enable
9098 stats show-node Europe-1
9099 stats uri /admin?stats
9100 stats refresh 5s
9101
9102 See also: "show-desc", "stats enable", "stats uri", and "node" in global
9103 section.
9104
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009105
9106stats uri <prefix>
9107 Enable statistics and define the URI prefix to access them
9108 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009109 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009110 Arguments :
9111 <prefix> is the prefix of any URI which will be redirected to stats. This
9112 prefix may contain a question mark ('?') to indicate part of a
9113 query string.
9114
9115 The statistics URI is intercepted on the relayed traffic, so it appears as a
9116 page within the normal application. It is strongly advised to ensure that the
9117 selected URI will never appear in the application, otherwise it will never be
9118 possible to reach it in the application.
9119
9120 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009121 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009122 It is generally a good idea to include a question mark in the URI so that
9123 intermediate proxies refrain from caching the results. Also, since any string
9124 beginning with the prefix will be accepted as a stats request, the question
9125 mark helps ensuring that no valid URI will begin with the same words.
9126
9127 It is sometimes very convenient to use "/" as the URI prefix, and put that
9128 statement in a "listen" instance of its own. That makes it easy to dedicate
9129 an address or a port to statistics only.
9130
9131 Though this statement alone is enough to enable statistics reporting, it is
9132 recommended to set all other settings in order to avoid relying on default
9133 unobvious parameters.
9134
9135 Example :
9136 # public access (limited to this backend only)
9137 backend public_www
9138 server srv1 192.168.0.1:80
9139 stats enable
9140 stats hide-version
9141 stats scope .
9142 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009143 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009144 stats auth admin1:AdMiN123
9145 stats auth admin2:AdMiN321
9146
9147 # internal monitoring access (unlimited)
9148 backend private_monitoring
9149 stats enable
9150 stats uri /admin?stats
9151 stats refresh 5s
9152
9153 See also : "stats auth", "stats enable", "stats realm"
9154
9155
Willy Tarreaud63335a2010-02-26 12:56:52 +01009156stick match <pattern> [table <table>] [{if | unless} <cond>]
9157 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009158 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01009159 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009160
9161 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009162 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009163 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009164 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009165 stickiness table. This rule is mandatory.
9166
9167 <table> is an optional stickiness table name. If unspecified, the same
9168 backend's table is used. A stickiness table is declared using
9169 the "stick-table" statement.
9170
9171 <cond> is an optional matching condition. It makes it possible to match
9172 on a certain criterion only when other conditions are met (or
9173 not met). For instance, it could be used to match on a source IP
9174 address except when a request passes through a known proxy, in
9175 which case we'd match on a header containing that IP address.
9176
9177 Some protocols or applications require complex stickiness rules and cannot
9178 always simply rely on cookies nor hashing. The "stick match" statement
9179 describes a rule to extract the stickiness criterion from an incoming request
9180 or connection. See section 7 for a complete list of possible patterns and
9181 transformation rules.
9182
9183 The table has to be declared using the "stick-table" statement. It must be of
9184 a type compatible with the pattern. By default it is the one which is present
9185 in the same backend. It is possible to share a table with other backends by
9186 referencing it using the "table" keyword. If another table is referenced,
9187 the server's ID inside the backends are used. By default, all server IDs
9188 start at 1 in each backend, so the server ordering is enough. But in case of
9189 doubt, it is highly recommended to force server IDs using their "id" setting.
9190
9191 It is possible to restrict the conditions where a "stick match" statement
9192 will apply, using "if" or "unless" followed by a condition. See section 7 for
9193 ACL based conditions.
9194
9195 There is no limit on the number of "stick match" statements. The first that
9196 applies and matches will cause the request to be directed to the same server
9197 as was used for the request which created the entry. That way, multiple
9198 matches can be used as fallbacks.
9199
9200 The stick rules are checked after the persistence cookies, so they will not
9201 affect stickiness if a cookie has already been used to select a server. That
9202 way, it becomes very easy to insert cookies and match on IP addresses in
9203 order to maintain stickiness between HTTP and HTTPS.
9204
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009205 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9206 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009207 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009208
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009209 Example :
9210 # forward SMTP users to the same server they just used for POP in the
9211 # last 30 minutes
9212 backend pop
9213 mode tcp
9214 balance roundrobin
9215 stick store-request src
9216 stick-table type ip size 200k expire 30m
9217 server s1 192.168.1.1:110
9218 server s2 192.168.1.1:110
9219
9220 backend smtp
9221 mode tcp
9222 balance roundrobin
9223 stick match src table pop
9224 server s1 192.168.1.1:25
9225 server s2 192.168.1.1:25
9226
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009227 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009228 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009229
9230
9231stick on <pattern> [table <table>] [{if | unless} <condition>]
9232 Define a request pattern to associate a user to a server
9233 May be used in sections : defaults | frontend | listen | backend
9234 no | no | yes | yes
9235
9236 Note : This form is exactly equivalent to "stick match" followed by
9237 "stick store-request", all with the same arguments. Please refer
9238 to both keywords for details. It is only provided as a convenience
9239 for writing more maintainable configurations.
9240
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009241 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9242 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009243 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009244
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009245 Examples :
9246 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01009247 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009248
9249 # ...is strictly equivalent to this one :
9250 stick match src table pop if !localhost
9251 stick store-request src table pop if !localhost
9252
9253
9254 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
9255 # well as HTTP without cookie. Share the same table between both accesses.
9256 backend http
9257 mode http
9258 balance roundrobin
9259 stick on src table https
9260 cookie SRV insert indirect nocache
9261 server s1 192.168.1.1:80 cookie s1
9262 server s2 192.168.1.1:80 cookie s2
9263
9264 backend https
9265 mode tcp
9266 balance roundrobin
9267 stick-table type ip size 200k expire 30m
9268 stick on src
9269 server s1 192.168.1.1:443
9270 server s2 192.168.1.1:443
9271
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009272 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009273
9274
9275stick store-request <pattern> [table <table>] [{if | unless} <condition>]
9276 Define a request pattern used to create an entry in a stickiness table
9277 May be used in sections : defaults | frontend | listen | backend
9278 no | no | yes | yes
9279
9280 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009281 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009282 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009283 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009284 server is selected.
9285
9286 <table> is an optional stickiness table name. If unspecified, the same
9287 backend's table is used. A stickiness table is declared using
9288 the "stick-table" statement.
9289
9290 <cond> is an optional storage condition. It makes it possible to store
9291 certain criteria only when some conditions are met (or not met).
9292 For instance, it could be used to store the source IP address
9293 except when the request passes through a known proxy, in which
9294 case we'd store a converted form of a header containing that IP
9295 address.
9296
9297 Some protocols or applications require complex stickiness rules and cannot
9298 always simply rely on cookies nor hashing. The "stick store-request" statement
9299 describes a rule to decide what to extract from the request and when to do
9300 it, in order to store it into a stickiness table for further requests to
9301 match it using the "stick match" statement. Obviously the extracted part must
9302 make sense and have a chance to be matched in a further request. Storing a
9303 client's IP address for instance often makes sense. Storing an ID found in a
9304 URL parameter also makes sense. Storing a source port will almost never make
9305 any sense because it will be randomly matched. See section 7 for a complete
9306 list of possible patterns and transformation rules.
9307
9308 The table has to be declared using the "stick-table" statement. It must be of
9309 a type compatible with the pattern. By default it is the one which is present
9310 in the same backend. It is possible to share a table with other backends by
9311 referencing it using the "table" keyword. If another table is referenced,
9312 the server's ID inside the backends are used. By default, all server IDs
9313 start at 1 in each backend, so the server ordering is enough. But in case of
9314 doubt, it is highly recommended to force server IDs using their "id" setting.
9315
9316 It is possible to restrict the conditions where a "stick store-request"
9317 statement will apply, using "if" or "unless" followed by a condition. This
9318 condition will be evaluated while parsing the request, so any criteria can be
9319 used. See section 7 for ACL based conditions.
9320
9321 There is no limit on the number of "stick store-request" statements, but
9322 there is a limit of 8 simultaneous stores per request or response. This
9323 makes it possible to store up to 8 criteria, all extracted from either the
9324 request or the response, regardless of the number of rules. Only the 8 first
9325 ones which match will be kept. Using this, it is possible to feed multiple
9326 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009327 another protocol or access method. Using multiple store-request rules with
9328 the same table is possible and may be used to find the best criterion to rely
9329 on, by arranging the rules by decreasing preference order. Only the first
9330 extracted criterion for a given table will be stored. All subsequent store-
9331 request rules referencing the same table will be skipped and their ACLs will
9332 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009333
9334 The "store-request" rules are evaluated once the server connection has been
9335 established, so that the table will contain the real server that processed
9336 the request.
9337
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009338 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9339 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009340 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009341
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009342 Example :
9343 # forward SMTP users to the same server they just used for POP in the
9344 # last 30 minutes
9345 backend pop
9346 mode tcp
9347 balance roundrobin
9348 stick store-request src
9349 stick-table type ip size 200k expire 30m
9350 server s1 192.168.1.1:110
9351 server s2 192.168.1.1:110
9352
9353 backend smtp
9354 mode tcp
9355 balance roundrobin
9356 stick match src table pop
9357 server s1 192.168.1.1:25
9358 server s2 192.168.1.1:25
9359
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009360 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009361 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009362
9363
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009364stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02009365 size <size> [expire <expire>] [nopurge] [peers <peersect>]
9366 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08009367 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009368 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009369 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009370
9371 Arguments :
9372 ip a table declared with "type ip" will only store IPv4 addresses.
9373 This form is very compact (about 50 bytes per entry) and allows
9374 very fast entry lookup and stores with almost no overhead. This
9375 is mainly used to store client source IP addresses.
9376
David du Colombier9a6d3c92011-03-17 10:40:24 +01009377 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
9378 This form is very compact (about 60 bytes per entry) and allows
9379 very fast entry lookup and stores with almost no overhead. This
9380 is mainly used to store client source IP addresses.
9381
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009382 integer a table declared with "type integer" will store 32bit integers
9383 which can represent a client identifier found in a request for
9384 instance.
9385
9386 string a table declared with "type string" will store substrings of up
9387 to <len> characters. If the string provided by the pattern
9388 extractor is larger than <len>, it will be truncated before
9389 being stored. During matching, at most <len> characters will be
9390 compared between the string in the table and the extracted
9391 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009392 to 32 characters.
9393
9394 binary a table declared with "type binary" will store binary blocks
9395 of <len> bytes. If the block provided by the pattern
9396 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02009397 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009398 is shorter than <len>, it will be padded by 0. When not
9399 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009400
9401 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009402 "string" type table (See type "string" above). Or the number
9403 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009404 changing this parameter as memory usage will proportionally
9405 increase.
9406
9407 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01009408 value directly impacts memory usage. Count approximately
9409 50 bytes per entry, plus the size of a string if any. The size
9410 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009411
9412 [nopurge] indicates that we refuse to purge older entries when the table
9413 is full. When not specified and the table is full when haproxy
9414 wants to store an entry in it, it will flush a few of the oldest
9415 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009416 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009417 be desirable to refuse new entries instead of purging the older
9418 ones. That may be the case when the amount of data to store is
9419 far above the hardware limits and we prefer not to offer access
9420 to new clients than to reject the ones already connected. When
9421 using this parameter, be sure to properly set the "expire"
9422 parameter (see below).
9423
Emeric Brunf099e792010-09-27 12:05:28 +02009424 <peersect> is the name of the peers section to use for replication. Entries
9425 which associate keys to server IDs are kept synchronized with
9426 the remote peers declared in this section. All entries are also
9427 automatically learned from the local peer (old process) during a
9428 soft restart.
9429
Willy Tarreau1abc6732015-05-01 19:21:02 +02009430 NOTE : each peers section may be referenced only by tables
9431 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009432
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009433 <expire> defines the maximum duration of an entry in the table since it
9434 was last created, refreshed or matched. The expiration delay is
9435 defined using the standard time format, similarly as the various
9436 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009437 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02009438 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009439 be removed once full. Be sure not to use the "nopurge" parameter
9440 if not expiration delay is specified.
9441
Willy Tarreau08d5f982010-06-06 13:34:54 +02009442 <data_type> is used to store additional information in the stick-table. This
9443 may be used by ACLs in order to control various criteria related
9444 to the activity of the client matching the stick-table. For each
9445 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02009446 that the additional data can fit. Several data types may be
9447 stored with an entry. Multiple data types may be specified after
9448 the "store" keyword, as a comma-separated list. Alternatively,
9449 it is possible to repeat the "store" keyword followed by one or
9450 several data types. Except for the "server_id" type which is
9451 automatically detected and enabled, all data types must be
9452 explicitly declared to be stored. If an ACL references a data
9453 type which is not stored, the ACL will simply not match. Some
9454 data types require an argument which must be passed just after
9455 the type between parenthesis. See below for the supported data
9456 types and their arguments.
9457
9458 The data types that can be stored with an entry are the following :
9459 - server_id : this is an integer which holds the numeric ID of the server a
9460 request was assigned to. It is used by the "stick match", "stick store",
9461 and "stick on" rules. It is automatically enabled when referenced.
9462
9463 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
9464 integer which may be used for anything. Most of the time it will be used
9465 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01009466 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009467
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009468 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
9469 over a period. It is a positive 32-bit integer integer which may be used
9470 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01009471 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009472 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01009473 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009474
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009475 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
9476 integer which may be used for anything. Most of the time it will be used
9477 to put a special tag on some entries, for instance to note that a
9478 specific behavior was detected and must be known for future matches.
9479
9480 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
9481 over a period. It is a positive 32-bit integer integer which may be used
9482 for anything. Just like <gpc1>, it counts events, but instead of keeping
9483 a cumulative number, it maintains the rate at which the counter is
9484 incremented. Most of the time it will be used to measure the frequency of
9485 occurrence of certain events (e.g. requests to a specific URL).
9486
Willy Tarreauc9705a12010-07-27 20:05:50 +02009487 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
9488 the absolute number of connections received from clients which matched
9489 this entry. It does not mean the connections were accepted, just that
9490 they were received.
9491
9492 - conn_cur : Current Connections. It is a positive 32-bit integer which
9493 stores the concurrent connection counts for the entry. It is incremented
9494 once an incoming connection matches the entry, and decremented once the
9495 connection leaves. That way it is possible to know at any time the exact
9496 number of concurrent connections for an entry.
9497
9498 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9499 integer parameter <period> which indicates in milliseconds the length
9500 of the period over which the average is measured. It reports the average
9501 incoming connection rate over that period, in connections per period. The
9502 result is an integer which can be matched using ACLs.
9503
9504 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
9505 the absolute number of sessions received from clients which matched this
9506 entry. A session is a connection that was accepted by the layer 4 rules.
9507
9508 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9509 integer parameter <period> which indicates in milliseconds the length
9510 of the period over which the average is measured. It reports the average
9511 incoming session rate over that period, in sessions per period. The
9512 result is an integer which can be matched using ACLs.
9513
9514 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
9515 counts the absolute number of HTTP requests received from clients which
9516 matched this entry. It does not matter whether they are valid requests or
9517 not. Note that this is different from sessions when keep-alive is used on
9518 the client side.
9519
9520 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9521 integer parameter <period> which indicates in milliseconds the length
9522 of the period over which the average is measured. It reports the average
9523 HTTP request rate over that period, in requests per period. The result is
9524 an integer which can be matched using ACLs. It does not matter whether
9525 they are valid requests or not. Note that this is different from sessions
9526 when keep-alive is used on the client side.
9527
9528 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
9529 counts the absolute number of HTTP requests errors induced by clients
9530 which matched this entry. Errors are counted on invalid and truncated
9531 requests, as well as on denied or tarpitted requests, and on failed
9532 authentications. If the server responds with 4xx, then the request is
9533 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01009534 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02009535
9536 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9537 integer parameter <period> which indicates in milliseconds the length
9538 of the period over which the average is measured. It reports the average
9539 HTTP request error rate over that period, in requests per period (see
9540 http_err_cnt above for what is accounted as an error). The result is an
9541 integer which can be matched using ACLs.
9542
9543 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009544 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02009545 which matched this entry. Headers are included in the count. This may be
9546 used to limit abuse of upload features on photo or video servers.
9547
9548 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9549 integer parameter <period> which indicates in milliseconds the length
9550 of the period over which the average is measured. It reports the average
9551 incoming bytes rate over that period, in bytes per period. It may be used
9552 to detect users which upload too much and too fast. Warning: with large
9553 uploads, it is possible that the amount of uploaded data will be counted
9554 once upon termination, thus causing spikes in the average transfer speed
9555 instead of having a smooth one. This may partially be smoothed with
9556 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
9557 recommended for better fairness.
9558
9559 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009560 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02009561 matched this entry. Headers are included in the count. This may be used
9562 to limit abuse of bots sucking the whole site.
9563
9564 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
9565 an integer parameter <period> which indicates in milliseconds the length
9566 of the period over which the average is measured. It reports the average
9567 outgoing bytes rate over that period, in bytes per period. It may be used
9568 to detect users which download too much and too fast. Warning: with large
9569 transfers, it is possible that the amount of transferred data will be
9570 counted once upon termination, thus causing spikes in the average
9571 transfer speed instead of having a smooth one. This may partially be
9572 smoothed with "option contstats" though this is not perfect yet. Use of
9573 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02009574
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009575 There is only one stick-table per proxy. At the moment of writing this doc,
9576 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009577 to be required, simply create a dummy backend with a stick-table in it and
9578 reference it.
9579
9580 It is important to understand that stickiness based on learning information
9581 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01009582 lost upon restart unless peers are properly configured to transfer such
9583 information upon restart (recommended). In general it can be good as a
9584 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009585
Willy Tarreauc9705a12010-07-27 20:05:50 +02009586 Last, memory requirements may be important when storing many data types.
9587 Indeed, storing all indicators above at once in each entry requires 116 bytes
9588 per entry, or 116 MB for a 1-million entries table. This is definitely not
9589 something that can be ignored.
9590
9591 Example:
9592 # Keep track of counters of up to 1 million IP addresses over 5 minutes
9593 # and store a general purpose counter and the average connection rate
9594 # computed over a sliding window of 30 seconds.
9595 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
9596
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009597 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01009598 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009599
9600
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009601stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01009602 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009603 May be used in sections : defaults | frontend | listen | backend
9604 no | no | yes | yes
9605
9606 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009607 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009608 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01009609 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009610 server is selected.
9611
9612 <table> is an optional stickiness table name. If unspecified, the same
9613 backend's table is used. A stickiness table is declared using
9614 the "stick-table" statement.
9615
9616 <cond> is an optional storage condition. It makes it possible to store
9617 certain criteria only when some conditions are met (or not met).
9618 For instance, it could be used to store the SSL session ID only
9619 when the response is a SSL server hello.
9620
9621 Some protocols or applications require complex stickiness rules and cannot
9622 always simply rely on cookies nor hashing. The "stick store-response"
9623 statement describes a rule to decide what to extract from the response and
9624 when to do it, in order to store it into a stickiness table for further
9625 requests to match it using the "stick match" statement. Obviously the
9626 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009627 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009628 See section 7 for a complete list of possible patterns and transformation
9629 rules.
9630
9631 The table has to be declared using the "stick-table" statement. It must be of
9632 a type compatible with the pattern. By default it is the one which is present
9633 in the same backend. It is possible to share a table with other backends by
9634 referencing it using the "table" keyword. If another table is referenced,
9635 the server's ID inside the backends are used. By default, all server IDs
9636 start at 1 in each backend, so the server ordering is enough. But in case of
9637 doubt, it is highly recommended to force server IDs using their "id" setting.
9638
9639 It is possible to restrict the conditions where a "stick store-response"
9640 statement will apply, using "if" or "unless" followed by a condition. This
9641 condition will be evaluated while parsing the response, so any criteria can
9642 be used. See section 7 for ACL based conditions.
9643
9644 There is no limit on the number of "stick store-response" statements, but
9645 there is a limit of 8 simultaneous stores per request or response. This
9646 makes it possible to store up to 8 criteria, all extracted from either the
9647 request or the response, regardless of the number of rules. Only the 8 first
9648 ones which match will be kept. Using this, it is possible to feed multiple
9649 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009650 another protocol or access method. Using multiple store-response rules with
9651 the same table is possible and may be used to find the best criterion to rely
9652 on, by arranging the rules by decreasing preference order. Only the first
9653 extracted criterion for a given table will be stored. All subsequent store-
9654 response rules referencing the same table will be skipped and their ACLs will
9655 not be evaluated. However, even if a store-request rule references a table, a
9656 store-response rule may also use the same table. This means that each table
9657 may learn exactly one element from the request and one element from the
9658 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009659
9660 The table will contain the real server that processed the request.
9661
9662 Example :
9663 # Learn SSL session ID from both request and response and create affinity.
9664 backend https
9665 mode tcp
9666 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02009667 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009668 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009669
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009670 acl clienthello req_ssl_hello_type 1
9671 acl serverhello rep_ssl_hello_type 2
9672
9673 # use tcp content accepts to detects ssl client and server hello.
9674 tcp-request inspect-delay 5s
9675 tcp-request content accept if clienthello
9676
9677 # no timeout on response inspect delay by default.
9678 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009679
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009680 # SSL session ID (SSLID) may be present on a client or server hello.
9681 # Its length is coded on 1 byte at offset 43 and its value starts
9682 # at offset 44.
9683
9684 # Match and learn on request if client hello.
9685 stick on payload_lv(43,1) if clienthello
9686
9687 # Learn on response if server hello.
9688 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02009689
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009690 server s1 192.168.1.1:443
9691 server s2 192.168.1.1:443
9692
9693 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
9694 extraction.
9695
9696
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009697tcp-check connect [params*]
9698 Opens a new connection
9699 May be used in sections: defaults | frontend | listen | backend
9700 no | no | yes | yes
9701
9702 When an application lies on more than a single TCP port or when HAProxy
9703 load-balance many services in a single backend, it makes sense to probe all
9704 the services individually before considering a server as operational.
9705
9706 When there are no TCP port configured on the server line neither server port
9707 directive, then the 'tcp-check connect port <port>' must be the first step
9708 of the sequence.
9709
9710 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9711 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9712 do.
9713
9714 Parameters :
9715 They are optional and can be used to describe how HAProxy should open and
9716 use the TCP connection.
9717
9718 port if not set, check port or server port is used.
9719 It tells HAProxy where to open the connection to.
9720 <port> must be a valid TCP port source integer, from 1 to 65535.
9721
9722 send-proxy send a PROXY protocol string
9723
9724 ssl opens a ciphered connection
9725
9726 Examples:
9727 # check HTTP and HTTPs services on a server.
9728 # first open port 80 thanks to server line port directive, then
9729 # tcp-check opens port 443, ciphered and run a request on it:
9730 option tcp-check
9731 tcp-check connect
9732 tcp-check send GET\ /\ HTTP/1.0\r\n
9733 tcp-check send Host:\ haproxy.1wt.eu\r\n
9734 tcp-check send \r\n
9735 tcp-check expect rstring (2..|3..)
9736 tcp-check connect port 443 ssl
9737 tcp-check send GET\ /\ HTTP/1.0\r\n
9738 tcp-check send Host:\ haproxy.1wt.eu\r\n
9739 tcp-check send \r\n
9740 tcp-check expect rstring (2..|3..)
9741 server www 10.0.0.1 check port 80
9742
9743 # check both POP and IMAP from a single server:
9744 option tcp-check
9745 tcp-check connect port 110
9746 tcp-check expect string +OK\ POP3\ ready
9747 tcp-check connect port 143
9748 tcp-check expect string *\ OK\ IMAP4\ ready
9749 server mail 10.0.0.1 check
9750
9751 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9752
9753
9754tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009755 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009756 May be used in sections: defaults | frontend | listen | backend
9757 no | no | yes | yes
9758
9759 Arguments :
9760 <match> is a keyword indicating how to look for a specific pattern in the
9761 response. The keyword may be one of "string", "rstring" or
9762 binary.
9763 The keyword may be preceded by an exclamation mark ("!") to negate
9764 the match. Spaces are allowed between the exclamation mark and the
9765 keyword. See below for more details on the supported keywords.
9766
9767 <pattern> is the pattern to look for. It may be a string or a regular
9768 expression. If the pattern contains spaces, they must be escaped
9769 with the usual backslash ('\').
9770 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009771 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009772 two digits will represent a byte. The hexadecimal digits may be
9773 used upper or lower case.
9774
9775
9776 The available matches are intentionally similar to their http-check cousins :
9777
9778 string <string> : test the exact string matches in the response buffer.
9779 A health check response will be considered valid if the
9780 response's buffer contains this exact string. If the
9781 "string" keyword is prefixed with "!", then the response
9782 will be considered invalid if the body contains this
9783 string. This can be used to look for a mandatory pattern
9784 in a protocol response, or to detect a failure when a
9785 specific error appears in a protocol banner.
9786
9787 rstring <regex> : test a regular expression on the response buffer.
9788 A health check response will be considered valid if the
9789 response's buffer matches this expression. If the
9790 "rstring" keyword is prefixed with "!", then the response
9791 will be considered invalid if the body matches the
9792 expression.
9793
9794 binary <hexstring> : test the exact string in its hexadecimal form matches
9795 in the response buffer. A health check response will
9796 be considered valid if the response's buffer contains
9797 this exact hexadecimal string.
9798 Purpose is to match data on binary protocols.
9799
9800 It is important to note that the responses will be limited to a certain size
9801 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9802 Thus, too large responses may not contain the mandatory pattern when using
9803 "string", "rstring" or binary. If a large response is absolutely required, it
9804 is possible to change the default max size by setting the global variable.
9805 However, it is worth keeping in mind that parsing very large responses can
9806 waste some CPU cycles, especially when regular expressions are used, and that
9807 it is always better to focus the checks on smaller resources. Also, in its
9808 current state, the check will not find any string nor regex past a null
9809 character in the response. Similarly it is not possible to request matching
9810 the null character.
9811
9812 Examples :
9813 # perform a POP check
9814 option tcp-check
9815 tcp-check expect string +OK\ POP3\ ready
9816
9817 # perform an IMAP check
9818 option tcp-check
9819 tcp-check expect string *\ OK\ IMAP4\ ready
9820
9821 # look for the redis master server
9822 option tcp-check
9823 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009824 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009825 tcp-check send info\ replication\r\n
9826 tcp-check expect string role:master
9827 tcp-check send QUIT\r\n
9828 tcp-check expect string +OK
9829
9830
9831 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9832 "tcp-check send-binary", "http-check expect", tune.chksize
9833
9834
9835tcp-check send <data>
9836 Specify a string to be sent as a question during a generic health check
9837 May be used in sections: defaults | frontend | listen | backend
9838 no | no | yes | yes
9839
9840 <data> : the data to be sent as a question during a generic health check
9841 session. For now, <data> must be a string.
9842
9843 Examples :
9844 # look for the redis master server
9845 option tcp-check
9846 tcp-check send info\ replication\r\n
9847 tcp-check expect string role:master
9848
9849 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9850 "tcp-check send-binary", tune.chksize
9851
9852
Davor Ocelice9ed2812017-12-25 17:49:28 +01009853tcp-check send-binary <hexstring>
9854 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009855 tcp health check
9856 May be used in sections: defaults | frontend | listen | backend
9857 no | no | yes | yes
9858
9859 <data> : the data to be sent as a question during a generic health check
9860 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009861 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009862 response buffer. A health check response will be considered
9863 valid if the response's buffer contains this exact
9864 hexadecimal string.
9865 Purpose is to send binary data to ask on binary protocols.
9866
9867 Examples :
9868 # redis check in binary
9869 option tcp-check
9870 tcp-check send-binary 50494e470d0a # PING\r\n
9871 tcp-check expect binary 2b504F4e47 # +PONG
9872
9873
9874 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9875 "tcp-check send", tune.chksize
9876
9877
Willy Tarreaue9656522010-08-17 15:40:09 +02009878tcp-request connection <action> [{if | unless} <condition>]
9879 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009880 May be used in sections : defaults | frontend | listen | backend
9881 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009882 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009883 <action> defines the action to perform if the condition applies. See
9884 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009885
Willy Tarreaue9656522010-08-17 15:40:09 +02009886 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009887
9888 Immediately after acceptance of a new incoming connection, it is possible to
9889 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009890 or dropped or have its counters tracked. Those conditions cannot make use of
9891 any data contents because the connection has not been read from yet, and the
9892 buffers are not yet allocated. This is used to selectively and very quickly
9893 accept or drop connections from various sources with a very low overhead. If
9894 some contents need to be inspected in order to take the decision, the
9895 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009896
Willy Tarreaue9656522010-08-17 15:40:09 +02009897 The "tcp-request connection" rules are evaluated in their exact declaration
9898 order. If no rule matches or if there is no rule, the default action is to
9899 accept the incoming connection. There is no specific limit to the number of
9900 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009901
Willy Tarreaua9083d02015-05-08 15:27:59 +02009902 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02009903 - accept :
9904 accepts the connection if the condition is true (when used with "if")
9905 or false (when used with "unless"). The first such rule executed ends
9906 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009907
Willy Tarreaue9656522010-08-17 15:40:09 +02009908 - reject :
9909 rejects the connection if the condition is true (when used with "if")
9910 or false (when used with "unless"). The first such rule executed ends
9911 the rules evaluation. Rejected connections do not even become a
9912 session, which is why they are accounted separately for in the stats,
9913 as "denied connections". They are not considered for the session
9914 rate-limit and are not logged either. The reason is that these rules
9915 should only be used to filter extremely high connection rates such as
9916 the ones encountered during a massive DDoS attack. Under these extreme
9917 conditions, the simple action of logging each event would make the
9918 system collapse and would considerably lower the filtering capacity. If
9919 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02009920 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009921
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009922 - expect-proxy layer4 :
9923 configures the client-facing connection to receive a PROXY protocol
9924 header before any byte is read from the socket. This is equivalent to
9925 having the "accept-proxy" keyword on the "bind" line, except that using
9926 the TCP rule allows the PROXY protocol to be accepted only for certain
9927 IP address ranges using an ACL. This is convenient when multiple layers
9928 of load balancers are passed through by traffic coming from public
9929 hosts.
9930
Bertrand Jacquin90759682016-06-06 15:35:39 +01009931 - expect-netscaler-cip layer4 :
9932 configures the client-facing connection to receive a NetScaler Client
9933 IP insertion protocol header before any byte is read from the socket.
9934 This is equivalent to having the "accept-netscaler-cip" keyword on the
9935 "bind" line, except that using the TCP rule allows the PROXY protocol
9936 to be accepted only for certain IP address ranges using an ACL. This
9937 is convenient when multiple layers of load balancers are passed
9938 through by traffic coming from public hosts.
9939
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009940 - capture <sample> len <length> :
9941 This only applies to "tcp-request content" rules. It captures sample
9942 expression <sample> from the request buffer, and converts it to a
9943 string of at most <len> characters. The resulting string is stored into
9944 the next request "capture" slot, so it will possibly appear next to
9945 some captured HTTP headers. It will then automatically appear in the
9946 logs, and it will be possible to extract it using sample fetch rules to
9947 feed it into headers or anything. The length should be limited given
9948 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02009949 session life. Please check section 7.3 (Fetching samples) and "capture
9950 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009951
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009952 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02009953 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009954 rules do not stop evaluation and do not change default action. The
9955 number of counters that may be simultaneously tracked by the same
9956 connection is set in MAX_SESS_STKCTR at build time (reported in
John Roeslerfb2fce12019-07-10 15:45:51 -05009957 haproxy -vv) which defaults to 3, so the track-sc number is between 0
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009958 and (MAX_SESS_STCKTR-1). The first "track-sc0" rule executed enables
9959 tracking of the counters of the specified table as the first set. The
9960 first "track-sc1" rule executed enables tracking of the counters of the
9961 specified table as the second set. The first "track-sc2" rule executed
9962 enables tracking of the counters of the specified table as the third
9963 set. It is a recommended practice to use the first set of counters for
9964 the per-frontend counters and the second set for the per-backend ones.
9965 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009966
Willy Tarreaue9656522010-08-17 15:40:09 +02009967 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009968 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02009969 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01009970 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009971 and used to select which table entry to update the counters.
9972 Note that "tcp-request connection" cannot use content-based
9973 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009974
Willy Tarreaue9656522010-08-17 15:40:09 +02009975 <table> is an optional table to be used instead of the default one,
9976 which is the stick-table declared in the current proxy. All
9977 the counters for the matches and updates for the key will
9978 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009979
Willy Tarreaue9656522010-08-17 15:40:09 +02009980 Once a "track-sc*" rule is executed, the key is looked up in the table
9981 and if it is not found, an entry is allocated for it. Then a pointer to
9982 that entry is kept during all the session's life, and this entry's
9983 counters are updated as often as possible, every time the session's
9984 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009985 Counters are only updated for events that happen after the tracking has
9986 been started. For example, connection counters will not be updated when
9987 tracking layer 7 information, since the connection event happens before
9988 layer7 information is extracted.
9989
Willy Tarreaue9656522010-08-17 15:40:09 +02009990 If the entry tracks concurrent connection counters, one connection is
9991 counted for as long as the entry is tracked, and the entry will not
9992 expire during that time. Tracking counters also provides a performance
9993 advantage over just checking the keys, because only one table lookup is
9994 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009995
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009996 - sc-inc-gpc0(<sc-id>):
9997 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
9998 counter designated by <sc-id>. If an error occurs, this action silently
9999 fails and the actions evaluation continues.
10000
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010001 - sc-inc-gpc1(<sc-id>):
10002 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
10003 counter designated by <sc-id>. If an error occurs, this action silently
10004 fails and the actions evaluation continues.
10005
Cédric Dufour0d7712d2019-11-06 18:38:53 +010010006 - sc-set-gpt0(<sc-id>) { <int> | <expr> }:
10007 This action sets the 32-bit unsigned GPT0 tag according to the sticky
10008 counter designated by <sc-id> and the value of <int>/<expr>. The
10009 expected result is a boolean. If an error occurs, this action silently
10010 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010011
William Lallemand2e785f22016-05-25 01:48:42 +020010012 - set-src <expr> :
10013 Is used to set the source IP address to the value of specified
10014 expression. Useful if you want to mask source IP for privacy.
10015 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010016 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +020010017
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010018 Arguments:
10019 <expr> Is a standard HAProxy expression formed by a sample-fetch
10020 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +020010021
10022 Example:
William Lallemand2e785f22016-05-25 01:48:42 +020010023 tcp-request connection set-src src,ipmask(24)
10024
Willy Tarreau0c630532016-10-21 17:52:58 +020010025 When possible, set-src preserves the original source port as long as the
10026 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +020010027
William Lallemand44be6402016-05-25 01:51:35 +020010028 - set-src-port <expr> :
10029 Is used to set the source port address to the value of specified
10030 expression.
10031
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010032 Arguments:
10033 <expr> Is a standard HAProxy expression formed by a sample-fetch
10034 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +020010035
10036 Example:
William Lallemand44be6402016-05-25 01:51:35 +020010037 tcp-request connection set-src-port int(4000)
10038
Willy Tarreau0c630532016-10-21 17:52:58 +020010039 When possible, set-src-port preserves the original source address as long
10040 as the address family supports a port, otherwise it forces the source
10041 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +020010042
William Lallemand13e9b0c2016-05-25 02:34:07 +020010043 - set-dst <expr> :
10044 Is used to set the destination IP address to the value of specified
10045 expression. Useful if you want to mask IP for privacy in log.
10046 If you want to provide an IP from a HTTP header use "http-request
10047 set-dst". If you want to connect to the new address/port, use
10048 '0.0.0.0:0' as a server address in the backend.
10049
10050 <expr> Is a standard HAProxy expression formed by a sample-fetch
10051 followed by some converters.
10052
10053 Example:
10054
10055 tcp-request connection set-dst dst,ipmask(24)
10056 tcp-request connection set-dst ipv4(10.0.0.1)
10057
Willy Tarreau0c630532016-10-21 17:52:58 +020010058 When possible, set-dst preserves the original destination port as long as
10059 the address family allows it, otherwise the destination port is set to 0.
10060
William Lallemand13e9b0c2016-05-25 02:34:07 +020010061 - set-dst-port <expr> :
10062 Is used to set the destination port address to the value of specified
10063 expression. If you want to connect to the new address/port, use
10064 '0.0.0.0:0' as a server address in the backend.
10065
10066
10067 <expr> Is a standard HAProxy expression formed by a sample-fetch
10068 followed by some converters.
10069
10070 Example:
10071
10072 tcp-request connection set-dst-port int(4000)
10073
Willy Tarreau0c630532016-10-21 17:52:58 +020010074 When possible, set-dst-port preserves the original destination address as
10075 long as the address family supports a port, otherwise it forces the
10076 destination address to IPv4 "0.0.0.0" before rewriting the port.
10077
Willy Tarreau2d392c22015-08-24 01:43:45 +020010078 - "silent-drop" :
10079 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010080 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010081 to prevent the client from being notified. The effect it then that the
10082 client still sees an established connection while there's none on
10083 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10084 except that it doesn't use any local resource at all on the machine
10085 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010086 slow down stronger attackers. It is important to understand the impact
10087 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010088 client and HAProxy (firewalls, proxies, load balancers) will also keep
10089 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010090 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010091 TCP_REPAIR socket option is used to block the emission of a TCP
10092 reset. On other systems, the socket's TTL is reduced to 1 so that the
10093 TCP reset doesn't pass the first router, though it's still delivered to
10094 local networks. Do not use it unless you fully understand how it works.
10095
Willy Tarreaue9656522010-08-17 15:40:09 +020010096 Note that the "if/unless" condition is optional. If no condition is set on
10097 the action, it is simply performed unconditionally. That can be useful for
10098 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010099
Willy Tarreaue9656522010-08-17 15:40:09 +020010100 Example: accept all connections from white-listed hosts, reject too fast
10101 connection without counting them, and track accepted connections.
10102 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010103
Willy Tarreaue9656522010-08-17 15:40:09 +020010104 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010105 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010106 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010107
Willy Tarreaue9656522010-08-17 15:40:09 +020010108 Example: accept all connections from white-listed hosts, count all other
10109 connections and reject too fast ones. This results in abusive ones
10110 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010111
Willy Tarreaue9656522010-08-17 15:40:09 +020010112 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010113 tcp-request connection track-sc0 src
10114 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010115
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010116 Example: enable the PROXY protocol for traffic coming from all known proxies.
10117
10118 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10119
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010120 See section 7 about ACL usage.
10121
Willy Tarreau4f614292016-10-21 17:49:36 +020010122 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010123
10124
Willy Tarreaue9656522010-08-17 15:40:09 +020010125tcp-request content <action> [{if | unless} <condition>]
10126 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +020010127 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010128 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +020010129 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010130 <action> defines the action to perform if the condition applies. See
10131 below.
Willy Tarreau62644772008-07-16 18:36:06 +020010132
Willy Tarreaue9656522010-08-17 15:40:09 +020010133 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +020010134
Davor Ocelice9ed2812017-12-25 17:49:28 +010010135 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +020010136 called "TCP content inspection". During this stage, ACL-based rules are
10137 evaluated every time the request contents are updated, until either an
10138 "accept" or a "reject" rule matches, or the TCP request inspection delay
10139 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +020010140
Willy Tarreaue9656522010-08-17 15:40:09 +020010141 The first difference between these rules and "tcp-request connection" rules
10142 is that "tcp-request content" rules can make use of contents to take a
10143 decision. Most often, these decisions will consider a protocol recognition or
10144 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +010010145 both frontends and backends. In case of HTTP keep-alive with the client, all
10146 tcp-request content rules are evaluated again, so haproxy keeps a record of
10147 what sticky counters were assigned by a "tcp-request connection" versus a
10148 "tcp-request content" rule, and flushes all the content-related ones after
10149 processing an HTTP request, so that they may be evaluated again by the rules
10150 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010151 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +010010152 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +020010153
Willy Tarreaue9656522010-08-17 15:40:09 +020010154 Content-based rules are evaluated in their exact declaration order. If no
10155 rule matches or if there is no rule, the default action is to accept the
10156 contents. There is no specific limit to the number of rules which may be
10157 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +020010158
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010159 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010160 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +010010161 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010162 - reject : the request is rejected and the connection is closed
10163 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -040010164 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010165 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010166 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010167 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +010010168 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010169 - set-dst <expr>
10170 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010171 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010172 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +020010173 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +010010174 - send-spoe-group <engine-name> <group-name>
Christopher Faulet579d83b2019-11-22 15:34:17 +010010175 - use-service <service-name>
Willy Tarreau62644772008-07-16 18:36:06 +020010176
Willy Tarreaue9656522010-08-17 15:40:09 +020010177 They have the same meaning as their counter-parts in "tcp-request connection"
10178 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +010010179 For "do-resolve" action, please check the "http-request do-resolve"
10180 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +020010181
Willy Tarreauf3338342014-01-28 21:40:28 +010010182 While there is nothing mandatory about it, it is recommended to use the
10183 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
10184 content" rules in the frontend, and track-sc2 for "tcp-request content"
10185 rules in the backend, because that makes the configuration more readable
10186 and easier to troubleshoot, but this is just a guideline and all counters
10187 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +020010188
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010189 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +020010190 the action, it is simply performed unconditionally. That can be useful for
10191 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +020010192
Willy Tarreaue9656522010-08-17 15:40:09 +020010193 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +020010194 rules, since HTTP-specific ACL matches are able to preliminarily parse the
10195 contents of a buffer before extracting the required data. If the buffered
10196 contents do not parse as a valid HTTP message, then the ACL does not match.
10197 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +010010198 processing, so there is no risk of parsing something differently. In an HTTP
10199 backend connected to from an HTTP frontend, it is guaranteed that HTTP
10200 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +020010201
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010202 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010203 are present when the rule is processed. The rule processing engine is able to
10204 wait until the inspect delay expires when the data to be tracked is not yet
10205 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010206
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010207 The "set-dst" and "set-dst-port" are used to set respectively the destination
10208 IP and port. More information on how to use it at "http-request set-dst".
10209
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010210 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +020010211 declared inline. For "tcp-request session" rules, only session-level
10212 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010213
Daniel Schneller0b547052016-03-21 20:46:57 +010010214 <var-name> The name of the variable starts with an indication about
10215 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010216 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010217 "sess" : the variable is shared with the whole session
10218 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010219 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010220 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010221 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010222 "res" : the variable is shared only during response
10223 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010224 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010225 The name may only contain characters 'a-z', 'A-Z', '0-9',
10226 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010227
10228 <expr> Is a standard HAProxy expression formed by a sample-fetch
10229 followed by some converters.
10230
Christopher Faulet85d79c92016-11-09 16:54:56 +010010231 The "unset-var" is used to unset a variable. See above for details about
10232 <var-name>.
10233
Patrick Hemmer268a7072018-05-11 12:52:31 -040010234 The "set-priority-class" is used to set the queue priority class of the
10235 current request. The value must be a sample expression which converts to an
10236 integer in the range -2047..2047. Results outside this range will be
10237 truncated. The priority class determines the order in which queued requests
10238 are processed. Lower values have higher priority.
10239
10240 The "set-priority-offset" is used to set the queue priority timestamp offset
10241 of the current request. The value must be a sample expression which converts
10242 to an integer in the range -524287..524287. Results outside this range will be
10243 truncated. When a request is queued, it is ordered first by the priority
10244 class, then by the current timestamp adjusted by the given offset in
10245 milliseconds. Lower values have higher priority.
10246 Note that the resulting timestamp is is only tracked with enough precision for
10247 524,287ms (8m44s287ms). If the request is queued long enough to where the
10248 adjusted timestamp exceeds this value, it will be misidentified as highest
10249 priority. Thus it is important to set "timeout queue" to a value, where when
10250 combined with the offset, does not exceed this limit.
10251
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010252 The "send-spoe-group" is used to trigger sending of a group of SPOE
10253 messages. To do so, the SPOE engine used to send messages must be defined, as
10254 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10255 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10256 the SPOE agent name must be used.
10257
10258 <engine-name> The SPOE engine name.
10259
10260 <group-name> The SPOE group name as specified in the engine configuration.
10261
Christopher Faulet579d83b2019-11-22 15:34:17 +010010262 The "use-service" is used to executes a TCP service which will reply to the
10263 request and stop the evaluation of the rules. This service may choose to
10264 reply by sending any valid response or it may immediately close the
10265 connection without sending anything. Outside natives services, it is possible
10266 to write your own services in Lua. No further "tcp-request" rules are
10267 evaluated.
10268
10269 Example:
10270 tcp-request content use-service lua.deny { src -f /etc/haproxy/blacklist.lst }
10271
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010272 Example:
10273
10274 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +010010275 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010276
Willy Tarreau62644772008-07-16 18:36:06 +020010277 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +020010278 # Accept HTTP requests containing a Host header saying "example.com"
10279 # and reject everything else.
10280 acl is_host_com hdr(Host) -i example.com
10281 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +020010282 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +020010283 tcp-request content reject
10284
10285 Example:
Willy Tarreau62644772008-07-16 18:36:06 +020010286 # reject SMTP connection if client speaks first
10287 tcp-request inspect-delay 30s
10288 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010289 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +020010290
10291 # Forward HTTPS connection only if client speaks
10292 tcp-request inspect-delay 30s
10293 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010294 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +020010295 tcp-request content reject
10296
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010297 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010298 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010299 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010300 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010301 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
10302 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010303
10304 Example:
10305 # track request counts per "base" (concatenation of Host+URL)
10306 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010307 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010308
Willy Tarreaue9656522010-08-17 15:40:09 +020010309 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010310 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +020010311
10312 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +010010313 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +020010314 # protecting all our sites
10315 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010316 tcp-request connection track-sc0 src
10317 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +020010318 ...
10319 use_backend http_dynamic if { path_end .php }
10320
10321 backend http_dynamic
10322 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010323 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +020010324 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010325 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010326 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010327 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +020010328 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +020010329
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010330 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +020010331
Jarno Huuskonen95b012b2017-04-06 13:59:14 +030010332 See also : "tcp-request connection", "tcp-request session",
10333 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +020010334
10335
10336tcp-request inspect-delay <timeout>
10337 Set the maximum allowed time to wait for data during content inspection
10338 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010339 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +020010340 Arguments :
10341 <timeout> is the timeout value specified in milliseconds by default, but
10342 can be in any other unit if the number is suffixed by the unit,
10343 as explained at the top of this document.
10344
10345 People using haproxy primarily as a TCP relay are often worried about the
10346 risk of passing any type of protocol to a server without any analysis. In
10347 order to be able to analyze the request contents, we must first withhold
10348 the data then analyze them. This statement simply enables withholding of
10349 data for at most the specified amount of time.
10350
Willy Tarreaufb356202010-08-03 14:02:05 +020010351 TCP content inspection applies very early when a connection reaches a
10352 frontend, then very early when the connection is forwarded to a backend. This
10353 means that a connection may experience a first delay in the frontend and a
10354 second delay in the backend if both have tcp-request rules.
10355
Willy Tarreau62644772008-07-16 18:36:06 +020010356 Note that when performing content inspection, haproxy will evaluate the whole
10357 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010358 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +020010359 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +010010360 contents are definitive. If no delay is set, haproxy will not wait at all
10361 and will immediately apply a verdict based on the available information.
10362 Obviously this is unlikely to be very useful and might even be racy, so such
10363 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020010364
10365 As soon as a rule matches, the request is released and continues as usual. If
10366 the timeout is reached and no rule matches, the default policy will be to let
10367 it pass through unaffected.
10368
10369 For most protocols, it is enough to set it to a few seconds, as most clients
10370 send the full request immediately upon connection. Add 3 or more seconds to
10371 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +010010372 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +010010373 before the server (e.g. SMTP), or to wait for a client to talk before passing
10374 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +020010375 least the inspection delay, otherwise it will expire first. If the client
10376 closes the connection or if the buffer is full, the delay immediately expires
10377 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +020010378
Willy Tarreau55165fe2009-05-10 12:02:55 +020010379 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +020010380 "timeout client".
10381
10382
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010383tcp-response content <action> [{if | unless} <condition>]
10384 Perform an action on a session response depending on a layer 4-7 condition
10385 May be used in sections : defaults | frontend | listen | backend
10386 no | no | yes | yes
10387 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010388 <action> defines the action to perform if the condition applies. See
10389 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010390
10391 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
10392
Davor Ocelice9ed2812017-12-25 17:49:28 +010010393 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010394 called "TCP content inspection". During this stage, ACL-based rules are
10395 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010396 "accept", "close" or a "reject" rule matches, or a TCP response inspection
10397 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010398
10399 Most often, these decisions will consider a protocol recognition or validity.
10400
10401 Content-based rules are evaluated in their exact declaration order. If no
10402 rule matches or if there is no rule, the default action is to accept the
10403 contents. There is no specific limit to the number of rules which may be
10404 inserted.
10405
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010406 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010407 - accept :
10408 accepts the response if the condition is true (when used with "if")
10409 or false (when used with "unless"). The first such rule executed ends
10410 the rules evaluation.
10411
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010412 - close :
10413 immediately closes the connection with the server if the condition is
10414 true (when used with "if"), or false (when used with "unless"). The
10415 first such rule executed ends the rules evaluation. The main purpose of
10416 this action is to force a connection to be finished between a client
10417 and a server after an exchange when the application protocol expects
10418 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010419 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010420 protocols.
10421
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010422 - reject :
10423 rejects the response if the condition is true (when used with "if")
10424 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010425 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010426
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010427 - set-var(<var-name>) <expr>
10428 Sets a variable.
10429
Christopher Faulet85d79c92016-11-09 16:54:56 +010010430 - unset-var(<var-name>)
10431 Unsets a variable.
10432
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010433 - sc-inc-gpc0(<sc-id>):
10434 This action increments the GPC0 counter according to the sticky
10435 counter designated by <sc-id>. If an error occurs, this action fails
10436 silently and the actions evaluation continues.
10437
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010438 - sc-inc-gpc1(<sc-id>):
10439 This action increments the GPC1 counter according to the sticky
10440 counter designated by <sc-id>. If an error occurs, this action fails
10441 silently and the actions evaluation continues.
10442
Cédric Dufour0d7712d2019-11-06 18:38:53 +010010443 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
10444 This action sets the 32-bit unsigned GPT0 tag according to the sticky
10445 counter designated by <sc-id> and the value of <int>/<expr>. The
10446 expected result is a boolean. If an error occurs, this action silently
10447 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010448
Willy Tarreau2d392c22015-08-24 01:43:45 +020010449 - "silent-drop" :
10450 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010451 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010452 to prevent the client from being notified. The effect it then that the
10453 client still sees an established connection while there's none on
10454 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10455 except that it doesn't use any local resource at all on the machine
10456 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010457 slow down stronger attackers. It is important to understand the impact
10458 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010459 client and HAProxy (firewalls, proxies, load balancers) will also keep
10460 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010461 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010462 TCP_REPAIR socket option is used to block the emission of a TCP
10463 reset. On other systems, the socket's TTL is reduced to 1 so that the
10464 TCP reset doesn't pass the first router, though it's still delivered to
10465 local networks. Do not use it unless you fully understand how it works.
10466
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010467 - send-spoe-group <engine-name> <group-name>
10468 Send a group of SPOE messages.
10469
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010470 Note that the "if/unless" condition is optional. If no condition is set on
10471 the action, it is simply performed unconditionally. That can be useful for
10472 for changing the default action to a reject.
10473
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010474 It is perfectly possible to match layer 7 contents with "tcp-response
10475 content" rules, but then it is important to ensure that a full response has
10476 been buffered, otherwise no contents will match. In order to achieve this,
10477 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010478 period.
10479
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010480 The "set-var" is used to set the content of a variable. The variable is
10481 declared inline.
10482
Daniel Schneller0b547052016-03-21 20:46:57 +010010483 <var-name> The name of the variable starts with an indication about
10484 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010485 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010486 "sess" : the variable is shared with the whole session
10487 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010488 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010489 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010490 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010491 "res" : the variable is shared only during response
10492 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010493 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010494 The name may only contain characters 'a-z', 'A-Z', '0-9',
10495 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010496
10497 <expr> Is a standard HAProxy expression formed by a sample-fetch
10498 followed by some converters.
10499
10500 Example:
10501
10502 tcp-request content set-var(sess.my_var) src
10503
Christopher Faulet85d79c92016-11-09 16:54:56 +010010504 The "unset-var" is used to unset a variable. See above for details about
10505 <var-name>.
10506
10507 Example:
10508
10509 tcp-request content unset-var(sess.my_var)
10510
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010511 The "send-spoe-group" is used to trigger sending of a group of SPOE
10512 messages. To do so, the SPOE engine used to send messages must be defined, as
10513 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10514 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10515 the SPOE agent name must be used.
10516
10517 <engine-name> The SPOE engine name.
10518
10519 <group-name> The SPOE group name as specified in the engine configuration.
10520
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010521 See section 7 about ACL usage.
10522
10523 See also : "tcp-request content", "tcp-response inspect-delay"
10524
10525
Willy Tarreau4f614292016-10-21 17:49:36 +020010526tcp-request session <action> [{if | unless} <condition>]
10527 Perform an action on a validated session depending on a layer 5 condition
10528 May be used in sections : defaults | frontend | listen | backend
10529 no | yes | yes | no
10530 Arguments :
10531 <action> defines the action to perform if the condition applies. See
10532 below.
10533
10534 <condition> is a standard layer5-only ACL-based condition (see section 7).
10535
Davor Ocelice9ed2812017-12-25 17:49:28 +010010536 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020010537 it is possible to evaluate some conditions to decide whether this session
10538 must be accepted or dropped or have its counters tracked. Those conditions
10539 cannot make use of any data contents because no buffers are allocated yet and
10540 the processing cannot wait at this stage. The main use case it to copy some
10541 early information into variables (since variables are accessible in the
10542 session), or to keep track of some information collected after the handshake,
10543 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010010544 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020010545 extracted information can thus be copied to a variable or tracked using
10546 "track-sc" rules. Of course it is also possible to decide to accept/reject as
10547 with other rulesets. Most operations performed here could also be performed
10548 in "tcp-request content" rules, except that in HTTP these rules are evaluated
10549 for each new request, and that might not always be acceptable. For example a
10550 rule might increment a counter on each evaluation. It would also be possible
10551 that a country is resolved by geolocation from the source IP address,
10552 assigned to a session-wide variable, then the source address rewritten from
10553 an HTTP header for all requests. If some contents need to be inspected in
10554 order to take the decision, the "tcp-request content" statements must be used
10555 instead.
10556
10557 The "tcp-request session" rules are evaluated in their exact declaration
10558 order. If no rule matches or if there is no rule, the default action is to
10559 accept the incoming session. There is no specific limit to the number of
10560 rules which may be inserted.
10561
10562 Several types of actions are supported :
10563 - accept : the request is accepted
10564 - reject : the request is rejected and the connection is closed
10565 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
10566 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010567 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +010010568 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Willy Tarreau4f614292016-10-21 17:49:36 +020010569 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010570 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010571 - silent-drop
10572
10573 These actions have the same meaning as their respective counter-parts in
10574 "tcp-request connection" and "tcp-request content", so please refer to these
10575 sections for a complete description.
10576
10577 Note that the "if/unless" condition is optional. If no condition is set on
10578 the action, it is simply performed unconditionally. That can be useful for
10579 "track-sc*" actions as well as for changing the default action to a reject.
10580
10581 Example: track the original source address by default, or the one advertised
10582 in the PROXY protocol header for connection coming from the local
10583 proxies. The first connection-level rule enables receipt of the
10584 PROXY protocol for these ones, the second rule tracks whatever
10585 address we decide to keep after optional decoding.
10586
10587 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10588 tcp-request session track-sc0 src
10589
10590 Example: accept all sessions from white-listed hosts, reject too fast
10591 sessions without counting them, and track accepted sessions.
10592 This results in session rate being capped from abusive sources.
10593
10594 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10595 tcp-request session reject if { src_sess_rate gt 10 }
10596 tcp-request session track-sc0 src
10597
10598 Example: accept all sessions from white-listed hosts, count all other
10599 sessions and reject too fast ones. This results in abusive ones
10600 being blocked as long as they don't slow down.
10601
10602 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10603 tcp-request session track-sc0 src
10604 tcp-request session reject if { sc0_sess_rate gt 10 }
10605
10606 See section 7 about ACL usage.
10607
10608 See also : "tcp-request connection", "tcp-request content", "stick-table"
10609
10610
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010611tcp-response inspect-delay <timeout>
10612 Set the maximum allowed time to wait for a response during content inspection
10613 May be used in sections : defaults | frontend | listen | backend
10614 no | no | yes | yes
10615 Arguments :
10616 <timeout> is the timeout value specified in milliseconds by default, but
10617 can be in any other unit if the number is suffixed by the unit,
10618 as explained at the top of this document.
10619
10620 See also : "tcp-response content", "tcp-request inspect-delay".
10621
10622
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010623timeout check <timeout>
10624 Set additional check timeout, but only after a connection has been already
10625 established.
10626
10627 May be used in sections: defaults | frontend | listen | backend
10628 yes | no | yes | yes
10629 Arguments:
10630 <timeout> is the timeout value specified in milliseconds by default, but
10631 can be in any other unit if the number is suffixed by the unit,
10632 as explained at the top of this document.
10633
10634 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
10635 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010010636 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010637 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010010638 (Please also note that there is no valid reason to have such long connect
10639 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
10640 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010641
10642 If "timeout check" is not set haproxy uses "inter" for complete check
10643 timeout (connect + read) exactly like all <1.3.15 version.
10644
10645 In most cases check request is much simpler and faster to handle than normal
10646 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010010647 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010648
10649 This parameter is specific to backends, but can be specified once for all in
10650 "defaults" sections. This is in fact one of the easiest solutions not to
10651 forget about it.
10652
Willy Tarreau41a340d2008-01-22 12:25:31 +010010653 See also: "timeout connect", "timeout queue", "timeout server",
10654 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010655
10656
Willy Tarreau0ba27502007-12-24 16:55:16 +010010657timeout client <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010010658 Set the maximum inactivity time on the client side.
10659 May be used in sections : defaults | frontend | listen | backend
10660 yes | yes | yes | no
10661 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010662 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010663 can be in any other unit if the number is suffixed by the unit,
10664 as explained at the top of this document.
10665
10666 The inactivity timeout applies when the client is expected to acknowledge or
10667 send data. In HTTP mode, this timeout is particularly important to consider
10668 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010669 response while it is reading data sent by the server. That said, for the
10670 first phase, it is preferable to set the "timeout http-request" to better
10671 protect HAProxy from Slowloris like attacks. The value is specified in
10672 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010010673 suffixed by the unit, as specified at the top of this document. In TCP mode
10674 (and to a lesser extent, in HTTP mode), it is highly recommended that the
10675 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010676 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010010677 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010010678 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
10679 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020010680 which overrides "timeout client" and "timeout server" for tunnels, as well as
10681 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010682
10683 This parameter is specific to frontends, but can be specified once for all in
10684 "defaults" sections. This is in fact one of the easiest solutions not to
10685 forget about it. An unspecified timeout results in an infinite timeout, which
10686 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010687 during startup because it may result in accumulation of expired sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010688 the system if the system's timeouts are not configured either.
10689
Willy Tarreau95c4e142017-11-26 12:18:55 +010010690 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010010691
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010692 See also : "timeout server", "timeout tunnel", "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010693
Willy Tarreau0ba27502007-12-24 16:55:16 +010010694
Willy Tarreau05cdd962014-05-10 14:30:07 +020010695timeout client-fin <timeout>
10696 Set the inactivity timeout on the client side for half-closed connections.
10697 May be used in sections : defaults | frontend | listen | backend
10698 yes | yes | yes | no
10699 Arguments :
10700 <timeout> is the timeout value specified in milliseconds by default, but
10701 can be in any other unit if the number is suffixed by the unit,
10702 as explained at the top of this document.
10703
10704 The inactivity timeout applies when the client is expected to acknowledge or
10705 send data while one direction is already shut down. This timeout is different
10706 from "timeout client" in that it only applies to connections which are closed
10707 in one direction. This is particularly useful to avoid keeping connections in
10708 FIN_WAIT state for too long when clients do not disconnect cleanly. This
10709 problem is particularly common long connections such as RDP or WebSocket.
10710 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010010711 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
10712 frame was sent, often indicating an expectation that the connection quickly
10713 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020010714
10715 This parameter is specific to frontends, but can be specified once for all in
10716 "defaults" sections. By default it is not set, so half-closed connections
10717 will use the other timeouts (timeout.client or timeout.tunnel).
10718
10719 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
10720
10721
Willy Tarreau0ba27502007-12-24 16:55:16 +010010722timeout connect <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010010723 Set the maximum time to wait for a connection attempt to a server to succeed.
10724 May be used in sections : defaults | frontend | listen | backend
10725 yes | no | yes | yes
10726 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010727 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010728 can be in any other unit if the number is suffixed by the unit,
10729 as explained at the top of this document.
10730
10731 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010732 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010010733 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010010734 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010735 connect timeout also presets both queue and tarpit timeouts to the same value
10736 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010737
10738 This parameter is specific to backends, but can be specified once for all in
10739 "defaults" sections. This is in fact one of the easiest solutions not to
10740 forget about it. An unspecified timeout results in an infinite timeout, which
10741 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010742 during startup because it may result in accumulation of failed sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010743 the system if the system's timeouts are not configured either.
10744
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010745 See also: "timeout check", "timeout queue", "timeout server", "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010746
Willy Tarreau0ba27502007-12-24 16:55:16 +010010747
Willy Tarreaub16a5742010-01-10 14:46:16 +010010748timeout http-keep-alive <timeout>
10749 Set the maximum allowed time to wait for a new HTTP request to appear
10750 May be used in sections : defaults | frontend | listen | backend
10751 yes | yes | yes | yes
10752 Arguments :
10753 <timeout> is the timeout value specified in milliseconds by default, but
10754 can be in any other unit if the number is suffixed by the unit,
10755 as explained at the top of this document.
10756
10757 By default, the time to wait for a new request in case of keep-alive is set
10758 by "timeout http-request". However this is not always convenient because some
10759 people want very short keep-alive timeouts in order to release connections
10760 faster, and others prefer to have larger ones but still have short timeouts
10761 once the request has started to present itself.
10762
10763 The "http-keep-alive" timeout covers these needs. It will define how long to
10764 wait for a new HTTP request to start coming after a response was sent. Once
10765 the first byte of request has been seen, the "http-request" timeout is used
10766 to wait for the complete request to come. Note that empty lines prior to a
10767 new request do not refresh the timeout and are not counted as a new request.
10768
10769 There is also another difference between the two timeouts : when a connection
10770 expires during timeout http-keep-alive, no error is returned, the connection
10771 just closes. If the connection expires in "http-request" while waiting for a
10772 connection to complete, a HTTP 408 error is returned.
10773
10774 In general it is optimal to set this value to a few tens to hundreds of
10775 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010776 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010777 1 millisecond) it will probably only accept pipelined requests but not the
10778 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010779 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010780
10781 If this parameter is not set, the "http-request" timeout applies, and if both
10782 are not set, "timeout client" still applies at the lower level. It should be
10783 set in the frontend to take effect, unless the frontend is in TCP mode, in
10784 which case the HTTP backend's timeout will be used.
10785
Willy Tarreau95c4e142017-11-26 12:18:55 +010010786 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10787 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010788 (where we only have one connection per client and a connection setup).
10789
Willy Tarreaub16a5742010-01-10 14:46:16 +010010790 See also : "timeout http-request", "timeout client".
10791
10792
Willy Tarreau036fae02008-01-06 13:24:40 +010010793timeout http-request <timeout>
10794 Set the maximum allowed time to wait for a complete HTTP request
10795 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010796 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010797 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010798 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010799 can be in any other unit if the number is suffixed by the unit,
10800 as explained at the top of this document.
10801
10802 In order to offer DoS protection, it may be required to lower the maximum
10803 accepted time to receive a complete HTTP request without affecting the client
10804 timeout. This helps protecting against established connections on which
10805 nothing is sent. The client timeout cannot offer a good protection against
10806 this abuse because it is an inactivity timeout, which means that if the
10807 attacker sends one character every now and then, the timeout will not
10808 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010809 types, the request will be aborted if it does not complete in time. When the
10810 timeout expires, an HTTP 408 response is sent to the client to inform it
10811 about the problem, and the connection is closed. The logs will report
10812 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010813 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010814 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10815 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010816
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010817 By default, this timeout only applies to the header part of the request,
10818 and not to any data. As soon as the empty line is received, this timeout is
10819 not used anymore. When combined with "option http-buffer-request", this
10820 timeout also applies to the body of the request..
10821 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010822 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010823
10824 Generally it is enough to set it to a few seconds, as most clients send the
10825 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010826 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010827 generally work on local networks as long as there are no packet losses. This
10828 will prevent people from sending bare HTTP requests using telnet.
10829
10830 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010831 chunk of the incoming request. It should be set in the frontend to take
10832 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10833 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010834
Willy Tarreau0f228a02015-05-01 15:37:53 +020010835 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010836 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010837
Willy Tarreau844e3c52008-01-11 16:28:18 +010010838
10839timeout queue <timeout>
10840 Set the maximum time to wait in the queue for a connection slot to be free
10841 May be used in sections : defaults | frontend | listen | backend
10842 yes | no | yes | yes
10843 Arguments :
10844 <timeout> is the timeout value specified in milliseconds by default, but
10845 can be in any other unit if the number is suffixed by the unit,
10846 as explained at the top of this document.
10847
10848 When a server's maxconn is reached, connections are left pending in a queue
10849 which may be server-specific or global to the backend. In order not to wait
10850 indefinitely, a timeout is applied to requests pending in the queue. If the
10851 timeout is reached, it is considered that the request will almost never be
10852 served, so it is dropped and a 503 error is returned to the client.
10853
10854 The "timeout queue" statement allows to fix the maximum time for a request to
10855 be left pending in a queue. If unspecified, the same value as the backend's
10856 connection timeout ("timeout connect") is used, for backwards compatibility
10857 with older versions with no "timeout queue" parameter.
10858
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010859 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010860
10861
10862timeout server <timeout>
Willy Tarreau844e3c52008-01-11 16:28:18 +010010863 Set the maximum inactivity time on the server side.
10864 May be used in sections : defaults | frontend | listen | backend
10865 yes | no | yes | yes
10866 Arguments :
10867 <timeout> is the timeout value specified in milliseconds by default, but
10868 can be in any other unit if the number is suffixed by the unit,
10869 as explained at the top of this document.
10870
10871 The inactivity timeout applies when the server is expected to acknowledge or
10872 send data. In HTTP mode, this timeout is particularly important to consider
10873 during the first phase of the server's response, when it has to send the
10874 headers, as it directly represents the server's processing time for the
10875 request. To find out what value to put there, it's often good to start with
10876 what would be considered as unacceptable response times, then check the logs
10877 to observe the response time distribution, and adjust the value accordingly.
10878
10879 The value is specified in milliseconds by default, but can be in any other
10880 unit if the number is suffixed by the unit, as specified at the top of this
10881 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
10882 recommended that the client timeout remains equal to the server timeout in
10883 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010884 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010010885 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010010886 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
10887 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020010888 "timeout tunnel", which overrides "timeout client" and "timeout server" for
10889 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010890
10891 This parameter is specific to backends, but can be specified once for all in
10892 "defaults" sections. This is in fact one of the easiest solutions not to
10893 forget about it. An unspecified timeout results in an infinite timeout, which
10894 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010895 during startup because it may result in accumulation of expired sessions in
Willy Tarreau844e3c52008-01-11 16:28:18 +010010896 the system if the system's timeouts are not configured either.
10897
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010898 See also : "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010899
Willy Tarreau05cdd962014-05-10 14:30:07 +020010900
10901timeout server-fin <timeout>
10902 Set the inactivity timeout on the server side for half-closed connections.
10903 May be used in sections : defaults | frontend | listen | backend
10904 yes | no | yes | yes
10905 Arguments :
10906 <timeout> is the timeout value specified in milliseconds by default, but
10907 can be in any other unit if the number is suffixed by the unit,
10908 as explained at the top of this document.
10909
10910 The inactivity timeout applies when the server is expected to acknowledge or
10911 send data while one direction is already shut down. This timeout is different
10912 from "timeout server" in that it only applies to connections which are closed
10913 in one direction. This is particularly useful to avoid keeping connections in
10914 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
10915 This problem is particularly common long connections such as RDP or WebSocket.
10916 Note that this timeout can override "timeout tunnel" when a connection shuts
10917 down in one direction. This setting was provided for completeness, but in most
10918 situations, it should not be needed.
10919
10920 This parameter is specific to backends, but can be specified once for all in
10921 "defaults" sections. By default it is not set, so half-closed connections
10922 will use the other timeouts (timeout.server or timeout.tunnel).
10923
10924 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
10925
Willy Tarreau844e3c52008-01-11 16:28:18 +010010926
10927timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010010928 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010010929 May be used in sections : defaults | frontend | listen | backend
10930 yes | yes | yes | yes
10931 Arguments :
10932 <timeout> is the tarpit duration specified in milliseconds by default, but
10933 can be in any other unit if the number is suffixed by the unit,
10934 as explained at the top of this document.
10935
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020010936 When a connection is tarpitted using "http-request tarpit", it is maintained
10937 open with no activity for a certain amount of time, then closed. "timeout
10938 tarpit" defines how long it will be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010939
10940 The value is specified in milliseconds by default, but can be in any other
10941 unit if the number is suffixed by the unit, as specified at the top of this
10942 document. If unspecified, the same value as the backend's connection timeout
10943 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010010944 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010945
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010946 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010947
10948
Willy Tarreauce887fd2012-05-12 12:50:00 +020010949timeout tunnel <timeout>
10950 Set the maximum inactivity time on the client and server side for tunnels.
10951 May be used in sections : defaults | frontend | listen | backend
10952 yes | no | yes | yes
10953 Arguments :
10954 <timeout> is the timeout value specified in milliseconds by default, but
10955 can be in any other unit if the number is suffixed by the unit,
10956 as explained at the top of this document.
10957
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010958 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020010959 between a client and a server, and the connection remains inactive in both
10960 directions. This timeout supersedes both the client and server timeouts once
10961 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010010962 analyzer remains attached to either connection (e.g. tcp content rules are
10963 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020010964 when switching to the WebSocket protocol, or forwarding a CONNECT request
10965 to a proxy), or after the first response when no keepalive/close option is
10966 specified.
10967
Willy Tarreau05cdd962014-05-10 14:30:07 +020010968 Since this timeout is usually used in conjunction with long-lived connections,
10969 it usually is a good idea to also set "timeout client-fin" to handle the
10970 situation where a client suddenly disappears from the net and does not
10971 acknowledge a close, or sends a shutdown and does not acknowledge pending
10972 data anymore. This can happen in lossy networks where firewalls are present,
10973 and is detected by the presence of large amounts of sessions in a FIN_WAIT
10974 state.
10975
Willy Tarreauce887fd2012-05-12 12:50:00 +020010976 The value is specified in milliseconds by default, but can be in any other
10977 unit if the number is suffixed by the unit, as specified at the top of this
10978 document. Whatever the expected normal idle time, it is a good practice to
10979 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010010980 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020010981
10982 This parameter is specific to backends, but can be specified once for all in
10983 "defaults" sections. This is in fact one of the easiest solutions not to
10984 forget about it.
10985
10986 Example :
10987 defaults http
10988 option http-server-close
10989 timeout connect 5s
10990 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020010991 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020010992 timeout server 30s
10993 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
10994
Willy Tarreau05cdd962014-05-10 14:30:07 +020010995 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020010996
10997
Willy Tarreau844e3c52008-01-11 16:28:18 +010010998transparent (deprecated)
10999 Enable client-side transparent proxying
11000 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010011001 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010011002 Arguments : none
11003
11004 This keyword was introduced in order to provide layer 7 persistence to layer
11005 3 load balancers. The idea is to use the OS's ability to redirect an incoming
11006 connection for a remote address to a local process (here HAProxy), and let
11007 this process know what address was initially requested. When this option is
11008 used, sessions without cookies will be forwarded to the original destination
11009 IP address of the incoming request (which should match that of another
11010 equipment), while requests with cookies will still be forwarded to the
11011 appropriate server.
11012
11013 The "transparent" keyword is deprecated, use "option transparent" instead.
11014
11015 Note that contrary to a common belief, this option does NOT make HAProxy
11016 present the client's IP to the server when establishing the connection.
11017
Willy Tarreau844e3c52008-01-11 16:28:18 +010011018 See also: "option transparent"
11019
William Lallemanda73203e2012-03-12 12:48:57 +010011020unique-id-format <string>
11021 Generate a unique ID for each request.
11022 May be used in sections : defaults | frontend | listen | backend
11023 yes | yes | yes | no
11024 Arguments :
11025 <string> is a log-format string.
11026
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011027 This keyword creates a ID for each request using the custom log format. A
11028 unique ID is useful to trace a request passing through many components of
11029 a complex infrastructure. The newly created ID may also be logged using the
11030 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010011031
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011032 The format should be composed from elements that are guaranteed to be
11033 unique when combined together. For instance, if multiple haproxy instances
11034 are involved, it might be important to include the node name. It is often
11035 needed to log the incoming connection's source and destination addresses
11036 and ports. Note that since multiple requests may be performed over the same
11037 connection, including a request counter may help differentiate them.
11038 Similarly, a timestamp may protect against a rollover of the counter.
11039 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010011040
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011041 It is recommended to use hexadecimal notation for many fields since it
11042 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010011043
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011044 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010011045
Julien Vehentf21be322014-03-07 08:27:34 -050011046 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010011047
11048 will generate:
11049
11050 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
11051
11052 See also: "unique-id-header"
11053
11054unique-id-header <name>
11055 Add a unique ID header in the HTTP request.
11056 May be used in sections : defaults | frontend | listen | backend
11057 yes | yes | yes | no
11058 Arguments :
11059 <name> is the name of the header.
11060
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011061 Add a unique-id header in the HTTP request sent to the server, using the
11062 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010011063
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011064 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010011065
Julien Vehentf21be322014-03-07 08:27:34 -050011066 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010011067 unique-id-header X-Unique-ID
11068
11069 will generate:
11070
11071 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
11072
11073 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010011074
Willy Tarreauf51658d2014-04-23 01:21:56 +020011075use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011076 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011077 May be used in sections : defaults | frontend | listen | backend
11078 no | yes | yes | no
11079 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011080 <backend> is the name of a valid backend or "listen" section, or a
11081 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011082
Willy Tarreauf51658d2014-04-23 01:21:56 +020011083 <condition> is a condition composed of ACLs, as described in section 7. If
11084 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011085
11086 When doing content-switching, connections arrive on a frontend and are then
11087 dispatched to various backends depending on a number of conditions. The
11088 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011089 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011090 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011091 source address validation) or combined with a "tcp-request" rule to wait for
11092 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011093
11094 There may be as many "use_backend" rules as desired. All of these rules are
11095 evaluated in their declaration order, and the first one which matches will
11096 assign the backend.
11097
11098 In the first form, the backend will be used if the condition is met. In the
11099 second form, the backend will be used if the condition is not met. If no
11100 condition is valid, the backend defined with "default_backend" will be used.
11101 If no default backend is defined, either the servers in the same section are
11102 used (in case of a "listen" section) or, in case of a frontend, no server is
11103 used and a 503 service unavailable response is returned.
11104
Willy Tarreau51aecc72009-07-12 09:47:04 +020011105 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011106 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020011107 and backend processing will immediately follow, or the backend will wait for
11108 a complete HTTP request to get in. This feature is useful when a frontend
11109 must decode several protocols on a unique port, one of them being HTTP.
11110
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011111 When <backend> is a simple name, it is resolved at configuration time, and an
11112 error is reported if the specified backend does not exist. If <backend> is
11113 a log-format string instead, no check may be done at configuration time, so
11114 the backend name is resolved dynamically at run time. If the resulting
11115 backend name does not correspond to any valid backend, no other rule is
11116 evaluated, and the default_backend directive is applied instead. Note that
11117 when using dynamic backend names, it is highly recommended to use a prefix
11118 that no other backend uses in order to ensure that an unauthorized backend
11119 cannot be forced from the request.
11120
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011121 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011122 used to detect the association between frontends and backends to compute the
11123 backend's "fullconn" setting. This cannot be done for dynamic names.
11124
11125 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
11126 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010011127
Christopher Fauletb30b3102019-09-12 23:03:09 +020011128use-fcgi-app <name>
11129 Defines the FastCGI application to use for the backend.
11130 May be used in sections : defaults | frontend | listen | backend
11131 no | no | yes | yes
11132 Arguments :
11133 <name> is the name of the FastCGI application to use.
11134
11135 See section 10.1 about FastCGI application setup for details.
Willy Tarreau036fae02008-01-06 13:24:40 +010011136
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011137use-server <server> if <condition>
11138use-server <server> unless <condition>
11139 Only use a specific server if/unless an ACL-based condition is matched.
11140 May be used in sections : defaults | frontend | listen | backend
11141 no | no | yes | yes
11142 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011143 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011144
11145 <condition> is a condition composed of ACLs, as described in section 7.
11146
11147 By default, connections which arrive to a backend are load-balanced across
11148 the available servers according to the configured algorithm, unless a
11149 persistence mechanism such as a cookie is used and found in the request.
11150
11151 Sometimes it is desirable to forward a particular request to a specific
11152 server without having to declare a dedicated backend for this server. This
11153 can be achieved using the "use-server" rules. These rules are evaluated after
11154 the "redirect" rules and before evaluating cookies, and they have precedence
11155 on them. There may be as many "use-server" rules as desired. All of these
11156 rules are evaluated in their declaration order, and the first one which
11157 matches will assign the server.
11158
11159 If a rule designates a server which is down, and "option persist" is not used
11160 and no force-persist rule was validated, it is ignored and evaluation goes on
11161 with the next rules until one matches.
11162
11163 In the first form, the server will be used if the condition is met. In the
11164 second form, the server will be used if the condition is not met. If no
11165 condition is valid, the processing continues and the server will be assigned
11166 according to other persistence mechanisms.
11167
11168 Note that even if a rule is matched, cookie processing is still performed but
11169 does not assign the server. This allows prefixed cookies to have their prefix
11170 stripped.
11171
11172 The "use-server" statement works both in HTTP and TCP mode. This makes it
11173 suitable for use with content-based inspection. For instance, a server could
11174 be selected in a farm according to the TLS SNI field. And if these servers
11175 have their weight set to zero, they will not be used for other traffic.
11176
11177 Example :
11178 # intercept incoming TLS requests based on the SNI field
11179 use-server www if { req_ssl_sni -i www.example.com }
11180 server www 192.168.0.1:443 weight 0
11181 use-server mail if { req_ssl_sni -i mail.example.com }
11182 server mail 192.168.0.1:587 weight 0
11183 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000011184 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011185 # all the rest is forwarded to this server
11186 server default 192.168.0.2:443 check
11187
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011188 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011189
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011190
Davor Ocelice9ed2812017-12-25 17:49:28 +0100111915. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011192--------------------------
11193
11194The "bind", "server" and "default-server" keywords support a number of settings
11195depending on some build options and on the system HAProxy was built on. These
11196settings generally each consist in one word sometimes followed by a value,
11197written on the same line as the "bind" or "server" line. All these options are
11198described in this section.
11199
11200
112015.1. Bind options
11202-----------------
11203
11204The "bind" keyword supports a certain number of settings which are all passed
11205as arguments on the same line. The order in which those arguments appear makes
11206no importance, provided that they appear after the bind address. All of these
11207parameters are optional. Some of them consist in a single words (booleans),
11208while other ones expect a value after them. In this case, the value must be
11209provided immediately after the setting name.
11210
11211The currently supported settings are the following ones.
11212
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011213accept-netscaler-cip <magic number>
11214 Enforces the use of the NetScaler Client IP insertion protocol over any
11215 connection accepted by any of the TCP sockets declared on the same line. The
11216 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
11217 the incoming connection to be used everywhere an address is used, with the
11218 only exception of "tcp-request connection" rules which will only see the
11219 real connection address. Logs will reflect the addresses indicated in the
11220 protocol, unless it is violated, in which case the real address will still
11221 be used. This keyword combined with support from external components can be
11222 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010011223 mechanism which is not always reliable and not even always usable. See also
11224 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
11225 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011226
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011227accept-proxy
11228 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020011229 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
11230 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011231 3/4 addresses of the incoming connection to be used everywhere an address is
11232 used, with the only exception of "tcp-request connection" rules which will
11233 only see the real connection address. Logs will reflect the addresses
11234 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010011235 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011236 components can be used as an efficient and reliable alternative to the
11237 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020011238 usable. See also "tcp-request connection expect-proxy" for a finer-grained
11239 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011240
Olivier Houchardc2aae742017-09-22 18:26:28 +020011241allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010011242 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010011243 due to security considerations. Because it is vulnerable to replay attacks,
John Roeslerfb2fce12019-07-10 15:45:51 -050011244 you should only allow if for requests that are safe to replay, i.e. requests
Olivier Houchard69752962019-01-08 15:35:32 +010011245 that are idempotent. You can use the "wait-for-handshake" action for any
11246 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020011247
Willy Tarreauab861d32013-04-02 02:30:41 +020011248alpn <protocols>
11249 This enables the TLS ALPN extension and advertises the specified protocol
11250 list as supported on top of ALPN. The protocol list consists in a comma-
11251 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050011252 quotes). This requires that the SSL library is built with support for TLS
Willy Tarreauab861d32013-04-02 02:30:41 +020011253 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010011254 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
11255 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11256 now obsolete NPN extension. At the time of writing this, most browsers still
11257 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
11258 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
11259 are expected to be supported, both versions can be advertised, in order of
11260 preference, like below :
11261
11262 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020011263
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011264backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010011265 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011266 backlog is used instead, which generally defaults to the maxconn value.
11267
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010011268curves <curves>
11269 This setting is only available when support for OpenSSL was built in. It sets
11270 the string describing the list of elliptic curves algorithms ("curve suite")
11271 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
11272 string is a colon-delimited list of curve name.
11273 Example: "X25519:P-256" (without quote)
11274 When "curves" is set, "ecdhe" parameter is ignored.
11275
Emeric Brun7fb34422012-09-28 15:26:15 +020011276ecdhe <named curve>
11277 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010011278 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
11279 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020011280
Emeric Brunfd33a262012-10-11 16:28:27 +020011281ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011282 This setting is only available when support for OpenSSL was built in. It
11283 designates a PEM file from which to load CA certificates used to verify
11284 client's certificate.
11285
Emeric Brunb6dc9342012-09-28 17:55:37 +020011286ca-ignore-err [all|<errorID>,...]
11287 This setting is only available when support for OpenSSL was built in.
11288 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
11289 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
11290 error is ignored.
11291
Christopher Faulet31af49d2015-06-09 17:29:50 +020011292ca-sign-file <cafile>
11293 This setting is only available when support for OpenSSL was built in. It
11294 designates a PEM file containing both the CA certificate and the CA private
11295 key used to create and sign server's certificates. This is a mandatory
11296 setting when the dynamic generation of certificates is enabled. See
11297 'generate-certificates' for details.
11298
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000011299ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020011300 This setting is only available when support for OpenSSL was built in. It is
11301 the CA private key passphrase. This setting is optional and used only when
11302 the dynamic generation of certificates is enabled. See
11303 'generate-certificates' for details.
11304
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011305ciphers <ciphers>
11306 This setting is only available when support for OpenSSL was built in. It sets
11307 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000011308 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011309 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011310 information and recommendations see e.g.
11311 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11312 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11313 cipher configuration, please check the "ciphersuites" keyword.
11314
11315ciphersuites <ciphersuites>
11316 This setting is only available when support for OpenSSL was built in and
11317 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
11318 the list of cipher algorithms ("cipher suite") that are negotiated during the
11319 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011320 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
11321 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011322
Emeric Brunfd33a262012-10-11 16:28:27 +020011323crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011324 This setting is only available when support for OpenSSL was built in. It
11325 designates a PEM file from which to load certificate revocation list used
11326 to verify client's certificate.
11327
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011328crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000011329 This setting is only available when support for OpenSSL was built in. It
11330 designates a PEM file containing both the required certificates and any
11331 associated private keys. This file can be built by concatenating multiple
11332 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
11333 requires an intermediate certificate, this can also be concatenated into this
Emmanuel Hocdet70df7bf2019-01-04 11:08:20 +010011334 file. Intermediate certificate can also be shared in a directory via
11335 "issuers-chain-path" directive.
Alex Davies0fbf0162013-03-02 16:04:50 +000011336
William Lallemand4c5adbf2020-02-24 14:23:22 +010011337 If the file does not contain a private key, HAProxy will try to load
11338 the key at the same path suffixed by a ".key".
11339
Alex Davies0fbf0162013-03-02 16:04:50 +000011340 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
11341 are loaded.
11342
11343 If a directory name is used instead of a PEM file, then all files found in
William Lallemand3f25ae32020-02-24 16:30:12 +010011344 that directory will be loaded in alphabetic order unless their name ends
11345 with '.key', '.issuer', '.ocsp' or '.sctl' (reserved extensions). This
11346 directive may be specified multiple times in order to load certificates from
11347 multiple files or directories. The certificates will be presented to clients
11348 who provide a valid TLS Server Name Indication field matching one of their
11349 CN or alt subjects. Wildcards are supported, where a wildcard character '*'
11350 is used instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011351 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000011352
11353 If no SNI is provided by the client or if the SSL library does not support
11354 TLS extensions, or if the client provides an SNI hostname which does not
11355 match any certificate, then the first loaded certificate will be presented.
11356 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011357 recommended to load the default one first as a file or to ensure that it will
11358 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000011359
Emeric Brune032bfa2012-09-28 13:01:45 +020011360 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011361
Davor Ocelice9ed2812017-12-25 17:49:28 +010011362 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000011363 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010011364 choose a web server that the CA believes requires an intermediate CA (for
11365 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000011366 others, e.g. nginx, result in a wrong bundle that will not work for some
11367 clients).
11368
Emeric Brun4147b2e2014-06-16 18:36:30 +020011369 For each PEM file, haproxy checks for the presence of file at the same path
11370 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
11371 Status Request extension (also known as "OCSP stapling") is automatically
11372 enabled. The content of this file is optional. If not empty, it must contain
11373 a valid OCSP Response in DER format. In order to be valid an OCSP Response
11374 must comply with the following rules: it has to indicate a good status,
11375 it has to be a single response for the certificate of the PEM file, and it
11376 has to be valid at the moment of addition. If these rules are not respected
11377 the OCSP Response is ignored and a warning is emitted. In order to identify
11378 which certificate an OCSP Response applies to, the issuer's certificate is
11379 necessary. If the issuer's certificate is not found in the PEM file, it will
11380 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
11381 if it exists otherwise it will fail with an error.
11382
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011383 For each PEM file, haproxy also checks for the presence of file at the same
11384 path suffixed by ".sctl". If such file is found, support for Certificate
11385 Transparency (RFC6962) TLS extension is enabled. The file must contain a
11386 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
11387 to check basic syntax, but no signatures are verified.
11388
yanbzhu6c25e9e2016-01-05 12:52:02 -050011389 There are cases where it is desirable to support multiple key types, e.g. RSA
11390 and ECDSA in the cipher suites offered to the clients. This allows clients
11391 that support EC certificates to be able to use EC ciphers, while
11392 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050011393
11394 In order to provide this functionality, multiple PEM files, each with a
11395 different key type, are required. To associate these PEM files into a
11396 "cert bundle" that is recognized by haproxy, they must be named in the
11397 following way: All PEM files that are to be bundled must have the same base
11398 name, with a suffix indicating the key type. Currently, three suffixes are
11399 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
11400 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
11401 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
11402 suffix matters. To load this bundle into haproxy, specify the base name only:
11403
11404 Example : bind :8443 ssl crt example.pem
11405
yanbzhu6c25e9e2016-01-05 12:52:02 -050011406 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050011407 a cert bundle.
11408
Davor Ocelice9ed2812017-12-25 17:49:28 +010011409 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050011410 support multiple key types. PEM files are combined based on Common Name
11411 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
11412 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
11413 entries in the certificates in that bundle, haproxy will not be able to
11414 provide multi-cert support.
11415
11416 Assuming bundle in the example above contained the following:
11417
11418 Filename | CN | SAN
11419 -------------------+-----------------+-------------------
11420 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050011421 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050011422 example.pem.ecdsa | www.example.com | ecdsa.example.com
11423 -------------------+-----------------+-------------------
11424
11425 Users connecting with an SNI of "www.example.com" will be able
11426 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
11427 "rsa.example.com" will only be able to use RSA cipher suites, and users
11428 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011429 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
11430 no need to bundle certificates. ECDSA certificate will be preferred if client
11431 support it.
yanbzhud19630c2015-12-14 15:10:25 -050011432
11433 If a directory name is given as the <cert> argument, haproxy will
11434 automatically search and load bundled files in that directory.
11435
11436 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
11437 bundling. Each certificate can have its own .ocsp and .issuer file. At this
11438 time, sctl is not supported in multi-certificate bundling.
11439
Emeric Brunb6dc9342012-09-28 17:55:37 +020011440crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000011441 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010011442 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011443 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000011444 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020011445
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011446crt-list <file>
11447 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011448 designates a list of PEM file with an optional ssl configuration and a SNI
11449 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011450
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011451 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
11452
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011453 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
11454 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011455 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011456 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011457
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020011458 Wildcards are supported in the SNI filter. Negative filter are also supported,
11459 only useful in combination with a wildcard filter to exclude a particular SNI.
11460 The certificates will be presented to clients who provide a valid TLS Server
11461 Name Indication field matching one of the SNI filters. If no SNI filter is
11462 specified, the CN and alt subjects are used. This directive may be specified
11463 multiple times. See the "crt" option for more information. The default
11464 certificate is still needed to meet OpenSSL expectations. If it is not used,
11465 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011466
yanbzhu6c25e9e2016-01-05 12:52:02 -050011467 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020011468 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011469 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
11470 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
11471 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050011472
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011473 crt-list file example:
11474 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011475 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011476 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011477 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011478
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011479defer-accept
11480 Is an optional keyword which is supported only on certain Linux kernels. It
11481 states that a connection will only be accepted once some data arrive on it,
11482 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010011483 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011484 performance by ensuring that most of the request is already available when
11485 the connection is accepted. On the other hand, it will not be able to detect
11486 connections which don't talk. It is important to note that this option is
11487 broken in all kernels up to 2.6.31, as the connection is never accepted until
11488 the client talks. This can cause issues with front firewalls which would see
11489 an established connection while the proxy will only see it in SYN_RECV. This
11490 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
11491
William Lallemandf6975e92017-05-26 17:42:10 +020011492expose-fd listeners
11493 This option is only usable with the stats socket. It gives your stats socket
11494 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020011495 During a reload with the master-worker mode, the process is automatically
11496 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011497 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020011498
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011499force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011500 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011501 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011502 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011503 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011504
11505force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011506 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011507 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011508 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011509
11510force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011511 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011512 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011513 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011514
11515force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011516 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011517 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011518 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011519
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011520force-tlsv13
11521 This option enforces use of TLSv1.3 only on SSL connections instantiated from
11522 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011523 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011524
Christopher Faulet31af49d2015-06-09 17:29:50 +020011525generate-certificates
11526 This setting is only available when support for OpenSSL was built in. It
11527 enables the dynamic SSL certificates generation. A CA certificate and its
11528 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
11529 a transparent forward proxy, SSL requests generate errors because of a common
11530 name mismatch on the certificate presented to the client. With this option
11531 enabled, HAProxy will try to forge a certificate using the SNI hostname
11532 indicated by the client. This is done only if no certificate matches the SNI
11533 hostname (see 'crt-list'). If an error occurs, the default certificate is
11534 used, else the 'strict-sni' option is set.
11535 It can also be used when HAProxy is configured as a reverse proxy to ease the
11536 deployment of an architecture with many backends.
11537
11538 Creating a SSL certificate is an expensive operation, so a LRU cache is used
11539 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010011540 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020011541 certificate is used many times.
11542
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011543gid <gid>
11544 Sets the group of the UNIX sockets to the designated system gid. It can also
11545 be set by default in the global section's "unix-bind" statement. Note that
11546 some platforms simply ignore this. This setting is equivalent to the "group"
11547 setting except that the group ID is used instead of its name. This setting is
11548 ignored by non UNIX sockets.
11549
11550group <group>
11551 Sets the group of the UNIX sockets to the designated system group. It can
11552 also be set by default in the global section's "unix-bind" statement. Note
11553 that some platforms simply ignore this. This setting is equivalent to the
11554 "gid" setting except that the group name is used instead of its gid. This
11555 setting is ignored by non UNIX sockets.
11556
11557id <id>
11558 Fixes the socket ID. By default, socket IDs are automatically assigned, but
11559 sometimes it is more convenient to fix them to ease monitoring. This value
11560 must be strictly positive and unique within the listener/frontend. This
11561 option can only be used when defining only a single socket.
11562
11563interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010011564 Restricts the socket to a specific interface. When specified, only packets
11565 received from that particular interface are processed by the socket. This is
11566 currently only supported on Linux. The interface must be a primary system
11567 interface, not an aliased interface. It is also possible to bind multiple
11568 frontends to the same address if they are bound to different interfaces. Note
11569 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010011570 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
11571 uses the same interface as inbound traffic, and its associated routing table,
11572 even if there are explicit routes through different interfaces configured.
11573 This can prove useful to address asymmetric routing issues when the same
11574 client IP addresses need to be able to reach frontends hosted on different
11575 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011576
Willy Tarreauabb175f2012-09-24 12:43:26 +020011577level <level>
11578 This setting is used with the stats sockets only to restrict the nature of
11579 the commands that can be issued on the socket. It is ignored by other
11580 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010011581 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020011582 read, and no change is allowed. It would make sense on systems where it
11583 is not easy to restrict access to the socket.
11584 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011585 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020011586 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010011587 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020011588 all counters).
11589
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020011590severity-output <format>
11591 This setting is used with the stats sockets only to configure severity
11592 level output prepended to informational feedback messages. Severity
11593 level of messages can range between 0 and 7, conforming to syslog
11594 rfc5424. Valid and successful socket commands requesting data
11595 (i.e. "show map", "get acl foo" etc.) will never have a severity level
11596 prepended. It is ignored by other sockets. <format> can be one of :
11597 - "none" (default) no severity level is prepended to feedback messages.
11598 - "number" severity level is prepended as a number.
11599 - "string" severity level is prepended as a string following the
11600 rfc5424 convention.
11601
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011602maxconn <maxconn>
11603 Limits the sockets to this number of concurrent connections. Extraneous
11604 connections will remain in the system's backlog until a connection is
11605 released. If unspecified, the limit will be the same as the frontend's
11606 maxconn. Note that in case of port ranges or multiple addresses, the same
11607 value will be applied to each socket. This setting enables different
11608 limitations on expensive sockets, for instance SSL entries which may easily
11609 eat all memory.
11610
11611mode <mode>
11612 Sets the octal mode used to define access permissions on the UNIX socket. It
11613 can also be set by default in the global section's "unix-bind" statement.
11614 Note that some platforms simply ignore this. This setting is ignored by non
11615 UNIX sockets.
11616
11617mss <maxseg>
11618 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
11619 connections. This can be used to force a lower MSS for certain specific
11620 ports, for instance for connections passing through a VPN. Note that this
11621 relies on a kernel feature which is theoretically supported under Linux but
11622 was buggy in all versions prior to 2.6.28. It may or may not work on other
11623 operating systems. It may also not change the advertised value but change the
11624 effective size of outgoing segments. The commonly advertised value for TCPv4
11625 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
11626 positive, it will be used as the advertised MSS. If it is negative, it will
11627 indicate by how much to reduce the incoming connection's advertised MSS for
11628 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
11629
11630name <name>
11631 Sets an optional name for these sockets, which will be reported on the stats
11632 page.
11633
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011634namespace <name>
11635 On Linux, it is possible to specify which network namespace a socket will
11636 belong to. This directive makes it possible to explicitly bind a listener to
11637 a namespace different from the default one. Please refer to your operating
11638 system's documentation to find more details about network namespaces.
11639
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011640nice <nice>
11641 Sets the 'niceness' of connections initiated from the socket. Value must be
11642 in the range -1024..1024 inclusive, and defaults to zero. Positive values
11643 means that such connections are more friendly to others and easily offer
11644 their place in the scheduler. On the opposite, negative values mean that
11645 connections want to run with a higher priority than others. The difference
11646 only happens under high loads when the system is close to saturation.
11647 Negative values are appropriate for low-latency or administration services,
11648 and high values are generally recommended for CPU intensive tasks such as SSL
11649 processing or bulk transfers which are less sensible to latency. For example,
11650 it may make sense to use a positive value for an SMTP socket and a negative
11651 one for an RDP socket.
11652
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011653no-ca-names
11654 This setting is only available when support for OpenSSL was built in. It
11655 prevents from send CA names in server hello message when ca-file is used.
11656
Emeric Brun9b3009b2012-10-05 11:55:06 +020011657no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011658 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011659 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011660 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011661 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011662 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
11663 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011664
Emeric Brun90ad8722012-10-02 14:00:59 +020011665no-tls-tickets
11666 This setting is only available when support for OpenSSL was built in. It
11667 disables the stateless session resumption (RFC 5077 TLS Ticket
11668 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011669 session resumption is more expensive in CPU usage. This option is also
11670 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020011671
Emeric Brun9b3009b2012-10-05 11:55:06 +020011672no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011673 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011674 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011675 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011676 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011677 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11678 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011679
Emeric Brun9b3009b2012-10-05 11:55:06 +020011680no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020011681 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011682 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011683 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011684 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011685 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11686 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011687
Emeric Brun9b3009b2012-10-05 11:55:06 +020011688no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020011689 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011690 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011691 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011692 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011693 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11694 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011695
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011696no-tlsv13
11697 This setting is only available when support for OpenSSL was built in. It
11698 disables support for TLSv1.3 on any sockets instantiated from the listener
11699 when SSL is supported. Note that SSLv2 is forced disabled in the code and
11700 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011701 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11702 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011703
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011704npn <protocols>
11705 This enables the NPN TLS extension and advertises the specified protocol list
11706 as supported on top of NPN. The protocol list consists in a comma-delimited
11707 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050011708 This requires that the SSL library is built with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020011709 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010011710 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11711 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
11712 version of OpenSSL, NPN might still be used as most clients still support it
11713 at the time of writing this. It is possible to enable both NPN and ALPN
11714 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011715
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011716prefer-client-ciphers
11717 Use the client's preference when selecting the cipher suite, by default
11718 the server's preference is enforced. This option is also available on
11719 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020011720 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
11721 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
11722 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011723
Christopher Fauletc644fa92017-11-23 22:44:11 +010011724process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010011725 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010011726 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010011727 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010011728 between the two is applied. If in the end the listener is not allowed to run
11729 on any remaining process, a warning is emitted, and the listener will either
11730 run on the first process of the listener if a single process was specified,
11731 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010011732 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010011733 connections for this listener, for the the process set. If multiple processes
11734 and threads are configured, a warning is emitted, as it either results from a
11735 configuration error or a misunderstanding of these models. For the unlikely
11736 case where several ranges are needed, this directive may be repeated.
11737 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010011738
11739 all | odd | even | number[-[number]]
11740
11741 Ranges can be partially defined. The higher bound can be omitted. In such
11742 case, it is replaced by the corresponding maximum value. The main purpose of
11743 this directive is to be used with the stats sockets and have one different
11744 socket per process. The second purpose is to have multiple bind lines sharing
11745 the same IP:port but not the same process in a listener, so that the system
11746 can distribute the incoming connections into multiple queues and allow a
11747 smoother inter-process load balancing. Currently Linux 3.9 and above is known
11748 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020011749
Christopher Fauleta717b992018-04-10 14:43:00 +020011750proto <name>
11751 Forces the multiplexer's protocol to use for the incoming connections. It
11752 must be compatible with the mode of the frontend (TCP or HTTP). It must also
11753 be usable on the frontend side. The list of available protocols is reported
11754 in haproxy -vv.
11755 Idea behind this optipon is to bypass the selection of the best multiplexer's
11756 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080011757 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020011758 h2" on the bind line.
11759
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011760ssl
11761 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011762 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011763 certificate is necessary (see "crt" above). All contents in the buffers will
11764 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011765 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11766 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011767
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011768ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11769 This option enforces use of <version> or lower on SSL connections instantiated
11770 from this listener. This option is also available on global statement
11771 "ssl-default-bind-options". See also "ssl-min-ver".
11772
11773ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11774 This option enforces use of <version> or upper on SSL connections instantiated
11775 from this listener. This option is also available on global statement
11776 "ssl-default-bind-options". See also "ssl-max-ver".
11777
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011778strict-sni
11779 This setting is only available when support for OpenSSL was built in. The
11780 SSL/TLS negotiation is allow only if the client provided an SNI which match
11781 a certificate. The default certificate is not used.
11782 See the "crt" option for more information.
11783
Willy Tarreau2af207a2015-02-04 00:45:58 +010011784tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011785 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011786 listening socket. This option is available on Linux since version 2.6.37. It
11787 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011788 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011789 useful on long-lived connections experiencing long idle periods such as
11790 remote terminals or database connection pools, where the client and server
11791 timeouts must remain high to allow a long period of idle, but where it is
11792 important to detect that the client has disappeared in order to release all
11793 resources associated with its connection (and the server's session). The
11794 argument is a delay expressed in milliseconds by default. This only works
11795 for regular TCP connections, and is ignored for other protocols.
11796
Willy Tarreau1c862c52012-10-05 16:21:00 +020011797tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011798 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011799 enables TCP Fast Open on the listening socket, which means that clients which
11800 support this feature will be able to send a request and receive a response
11801 during the 3-way handshake starting from second connection, thus saving one
11802 round-trip after the first connection. This only makes sense with protocols
11803 that use high connection rates and where each round trip matters. This can
11804 possibly cause issues with many firewalls which do not accept data on SYN
11805 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011806 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11807 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11808 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011809
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011810tls-ticket-keys <keyfile>
11811 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010011812 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
11813 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
11814 The first key determines the key length used for next keys: you can't mix
11815 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
11816 build option (default 3) and at least as many keys need to be present in
11817 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
11818 penultimate one for encryption. This enables easy key rotation by just
11819 appending new key to the file and reloading the process. Keys must be
11820 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
11821 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011822 storage such as hard drives (hint: use tmpfs and don't swap those files).
11823 Lifetime hint can be changed using tune.ssl.timeout.
11824
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011825transparent
11826 Is an optional keyword which is supported only on certain Linux kernels. It
11827 indicates that the addresses will be bound even if they do not belong to the
11828 local machine, and that packets targeting any of these addresses will be
11829 intercepted just as if the addresses were locally configured. This normally
11830 requires that IP forwarding is enabled. Caution! do not use this with the
11831 default address '*', as it would redirect any traffic for the specified port.
11832 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11833 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11834 kernel version. Some distribution kernels include backports of the feature,
11835 so check for support with your vendor.
11836
Willy Tarreau77e3af92012-11-24 15:07:23 +010011837v4v6
11838 Is an optional keyword which is supported only on most recent systems
11839 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11840 and IPv6 when it uses the default address. Doing so is sometimes necessary
11841 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011842 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011843
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011844v6only
11845 Is an optional keyword which is supported only on most recent systems
11846 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11847 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011848 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11849 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011850
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011851uid <uid>
11852 Sets the owner of the UNIX sockets to the designated system uid. It can also
11853 be set by default in the global section's "unix-bind" statement. Note that
11854 some platforms simply ignore this. This setting is equivalent to the "user"
11855 setting except that the user numeric ID is used instead of its name. This
11856 setting is ignored by non UNIX sockets.
11857
11858user <user>
11859 Sets the owner of the UNIX sockets to the designated system user. It can also
11860 be set by default in the global section's "unix-bind" statement. Note that
11861 some platforms simply ignore this. This setting is equivalent to the "uid"
11862 setting except that the user name is used instead of its uid. This setting is
11863 ignored by non UNIX sockets.
11864
Emeric Brun1a073b42012-09-28 17:07:34 +020011865verify [none|optional|required]
11866 This setting is only available when support for OpenSSL was built in. If set
11867 to 'none', client certificate is not requested. This is the default. In other
11868 cases, a client certificate is requested. If the client does not provide a
11869 certificate after the request and if 'verify' is set to 'required', then the
11870 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011871 certificate provided by the client is always verified using CAs from
11872 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11873 is aborted, regardless of the 'verify' option, unless the error code exactly
11874 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011875
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200118765.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011877------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011878
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011879The "server" and "default-server" keywords support a certain number of settings
11880which are all passed as arguments on the server line. The order in which those
11881arguments appear does not count, and they are all optional. Some of those
11882settings are single words (booleans) while others expect one or several values
11883after them. In this case, the values must immediately follow the setting name.
11884Except default-server, all those settings must be specified after the server's
11885address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011886
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011887 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011888 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020011889
Frédéric Lécailled2376272017-03-21 18:52:12 +010011890Note that all these settings are supported both by "server" and "default-server"
11891keywords, except "id" which is only supported by "server".
11892
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011893The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011894
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011895addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011896 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010011897 to send health-checks or to probe the agent-check. On some servers, it may be
11898 desirable to dedicate an IP address to specific component able to perform
11899 complex tests which are more suitable to health-checks than the application.
11900 This parameter is ignored if the "check" parameter is not set. See also the
11901 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011902
Simon Hormand60d6912013-11-25 10:46:36 +090011903agent-check
11904 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010011905 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010011906 to the port set by the "agent-port" parameter and reading an ASCII string
11907 terminated by the first '\r' or '\n' met. The string is made of a series of
11908 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090011909
Willy Tarreau81f5d942013-12-09 20:51:51 +010011910 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090011911 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020011912 weight of a server as configured when haproxy starts. Note that a zero
11913 weight is reported on the stats page as "DRAIN" since it has the same
11914 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090011915
Davor Ocelice9ed2812017-12-25 17:49:28 +010011916 - The string "maxconn:" followed by an integer (no space between). Values
11917 in this format will set the maxconn of a server. The maximum number of
11918 connections advertised needs to be multiplied by the number of load
11919 balancers and different backends that use this health check to get the
11920 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020011921
Willy Tarreau81f5d942013-12-09 20:51:51 +010011922 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011923 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090011924
Willy Tarreau81f5d942013-12-09 20:51:51 +010011925 - The word "drain". This will turn the server's administrative state to the
11926 DRAIN mode, thus it will not accept any new connections other than those
11927 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090011928
Willy Tarreau81f5d942013-12-09 20:51:51 +010011929 - The word "maint". This will turn the server's administrative state to the
11930 MAINT mode, thus it will not accept any new connections at all, and health
11931 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090011932
Willy Tarreau81f5d942013-12-09 20:51:51 +010011933 - The words "down", "failed", or "stopped", optionally followed by a
11934 description string after a sharp ('#'). All of these mark the server's
11935 operating state as DOWN, but since the word itself is reported on the stats
11936 page, the difference allows an administrator to know if the situation was
11937 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010011938 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010011939 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090011940
Willy Tarreau81f5d942013-12-09 20:51:51 +010011941 - The word "up" sets back the server's operating state as UP if health checks
11942 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090011943
Willy Tarreau81f5d942013-12-09 20:51:51 +010011944 Parameters which are not advertised by the agent are not changed. For
11945 example, an agent might be designed to monitor CPU usage and only report a
11946 relative weight and never interact with the operating status. Similarly, an
11947 agent could be designed as an end-user interface with 3 radio buttons
11948 allowing an administrator to change only the administrative state. However,
11949 it is important to consider that only the agent may revert its own actions,
11950 so if a server is set to DRAIN mode or to DOWN state using the agent, the
11951 agent must implement the other equivalent actions to bring the service into
11952 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090011953
Simon Horman2f1f9552013-11-25 10:46:37 +090011954 Failure to connect to the agent is not considered an error as connectivity
11955 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010011956 parameter. Warning though, it is not a good idea to stop an agent after it
11957 reports "down", since only an agent reporting "up" will be able to turn the
11958 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010011959 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090011960
Willy Tarreau81f5d942013-12-09 20:51:51 +010011961 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010011962 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090011963
James Brown55f9ff12015-10-21 18:19:05 -070011964agent-send <string>
11965 If this option is specified, haproxy will send the given string (verbatim)
11966 to the agent server upon connection. You could, for example, encode
11967 the backend name into this string, which would enable your agent to send
11968 different responses based on the backend. Make sure to include a '\n' if
11969 you want to terminate your request with a newline.
11970
Simon Hormand60d6912013-11-25 10:46:36 +090011971agent-inter <delay>
11972 The "agent-inter" parameter sets the interval between two agent checks
11973 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11974
11975 Just as with every other time-based parameter, it may be entered in any
11976 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
11977 parameter also serves as a timeout for agent checks "timeout check" is
11978 not set. In order to reduce "resonance" effects when multiple servers are
11979 hosted on the same hardware, the agent and health checks of all servers
11980 are started with a small time offset between them. It is also possible to
11981 add some random noise in the agent and health checks interval using the
11982 global "spread-checks" keyword. This makes sense for instance when a lot
11983 of backends use the same servers.
11984
11985 See also the "agent-check" and "agent-port" parameters.
11986
Misiek768d8602017-01-09 09:52:43 +010011987agent-addr <addr>
11988 The "agent-addr" parameter sets address for agent check.
11989
11990 You can offload agent-check to another target, so you can make single place
11991 managing status and weights of servers defined in haproxy in case you can't
11992 make self-aware and self-managing services. You can specify both IP or
11993 hostname, it will be resolved.
11994
Simon Hormand60d6912013-11-25 10:46:36 +090011995agent-port <port>
11996 The "agent-port" parameter sets the TCP port used for agent checks.
11997
11998 See also the "agent-check" and "agent-inter" parameters.
11999
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020012000allow-0rtt
12001 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020012002 Note that early data will be sent only if the client used early data, or
12003 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020012004
Olivier Houchardc7566002018-11-20 23:33:50 +010012005alpn <protocols>
12006 This enables the TLS ALPN extension and advertises the specified protocol
12007 list as supported on top of ALPN. The protocol list consists in a comma-
12008 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050012009 quotes). This requires that the SSL library is built with support for TLS
Olivier Houchardc7566002018-11-20 23:33:50 +010012010 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
12011 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
12012 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
12013 now obsolete NPN extension.
12014 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
12015 be advertised, in order of preference, like below :
12016
12017 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
12018
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012019backup
12020 When "backup" is present on a server line, the server is only used in load
12021 balancing when all other non-backup servers are unavailable. Requests coming
12022 with a persistence cookie referencing the server will always be served
12023 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010012024 the "allbackups" option is set in the backend. See also the "no-backup" and
12025 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012026
Emeric Brunef42d922012-10-11 16:11:36 +020012027ca-file <cafile>
12028 This setting is only available when support for OpenSSL was built in. It
12029 designates a PEM file from which to load CA certificates used to verify
12030 server's certificate.
12031
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012032check
12033 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010012034 always considered available. If "check" is set, the server is available when
12035 accepting periodic TCP connections, to ensure that it is really able to serve
12036 requests. The default address and port to send the tests to are those of the
12037 server, and the default source is the same as the one defined in the
12038 backend. It is possible to change the address using the "addr" parameter, the
12039 port using the "port" parameter, the source address using the "source"
12040 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090012041 parameters. The request method is define in the backend using the "httpchk",
12042 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010012043 refer to those options and parameters for more information. See also
12044 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012045
Willy Tarreau6c16adc2012-10-05 00:04:16 +020012046check-send-proxy
12047 This option forces emission of a PROXY protocol line with outgoing health
12048 checks, regardless of whether the server uses send-proxy or not for the
12049 normal traffic. By default, the PROXY protocol is enabled for health checks
12050 if it is already enabled for normal traffic and if no "port" nor "addr"
12051 directive is present. However, if such a directive is present, the
12052 "check-send-proxy" option needs to be used to force the use of the
12053 protocol. See also the "send-proxy" option for more information.
12054
Olivier Houchard92150142018-12-21 19:47:01 +010012055check-alpn <protocols>
12056 Defines which protocols to advertise with ALPN. The protocol list consists in
12057 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
12058 (without quotes). If it is not set, the server ALPN is used.
12059
Jérôme Magninae9bb762018-12-09 16:08:26 +010012060check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020012061 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010012062 over SSL. It is only possible to use a string to set <sni>. If you want to
12063 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020012064
Willy Tarreau763a95b2012-10-04 23:15:39 +020012065check-ssl
12066 This option forces encryption of all health checks over SSL, regardless of
12067 whether the server uses SSL or not for the normal traffic. This is generally
12068 used when an explicit "port" or "addr" directive is specified and SSL health
12069 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012070 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020012071 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
12072 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012073 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010012074 See the "ssl" option for more information and "no-check-ssl" to disable
12075 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012076
Alexander Liu2a54bb72019-05-22 19:44:48 +080012077check-via-socks4
John Roeslerfb2fce12019-07-10 15:45:51 -050012078 This option enables outgoing health checks using upstream socks4 proxy. By
Alexander Liu2a54bb72019-05-22 19:44:48 +080012079 default, the health checks won't go through socks tunnel even it was enabled
12080 for normal traffic.
12081
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012082ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012083 This setting is only available when support for OpenSSL was built in. This
12084 option sets the string describing the list of cipher algorithms that is
12085 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000012086 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
12087 information and recommendations see e.g.
12088 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
12089 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
12090 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012091
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012092ciphersuites <ciphersuites>
12093 This setting is only available when support for OpenSSL was built in and
12094 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
12095 describing the list of cipher algorithms that is negotiated during the TLS
12096 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000012097 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
12098 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
12099 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012100
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012101cookie <value>
12102 The "cookie" parameter sets the cookie value assigned to the server to
12103 <value>. This value will be checked in incoming requests, and the first
12104 operational server possessing the same value will be selected. In return, in
12105 cookie insertion or rewrite modes, this value will be assigned to the cookie
12106 sent to the client. There is nothing wrong in having several servers sharing
12107 the same cookie value, and it is in fact somewhat common between normal and
12108 backup servers. See also the "cookie" keyword in backend section.
12109
Emeric Brunef42d922012-10-11 16:11:36 +020012110crl-file <crlfile>
12111 This setting is only available when support for OpenSSL was built in. It
12112 designates a PEM file from which to load certificate revocation list used
12113 to verify server's certificate.
12114
Emeric Bruna7aa3092012-10-26 12:58:00 +020012115crt <cert>
12116 This setting is only available when support for OpenSSL was built in.
12117 It designates a PEM file from which to load both a certificate and the
12118 associated private key. This file can be built by concatenating both PEM
12119 files into one. This certificate will be sent if the server send a client
12120 certificate request.
12121
Willy Tarreau96839092010-03-29 10:02:24 +020012122disabled
12123 The "disabled" keyword starts the server in the "disabled" state. That means
12124 that it is marked down in maintenance mode, and no connection other than the
12125 ones allowed by persist mode will reach it. It is very well suited to setup
12126 new servers, because normal traffic will never reach them, while it is still
12127 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012128 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012129
Frédéric Lécailled2376272017-03-21 18:52:12 +010012130enabled
12131 This option may be used as 'server' setting to reset any 'disabled'
12132 setting which would have been inherited from 'default-server' directive as
12133 default value.
12134 It may also be used as 'default-server' setting to reset any previous
12135 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012136
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012137error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010012138 If health observing is enabled, the "error-limit" parameter specifies the
12139 number of consecutive errors that triggers event selected by the "on-error"
12140 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012141
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012142 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012143
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012144fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012145 The "fall" parameter states that a server will be considered as dead after
12146 <count> consecutive unsuccessful health checks. This value defaults to 3 if
12147 unspecified. See also the "check", "inter" and "rise" parameters.
12148
Emeric Brun8694b9a2012-10-05 14:39:07 +020012149force-sslv3
12150 This option enforces use of SSLv3 only when SSL is used to communicate with
12151 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012152 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012153 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012154
12155force-tlsv10
12156 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012157 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012158 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012159
12160force-tlsv11
12161 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012162 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012163 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012164
12165force-tlsv12
12166 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012167 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012168 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012169
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012170force-tlsv13
12171 This option enforces use of TLSv1.3 only when SSL is used to communicate with
12172 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012173 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012174
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012175id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020012176 Set a persistent ID for the server. This ID must be positive and unique for
12177 the proxy. An unused ID will automatically be assigned if unset. The first
12178 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012179
Willy Tarreau6a031d12016-11-07 19:42:35 +010012180init-addr {last | libc | none | <ip>},[...]*
12181 Indicate in what order the server's address should be resolved upon startup
12182 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012183 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010012184 method which succeeds is used. If the end of the list is reached without
12185 finding a working method, an error is thrown. Method "last" suggests to pick
12186 the address which appears in the state file (see "server-state-file"). Method
12187 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
12188 depending on the operating system and build options). Method "none"
12189 specifically indicates that the server should start without any valid IP
12190 address in a down state. It can be useful to ignore some DNS issues upon
12191 startup, waiting for the situation to get fixed later. Finally, an IP address
12192 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012193 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010012194 server used to catch old sessions and present them with a decent error
12195 message for example. When the "first" load balancing algorithm is used, this
12196 IP address could point to a fake server used to trigger the creation of new
12197 instances on the fly. This option defaults to "last,libc" indicating that the
12198 previous address found in the state file (if any) is used first, otherwise
12199 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012200 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010012201
12202 Example:
12203 defaults
12204 # never fail on address resolution
12205 default-server init-addr last,libc,none
12206
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012207inter <delay>
12208fastinter <delay>
12209downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012210 The "inter" parameter sets the interval between two consecutive health checks
12211 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
12212 It is also possible to use "fastinter" and "downinter" to optimize delays
12213 between checks depending on the server state :
12214
Pieter Baauw44fc9df2015-09-17 21:30:46 +020012215 Server state | Interval used
12216 ----------------------------------------+----------------------------------
12217 UP 100% (non-transitional) | "inter"
12218 ----------------------------------------+----------------------------------
12219 Transitionally UP (going down "fall"), | "fastinter" if set,
12220 Transitionally DOWN (going up "rise"), | "inter" otherwise.
12221 or yet unchecked. |
12222 ----------------------------------------+----------------------------------
12223 DOWN 100% (non-transitional) | "downinter" if set,
12224 | "inter" otherwise.
12225 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010012226
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012227 Just as with every other time-based parameter, they can be entered in any
12228 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
12229 serves as a timeout for health checks sent to servers if "timeout check" is
12230 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090012231 hosted on the same hardware, the agent and health checks of all servers
12232 are started with a small time offset between them. It is also possible to
12233 add some random noise in the agent and health checks interval using the
12234 global "spread-checks" keyword. This makes sense for instance when a lot
12235 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012236
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012237maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012238 The "maxconn" parameter specifies the maximal number of concurrent
12239 connections that will be sent to this server. If the number of incoming
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010012240 concurrent connections goes higher than this value, they will be queued,
12241 waiting for a slot to be released. This parameter is very important as it can
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012242 save fragile servers from going down under extreme loads. If a "minconn"
12243 parameter is specified, the limit becomes dynamic. The default value is "0"
12244 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
12245 the backend's "fullconn" keyword.
12246
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010012247 In HTTP mode this parameter limits the number of concurrent requests instead
12248 of the number of connections. Multiple requests might be multiplexed over a
12249 single TCP connection to the server. As an example if you specify a maxconn
12250 of 50 you might see between 1 and 50 actual server connections, but no more
12251 than 50 concurrent requests.
12252
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012253maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012254 The "maxqueue" parameter specifies the maximal number of connections which
12255 will wait in the queue for this server. If this limit is reached, next
12256 requests will be redispatched to other servers instead of indefinitely
12257 waiting to be served. This will break persistence but may allow people to
12258 quickly re-log in when the server they try to connect to is dying. The
12259 default value is "0" which means the queue is unlimited. See also the
12260 "maxconn" and "minconn" parameters.
12261
Willy Tarreau9c538e02019-01-23 10:21:49 +010012262max-reuse <count>
12263 The "max-reuse" argument indicates the HTTP connection processors that they
12264 should not reuse a server connection more than this number of times to send
12265 new requests. Permitted values are -1 (the default), which disables this
12266 limit, or any positive value. Value zero will effectively disable keep-alive.
12267 This is only used to work around certain server bugs which cause them to leak
12268 resources over time. The argument is not necessarily respected by the lower
12269 layers as there might be technical limitations making it impossible to
12270 enforce. At least HTTP/2 connections to servers will respect it.
12271
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012272minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012273 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
12274 limit following the backend's load. The server will always accept at least
12275 <minconn> connections, never more than <maxconn>, and the limit will be on
12276 the ramp between both values when the backend has less than <fullconn>
12277 concurrent connections. This makes it possible to limit the load on the
12278 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012279 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012280 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012281
Willy Tarreaud72f0f32015-10-13 14:50:22 +020012282namespace <name>
12283 On Linux, it is possible to specify which network namespace a socket will
12284 belong to. This directive makes it possible to explicitly bind a server to
12285 a namespace different from the default one. Please refer to your operating
12286 system's documentation to find more details about network namespaces.
12287
Frédéric Lécailled2376272017-03-21 18:52:12 +010012288no-agent-check
12289 This option may be used as "server" setting to reset any "agent-check"
12290 setting which would have been inherited from "default-server" directive as
12291 default value.
12292 It may also be used as "default-server" setting to reset any previous
12293 "default-server" "agent-check" setting.
12294
12295no-backup
12296 This option may be used as "server" setting to reset any "backup"
12297 setting which would have been inherited from "default-server" directive as
12298 default value.
12299 It may also be used as "default-server" setting to reset any previous
12300 "default-server" "backup" setting.
12301
12302no-check
12303 This option may be used as "server" setting to reset any "check"
12304 setting which would have been inherited from "default-server" directive as
12305 default value.
12306 It may also be used as "default-server" setting to reset any previous
12307 "default-server" "check" setting.
12308
12309no-check-ssl
12310 This option may be used as "server" setting to reset any "check-ssl"
12311 setting which would have been inherited from "default-server" directive as
12312 default value.
12313 It may also be used as "default-server" setting to reset any previous
12314 "default-server" "check-ssl" setting.
12315
Frédéric Lécailled2376272017-03-21 18:52:12 +010012316no-send-proxy
12317 This option may be used as "server" setting to reset any "send-proxy"
12318 setting which would have been inherited from "default-server" directive as
12319 default value.
12320 It may also be used as "default-server" setting to reset any previous
12321 "default-server" "send-proxy" setting.
12322
12323no-send-proxy-v2
12324 This option may be used as "server" setting to reset any "send-proxy-v2"
12325 setting which would have been inherited from "default-server" directive as
12326 default value.
12327 It may also be used as "default-server" setting to reset any previous
12328 "default-server" "send-proxy-v2" setting.
12329
12330no-send-proxy-v2-ssl
12331 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
12332 setting which would have been inherited from "default-server" directive as
12333 default value.
12334 It may also be used as "default-server" setting to reset any previous
12335 "default-server" "send-proxy-v2-ssl" setting.
12336
12337no-send-proxy-v2-ssl-cn
12338 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
12339 setting which would have been inherited from "default-server" directive as
12340 default value.
12341 It may also be used as "default-server" setting to reset any previous
12342 "default-server" "send-proxy-v2-ssl-cn" setting.
12343
12344no-ssl
12345 This option may be used as "server" setting to reset any "ssl"
12346 setting which would have been inherited from "default-server" directive as
12347 default value.
12348 It may also be used as "default-server" setting to reset any previous
12349 "default-server" "ssl" setting.
12350
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010012351no-ssl-reuse
12352 This option disables SSL session reuse when SSL is used to communicate with
12353 the server. It will force the server to perform a full handshake for every
12354 new connection. It's probably only useful for benchmarking, troubleshooting,
12355 and for paranoid users.
12356
Emeric Brun9b3009b2012-10-05 11:55:06 +020012357no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012358 This option disables support for SSLv3 when SSL is used to communicate with
12359 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012360 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012361
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012362 Supported in default-server: No
12363
Emeric Brunf9c5c472012-10-11 15:28:34 +020012364no-tls-tickets
12365 This setting is only available when support for OpenSSL was built in. It
12366 disables the stateless session resumption (RFC 5077 TLS Ticket
12367 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012368 session resumption is more expensive in CPU usage for servers. This option
12369 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010012370 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020012371
Emeric Brun9b3009b2012-10-05 11:55:06 +020012372no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020012373 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012374 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12375 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012376 often makes sense to disable it when communicating with local servers. This
12377 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012378 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012379
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012380 Supported in default-server: No
12381
Emeric Brun9b3009b2012-10-05 11:55:06 +020012382no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020012383 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012384 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12385 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012386 often makes sense to disable it when communicating with local servers. This
12387 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012388 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012389
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012390 Supported in default-server: No
12391
Emeric Brun9b3009b2012-10-05 11:55:06 +020012392no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020012393 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012394 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12395 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012396 often makes sense to disable it when communicating with local servers. This
12397 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012398 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012399
12400 Supported in default-server: No
12401
12402no-tlsv13
12403 This option disables support for TLSv1.3 when SSL is used to communicate with
12404 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12405 using any configuration option. TLSv1 is more expensive than SSLv3 so it
12406 often makes sense to disable it when communicating with local servers. This
12407 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012408 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012409
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012410 Supported in default-server: No
12411
Frédéric Lécailled2376272017-03-21 18:52:12 +010012412no-verifyhost
12413 This option may be used as "server" setting to reset any "verifyhost"
12414 setting which would have been inherited from "default-server" directive as
12415 default value.
12416 It may also be used as "default-server" setting to reset any previous
12417 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012418
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020012419no-tfo
12420 This option may be used as "server" setting to reset any "tfo"
12421 setting which would have been inherited from "default-server" directive as
12422 default value.
12423 It may also be used as "default-server" setting to reset any previous
12424 "default-server" "tfo" setting.
12425
Simon Hormanfa461682011-06-25 09:39:49 +090012426non-stick
12427 Never add connections allocated to this sever to a stick-table.
12428 This may be used in conjunction with backup to ensure that
12429 stick-table persistence is disabled for backup servers.
12430
Olivier Houchardc7566002018-11-20 23:33:50 +010012431npn <protocols>
12432 This enables the NPN TLS extension and advertises the specified protocol list
12433 as supported on top of NPN. The protocol list consists in a comma-delimited
12434 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050012435 This requires that the SSL library is built with support for TLS extensions
Olivier Houchardc7566002018-11-20 23:33:50 +010012436 enabled (check with haproxy -vv). Note that the NPN extension has been
12437 replaced with the ALPN extension (see the "alpn" keyword), though this one is
12438 only available starting with OpenSSL 1.0.2.
12439
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012440observe <mode>
12441 This option enables health adjusting based on observing communication with
12442 the server. By default this functionality is disabled and enabling it also
12443 requires to enable health checks. There are two supported modes: "layer4" and
12444 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
12445 significant. In layer7, which is only allowed for http proxies, responses
12446 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010012447 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012448
12449 See also the "check", "on-error" and "error-limit".
12450
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012451on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012452 Select what should happen when enough consecutive errors are detected.
12453 Currently, four modes are available:
12454 - fastinter: force fastinter
12455 - fail-check: simulate a failed check, also forces fastinter (default)
12456 - sudden-death: simulate a pre-fatal failed health check, one more failed
12457 check will mark a server down, forces fastinter
12458 - mark-down: mark the server immediately down and force fastinter
12459
12460 See also the "check", "observe" and "error-limit".
12461
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012462on-marked-down <action>
12463 Modify what occurs when a server is marked down.
12464 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012465 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
12466 all connections to the server are immediately terminated when the server
12467 goes down. It might be used if the health check detects more complex cases
12468 than a simple connection status, and long timeouts would cause the service
12469 to remain unresponsive for too long a time. For instance, a health check
12470 might detect that a database is stuck and that there's no chance to reuse
12471 existing connections anymore. Connections killed this way are logged with
12472 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012473
12474 Actions are disabled by default
12475
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012476on-marked-up <action>
12477 Modify what occurs when a server is marked up.
12478 Currently one action is available:
12479 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
12480 done only if the server is not in backup state and if it is not disabled
12481 (it must have an effective weight > 0). This can be used sometimes to force
12482 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012483 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
12484 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012485 with extreme care. Sessions killed because a server comes up are logged
12486 with an 'U' termination code (for "Up").
12487
12488 Actions are disabled by default
12489
Olivier Houchard006e3102018-12-10 18:30:32 +010012490pool-max-conn <max>
12491 Set the maximum number of idling connections for a server. -1 means unlimited
12492 connections, 0 means no idle connections. The default is -1. When idle
12493 connections are enabled, orphaned idle connections which do not belong to any
12494 client session anymore are moved to a dedicated pool so that they remain
12495 usable by future clients. This only applies to connections that can be shared
12496 according to the same principles as those applying to "http-reuse".
12497
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012498pool-purge-delay <delay>
12499 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010012500 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020012501 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012502
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012503port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012504 Using the "port" parameter, it becomes possible to use a different port to
12505 send health-checks. On some servers, it may be desirable to dedicate a port
12506 to a specific component able to perform complex tests which are more suitable
12507 to health-checks than the application. It is common to run a simple script in
12508 inetd for instance. This parameter is ignored if the "check" parameter is not
12509 set. See also the "addr" parameter.
12510
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020012511proto <name>
12512
12513 Forces the multiplexer's protocol to use for the outgoing connections to this
12514 server. It must be compatible with the mode of the backend (TCP or HTTP). It
12515 must also be usable on the backend side. The list of available protocols is
12516 reported in haproxy -vv.
12517 Idea behind this optipon is to bypass the selection of the best multiplexer's
12518 protocol for all connections established to this server.
12519
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012520redir <prefix>
12521 The "redir" parameter enables the redirection mode for all GET and HEAD
12522 requests addressing this server. This means that instead of having HAProxy
12523 forward the request to the server, it will send an "HTTP 302" response with
12524 the "Location" header composed of this prefix immediately followed by the
12525 requested URI beginning at the leading '/' of the path component. That means
12526 that no trailing slash should be used after <prefix>. All invalid requests
12527 will be rejected, and all non-GET or HEAD requests will be normally served by
12528 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012529 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012530 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012531 users to a remote location in case of local disaster. Main use consists in
12532 increasing bandwidth for static servers by having the clients directly
12533 connect to them. Note: never use a relative location here, it would cause a
12534 loop between the client and HAProxy!
12535
12536 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
12537
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012538rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012539 The "rise" parameter states that a server will be considered as operational
12540 after <count> consecutive successful health checks. This value defaults to 2
12541 if unspecified. See also the "check", "inter" and "fall" parameters.
12542
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012543resolve-opts <option>,<option>,...
12544 Comma separated list of options to apply to DNS resolution linked to this
12545 server.
12546
12547 Available options:
12548
12549 * allow-dup-ip
12550 By default, HAProxy prevents IP address duplication in a backend when DNS
12551 resolution at runtime is in operation.
12552 That said, for some cases, it makes sense that two servers (in the same
12553 backend, being resolved by the same FQDN) have the same IP address.
12554 For such case, simply enable this option.
12555 This is the opposite of prevent-dup-ip.
12556
Daniel Corbettf8716912019-11-17 09:48:56 -050012557 * ignore-weight
12558 Ignore any weight that is set within an SRV record. This is useful when
12559 you would like to control the weights using an alternate method, such as
12560 using an "agent-check" or through the runtime api.
12561
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012562 * prevent-dup-ip
12563 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
12564 an IP address already set to a server in the same backend and sharing the
12565 same fqdn.
12566 This is the opposite of allow-dup-ip.
12567
12568 Example:
12569 backend b_myapp
12570 default-server init-addr none resolvers dns
12571 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
12572 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
12573
12574 With the option allow-dup-ip set:
12575 * if the nameserver returns a single IP address, then both servers will use
12576 it
12577 * If the nameserver returns 2 IP addresses, then each server will pick up a
12578 different address
12579
12580 Default value: not set
12581
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012582resolve-prefer <family>
12583 When DNS resolution is enabled for a server and multiple IP addresses from
12584 different families are returned, HAProxy will prefer using an IP address
12585 from the family mentioned in the "resolve-prefer" parameter.
12586 Available families: "ipv4" and "ipv6"
12587
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020012588 Default value: ipv6
12589
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012590 Example:
12591
12592 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012593
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012594resolve-net <network>[,<network[,...]]
John Roeslerfb2fce12019-07-10 15:45:51 -050012595 This option prioritizes the choice of an ip address matching a network. This is
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012596 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010012597 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010012598 different datacenters. The latency between datacenter is not negligible, so
12599 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012600 configured network, another address is selected.
12601
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012602 Example:
12603
12604 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012605
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012606resolvers <id>
12607 Points to an existing "resolvers" section to resolve current server's
12608 hostname.
12609
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012610 Example:
12611
12612 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012613
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012614 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012615
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012616send-proxy
12617 The "send-proxy" parameter enforces use of the PROXY protocol over any
12618 connection established to this server. The PROXY protocol informs the other
12619 end about the layer 3/4 addresses of the incoming connection, so that it can
12620 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010012621 upper layer protocol. For connections accepted by an "accept-proxy" or
12622 "accept-netscaler-cip" listener, the advertised address will be used. Only
12623 TCPv4 and TCPv6 address families are supported. Other families such as
12624 Unix sockets, will report an UNKNOWN family. Servers using this option can
12625 fully be chained to another instance of haproxy listening with an
12626 "accept-proxy" setting. This setting must not be used if the server isn't
12627 aware of the protocol. When health checks are sent to the server, the PROXY
12628 protocol is automatically used when this option is set, unless there is an
12629 explicit "port" or "addr" directive, in which case an explicit
12630 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012631 See also the "no-send-proxy" option of this section and "accept-proxy" and
12632 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012633
David Safb76832014-05-08 23:42:08 -040012634send-proxy-v2
12635 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
12636 over any connection established to this server. The PROXY protocol informs
12637 the other end about the layer 3/4 addresses of the incoming connection, so
12638 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020012639 whatever the upper layer protocol. It also send ALPN information if an alpn
12640 have been negotiated. This setting must not be used if the server isn't aware
12641 of this version of the protocol. See also the "no-send-proxy-v2" option of
12642 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012643
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012644proxy-v2-options <option>[,<option>]*
12645 The "proxy-v2-options" parameter add option to send in PROXY protocol version
12646 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010012647 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
12648 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010012649 certificate, "cert-key": key algorithm of the used certificate), "authority":
12650 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010012651 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012652
David Safb76832014-05-08 23:42:08 -040012653send-proxy-v2-ssl
12654 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12655 2 over any connection established to this server. The PROXY protocol informs
12656 the other end about the layer 3/4 addresses of the incoming connection, so
12657 that it can know the client's address or the public address it accessed to,
12658 whatever the upper layer protocol. In addition, the SSL information extension
12659 of the PROXY protocol is added to the PROXY protocol header. This setting
12660 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012661 See also the "no-send-proxy-v2-ssl" option of this section and the
12662 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012663
12664send-proxy-v2-ssl-cn
12665 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12666 2 over any connection established to this server. The PROXY protocol informs
12667 the other end about the layer 3/4 addresses of the incoming connection, so
12668 that it can know the client's address or the public address it accessed to,
12669 whatever the upper layer protocol. In addition, the SSL information extension
12670 of the PROXY protocol, along along with the Common Name from the subject of
12671 the client certificate (if any), is added to the PROXY protocol header. This
12672 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012673 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
12674 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012675
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012676slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012677 The "slowstart" parameter for a server accepts a value in milliseconds which
12678 indicates after how long a server which has just come back up will run at
12679 full speed. Just as with every other time-based parameter, it can be entered
12680 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
12681 linearly from 0 to 100% during this time. The limitation applies to two
12682 parameters :
12683
12684 - maxconn: the number of connections accepted by the server will grow from 1
12685 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
12686
12687 - weight: when the backend uses a dynamic weighted algorithm, the weight
12688 grows linearly from 1 to 100%. In this case, the weight is updated at every
12689 health-check. For this reason, it is important that the "inter" parameter
12690 is smaller than the "slowstart", in order to maximize the number of steps.
12691
12692 The slowstart never applies when haproxy starts, otherwise it would cause
12693 trouble to running servers. It only applies when a server has been previously
12694 seen as failed.
12695
Willy Tarreau732eac42015-07-09 11:40:25 +020012696sni <expression>
12697 The "sni" parameter evaluates the sample fetch expression, converts it to a
12698 string and uses the result as the host name sent in the SNI TLS extension to
12699 the server. A typical use case is to send the SNI received from the client in
12700 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020012701 expression, though alternatives such as req.hdr(host) can also make sense. If
12702 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012703 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010012704 "verify" directive for more details. If you want to set a SNI for health
12705 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020012706
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012707source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020012708source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012709source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012710 The "source" parameter sets the source address which will be used when
12711 connecting to the server. It follows the exact same parameters and principle
12712 as the backend "source" keyword, except that it only applies to the server
12713 referencing it. Please consult the "source" keyword for details.
12714
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012715 Additionally, the "source" statement on a server line allows one to specify a
12716 source port range by indicating the lower and higher bounds delimited by a
12717 dash ('-'). Some operating systems might require a valid IP address when a
12718 source port range is specified. It is permitted to have the same IP/range for
12719 several servers. Doing so makes it possible to bypass the maximum of 64k
12720 total concurrent connections. The limit will then reach 64k connections per
12721 server.
12722
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000012723 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
12724 specifying the source address without port(s).
12725
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012726ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020012727 This option enables SSL ciphering on outgoing connections to the server. It
12728 is critical to verify server certificates using "verify" when using SSL to
12729 connect to servers, otherwise the communication is prone to trivial man in
12730 the-middle attacks rendering SSL useless. When this option is used, health
12731 checks are automatically sent in SSL too unless there is a "port" or an
12732 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012733 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
12734 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012735
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012736ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12737 This option enforces use of <version> or lower when SSL is used to communicate
12738 with the server. This option is also available on global statement
12739 "ssl-default-server-options". See also "ssl-min-ver".
12740
12741ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12742 This option enforces use of <version> or upper when SSL is used to communicate
12743 with the server. This option is also available on global statement
12744 "ssl-default-server-options". See also "ssl-max-ver".
12745
Frédéric Lécailled2376272017-03-21 18:52:12 +010012746ssl-reuse
12747 This option may be used as "server" setting to reset any "no-ssl-reuse"
12748 setting which would have been inherited from "default-server" directive as
12749 default value.
12750 It may also be used as "default-server" setting to reset any previous
12751 "default-server" "no-ssl-reuse" setting.
12752
12753stick
12754 This option may be used as "server" setting to reset any "non-stick"
12755 setting which would have been inherited from "default-server" directive as
12756 default value.
12757 It may also be used as "default-server" setting to reset any previous
12758 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012759
Alexander Liu2a54bb72019-05-22 19:44:48 +080012760socks4 <addr>:<port>
John Roeslerfb2fce12019-07-10 15:45:51 -050012761 This option enables upstream socks4 tunnel for outgoing connections to the
Alexander Liu2a54bb72019-05-22 19:44:48 +080012762 server. Using this option won't force the health check to go via socks4 by
12763 default. You will have to use the keyword "check-via-socks4" to enable it.
12764
Willy Tarreau163d4622015-10-13 16:16:41 +020012765tcp-ut <delay>
12766 Sets the TCP User Timeout for all outgoing connections to this server. This
12767 option is available on Linux since version 2.6.37. It allows haproxy to
12768 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012769 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020012770 long-lived connections experiencing long idle periods such as remote
12771 terminals or database connection pools, where the client and server timeouts
12772 must remain high to allow a long period of idle, but where it is important to
12773 detect that the server has disappeared in order to release all resources
12774 associated with its connection (and the client's session). One typical use
12775 case is also to force dead server connections to die when health checks are
12776 too slow or during a soft reload since health checks are then disabled. The
12777 argument is a delay expressed in milliseconds by default. This only works for
12778 regular TCP connections, and is ignored for other protocols.
12779
Willy Tarreau034c88c2017-01-23 23:36:45 +010012780tfo
12781 This option enables using TCP fast open when connecting to servers, on
12782 systems that support it (currently only the Linux kernel >= 4.11).
12783 See the "tfo" bind option for more information about TCP fast open.
12784 Please note that when using tfo, you should also use the "conn-failure",
12785 "empty-response" and "response-timeout" keywords for "retry-on", or haproxy
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020012786 won't be able to retry the connection on failure. See also "no-tfo".
Willy Tarreau034c88c2017-01-23 23:36:45 +010012787
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012788track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020012789 This option enables ability to set the current state of the server by tracking
12790 another one. It is possible to track a server which itself tracks another
12791 server, provided that at the end of the chain, a server has health checks
12792 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012793 used, it has to be enabled on both proxies.
12794
Frédéric Lécailled2376272017-03-21 18:52:12 +010012795tls-tickets
12796 This option may be used as "server" setting to reset any "no-tls-tickets"
12797 setting which would have been inherited from "default-server" directive as
12798 default value.
12799 It may also be used as "default-server" setting to reset any previous
12800 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012801
Emeric Brunef42d922012-10-11 16:11:36 +020012802verify [none|required]
12803 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010012804 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012805 certificate provided by the server is verified using CAs from 'ca-file' and
12806 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012807 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012808 the name passed using the "sni" directive, or if not provided, the static
12809 host name passed using the "verifyhost" directive. When no name is found, the
12810 certificate's names are ignored. For this reason, without SNI it's important
12811 to use "verifyhost". On verification failure the handshake is aborted. It is
12812 critically important to verify server certificates when using SSL to connect
12813 to servers, otherwise the communication is prone to trivial man-in-the-middle
12814 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
12815 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020012816
Evan Broderbe554312013-06-27 00:05:25 -070012817verifyhost <hostname>
12818 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012819 only takes effect if 'verify required' is also specified. This directive sets
12820 a default static hostname to check the server's certificate against when no
12821 SNI was used to connect to the server. If SNI is not used, this is the only
12822 way to enable hostname verification. This static hostname, when set, will
12823 also be used for health checks (which cannot provide an SNI value). If none
12824 of the hostnames in the certificate match the specified hostname, the
12825 handshake is aborted. The hostnames in the server-provided certificate may
12826 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070012827
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012828weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012829 The "weight" parameter is used to adjust the server's weight relative to
12830 other servers. All servers will receive a load proportional to their weight
12831 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020012832 load. The default weight is 1, and the maximal value is 256. A value of 0
12833 means the server will not participate in load-balancing but will still accept
12834 persistent connections. If this parameter is used to distribute the load
12835 according to server's capacity, it is recommended to start with values which
12836 can both grow and shrink, for instance between 10 and 100 to leave enough
12837 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012838
12839
Cyril Bonté46175dd2015-07-02 22:45:32 +0200128405.3. Server IP address resolution using DNS
12841-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012842
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012843HAProxy allows using a host name on the server line to retrieve its IP address
12844using name servers. By default, HAProxy resolves the name when parsing the
12845configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012846This is not sufficient in some cases, such as in Amazon where a server's IP
12847can change after a reboot or an ELB Virtual IP can change based on current
12848workload.
12849This chapter describes how HAProxy can be configured to process server's name
12850resolution at run time.
12851Whether run time server name resolution has been enable or not, HAProxy will
12852carry on doing the first resolution when parsing the configuration.
12853
12854
Cyril Bonté46175dd2015-07-02 22:45:32 +0200128555.3.1. Global overview
12856----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012857
12858As we've seen in introduction, name resolution in HAProxy occurs at two
12859different steps of the process life:
12860
12861 1. when starting up, HAProxy parses the server line definition and matches a
12862 host name. It uses libc functions to get the host name resolved. This
12863 resolution relies on /etc/resolv.conf file.
12864
Christopher Faulet67957bd2017-09-27 11:00:59 +020012865 2. at run time, HAProxy performs periodically name resolutions for servers
12866 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012867
12868A few other events can trigger a name resolution at run time:
12869 - when a server's health check ends up in a connection timeout: this may be
12870 because the server has a new IP address. So we need to trigger a name
12871 resolution to know this new IP.
12872
Christopher Faulet67957bd2017-09-27 11:00:59 +020012873When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012874HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020012875SRV label is specified, then the corresponding SRV records will be retrieved
12876from the DNS server, and the provided hostnames will be used. The SRV label
12877will be checked periodically, and if any server are added or removed, haproxy
12878will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020012879
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012880A few things important to notice:
John Roeslerfb2fce12019-07-10 15:45:51 -050012881 - all the name servers are queried in the meantime. HAProxy will process the
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012882 first valid response.
12883
12884 - a resolution is considered as invalid (NX, timeout, refused), when all the
12885 servers return an error.
12886
12887
Cyril Bonté46175dd2015-07-02 22:45:32 +0200128885.3.2. The resolvers section
12889----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012890
12891This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020012892HAProxy. There can be as many as resolvers section as needed. Each section can
12893contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012894
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012895When multiple name servers are configured in a resolvers section, then HAProxy
12896uses the first valid response. In case of invalid responses, only the last one
12897is treated. Purpose is to give the chance to a slow server to deliver a valid
12898answer after a fast faulty or outdated server.
12899
12900When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020012901used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012902
Christopher Faulet67957bd2017-09-27 11:00:59 +020012903 1. HAProxy retries the same DNS query with a new query type. The A queries are
12904 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
12905 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012906
Christopher Faulet67957bd2017-09-27 11:00:59 +020012907 2. When the fallback on the query type was done (or not applicable), HAProxy
12908 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012909
Christopher Faulet67957bd2017-09-27 11:00:59 +020012910 3. HAProxy retries previous steps <resolve_retires> times. If no valid
12911 response is received after that, it stops the DNS resolution and reports
12912 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012913
Christopher Faulet67957bd2017-09-27 11:00:59 +020012914For example, with 2 name servers configured in a resolvers section, the
12915following scenarios are possible:
12916
12917 - First response is valid and is applied directly, second response is
12918 ignored
12919
12920 - First response is invalid and second one is valid, then second response is
12921 applied
12922
12923 - First response is a NX domain and second one a truncated response, then
12924 HAProxy retries the query with a new type
12925
12926 - First response is a NX domain and second one is a timeout, then HAProxy
12927 retries the query with a new type
12928
12929 - Query timed out for both name servers, then HAProxy retries it with the
12930 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012931
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012932As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
12933a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020012934<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012935
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012936
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012937resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012938 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012939
12940A resolvers section accept the following parameters:
12941
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012942accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012943 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020012944 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012945 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
12946 by RFC 6891)
12947
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020012948 Note: the maximum allowed value is 8192.
12949
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012950nameserver <id> <ip>:<port>
12951 DNS server description:
12952 <id> : label of the server, should be unique
12953 <ip> : IP address of the server
12954 <port> : port where the DNS service actually runs
12955
Ben Draut44e609b2018-05-29 15:40:08 -060012956parse-resolv-conf
12957 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
12958 list. Ordered as if each nameserver in /etc/resolv.conf was individually
12959 placed in the resolvers section in place of this directive.
12960
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012961hold <status> <period>
12962 Defines <period> during which the last name resolution should be kept based
12963 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012964 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012965 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012966 <period> : interval between two successive name resolution when the last
12967 answer was in <status>. It follows the HAProxy time format.
12968 <period> is in milliseconds by default.
12969
Baptiste Assmann686408b2017-08-18 10:15:42 +020012970 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012971
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012972resolve_retries <nb>
12973 Defines the number <nb> of queries to send to resolve a server name before
12974 giving up.
12975 Default value: 3
12976
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012977 A retry occurs on name server timeout or when the full sequence of DNS query
12978 type failover is over and we need to start up from the default ANY query
12979 type.
12980
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012981timeout <event> <time>
12982 Defines timeouts related to name resolution
12983 <event> : the event on which the <time> timeout period applies to.
12984 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012985 - resolve : default time to trigger name resolutions when no
12986 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012987 Default value: 1s
12988 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012989 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012990 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012991 <time> : time related to the event. It follows the HAProxy time format.
12992 <time> is expressed in milliseconds.
12993
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012994 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012995
12996 resolvers mydns
12997 nameserver dns1 10.0.0.1:53
12998 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060012999 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013000 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020013001 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013002 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010013003 hold other 30s
13004 hold refused 30s
13005 hold nx 30s
13006 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013007 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020013008 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013009
13010
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200130116. Cache
13012---------
13013
13014HAProxy provides a cache, which was designed to perform cache on small objects
13015(favicon, css...). This is a minimalist low-maintenance cache which runs in
13016RAM.
13017
13018The cache is based on a memory which is shared between processes and threads,
13019this memory is split in blocks of 1k.
13020
13021If an object is not used anymore, it can be deleted to store a new object
13022independently of its expiration date. The oldest objects are deleted first
13023when we try to allocate a new one.
13024
13025The cache uses a hash of the host header and the URI as the key.
13026
13027It's possible to view the status of a cache using the Unix socket command
13028"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
13029for more details.
13030
13031When an object is delivered from the cache, the server name in the log is
13032replaced by "<CACHE>".
13033
13034
130356.1. Limitation
13036----------------
13037
13038The cache won't store and won't deliver objects in these cases:
13039
13040- If the response is not a 200
13041- If the response contains a Vary header
13042- If the Content-Length + the headers size is greater than "max-object-size"
13043- If the response is not cacheable
13044
13045- If the request is not a GET
13046- If the HTTP version of the request is smaller than 1.1
13047- If the request contains an Authorization header
13048
13049
130506.2. Setup
13051-----------
13052
13053To setup a cache, you must define a cache section and use it in a proxy with
13054the corresponding http-request and response actions.
13055
13056
130576.2.1. Cache section
13058---------------------
13059
13060cache <name>
13061 Declare a cache section, allocate a shared cache memory named <name>, the
13062 size of cache is mandatory.
13063
13064total-max-size <megabytes>
13065 Define the size in RAM of the cache in megabytes. This size is split in
13066 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
13067
13068max-object-size <bytes>
13069 Define the maximum size of the objects to be cached. Must not be greater than
13070 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
13071 All objects with sizes larger than "max-object-size" will not be cached.
13072
13073max-age <seconds>
13074 Define the maximum expiration duration. The expiration is set has the lowest
13075 value between the s-maxage or max-age (in this order) directive in the
13076 Cache-Control response header and this value. The default value is 60
13077 seconds, which means that you can't cache an object more than 60 seconds by
13078 default.
13079
13080
130816.2.2. Proxy section
13082---------------------
13083
13084http-request cache-use <name> [ { if | unless } <condition> ]
13085 Try to deliver a cached object from the cache <name>. This directive is also
13086 mandatory to store the cache as it calculates the cache hash. If you want to
13087 use a condition for both storage and delivering that's a good idea to put it
13088 after this one.
13089
13090http-response cache-store <name> [ { if | unless } <condition> ]
13091 Store an http-response within the cache. The storage of the response headers
13092 is done at this step, which means you can use others http-response actions
13093 to modify headers before or after the storage of the response. This action
13094 is responsible for the setup of the cache storage filter.
13095
13096
13097Example:
13098
13099 backend bck1
13100 mode http
13101
13102 http-request cache-use foobar
13103 http-response cache-store foobar
13104 server srv1 127.0.0.1:80
13105
13106 cache foobar
13107 total-max-size 4
13108 max-age 240
13109
13110
Willy Tarreau74ca5042013-06-11 23:12:07 +0200131117. Using ACLs and fetching samples
13112----------------------------------
13113
Davor Ocelice9ed2812017-12-25 17:49:28 +010013114HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020013115client or server information, from tables, environmental information etc...
13116The action of extracting such data is called fetching a sample. Once retrieved,
13117these samples may be used for various purposes such as a key to a stick-table,
13118but most common usages consist in matching them against predefined constant
13119data called patterns.
13120
13121
131227.1. ACL basics
13123---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013124
13125The use of Access Control Lists (ACL) provides a flexible solution to perform
13126content switching and generally to take decisions based on content extracted
13127from the request, the response or any environmental status. The principle is
13128simple :
13129
Willy Tarreau74ca5042013-06-11 23:12:07 +020013130 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013131 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020013132 - apply one or multiple pattern matching methods on this sample
13133 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013134
Willy Tarreau74ca5042013-06-11 23:12:07 +020013135The actions generally consist in blocking a request, selecting a backend, or
13136adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013137
13138In order to define a test, the "acl" keyword is used. The syntax is :
13139
Willy Tarreau74ca5042013-06-11 23:12:07 +020013140 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013141
13142This creates a new ACL <aclname> or completes an existing one with new tests.
13143Those tests apply to the portion of request/response specified in <criterion>
13144and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013145an operator which may be specified before the set of values. Optionally some
13146conversion operators may be applied to the sample, and they will be specified
13147as a comma-delimited list of keywords just after the first keyword. The values
13148are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013149
13150ACL names must be formed from upper and lower case letters, digits, '-' (dash),
13151'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
13152which means that "my_acl" and "My_Acl" are two different ACLs.
13153
13154There is no enforced limit to the number of ACLs. The unused ones do not affect
13155performance, they just consume a small amount of memory.
13156
Willy Tarreau74ca5042013-06-11 23:12:07 +020013157The criterion generally is the name of a sample fetch method, or one of its ACL
13158specific declinations. The default test method is implied by the output type of
13159this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013160methods of a same sample fetch method. The sample fetch methods are the only
13161ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013162
13163Sample fetch methods return data which can be of the following types :
13164 - boolean
13165 - integer (signed or unsigned)
13166 - IPv4 or IPv6 address
13167 - string
13168 - data block
13169
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013170Converters transform any of these data into any of these. For example, some
13171converters might convert a string to a lower-case string while other ones
13172would turn a string to an IPv4 address, or apply a netmask to an IP address.
13173The resulting sample is of the type of the last converter applied to the list,
13174which defaults to the type of the sample fetch method.
13175
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013176Each sample or converter returns data of a specific type, specified with its
13177keyword in this documentation. When an ACL is declared using a standard sample
13178fetch method, certain types automatically involved a default matching method
13179which are summarized in the table below :
13180
13181 +---------------------+-----------------+
13182 | Sample or converter | Default |
13183 | output type | matching method |
13184 +---------------------+-----------------+
13185 | boolean | bool |
13186 +---------------------+-----------------+
13187 | integer | int |
13188 +---------------------+-----------------+
13189 | ip | ip |
13190 +---------------------+-----------------+
13191 | string | str |
13192 +---------------------+-----------------+
13193 | binary | none, use "-m" |
13194 +---------------------+-----------------+
13195
13196Note that in order to match a binary samples, it is mandatory to specify a
13197matching method, see below.
13198
Willy Tarreau74ca5042013-06-11 23:12:07 +020013199The ACL engine can match these types against patterns of the following types :
13200 - boolean
13201 - integer or integer range
13202 - IP address / network
13203 - string (exact, substring, suffix, prefix, subdir, domain)
13204 - regular expression
13205 - hex block
13206
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013207The following ACL flags are currently supported :
13208
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013209 -i : ignore case during matching of all subsequent patterns.
13210 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013211 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013212 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013213 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013214 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020013215 -- : force end of flags. Useful when a string looks like one of the flags.
13216
Willy Tarreau74ca5042013-06-11 23:12:07 +020013217The "-f" flag is followed by the name of a file from which all lines will be
13218read as individual values. It is even possible to pass multiple "-f" arguments
13219if the patterns are to be loaded from multiple files. Empty lines as well as
13220lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
13221will be stripped. If it is absolutely necessary to insert a valid pattern
13222beginning with a sharp, just prefix it with a space so that it is not taken for
13223a comment. Depending on the data type and match method, haproxy may load the
13224lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
13225exact string matching. In this case, duplicates will automatically be removed.
13226
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013227The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
13228parsed as two column file. The first column contains the patterns used by the
13229ACL, and the second column contain the samples. The sample can be used later by
13230a map. This can be useful in some rare cases where an ACL would just be used to
13231check for the existence of a pattern in a map before a mapping is applied.
13232
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013233The "-u" flag forces the unique id of the ACL. This unique id is used with the
13234socket interface to identify ACL and dynamically change its values. Note that a
13235file is always identified by its name even if an id is set.
13236
Willy Tarreau74ca5042013-06-11 23:12:07 +020013237Also, note that the "-i" flag applies to subsequent entries and not to entries
13238loaded from files preceding it. For instance :
13239
13240 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
13241
13242In this example, each line of "exact-ua.lst" will be exactly matched against
13243the "user-agent" header of the request. Then each line of "generic-ua" will be
13244case-insensitively matched. Then the word "test" will be insensitively matched
13245as well.
13246
13247The "-m" flag is used to select a specific pattern matching method on the input
13248sample. All ACL-specific criteria imply a pattern matching method and generally
13249do not need this flag. However, this flag is useful with generic sample fetch
13250methods to describe how they're going to be matched against the patterns. This
13251is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010013252obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013253followed by a pattern matching method name, this method is used instead of the
13254default one for the criterion. This makes it possible to match contents in ways
13255that were not initially planned, or with sample fetch methods which return a
13256string. The matching method also affects the way the patterns are parsed.
13257
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013258The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
13259By default, if the parser cannot parse ip address it considers that the parsed
13260string is maybe a domain name and try dns resolution. The flag "-n" disable this
13261resolution. It is useful for detecting malformed ip lists. Note that if the DNS
13262server is not reachable, the haproxy configuration parsing may last many minutes
John Roeslerfb2fce12019-07-10 15:45:51 -050013263waiting for the timeout. During this time no error messages are displayed. The
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013264flag "-n" disable this behavior. Note also that during the runtime, this
13265function is disabled for the dynamic acl modifications.
13266
Willy Tarreau74ca5042013-06-11 23:12:07 +020013267There are some restrictions however. Not all methods can be used with all
13268sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
13269be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020013270
13271 - "found" : only check if the requested sample could be found in the stream,
13272 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020013273 to pass any pattern to avoid confusion. This matching method is
13274 particularly useful to detect presence of certain contents such
13275 as headers, cookies, etc... even if they are empty and without
13276 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013277
13278 - "bool" : check the value as a boolean. It can only be applied to fetches
13279 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013280 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013281
13282 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013283 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013284
13285 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020013286 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013287
Davor Ocelice9ed2812017-12-25 17:49:28 +010013288 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020013289 binary sequence. This may be used with binary or string samples.
13290
13291 - "len" : match the sample's length as an integer. This may be used with
13292 binary or string samples.
13293
Willy Tarreau74ca5042013-06-11 23:12:07 +020013294 - "str" : exact match : match the contents against a string. This may be
13295 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013296
Willy Tarreau74ca5042013-06-11 23:12:07 +020013297 - "sub" : substring match : check that the contents contain at least one of
13298 the provided string patterns. This may be used with binary or
13299 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013300
Willy Tarreau74ca5042013-06-11 23:12:07 +020013301 - "reg" : regex match : match the contents against a list of regular
13302 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013303
Willy Tarreau74ca5042013-06-11 23:12:07 +020013304 - "beg" : prefix match : check that the contents begin like the provided
13305 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013306
Willy Tarreau74ca5042013-06-11 23:12:07 +020013307 - "end" : suffix match : check that the contents end like the provided
13308 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013309
Willy Tarreau74ca5042013-06-11 23:12:07 +020013310 - "dir" : subdir match : check that a slash-delimited portion of the
13311 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013312 This may be used with binary or string samples.
13313
Willy Tarreau74ca5042013-06-11 23:12:07 +020013314 - "dom" : domain match : check that a dot-delimited portion of the contents
13315 exactly match one of the provided string patterns. This may be
13316 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013317
13318For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
13319request, it is possible to do :
13320
13321 acl jsess_present cook(JSESSIONID) -m found
13322
13323In order to apply a regular expression on the 500 first bytes of data in the
13324buffer, one would use the following acl :
13325
13326 acl script_tag payload(0,500) -m reg -i <script>
13327
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013328On systems where the regex library is much slower when using "-i", it is
13329possible to convert the sample to lowercase before matching, like this :
13330
13331 acl script_tag payload(0,500),lower -m reg <script>
13332
Willy Tarreau74ca5042013-06-11 23:12:07 +020013333All ACL-specific criteria imply a default matching method. Most often, these
13334criteria are composed by concatenating the name of the original sample fetch
13335method and the matching method. For example, "hdr_beg" applies the "beg" match
13336to samples retrieved using the "hdr" fetch method. Since all ACL-specific
13337criteria rely on a sample fetch method, it is always possible instead to use
13338the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013339
Willy Tarreau74ca5042013-06-11 23:12:07 +020013340If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013341the matching method is simply applied to the underlying sample fetch method.
13342For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013343
Willy Tarreau74ca5042013-06-11 23:12:07 +020013344 acl short_form hdr_beg(host) www.
13345 acl alternate1 hdr_beg(host) -m beg www.
13346 acl alternate2 hdr_dom(host) -m beg www.
13347 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013348
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013349
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013350The table below summarizes the compatibility matrix between sample or converter
13351types and the pattern types to fetch against. It indicates for each compatible
13352combination the name of the matching method to be used, surrounded with angle
13353brackets ">" and "<" when the method is the default one and will work by
13354default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013355
Willy Tarreau74ca5042013-06-11 23:12:07 +020013356 +-------------------------------------------------+
13357 | Input sample type |
13358 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013359 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013360 +----------------------+---------+---------+---------+---------+---------+
13361 | none (presence only) | found | found | found | found | found |
13362 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013363 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013364 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013365 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013366 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013367 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013368 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013369 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013370 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013371 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013372 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013373 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013374 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013375 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013376 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013377 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013378 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013379 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013380 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013381 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013382 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013383 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013384 +----------------------+---------+---------+---------+---------+---------+
13385 | hex block | | | | bin | bin |
13386 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020013387
13388
Willy Tarreau74ca5042013-06-11 23:12:07 +0200133897.1.1. Matching booleans
13390------------------------
13391
13392In order to match a boolean, no value is needed and all values are ignored.
13393Boolean matching is used by default for all fetch methods of type "boolean".
13394When boolean matching is used, the fetched value is returned as-is, which means
13395that a boolean "true" will always match and a boolean "false" will never match.
13396
13397Boolean matching may also be enforced using "-m bool" on fetch methods which
13398return an integer value. Then, integer value 0 is converted to the boolean
13399"false" and all other values are converted to "true".
13400
Willy Tarreau6a06a402007-07-15 20:15:28 +020013401
Willy Tarreau74ca5042013-06-11 23:12:07 +0200134027.1.2. Matching integers
13403------------------------
13404
13405Integer matching applies by default to integer fetch methods. It can also be
13406enforced on boolean fetches using "-m int". In this case, "false" is converted
13407to the integer 0, and "true" is converted to the integer 1.
13408
13409Integer matching also supports integer ranges and operators. Note that integer
13410matching only applies to positive values. A range is a value expressed with a
13411lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013412
13413For instance, "1024:65535" is a valid range to represent a range of
13414unprivileged ports, and "1024:" would also work. "0:1023" is a valid
13415representation of privileged ports, and ":1023" would also work.
13416
Willy Tarreau62644772008-07-16 18:36:06 +020013417As a special case, some ACL functions support decimal numbers which are in fact
13418two integers separated by a dot. This is used with some version checks for
13419instance. All integer properties apply to those decimal numbers, including
13420ranges and operators.
13421
Willy Tarreau6a06a402007-07-15 20:15:28 +020013422For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010013423operators with ranges does not make much sense and is strongly discouraged.
13424Similarly, it does not make much sense to perform order comparisons with a set
13425of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013426
Willy Tarreau0ba27502007-12-24 16:55:16 +010013427Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013428
13429 eq : true if the tested value equals at least one value
13430 ge : true if the tested value is greater than or equal to at least one value
13431 gt : true if the tested value is greater than at least one value
13432 le : true if the tested value is less than or equal to at least one value
13433 lt : true if the tested value is less than at least one value
13434
Willy Tarreau0ba27502007-12-24 16:55:16 +010013435For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013436
13437 acl negative-length hdr_val(content-length) lt 0
13438
Willy Tarreau62644772008-07-16 18:36:06 +020013439This one matches SSL versions between 3.0 and 3.1 (inclusive) :
13440
13441 acl sslv3 req_ssl_ver 3:3.1
13442
Willy Tarreau6a06a402007-07-15 20:15:28 +020013443
Willy Tarreau74ca5042013-06-11 23:12:07 +0200134447.1.3. Matching strings
13445-----------------------
13446
13447String matching applies to string or binary fetch methods, and exists in 6
13448different forms :
13449
13450 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013451 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013452
13453 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013454 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013455
13456 - prefix match (-m beg) : the patterns are compared with the beginning of
13457 the extracted string, and the ACL matches if any of them matches.
13458
13459 - suffix match (-m end) : the patterns are compared with the end of the
13460 extracted string, and the ACL matches if any of them matches.
13461
Baptiste Assmann33db6002016-03-06 23:32:10 +010013462 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020013463 string, delimited with slashes ("/"), and the ACL matches if any of them
13464 matches.
13465
13466 - domain match (-m dom) : the patterns are looked up inside the extracted
13467 string, delimited with dots ("."), and the ACL matches if any of them
13468 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013469
13470String matching applies to verbatim strings as they are passed, with the
13471exception of the backslash ("\") which makes it possible to escape some
13472characters such as the space. If the "-i" flag is passed before the first
13473string, then the matching will be performed ignoring the case. In order
13474to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010013475before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013476
Mathias Weiersmuellercb250fc2019-12-02 09:43:40 +010013477Do not use string matches for binary fetches which might contain null bytes
13478(0x00), as the comparison stops at the occurrence of the first null byte.
13479Instead, convert the binary fetch to a hex string with the hex converter first.
13480
13481Example:
13482 # matches if the string <tag> is present in the binary sample
13483 acl tag_found req.payload(0,0),hex -m sub 3C7461673E
13484
Willy Tarreau6a06a402007-07-15 20:15:28 +020013485
Willy Tarreau74ca5042013-06-11 23:12:07 +0200134867.1.4. Matching regular expressions (regexes)
13487---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013488
13489Just like with string matching, regex matching applies to verbatim strings as
13490they are passed, with the exception of the backslash ("\") which makes it
13491possible to escape some characters such as the space. If the "-i" flag is
13492passed before the first regex, then the matching will be performed ignoring
13493the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010013494the "--" flag before the first string. Same principle applies of course to
13495match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013496
13497
Willy Tarreau74ca5042013-06-11 23:12:07 +0200134987.1.5. Matching arbitrary data blocks
13499-------------------------------------
13500
13501It is possible to match some extracted samples against a binary block which may
13502not safely be represented as a string. For this, the patterns must be passed as
13503a series of hexadecimal digits in an even number, when the match method is set
13504to binary. Each sequence of two digits will represent a byte. The hexadecimal
13505digits may be used upper or lower case.
13506
13507Example :
13508 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
13509 acl hello payload(0,6) -m bin 48656c6c6f0a
13510
13511
135127.1.6. Matching IPv4 and IPv6 addresses
13513---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013514
13515IPv4 addresses values can be specified either as plain addresses or with a
13516netmask appended, in which case the IPv4 address matches whenever it is
13517within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010013518host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010013519difficult to read and debug configurations. If hostnames are used, you should
13520at least ensure that they are present in /etc/hosts so that the configuration
13521does not depend on any random DNS match at the moment the configuration is
13522parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013523
Daniel Schnellereba56342016-04-13 00:26:52 +020013524The dotted IPv4 address notation is supported in both regular as well as the
13525abbreviated form with all-0-octets omitted:
13526
13527 +------------------+------------------+------------------+
13528 | Example 1 | Example 2 | Example 3 |
13529 +------------------+------------------+------------------+
13530 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
13531 | 192.168.1 | 10.12 | 127.1 |
13532 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
13533 | 192.168.1/22 | 10.12/8 | 127.1/8 |
13534 +------------------+------------------+------------------+
13535
13536Notice that this is different from RFC 4632 CIDR address notation in which
13537192.168.42/24 would be equivalent to 192.168.42.0/24.
13538
Willy Tarreauceb4ac92012-04-28 00:41:46 +020013539IPv6 may be entered in their usual form, with or without a netmask appended.
13540Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
13541trouble with randomly resolved IP addresses, host names are never allowed in
13542IPv6 patterns.
13543
13544HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
13545following situations :
13546 - tested address is IPv4, pattern address is IPv4, the match applies
13547 in IPv4 using the supplied mask if any.
13548 - tested address is IPv6, pattern address is IPv6, the match applies
13549 in IPv6 using the supplied mask if any.
13550 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
13551 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
13552 ::IPV4 or ::ffff:IPV4, otherwise it fails.
13553 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
13554 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
13555 applied in IPv6 using the supplied IPv6 mask.
13556
Willy Tarreau74ca5042013-06-11 23:12:07 +020013557
135587.2. Using ACLs to form conditions
13559----------------------------------
13560
13561Some actions are only performed upon a valid condition. A condition is a
13562combination of ACLs with operators. 3 operators are supported :
13563
13564 - AND (implicit)
13565 - OR (explicit with the "or" keyword or the "||" operator)
13566 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020013567
Willy Tarreau74ca5042013-06-11 23:12:07 +020013568A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020013569
Willy Tarreau74ca5042013-06-11 23:12:07 +020013570 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020013571
Willy Tarreau74ca5042013-06-11 23:12:07 +020013572Such conditions are generally used after an "if" or "unless" statement,
13573indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020013574
Willy Tarreau74ca5042013-06-11 23:12:07 +020013575For instance, to block HTTP requests to the "*" URL with methods other than
13576"OPTIONS", as well as POST requests without content-length, and GET or HEAD
13577requests with a content-length greater than 0, and finally every request which
13578is not either GET/HEAD/POST/OPTIONS !
13579
13580 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013581 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
13582 http-request deny if METH_GET HTTP_CONTENT
13583 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020013584
13585To select a different backend for requests to static contents on the "www" site
13586and to every request on the "img", "video", "download" and "ftp" hosts :
13587
13588 acl url_static path_beg /static /images /img /css
13589 acl url_static path_end .gif .png .jpg .css .js
13590 acl host_www hdr_beg(host) -i www
13591 acl host_static hdr_beg(host) -i img. video. download. ftp.
13592
Davor Ocelice9ed2812017-12-25 17:49:28 +010013593 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020013594 # of host "www". Use backend "www" for the rest.
13595 use_backend static if host_static or host_www url_static
13596 use_backend www if host_www
13597
13598It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
13599expressions that are built on the fly without needing to be declared. They must
13600be enclosed between braces, with a space before and after each brace (because
13601the braces must be seen as independent words). Example :
13602
13603 The following rule :
13604
13605 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013606 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020013607
13608 Can also be written that way :
13609
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013610 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020013611
13612It is generally not recommended to use this construct because it's a lot easier
13613to leave errors in the configuration when written that way. However, for very
13614simple rules matching only one source IP address for instance, it can make more
13615sense to use them than to declare ACLs with random names. Another example of
13616good use is the following :
13617
13618 With named ACLs :
13619
13620 acl site_dead nbsrv(dynamic) lt 2
13621 acl site_dead nbsrv(static) lt 2
13622 monitor fail if site_dead
13623
13624 With anonymous ACLs :
13625
13626 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
13627
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013628See section 4.2 for detailed help on the "http-request deny" and "use_backend"
13629keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013630
13631
136327.3. Fetching samples
13633---------------------
13634
13635Historically, sample fetch methods were only used to retrieve data to match
13636against patterns using ACLs. With the arrival of stick-tables, a new class of
13637sample fetch methods was created, most often sharing the same syntax as their
13638ACL counterpart. These sample fetch methods are also known as "fetches". As
13639of now, ACLs and fetches have converged. All ACL fetch methods have been made
13640available as fetch methods, and ACLs may use any sample fetch method as well.
13641
13642This section details all available sample fetch methods and their output type.
13643Some sample fetch methods have deprecated aliases that are used to maintain
13644compatibility with existing configurations. They are then explicitly marked as
13645deprecated and should not be used in new setups.
13646
13647The ACL derivatives are also indicated when available, with their respective
13648matching methods. These ones all have a well defined default pattern matching
13649method, so it is never necessary (though allowed) to pass the "-m" option to
13650indicate how the sample will be matched using ACLs.
13651
13652As indicated in the sample type versus matching compatibility matrix above,
13653when using a generic sample fetch method in an ACL, the "-m" option is
13654mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
13655the same keyword exists as an ACL keyword and as a standard fetch method, the
13656ACL engine will automatically pick the ACL-only one by default.
13657
13658Some of these keywords support one or multiple mandatory arguments, and one or
13659multiple optional arguments. These arguments are strongly typed and are checked
13660when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010013661incorrect argument (e.g. an unresolved backend name). Fetch function arguments
13662are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020013663is optional, it will be indicated below between square brackets ('[ ]'). When
13664all arguments are optional, the parenthesis may be omitted.
13665
13666Thus, the syntax of a standard sample fetch method is one of the following :
13667 - name
13668 - name(arg1)
13669 - name(arg1,arg2)
13670
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013671
136727.3.1. Converters
13673-----------------
13674
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013675Sample fetch methods may be combined with transformations to be applied on top
13676of the fetched sample (also called "converters"). These combinations form what
13677is called "sample expressions" and the result is a "sample". Initially this
13678was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010013679has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013680unique-id-format, add-header, ...).
13681
13682These transformations are enumerated as a series of specific keywords after the
13683sample fetch method. These keywords may equally be appended immediately after
13684the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010013685support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013686
Willy Tarreau97707872015-01-27 15:12:13 +010013687A certain category of converters are bitwise and arithmetic operators which
13688support performing basic operations on integers. Some bitwise operations are
13689supported (and, or, xor, cpl) and some arithmetic operations are supported
13690(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
13691bool) which make it possible to report a match without having to write an ACL.
13692
Willy Tarreau74ca5042013-06-11 23:12:07 +020013693The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010013694
Ben Shillitof25e8e52016-12-02 14:25:37 +00001369551d.single(<prop>[,<prop>*])
13696 Returns values for the properties requested as a string, where values are
13697 separated by the delimiter specified with "51degrees-property-separator".
13698 The device is identified using the User-Agent header passed to the
13699 converter. The function can be passed up to five property names, and if a
13700 property name can't be found, the value "NoData" is returned.
13701
13702 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013703 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
13704 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000013705 # User-Agent passed to the converter.
13706 frontend http-in
13707 bind *:8081
13708 default_backend servers
13709 http-request set-header X-51D-DeviceTypeMobileTablet \
13710 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
13711
Willy Tarreau97707872015-01-27 15:12:13 +010013712add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013713 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013714 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013715 name. The name of the variable starts with an indication about its scope. The
13716 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013717 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013718 "sess" : the variable is shared with the whole session
13719 "txn" : the variable is shared with the transaction (request and response)
13720 "req" : the variable is shared only during request processing
13721 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013722 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013723 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013724
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010013725aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
13726 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
13727 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
13728 need to be base64 encoded and the returned result is in raw byte format.
13729 If the <aead_tag> validation fails, the converter doesn't return any data.
13730 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
13731 converter requires at least OpenSSL 1.0.1.
13732
13733 Example:
13734 http-response set-header X-Decrypted-Text %[var(txn.enc),\
13735 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
13736
Willy Tarreau97707872015-01-27 15:12:13 +010013737and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013738 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013739 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013740 numeric value or a variable name. The name of the variable starts with an
13741 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013742 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013743 "sess" : the variable is shared with the whole session
13744 "txn" : the variable is shared with the transaction (request and response)
13745 "req" : the variable is shared only during request processing
13746 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013747 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013748 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013749
Holger Just1bfc24b2017-05-06 00:56:53 +020013750b64dec
13751 Converts (decodes) a base64 encoded input string to its binary
13752 representation. It performs the inverse operation of base64().
13753
Emeric Brun53d1a982014-04-30 18:21:37 +020013754base64
13755 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010013756 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020013757 an SSL ID can be copied in a header).
13758
Willy Tarreau97707872015-01-27 15:12:13 +010013759bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013760 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013761 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013762 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013763 presence of a flag).
13764
Emeric Brun54c4ac82014-11-03 15:32:43 +010013765bytes(<offset>[,<length>])
13766 Extracts some bytes from an input binary sample. The result is a binary
13767 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010013768 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010013769
Willy Tarreau280f42b2018-02-19 15:34:12 +010013770concat([<start>],[<var>],[<end>])
13771 Concatenates up to 3 fields after the current sample which is then turned to
13772 a string. The first one, <start>, is a constant string, that will be appended
13773 immediately after the existing sample. It may be omitted if not used. The
13774 second one, <var>, is a variable name. The variable will be looked up, its
13775 contents converted to a string, and it will be appended immediately after the
13776 <first> part. If the variable is not found, nothing is appended. It may be
13777 omitted as well. The third field, <end> is a constant string that will be
13778 appended after the variable. It may also be omitted. Together, these elements
13779 allow to concatenate variables with delimiters to an existing set of
13780 variables. This can be used to build new variables made of a succession of
Willy Tarreauef21fac2020-02-14 13:37:20 +010013781 other variables, such as colon-delimited values. If commas or closing
13782 parethesis are needed as delimiters, they must be protected by quotes or
13783 backslashes, themselves protected so that they are not stripped by the first
13784 level parser. See examples below.
Willy Tarreau280f42b2018-02-19 15:34:12 +010013785
13786 Example:
13787 tcp-request session set-var(sess.src) src
13788 tcp-request session set-var(sess.dn) ssl_c_s_dn
13789 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
Willy Tarreauef21fac2020-02-14 13:37:20 +010013790 tcp-request session set-var(txn.ipport) "str(),concat('addr=(',sess.ip),concat(',',sess.port,')')"
Willy Tarreau280f42b2018-02-19 15:34:12 +010013791 http-request set-header x-hap-sig %[var(txn.sig)]
13792
Willy Tarreau97707872015-01-27 15:12:13 +010013793cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013794 Takes the input value of type signed integer, applies a ones-complement
13795 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010013796
Willy Tarreau80599772015-01-20 19:35:24 +010013797crc32([<avalanche>])
13798 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
13799 hash function. Optionally, it is possible to apply a full avalanche hash
13800 function to the output if the optional <avalanche> argument equals 1. This
13801 converter uses the same functions as used by the various hash-based load
13802 balancing algorithms, so it will provide exactly the same results. It is
13803 provided for compatibility with other software which want a CRC32 to be
13804 computed on some input keys, so it follows the most common implementation as
13805 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
13806 but may provide a better or at least less predictable distribution. It must
13807 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013808 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
13809
13810crc32c([<avalanche>])
13811 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
13812 hash function. Optionally, it is possible to apply a full avalanche hash
13813 function to the output if the optional <avalanche> argument equals 1. This
13814 converter uses the same functions as described in RFC4960, Appendix B [8].
13815 It is provided for compatibility with other software which want a CRC32C to be
13816 computed on some input keys. It is slower than the other algorithms and it must
13817 not be used for security purposes as a 32-bit hash is trivial to break. See
13818 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010013819
David Carlier29b3ca32015-09-25 14:09:21 +010013820da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020013821 Asks the DeviceAtlas converter to identify the User Agent string passed on
13822 input, and to emit a string made of the concatenation of the properties
13823 enumerated in argument, delimited by the separator defined by the global
13824 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000013825 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020013826 configuration language.
13827
13828 Example:
13829 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013830 bind *:8881
13831 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000013832 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 +020013833
Willy Tarreau0851fd52019-12-17 10:07:25 +010013834debug([<prefix][,<destination>])
13835 This converter is used as debug tool. It takes a capture of the input sample
13836 and sends it to event sink <destination>, which may designate a ring buffer
13837 such as "buf0", as well as "stdout", or "stderr". Available sinks may be
13838 checked at run time by issuing "show events" on the CLI. When not specified,
13839 the output will be "buf0", which may be consulted via the CLI's "show events"
13840 command. An optional prefix <prefix> may be passed to help distinguish
13841 outputs from multiple expressions. It will then appear before the colon in
13842 the output message. The input sample is passed as-is on the output, so that
13843 it is safe to insert the debug converter anywhere in a chain, even with non-
13844 printable sample types.
13845
13846 Example:
13847 tcp-request connection track-sc0 src,debug(track-sc)
Thierry FOURNIER9687c772015-05-07 15:46:29 +020013848
Willy Tarreau97707872015-01-27 15:12:13 +010013849div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013850 Divides the input value of type signed integer by <value>, and returns the
13851 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013852 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010013853 variable name. The name of the variable starts with an indication about its
13854 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013855 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013856 "sess" : the variable is shared with the whole session
13857 "txn" : the variable is shared with the transaction (request and response)
13858 "req" : the variable is shared only during request processing
13859 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013860 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013861 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013862
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013863djb2([<avalanche>])
13864 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
13865 hash function. Optionally, it is possible to apply a full avalanche hash
13866 function to the output if the optional <avalanche> argument equals 1. This
13867 converter uses the same functions as used by the various hash-based load
13868 balancing algorithms, so it will provide exactly the same results. It is
13869 mostly intended for debugging, but can be used as a stick-table entry to
13870 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013871 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
13872 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013873
Willy Tarreau97707872015-01-27 15:12:13 +010013874even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013875 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010013876 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
13877
Marcin Deranek9631a282018-04-16 14:30:46 +020013878field(<index>,<delimiters>[,<count>])
13879 Extracts the substring at the given index counting from the beginning
13880 (positive index) or from the end (negative index) considering given delimiters
13881 from an input string. Indexes start at 1 or -1 and delimiters are a string
13882 formatted list of chars. Optionally you can specify <count> of fields to
13883 extract (default: 1). Value of 0 indicates extraction of all remaining
13884 fields.
13885
13886 Example :
13887 str(f1_f2_f3__f5),field(5,_) # f5
13888 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
13889 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
13890 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
13891 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010013892
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013893hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010013894 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013895 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010013896 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 +020013897 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010013898
Dragan Dosen3f957b22017-10-24 09:27:34 +020013899hex2i
13900 Converts a hex string containing two hex digits per input byte to an
John Roeslerfb2fce12019-07-10 15:45:51 -050013901 integer. If the input value cannot be converted, then zero is returned.
Dragan Dosen3f957b22017-10-24 09:27:34 +020013902
Cyril Bonté6bcd1822019-11-05 23:13:59 +010013903http_date([<offset],[<unit>])
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013904 Converts an integer supposed to contain a date since epoch to a string
13905 representing this date in a format suitable for use in HTTP header fields. If
Damien Claisseae6f1252019-10-30 15:57:28 +000013906 an offset value is specified, then it is added to the date before the
13907 conversion is operated. This is particularly useful to emit Date header fields,
13908 Expires values in responses when combined with a positive offset, or
13909 Last-Modified values when the offset is negative.
13910 If a unit value is specified, then consider the timestamp as either
13911 "s" for seconds (default behavior), "ms" for milliseconds, or "us" for
13912 microseconds since epoch. Offset is assumed to have the same unit as
13913 input timestamp.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013914
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013915in_table(<table>)
13916 Uses the string representation of the input sample to perform a look up in
13917 the specified table. If the key is not found in the table, a boolean false
13918 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010013919 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013920 or not a source IP address or an Authorization header was already seen).
13921
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013922ipmask(<mask4>, [<mask6>])
13923 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013924 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013925 table entries and as such use the same server. The mask4 can be passed in
13926 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
13927 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
13928 If no mask6 is given IPv6 addresses will fail to convert for backwards
13929 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013930
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013931json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010013932 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013933 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013934 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013935 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
13936 of errors:
13937 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
13938 bytes, ...)
13939 - invalid range (the decoded value is within a UTF-8 prohibited range),
13940 - code overlong (the value is encoded with more bytes than necessary).
13941
13942 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
13943 character is greater than 0xffff because the JSON string escape specification
13944 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
13945 in 4 variants designated by a combination of two suffix letters : "p" for
13946 "permissive" and "s" for "silently ignore". The behaviors of the decoders
13947 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013948 - "ascii" : never fails;
13949 - "utf8" : fails on any detected errors;
13950 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013951 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010013952 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013953 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
13954 characters corresponding to the other errors.
13955
13956 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010013957 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013958
13959 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013960 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013961 capture request header user-agent len 150
13962 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013963
13964 Input request from client 127.0.0.1:
13965 GET / HTTP/1.0
13966 User-Agent: Very "Ugly" UA 1/2
13967
13968 Output log:
13969 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
13970
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013971language(<value>[,<default>])
13972 Returns the value with the highest q-factor from a list as extracted from the
13973 "accept-language" header using "req.fhdr". Values with no q-factor have a
13974 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
13975 belong to the list of semi-colon delimited <values> will be considered. The
13976 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
13977 given list and a default value is provided, it is returned. Note that language
13978 names may have a variant after a dash ('-'). If this variant is present in the
13979 list, it will be matched, but if it is not, only the base language is checked.
13980 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010013981 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013982 ordering of the values in the request counts, as the first value among
13983 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020013984
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013985 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013986
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013987 # this configuration switches to the backend matching a
13988 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013989
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013990 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
13991 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
13992 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
13993 use_backend spanish if es
13994 use_backend french if fr
13995 use_backend english if en
13996 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020013997
Willy Tarreau60a2ee72017-12-15 07:13:48 +010013998length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010013999 Get the length of the string. This can only be placed after a string
14000 sample fetch function or after a transformation keyword returning a string
14001 type. The result is of type integer.
14002
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014003lower
14004 Convert a string sample to lower case. This can only be placed after a string
14005 sample fetch function or after a transformation keyword returning a string
14006 type. The result is of type string.
14007
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014008ltime(<format>[,<offset>])
14009 Converts an integer supposed to contain a date since epoch to a string
14010 representing this date in local time using a format defined by the <format>
14011 string using strftime(3). The purpose is to allow any date format to be used
14012 in logs. An optional <offset> in seconds may be applied to the input date
14013 (positive or negative). See the strftime() man page for the format supported
14014 by your operating system. See also the utime converter.
14015
14016 Example :
14017
14018 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014019 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014020 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
14021
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014022map(<map_file>[,<default_value>])
14023map_<match_type>(<map_file>[,<default_value>])
14024map_<match_type>_<output_type>(<map_file>[,<default_value>])
14025 Search the input value from <map_file> using the <match_type> matching method,
14026 and return the associated value converted to the type <output_type>. If the
14027 input value cannot be found in the <map_file>, the converter returns the
14028 <default_value>. If the <default_value> is not set, the converter fails and
14029 acts as if no input value could be fetched. If the <match_type> is not set, it
14030 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
14031 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
14032 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014033
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014034 It is important to avoid overlapping between the keys : IP addresses and
14035 strings are stored in trees, so the first of the finest match will be used.
14036 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014037
Tim Düsterhus4896c442016-11-29 02:15:19 +010014038 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014039 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014040
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014041 input type | match method | output type str | output type int | output type ip
14042 -----------+--------------+-----------------+-----------------+---------------
14043 str | str | map_str | map_str_int | map_str_ip
14044 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020014045 str | beg | map_beg | map_beg_int | map_end_ip
14046 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014047 str | sub | map_sub | map_sub_int | map_sub_ip
14048 -----------+--------------+-----------------+-----------------+---------------
14049 str | dir | map_dir | map_dir_int | map_dir_ip
14050 -----------+--------------+-----------------+-----------------+---------------
14051 str | dom | map_dom | map_dom_int | map_dom_ip
14052 -----------+--------------+-----------------+-----------------+---------------
14053 str | end | map_end | map_end_int | map_end_ip
14054 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080014055 str | reg | map_reg | map_reg_int | map_reg_ip
14056 -----------+--------------+-----------------+-----------------+---------------
14057 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014058 -----------+--------------+-----------------+-----------------+---------------
14059 int | int | map_int | map_int_int | map_int_ip
14060 -----------+--------------+-----------------+-----------------+---------------
14061 ip | ip | map_ip | map_ip_int | map_ip_ip
14062 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014063
Thierry Fournier8feaa662016-02-10 22:55:20 +010014064 The special map called "map_regm" expect matching zone in the regular
14065 expression and modify the output replacing back reference (like "\1") by
14066 the corresponding match text.
14067
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014068 The file contains one key + value per line. Lines which start with '#' are
14069 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
14070 is then the first "word" (series of non-space/tabs characters), and the value
14071 is what follows this series of space/tab till the end of the line excluding
14072 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014073
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014074 Example :
14075
14076 # this is a comment and is ignored
14077 2.22.246.0/23 United Kingdom \n
14078 <-><-----------><--><------------><---->
14079 | | | | `- trailing spaces ignored
14080 | | | `---------- value
14081 | | `-------------------- middle spaces ignored
14082 | `---------------------------- key
14083 `------------------------------------ leading spaces ignored
14084
Willy Tarreau97707872015-01-27 15:12:13 +010014085mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014086 Divides the input value of type signed integer by <value>, and returns the
14087 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014088 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014089 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014090 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014091 "sess" : the variable is shared with the whole session
14092 "txn" : the variable is shared with the transaction (request and response)
14093 "req" : the variable is shared only during request processing
14094 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014095 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014096 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014097
14098mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014099 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020014100 the product as an signed integer. In case of overflow, the largest possible
14101 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014102 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014103 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014104 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014105 "sess" : the variable is shared with the whole session
14106 "txn" : the variable is shared with the transaction (request and response)
14107 "req" : the variable is shared only during request processing
14108 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014109 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014110 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014111
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010014112nbsrv
14113 Takes an input value of type string, interprets it as a backend name and
14114 returns the number of usable servers in that backend. Can be used in places
14115 where we want to look up a backend from a dynamic name, like a result of a
14116 map lookup.
14117
Willy Tarreau97707872015-01-27 15:12:13 +010014118neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014119 Takes the input value of type signed integer, computes the opposite value,
14120 and returns the remainder as an signed integer. 0 is identity. This operator
14121 is provided for reversed subtracts : in order to subtract the input from a
14122 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010014123
14124not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014125 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010014126 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010014127 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010014128 absence of a flag).
14129
14130odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014131 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010014132 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
14133
14134or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014135 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014136 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010014137 numeric value or a variable name. The name of the variable starts with an
14138 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014139 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014140 "sess" : the variable is shared with the whole session
14141 "txn" : the variable is shared with the transaction (request and response)
14142 "req" : the variable is shared only during request processing
14143 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014144 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014145 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014146
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014147protobuf(<field_number>,[<field_type>])
14148 This extracts the protocol buffers message field in raw mode of an input binary
14149 sample representation of a protocol buffer message with <field_number> as field
14150 number (dotted notation) if <field_type> is not present, or as an integer sample
14151 if this field is present (see also "ungrpc" below).
14152 The list of the authorized types is the following one: "int32", "int64", "uint32",
14153 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14154 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14155 "float" for the wire type 5. Note that "string" is considered as a length-delimited
14156 type, so it does not require any <field_type> argument to be extracted.
14157 More information may be found here about the protocol buffers message field types:
14158 https://developers.google.com/protocol-buffers/docs/encoding
14159
Willy Tarreauc4dc3502015-01-23 20:39:28 +010014160regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010014161 Applies a regex-based substitution to the input string. It does the same
14162 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
14163 default it will replace in the input string the first occurrence of the
14164 largest part matching the regular expression <regex> with the substitution
14165 string <subst>. It is possible to replace all occurrences instead by adding
14166 the flag "g" in the third argument <flags>. It is also possible to make the
14167 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
14168 string, it is made up from the concatenation of all desired flags. Thus if
14169 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
Willy Tarreauef21fac2020-02-14 13:37:20 +010014170 The first use of this converter is to replace certain characters or sequence
14171 of characters with other ones.
14172
14173 It is highly recommended to enclose the regex part using protected quotes to
14174 improve clarity and never have a closing parenthesis from the regex mixed up
14175 with the parenthesis from the function. Just like in Bourne shell, the first
14176 level of quotes is processed when delimiting word groups on the line, a
14177 second level is usable for argument. It is recommended to use single quotes
14178 outside since these ones do not try to resolve backslashes nor dollar signs.
Willy Tarreau7eda8492015-01-20 19:47:06 +010014179
Willy Tarreaucd0d2ed2020-02-14 17:33:06 +010014180 Examples:
Willy Tarreau7eda8492015-01-20 19:47:06 +010014181
14182 # de-duplicate "/" in header "x-path".
14183 # input: x-path: /////a///b/c/xzxyz/
14184 # output: x-path: /a/b/c/xzxyz/
Willy Tarreauef21fac2020-02-14 13:37:20 +010014185 http-request set-header x-path "%[hdr(x-path),regsub('/+','/','g')]"
Willy Tarreau7eda8492015-01-20 19:47:06 +010014186
Willy Tarreaucd0d2ed2020-02-14 17:33:06 +010014187 # copy query string to x-query and drop all leading '?', ';' and '&'
14188 http-request set-header x-query "%[query,regsub([?;&]*,'')]"
14189
Jerome Magnin07e1e3c2020-02-16 19:20:19 +010014190 # capture groups and backreferences
14191 # both lines do the same.
14192 http-request redirect location %[url,'regsub("(foo|bar)([0-9]+)?","\2\1",i)]'
14193 http-request redirect location %[url,regsub(\"(foo|bar)([0-9]+)?\",\"\2\1\",i)]
14194
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014195capture-req(<id>)
14196 Capture the string entry in the request slot <id> and returns the entry as
14197 is. If the slot doesn't exist, the capture fails silently.
14198
14199 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014200 "http-response capture", "capture.req.hdr" and
14201 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014202
14203capture-res(<id>)
14204 Capture the string entry in the response slot <id> and returns the entry as
14205 is. If the slot doesn't exist, the capture fails silently.
14206
14207 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014208 "http-response capture", "capture.req.hdr" and
14209 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014210
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014211sdbm([<avalanche>])
14212 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
14213 hash function. Optionally, it is possible to apply a full avalanche hash
14214 function to the output if the optional <avalanche> argument equals 1. This
14215 converter uses the same functions as used by the various hash-based load
14216 balancing algorithms, so it will provide exactly the same results. It is
14217 mostly intended for debugging, but can be used as a stick-table entry to
14218 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014219 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
14220 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014221
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014222set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010014223 Sets a variable with the input content and returns the content on the output
14224 as-is. The variable keeps the value and the associated input type. The name of
14225 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014226 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014227 "sess" : the variable is shared with the whole session
14228 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014229 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014230 "req" : the variable is shared only during request processing,
14231 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014232 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014233 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014234
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020014235sha1
Tim Duesterhusd4376302019-06-17 12:41:44 +020014236 Converts a binary input sample to a SHA-1 digest. The result is a binary
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020014237 sample with length of 20 bytes.
14238
Tim Duesterhusd4376302019-06-17 12:41:44 +020014239sha2([<bits>])
14240 Converts a binary input sample to a digest in the SHA-2 family. The result
14241 is a binary sample with length of <bits>/8 bytes.
14242
14243 Valid values for <bits> are 224, 256, 384, 512, each corresponding to
14244 SHA-<bits>. The default value is 256.
14245
14246 Please note that this converter is only available when haproxy has been
14247 compiled with USE_OPENSSL.
14248
Nenad Merdanovic177adc92019-08-27 01:58:13 +020014249srv_queue
14250 Takes an input value of type string, either a server name or <backend>/<server>
14251 format and returns the number of queued sessions on that server. Can be used
14252 in places where we want to look up queued sessions from a dynamic name, like a
14253 cookie value (e.g. req.cook(SRVID),srv_queue) and then make a decision to break
14254 persistence or direct a request elsewhere.
14255
Tim Duesterhusca097c12018-04-27 21:18:45 +020014256strcmp(<var>)
14257 Compares the contents of <var> with the input value of type string. Returns
14258 the result as a signed integer compatible with strcmp(3): 0 if both strings
14259 are identical. A value less than 0 if the left string is lexicographically
14260 smaller than the right string or if the left string is shorter. A value greater
14261 than 0 otherwise (right string greater than left string or the right string is
14262 shorter).
14263
14264 Example :
14265
14266 http-request set-var(txn.host) hdr(host)
14267 # Check whether the client is attempting domain fronting.
14268 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
14269
14270
Willy Tarreau97707872015-01-27 15:12:13 +010014271sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014272 Subtracts <value> from the input value of type signed integer, and returns
14273 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014274 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010014275 or a variable name. The name of the variable starts with an indication about
14276 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014277 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014278 "sess" : the variable is shared with the whole session
14279 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014280 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014281 "req" : the variable is shared only during request processing,
14282 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014283 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014284 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014285
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014286table_bytes_in_rate(<table>)
14287 Uses the string representation of the input sample to perform a look up in
14288 the specified table. If the key is not found in the table, integer value zero
14289 is returned. Otherwise the converter returns the average client-to-server
14290 bytes rate associated with the input sample in the designated table, measured
14291 in amount of bytes over the period configured in the table. See also the
14292 sc_bytes_in_rate sample fetch keyword.
14293
14294
14295table_bytes_out_rate(<table>)
14296 Uses the string representation of the input sample to perform a look up in
14297 the specified table. If the key is not found in the table, integer value zero
14298 is returned. Otherwise the converter returns the average server-to-client
14299 bytes rate associated with the input sample in the designated table, measured
14300 in amount of bytes over the period configured in the table. See also the
14301 sc_bytes_out_rate sample fetch keyword.
14302
14303table_conn_cnt(<table>)
14304 Uses the string representation of the input sample to perform a look up in
14305 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014306 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014307 connections associated with the input sample in the designated table. See
14308 also the sc_conn_cnt sample fetch keyword.
14309
14310table_conn_cur(<table>)
14311 Uses the string representation of the input sample to perform a look up in
14312 the specified table. If the key is not found in the table, integer value zero
14313 is returned. Otherwise the converter returns the current amount of concurrent
14314 tracked connections associated with the input sample in the designated table.
14315 See also the sc_conn_cur sample fetch keyword.
14316
14317table_conn_rate(<table>)
14318 Uses the string representation of the input sample to perform a look up in
14319 the specified table. If the key is not found in the table, integer value zero
14320 is returned. Otherwise the converter returns the average incoming connection
14321 rate associated with the input sample in the designated table. See also the
14322 sc_conn_rate sample fetch keyword.
14323
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014324table_gpt0(<table>)
14325 Uses the string representation of the input sample to perform a look up in
14326 the specified table. If the key is not found in the table, boolean value zero
14327 is returned. Otherwise the converter returns the current value of the first
14328 general purpose tag associated with the input sample in the designated table.
14329 See also the sc_get_gpt0 sample fetch keyword.
14330
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014331table_gpc0(<table>)
14332 Uses the string representation of the input sample to perform a look up in
14333 the specified table. If the key is not found in the table, integer value zero
14334 is returned. Otherwise the converter returns the current value of the first
14335 general purpose counter associated with the input sample in the designated
14336 table. See also the sc_get_gpc0 sample fetch keyword.
14337
14338table_gpc0_rate(<table>)
14339 Uses the string representation of the input sample to perform a look up in
14340 the specified table. If the key is not found in the table, integer value zero
14341 is returned. Otherwise the converter returns the frequency which the gpc0
14342 counter was incremented over the configured period in the table, associated
14343 with the input sample in the designated table. See also the sc_get_gpc0_rate
14344 sample fetch keyword.
14345
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014346table_gpc1(<table>)
14347 Uses the string representation of the input sample to perform a look up in
14348 the specified table. If the key is not found in the table, integer value zero
14349 is returned. Otherwise the converter returns the current value of the second
14350 general purpose counter associated with the input sample in the designated
14351 table. See also the sc_get_gpc1 sample fetch keyword.
14352
14353table_gpc1_rate(<table>)
14354 Uses the string representation of the input sample to perform a look up in
14355 the specified table. If the key is not found in the table, integer value zero
14356 is returned. Otherwise the converter returns the frequency which the gpc1
14357 counter was incremented over the configured period in the table, associated
14358 with the input sample in the designated table. See also the sc_get_gpc1_rate
14359 sample fetch keyword.
14360
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014361table_http_err_cnt(<table>)
14362 Uses the string representation of the input sample to perform a look up in
14363 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014364 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014365 errors associated with the input sample in the designated table. See also the
14366 sc_http_err_cnt sample fetch keyword.
14367
14368table_http_err_rate(<table>)
14369 Uses the string representation of the input sample to perform a look up in
14370 the specified table. If the key is not found in the table, integer value zero
14371 is returned. Otherwise the average rate of HTTP errors associated with the
14372 input sample in the designated table, measured in amount of errors over the
14373 period configured in the table. See also the sc_http_err_rate sample fetch
14374 keyword.
14375
14376table_http_req_cnt(<table>)
14377 Uses the string representation of the input sample to perform a look up in
14378 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014379 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014380 requests associated with the input sample in the designated table. See also
14381 the sc_http_req_cnt sample fetch keyword.
14382
14383table_http_req_rate(<table>)
14384 Uses the string representation of the input sample to perform a look up in
14385 the specified table. If the key is not found in the table, integer value zero
14386 is returned. Otherwise the average rate of HTTP requests associated with the
14387 input sample in the designated table, measured in amount of requests over the
14388 period configured in the table. See also the sc_http_req_rate sample fetch
14389 keyword.
14390
14391table_kbytes_in(<table>)
14392 Uses the string representation of the input sample to perform a look up in
14393 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014394 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014395 to-server data associated with the input sample in the designated table,
14396 measured in kilobytes. The test is currently performed on 32-bit integers,
14397 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
14398 keyword.
14399
14400table_kbytes_out(<table>)
14401 Uses the string representation of the input sample to perform a look up in
14402 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014403 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014404 to-client data associated with the input sample in the designated table,
14405 measured in kilobytes. The test is currently performed on 32-bit integers,
14406 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
14407 keyword.
14408
14409table_server_id(<table>)
14410 Uses the string representation of the input sample to perform a look up in
14411 the specified table. If the key is not found in the table, integer value zero
14412 is returned. Otherwise the converter returns the server ID associated with
14413 the input sample in the designated table. A server ID is associated to a
14414 sample by a "stick" rule when a connection to a server succeeds. A server ID
14415 zero means that no server is associated with this key.
14416
14417table_sess_cnt(<table>)
14418 Uses the string representation of the input sample to perform a look up in
14419 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014420 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014421 sessions associated with the input sample in the designated table. Note that
14422 a session here refers to an incoming connection being accepted by the
14423 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
14424 keyword.
14425
14426table_sess_rate(<table>)
14427 Uses the string representation of the input sample to perform a look up in
14428 the specified table. If the key is not found in the table, integer value zero
14429 is returned. Otherwise the converter returns the average incoming session
14430 rate associated with the input sample in the designated table. Note that a
14431 session here refers to an incoming connection being accepted by the
14432 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
14433 keyword.
14434
14435table_trackers(<table>)
14436 Uses the string representation of the input sample to perform a look up in
14437 the specified table. If the key is not found in the table, integer value zero
14438 is returned. Otherwise the converter returns the current amount of concurrent
14439 connections tracking the same key as the input sample in the designated
14440 table. It differs from table_conn_cur in that it does not rely on any stored
14441 information but on the table's reference count (the "use" value which is
14442 returned by "show table" on the CLI). This may sometimes be more suited for
14443 layer7 tracking. It can be used to tell a server how many concurrent
14444 connections there are from a given address for example. See also the
14445 sc_trackers sample fetch keyword.
14446
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014447upper
14448 Convert a string sample to upper case. This can only be placed after a string
14449 sample fetch function or after a transformation keyword returning a string
14450 type. The result is of type string.
14451
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020014452url_dec
14453 Takes an url-encoded string provided as input and returns the decoded
14454 version as output. The input and the output are of type string.
14455
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014456ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014457 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014458 sample representation of a gRPC message with <field_number> as field number
14459 (dotted notation) if <field_type> is not present, or as an integer sample if this
14460 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014461 The list of the authorized types is the following one: "int32", "int64", "uint32",
14462 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14463 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14464 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014465 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014466 More information may be found here about the protocol buffers message field types:
14467 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014468
14469 Example:
14470 // with such a protocol buffer .proto file content adapted from
14471 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
14472
14473 message Point {
14474 int32 latitude = 1;
14475 int32 longitude = 2;
14476 }
14477
14478 message PPoint {
14479 Point point = 59;
14480 }
14481
14482 message Rectangle {
14483 // One corner of the rectangle.
14484 PPoint lo = 48;
14485 // The other corner of the rectangle.
14486 PPoint hi = 49;
14487 }
14488
14489 let's say a body request is made of a "Rectangle" object value (two PPoint
14490 protocol buffers messages), the four protocol buffers fields could be
14491 extracted with these "ungrpc" directives:
14492
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014493 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
14494 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
John Roeslerfb2fce12019-07-10 15:45:51 -050014495 req.body,ungrpc(49.59.1,int32) # "latitude" of "hi" second PPoint
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014496 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
14497
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014498 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014499
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014500 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014501
John Roeslerfb2fce12019-07-10 15:45:51 -050014502 As a gRPC message is always made of a gRPC header followed by protocol buffers
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014503 messages, in the previous example the "latitude" of "lo" first PPoint
14504 could be extracted with these equivalent directives:
14505
14506 req.body,ungrpc(48.59),protobuf(1,int32)
14507 req.body,ungrpc(48),protobuf(59.1,int32)
14508 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
14509
14510 Note that the first convert must be "ungrpc", the remaining ones must be
14511 "protobuf" and only the last one may have or not a second argument to
14512 interpret the previous binary sample.
14513
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014514
Christopher Faulet85d79c92016-11-09 16:54:56 +010014515unset-var(<var name>)
14516 Unsets a variable if the input content is defined. The name of the variable
14517 starts with an indication about its scope. The scopes allowed are:
14518 "proc" : the variable is shared with the whole process
14519 "sess" : the variable is shared with the whole session
14520 "txn" : the variable is shared with the transaction (request and
14521 response),
14522 "req" : the variable is shared only during request processing,
14523 "res" : the variable is shared only during response processing.
14524 This prefix is followed by a name. The separator is a '.'. The name may only
14525 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
14526
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014527utime(<format>[,<offset>])
14528 Converts an integer supposed to contain a date since epoch to a string
14529 representing this date in UTC time using a format defined by the <format>
14530 string using strftime(3). The purpose is to allow any date format to be used
14531 in logs. An optional <offset> in seconds may be applied to the input date
14532 (positive or negative). See the strftime() man page for the format supported
14533 by your operating system. See also the ltime converter.
14534
14535 Example :
14536
14537 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014538 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014539 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
14540
Marcin Deranek9631a282018-04-16 14:30:46 +020014541word(<index>,<delimiters>[,<count>])
14542 Extracts the nth word counting from the beginning (positive index) or from
14543 the end (negative index) considering given delimiters from an input string.
14544 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
Jerome Magnin88209322020-01-28 13:33:44 +010014545 Delimiters at the beginning or end of the input string are ignored.
Marcin Deranek9631a282018-04-16 14:30:46 +020014546 Optionally you can specify <count> of words to extract (default: 1).
14547 Value of 0 indicates extraction of all remaining words.
14548
14549 Example :
14550 str(f1_f2_f3__f5),word(4,_) # f5
14551 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
14552 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
14553 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
14554 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Jerome Magnin88209322020-01-28 13:33:44 +010014555 str(/f1/f2/f3/f4),word(1,/) # f1
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010014556
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014557wt6([<avalanche>])
14558 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
14559 hash function. Optionally, it is possible to apply a full avalanche hash
14560 function to the output if the optional <avalanche> argument equals 1. This
14561 converter uses the same functions as used by the various hash-based load
14562 balancing algorithms, so it will provide exactly the same results. It is
14563 mostly intended for debugging, but can be used as a stick-table entry to
14564 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014565 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
14566 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014567
Willy Tarreau97707872015-01-27 15:12:13 +010014568xor(<value>)
14569 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014570 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014571 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014572 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014573 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014574 "sess" : the variable is shared with the whole session
14575 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014576 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014577 "req" : the variable is shared only during request processing,
14578 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014579 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014580 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014581
Thierry FOURNIER01e09742016-12-26 11:46:11 +010014582xxh32([<seed>])
14583 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
14584 variant of the XXHash hash function. This hash supports a seed which defaults
14585 to zero but a different value maybe passed as the <seed> argument. This hash
14586 is known to be very good and very fast so it can be used to hash URLs and/or
14587 URL parameters for use as stick-table keys to collect statistics with a low
14588 collision rate, though care must be taken as the algorithm is not considered
14589 as cryptographically secure.
14590
14591xxh64([<seed>])
14592 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
14593 variant of the XXHash hash function. This hash supports a seed which defaults
14594 to zero but a different value maybe passed as the <seed> argument. This hash
14595 is known to be very good and very fast so it can be used to hash URLs and/or
14596 URL parameters for use as stick-table keys to collect statistics with a low
14597 collision rate, though care must be taken as the algorithm is not considered
14598 as cryptographically secure.
14599
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014600
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200146017.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020014602--------------------------------------------
14603
14604A first set of sample fetch methods applies to internal information which does
14605not even relate to any client information. These ones are sometimes used with
14606"monitor-fail" directives to report an internal status to external watchers.
14607The sample fetch methods described in this section are usable anywhere.
14608
14609always_false : boolean
14610 Always returns the boolean "false" value. It may be used with ACLs as a
14611 temporary replacement for another one when adjusting configurations.
14612
14613always_true : boolean
14614 Always returns the boolean "true" value. It may be used with ACLs as a
14615 temporary replacement for another one when adjusting configurations.
14616
14617avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014618 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020014619 divided by the number of active servers. The current backend is used if no
14620 backend is specified. This is very similar to "queue" except that the size of
14621 the farm is considered, in order to give a more accurate measurement of the
14622 time it may take for a new connection to be processed. The main usage is with
14623 ACL to return a sorry page to new users when it becomes certain they will get
14624 a degraded service, or to pass to the backend servers in a header so that
14625 they decide to work in degraded mode or to disable some functions to speed up
14626 the processing a bit. Note that in the event there would not be any active
14627 server anymore, twice the number of queued connections would be considered as
14628 the measured value. This is a fair estimate, as we expect one server to get
14629 back soon anyway, but we still prefer to send new traffic to another backend
14630 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
14631 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010014632
Willy Tarreau74ca5042013-06-11 23:12:07 +020014633be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020014634 Applies to the number of currently established connections on the backend,
14635 possibly including the connection being evaluated. If no backend name is
14636 specified, the current one is used. But it is also possible to check another
14637 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014638 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
14639
14640be_conn_free([<backend>]) : integer
14641 Returns an integer value corresponding to the number of available connections
14642 across available servers in the backend. Queue slots are not included. Backup
14643 servers are also not included, unless all other servers are down. If no
14644 backend name is specified, the current one is used. But it is also possible
14645 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014646 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
14647 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014648
14649 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
14650 (meaning unlimited), then this fetch clearly does not make sense, in which
14651 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014652
Willy Tarreau74ca5042013-06-11 23:12:07 +020014653be_sess_rate([<backend>]) : integer
14654 Returns an integer value corresponding to the sessions creation rate on the
14655 backend, in number of new sessions per second. This is used with ACLs to
14656 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010014657 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014658 online dictionary). It can also be useful to add this element to logs using a
14659 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014660
14661 Example :
14662 # Redirect to an error page if the dictionary is requested too often
14663 backend dynamic
14664 mode http
14665 acl being_scanned be_sess_rate gt 100
14666 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010014667
Davor Ocelice9ed2812017-12-25 17:49:28 +010014668bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014669 Returns a binary chain. The input is the hexadecimal representation
14670 of the string.
14671
14672bool(<bool>) : bool
14673 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
14674 'false' and '0' are the same. 'true' and '1' are the same.
14675
Willy Tarreau74ca5042013-06-11 23:12:07 +020014676connslots([<backend>]) : integer
14677 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014678 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020014679 connections on all servers and the maximum queue size. This is probably only
14680 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050014681
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014682 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020014683 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014684 usage; see "use_backend" keyword) can be redirected to a different backend.
14685
Willy Tarreau55165fe2009-05-10 12:02:55 +020014686 'connslots' = number of available server connection slots, + number of
14687 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014688
Willy Tarreaua36af912009-10-10 12:02:45 +020014689 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020014690 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020014691 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020014692 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010014693 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020014694 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020014695 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014696
Willy Tarreau55165fe2009-05-10 12:02:55 +020014697 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
14698 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020014699 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020014700 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014701
Willy Tarreau70fe9442018-11-22 16:07:39 +010014702cpu_calls : integer
14703 Returns the number of calls to the task processing the stream or current
14704 request since it was allocated. This number is reset for each new request on
14705 the same connections in case of HTTP keep-alive. This value should usually be
14706 low and stable (around 2 calls for a typically simple request) but may become
14707 high if some processing (compression, caching or analysis) is performed. This
14708 is purely for performance monitoring purposes.
14709
14710cpu_ns_avg : integer
14711 Returns the average number of nanoseconds spent in each call to the task
14712 processing the stream or current request. This number is reset for each new
14713 request on the same connections in case of HTTP keep-alive. This value
14714 indicates the overall cost of processing the request or the connection for
14715 each call. There is no good nor bad value but the time spent in a call
14716 automatically causes latency for other processing (see lat_ns_avg below),
14717 and may affect other connection's apparent response time. Certain operations
14718 like compression, complex regex matching or heavy Lua operations may directly
14719 affect this value, and having it in the logs will make it easier to spot the
14720 faulty processing that needs to be fixed to recover decent performance.
14721 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
14722
14723cpu_ns_tot : integer
14724 Returns the total number of nanoseconds spent in each call to the task
14725 processing the stream or current request. This number is reset for each new
14726 request on the same connections in case of HTTP keep-alive. This value
14727 indicates the overall cost of processing the request or the connection for
14728 each call. There is no good nor bad value but the time spent in a call
14729 automatically causes latency for other processing (see lat_ns_avg below),
14730 induces CPU costs on the machine, and may affect other connection's apparent
14731 response time. Certain operations like compression, complex regex matching or
14732 heavy Lua operations may directly affect this value, and having it in the
14733 logs will make it easier to spot the faulty processing that needs to be fixed
14734 to recover decent performance. The value may be artificially high due to a
14735 high cpu_calls count, for example when processing many HTTP chunks, and for
14736 this reason it is often preferred to log cpu_ns_avg instead.
14737
Cyril Bonté6bcd1822019-11-05 23:13:59 +010014738date([<offset>],[<unit>]) : integer
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014739 Returns the current date as the epoch (number of seconds since 01/01/1970).
Damien Claisseae6f1252019-10-30 15:57:28 +000014740
14741 If an offset value is specified, then it is added to the current date before
14742 returning the value. This is particularly useful to compute relative dates,
14743 as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020014744 It is useful combined with the http_date converter.
14745
Damien Claisseae6f1252019-10-30 15:57:28 +000014746 <unit> is facultative, and can be set to "s" for seconds (default behavior),
14747 "ms" for milliseconds or "us" for microseconds.
14748 If unit is set, return value is an integer reflecting either seconds,
14749 milliseconds or microseconds since epoch, plus offset.
14750 It is useful when a time resolution of less than a second is needed.
14751
Willy Tarreau276fae92013-07-25 14:36:01 +020014752 Example :
14753
14754 # set an expires header to now+1 hour in every response
14755 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014756
Damien Claisseae6f1252019-10-30 15:57:28 +000014757 # set an expires header to now+1 hour in every response, with
14758 # millisecond granularity
14759 http-response set-header Expires %[date(3600000,ms),http_date(0,ms)]
14760
Etienne Carrierea792a0a2018-01-17 13:43:24 +010014761date_us : integer
14762 Return the microseconds part of the date (the "second" part is returned by
14763 date sample). This sample is coherent with the date sample as it is comes
14764 from the same timeval structure.
14765
Willy Tarreaud716f9b2017-10-13 11:03:15 +020014766distcc_body(<token>[,<occ>]) : binary
14767 Parses a distcc message and returns the body associated to occurrence #<occ>
14768 of the token <token>. Occurrences start at 1, and when unspecified, any may
14769 match though in practice only the first one is checked for now. This can be
14770 used to extract file names or arguments in files built using distcc through
14771 haproxy. Please refer to distcc's protocol documentation for the complete
14772 list of supported tokens.
14773
14774distcc_param(<token>[,<occ>]) : integer
14775 Parses a distcc message and returns the parameter associated to occurrence
14776 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
14777 any may match though in practice only the first one is checked for now. This
14778 can be used to extract certain information such as the protocol version, the
14779 file size or the argument in files built using distcc through haproxy.
14780 Another use case consists in waiting for the start of the preprocessed file
14781 contents before connecting to the server to avoid keeping idle connections.
14782 Please refer to distcc's protocol documentation for the complete list of
14783 supported tokens.
14784
14785 Example :
14786 # wait up to 20s for the pre-processed file to be uploaded
14787 tcp-request inspect-delay 20s
14788 tcp-request content accept if { distcc_param(DOTI) -m found }
14789 # send large files to the big farm
14790 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
14791
Willy Tarreau595ec542013-06-12 21:34:28 +020014792env(<name>) : string
14793 Returns a string containing the value of environment variable <name>. As a
14794 reminder, environment variables are per-process and are sampled when the
14795 process starts. This can be useful to pass some information to a next hop
14796 server, or with ACLs to take specific action when the process is started a
14797 certain way.
14798
14799 Examples :
14800 # Pass the Via header to next hop with the local hostname in it
14801 http-request add-header Via 1.1\ %[env(HOSTNAME)]
14802
14803 # reject cookie-less requests when the STOP environment variable is set
14804 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
14805
Willy Tarreau74ca5042013-06-11 23:12:07 +020014806fe_conn([<frontend>]) : integer
14807 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010014808 possibly including the connection being evaluated. If no frontend name is
14809 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014810 frontend. It can be used to return a sorry page before hard-blocking, or to
14811 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010014812 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020014813 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
14814 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020014815
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020014816fe_req_rate([<frontend>]) : integer
14817 Returns an integer value corresponding to the number of HTTP requests per
14818 second sent to a frontend. This number can differ from "fe_sess_rate" in
14819 situations where client-side keep-alive is enabled.
14820
Willy Tarreau74ca5042013-06-11 23:12:07 +020014821fe_sess_rate([<frontend>]) : integer
14822 Returns an integer value corresponding to the sessions creation rate on the
14823 frontend, in number of new sessions per second. This is used with ACLs to
14824 limit the incoming session rate to an acceptable range in order to prevent
14825 abuse of service at the earliest moment, for example when combined with other
14826 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
14827 down below the limit. It can also be useful to add this element to logs using
14828 a log-format directive. See also the "rate-limit sessions" directive for use
14829 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014830
14831 Example :
14832 # This frontend limits incoming mails to 10/s with a max of 100
14833 # concurrent connections. We accept any connection below 10/s, and
14834 # force excess clients to wait for 100 ms. Since clients are limited to
14835 # 100 max, there cannot be more than 10 incoming mails per second.
14836 frontend mail
14837 bind :25
14838 mode tcp
14839 maxconn 100
14840 acl too_fast fe_sess_rate ge 10
14841 tcp-request inspect-delay 100ms
14842 tcp-request content accept if ! too_fast
14843 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010014844
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010014845hostname : string
14846 Returns the system hostname.
14847
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014848int(<integer>) : signed integer
14849 Returns a signed integer.
14850
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014851ipv4(<ipv4>) : ipv4
14852 Returns an ipv4.
14853
14854ipv6(<ipv6>) : ipv6
14855 Returns an ipv6.
14856
Willy Tarreau70fe9442018-11-22 16:07:39 +010014857lat_ns_avg : integer
14858 Returns the average number of nanoseconds spent between the moment the task
14859 handling the stream is woken up and the moment it is effectively called. This
14860 number is reset for each new request on the same connections in case of HTTP
14861 keep-alive. This value indicates the overall latency inflicted to the current
14862 request by all other requests being processed in parallel, and is a direct
14863 indicator of perceived performance due to noisy neighbours. In order to keep
14864 the value low, it is possible to reduce the scheduler's run queue depth using
14865 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14866 once using "tune.maxpollevents", to decrease the stream's nice value using
14867 the "nice" option on the "bind" lines or in the frontend, or to look for
14868 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14869 whose processing needs to be adjusted or fixed. Compression of large buffers
14870 could be a culprit, like heavy regex or long lists of regex.
14871 Note: this value is exactly lat_ns_tot divided by cpu_calls.
14872
14873lat_ns_tot : integer
14874 Returns the total number of nanoseconds spent between the moment the task
14875 handling the stream is woken up and the moment it is effectively called. This
14876 number is reset for each new request on the same connections in case of HTTP
14877 keep-alive. This value indicates the overall latency inflicted to the current
14878 request by all other requests being processed in parallel, and is a direct
14879 indicator of perceived performance due to noisy neighbours. In order to keep
14880 the value low, it is possible to reduce the scheduler's run queue depth using
14881 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14882 once using "tune.maxpollevents", to decrease the stream's nice value using
14883 the "nice" option on the "bind" lines or in the frontend, or to look for
14884 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14885 whose processing needs to be adjusted or fixed. Compression of large buffers
14886 could be a culprit, like heavy regex or long lists of regex. Note: while it
14887 may intuitively seem that the total latency adds to a transfer time, it is
14888 almost never true because while a task waits for the CPU, network buffers
14889 continue to fill up and the next call will process more at once. The value
14890 may be artificially high due to a high cpu_calls count, for example when
14891 processing many HTTP chunks, and for this reason it is often preferred to log
14892 lat_ns_avg instead, which is a more relevant performance indicator.
14893
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014894meth(<method>) : method
14895 Returns a method.
14896
Willy Tarreau0f30d262014-11-24 16:02:05 +010014897nbproc : integer
14898 Returns an integer value corresponding to the number of processes that were
14899 started (it equals the global "nbproc" setting). This is useful for logging
14900 and debugging purposes.
14901
Willy Tarreau74ca5042013-06-11 23:12:07 +020014902nbsrv([<backend>]) : integer
14903 Returns an integer value corresponding to the number of usable servers of
14904 either the current backend or the named backend. This is mostly used with
14905 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010014906 switch to an alternate backend when the number of servers is too low to
14907 to handle some load. It is useful to report a failure when combined with
14908 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014909
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040014910prio_class : integer
14911 Returns the priority class of the current session for http mode or connection
14912 for tcp mode. The value will be that set by the last call to "http-request
14913 set-priority-class" or "tcp-request content set-priority-class".
14914
14915prio_offset : integer
14916 Returns the priority offset of the current session for http mode or
14917 connection for tcp mode. The value will be that set by the last call to
14918 "http-request set-priority-offset" or "tcp-request content
14919 set-priority-offset".
14920
Willy Tarreau0f30d262014-11-24 16:02:05 +010014921proc : integer
14922 Returns an integer value corresponding to the position of the process calling
14923 the function, between 1 and global.nbproc. This is useful for logging and
14924 debugging purposes.
14925
Willy Tarreau74ca5042013-06-11 23:12:07 +020014926queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014927 Returns the total number of queued connections of the designated backend,
14928 including all the connections in server queues. If no backend name is
14929 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014930 one. This is useful with ACLs or to pass statistics to backend servers. This
14931 can be used to take actions when queuing goes above a known level, generally
14932 indicating a surge of traffic or a massive slowdown on the servers. One
14933 possible action could be to reject new users but still accept old ones. See
14934 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
14935
Willy Tarreau84310e22014-02-14 11:59:04 +010014936rand([<range>]) : integer
14937 Returns a random integer value within a range of <range> possible values,
14938 starting at zero. If the range is not specified, it defaults to 2^32, which
14939 gives numbers between 0 and 4294967295. It can be useful to pass some values
14940 needed to take some routing decisions for example, or just for debugging
14941 purposes. This random must not be used for security purposes.
14942
Luca Schimweg8a694b82019-09-10 15:42:52 +020014943uuid([<version>]) : string
14944 Returns a UUID following the RFC4122 standard. If the version is not
14945 specified, a UUID version 4 (fully random) is returned.
14946 Currently, only version 4 is supported.
14947
Willy Tarreau74ca5042013-06-11 23:12:07 +020014948srv_conn([<backend>/]<server>) : integer
14949 Returns an integer value corresponding to the number of currently established
14950 connections on the designated server, possibly including the connection being
14951 evaluated. If <backend> is omitted, then the server is looked up in the
14952 current backend. It can be used to use a specific farm when one server is
14953 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014954 connections with it. See also the "fe_conn", "be_conn", "queue", and
14955 "srv_conn_free" fetch methods.
14956
14957srv_conn_free([<backend>/]<server>) : integer
14958 Returns an integer value corresponding to the number of available connections
14959 on the designated server, possibly including the connection being evaluated.
14960 The value does not include queue slots. If <backend> is omitted, then the
14961 server is looked up in the current backend. It can be used to use a specific
14962 farm when one server is full, or to inform the server about our view of the
14963 number of active connections with it. See also the "be_conn_free" and
14964 "srv_conn" fetch methods.
14965
14966 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
14967 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014968
14969srv_is_up([<backend>/]<server>) : boolean
14970 Returns true when the designated server is UP, and false when it is either
14971 DOWN or in maintenance mode. If <backend> is omitted, then the server is
14972 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010014973 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014974 availability). Another possible use which is more of a hack consists in
14975 using dummy servers as boolean variables that can be enabled or disabled from
14976 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
14977
Willy Tarreauff2b7af2017-10-13 11:46:26 +020014978srv_queue([<backend>/]<server>) : integer
14979 Returns an integer value corresponding to the number of connections currently
14980 pending in the designated server's queue. If <backend> is omitted, then the
14981 server is looked up in the current backend. It can sometimes be used together
14982 with the "use-server" directive to force to use a known faster server when it
14983 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
14984 fetch methods.
14985
Willy Tarreau74ca5042013-06-11 23:12:07 +020014986srv_sess_rate([<backend>/]<server>) : integer
14987 Returns an integer corresponding to the sessions creation rate on the
14988 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014989 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020014990 used with ACLs but can make sense with logs too. This is used to switch to an
14991 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010014992 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020014993 overloading servers).
14994
14995 Example :
14996 # Redirect to a separate back
14997 acl srv1_full srv_sess_rate(be1/srv1) gt 50
14998 acl srv2_full srv_sess_rate(be1/srv2) gt 50
14999 use_backend be2 if srv1_full or srv2_full
15000
Willy Tarreau0f30d262014-11-24 16:02:05 +010015001stopping : boolean
15002 Returns TRUE if the process calling the function is currently stopping. This
15003 can be useful for logging, or for relaxing certain checks or helping close
15004 certain connections upon graceful shutdown.
15005
Thierry FOURNIERcc103292015-06-06 19:30:17 +020015006str(<string>) : string
15007 Returns a string.
15008
Willy Tarreau74ca5042013-06-11 23:12:07 +020015009table_avl([<table>]) : integer
15010 Returns the total number of available entries in the current proxy's
15011 stick-table or in the designated stick-table. See also table_cnt.
15012
15013table_cnt([<table>]) : integer
15014 Returns the total number of entries currently in use in the current proxy's
15015 stick-table or in the designated stick-table. See also src_conn_cnt and
15016 table_avl for other entry counting methods.
15017
Christopher Faulet34adb2a2017-11-21 21:45:38 +010015018thread : integer
15019 Returns an integer value corresponding to the position of the thread calling
15020 the function, between 0 and (global.nbthread-1). This is useful for logging
15021 and debugging purposes.
15022
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015023var(<var-name>) : undefined
15024 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010015025 sample fetch fails. The name of the variable starts with an indication
15026 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010015027 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010015028 "sess" : the variable is shared with the whole session
15029 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015030 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010015031 "req" : the variable is shared only during request processing,
15032 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015033 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010015034 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015035
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200150367.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020015037----------------------------------
15038
15039The layer 4 usually describes just the transport layer which in haproxy is
15040closest to the connection, where no content is yet made available. The fetch
15041methods described here are usable as low as the "tcp-request connection" rule
15042sets unless they require some future information. Those generally include
15043TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015044the incoming connection. For retrieving a value from a sticky counters, the
15045counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020015046"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
15047used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
15048can be specified as the first integer argument when using the "sc_" prefix.
15049Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
15050table may be specified with the "sc*" form, in which case the currently
15051tracked key will be looked up into this alternate table instead of the table
15052currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015053
Jérôme Magnin35e53a62019-01-16 14:38:37 +010015054bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010015055 Returns the backend connection's HTTP major version encoding, which may be 1
15056 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
15057 encoding and not the version present in the request header.
15058
Willy Tarreau74ca5042013-06-11 23:12:07 +020015059be_id : integer
15060 Returns an integer containing the current backend's id. It can be used in
15061 frontends with responses to check which backend processed the request.
15062
Marcin Deranekd2471c22016-12-12 14:08:05 +010015063be_name : string
15064 Returns a string containing the current backend's name. It can be used in
15065 frontends with responses to check which backend processed the request.
15066
Willy Tarreau74ca5042013-06-11 23:12:07 +020015067dst : ip
15068 This is the destination IPv4 address of the connection on the client side,
15069 which is the address the client connected to. It can be useful when running
15070 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
15071 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015072 RFC 4291. When the incoming connection passed through address translation or
15073 redirection involving connection tracking, the original destination address
15074 before the redirection will be reported. On Linux systems, the source and
15075 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
15076 is set, because a late response may reopen a timed out connection and switch
15077 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015078
15079dst_conn : integer
15080 Returns an integer value corresponding to the number of currently established
15081 connections on the same socket including the one being evaluated. It is
15082 normally used with ACLs but can as well be used to pass the information to
15083 servers in an HTTP header or in logs. It can be used to either return a sorry
15084 page before hard-blocking, or to use a specific backend to drain new requests
15085 when the socket is considered saturated. This offers the ability to assign
15086 different limits to different listening ports or addresses. See also the
15087 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015088
Willy Tarreau16e01562016-08-09 16:46:18 +020015089dst_is_local : boolean
15090 Returns true if the destination address of the incoming connection is local
15091 to the system, or false if the address doesn't exist on the system, meaning
15092 that it was intercepted in transparent mode. It can be useful to apply
15093 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010015094 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020015095 be delivered only on this address, or SSH access could be locally redirected.
15096 Please note that the check involves a few system calls, so it's better to do
15097 it only once per connection.
15098
Willy Tarreau74ca5042013-06-11 23:12:07 +020015099dst_port : integer
15100 Returns an integer value corresponding to the destination TCP port of the
15101 connection on the client side, which is the port the client connected to.
15102 This might be used when running in transparent mode, when assigning dynamic
15103 ports to some clients for a whole application session, to stick all users to
15104 a same server, or to pass the destination port information to a server using
15105 an HTTP header.
15106
Willy Tarreau60ca10a2017-08-18 15:26:54 +020015107fc_http_major : integer
15108 Reports the front connection's HTTP major version encoding, which may be 1
15109 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
15110 encoding and not on the version present in the request header.
15111
Geoff Simmons7185b782019-08-27 18:31:16 +020015112fc_pp_authority : string
15113 Returns the authority TLV sent by the client in the PROXY protocol header,
15114 if any.
15115
Emeric Brun4f603012017-01-05 15:11:44 +010015116fc_rcvd_proxy : boolean
15117 Returns true if the client initiated the connection with a PROXY protocol
15118 header.
15119
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020015120fc_rtt(<unit>) : integer
15121 Returns the Round Trip Time (RTT) measured by the kernel for the client
15122 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
15123 can be set to "ms" for milliseconds or "us" for microseconds. If the server
15124 connection is not established, if the connection is not TCP or if the
15125 operating system does not support TCP_INFO, for example Linux kernels before
15126 2.4, the sample fetch fails.
15127
15128fc_rttvar(<unit>) : integer
15129 Returns the Round Trip Time (RTT) variance measured by the kernel for the
15130 client connection. <unit> is facultative, by default the unit is milliseconds.
15131 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
15132 server connection is not established, if the connection is not TCP or if the
15133 operating system does not support TCP_INFO, for example Linux kernels before
15134 2.4, the sample fetch fails.
15135
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015136fc_unacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015137 Returns the unacked counter measured by the kernel for the client connection.
15138 If the server connection is not established, if the connection is not TCP or
15139 if the operating system does not support TCP_INFO, for example Linux kernels
15140 before 2.4, the sample fetch fails.
15141
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015142fc_sacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015143 Returns the sacked counter measured by the kernel for the client connection.
15144 If the server connection is not established, if the connection is not TCP or
15145 if the operating system does not support TCP_INFO, for example Linux kernels
15146 before 2.4, the sample fetch fails.
15147
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015148fc_retrans : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015149 Returns the retransmits counter measured by the kernel for the client
15150 connection. If the server connection is not established, if the connection is
15151 not TCP or if the operating system does not support TCP_INFO, for example
15152 Linux kernels before 2.4, the sample fetch fails.
15153
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015154fc_fackets : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015155 Returns the fack counter measured by the kernel for the client
15156 connection. If the server connection is not established, if the connection is
15157 not TCP or if the operating system does not support TCP_INFO, for example
15158 Linux kernels before 2.4, the sample fetch fails.
15159
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015160fc_lost : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015161 Returns the lost counter measured by the kernel for the client
15162 connection. If the server connection is not established, if the connection is
15163 not TCP or if the operating system does not support TCP_INFO, for example
15164 Linux kernels before 2.4, the sample fetch fails.
15165
Christopher Fauletba0c53e2019-10-17 14:40:48 +020015166fc_reordering : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015167 Returns the reordering counter measured by the kernel for the client
15168 connection. If the server connection is not established, if the connection is
15169 not TCP or if the operating system does not support TCP_INFO, for example
15170 Linux kernels before 2.4, the sample fetch fails.
15171
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020015172fe_defbe : string
15173 Returns a string containing the frontend's default backend name. It can be
15174 used in frontends to check which backend will handle requests by default.
15175
Willy Tarreau74ca5042013-06-11 23:12:07 +020015176fe_id : integer
15177 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010015178 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020015179 coming via a same frontend to the same server.
15180
Marcin Deranekd2471c22016-12-12 14:08:05 +010015181fe_name : string
15182 Returns a string containing the current frontend's name. It can be used in
15183 backends to check from which frontend it was called, or to stick all users
15184 coming via a same frontend to the same server.
15185
Cyril Bonté62ba8702014-04-22 23:52:25 +020015186sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015187sc0_bytes_in_rate([<table>]) : integer
15188sc1_bytes_in_rate([<table>]) : integer
15189sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015190 Returns the average client-to-server bytes rate from the currently tracked
15191 counters, measured in amount of bytes over the period configured in the
15192 table. See also src_bytes_in_rate.
15193
Cyril Bonté62ba8702014-04-22 23:52:25 +020015194sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015195sc0_bytes_out_rate([<table>]) : integer
15196sc1_bytes_out_rate([<table>]) : integer
15197sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015198 Returns the average server-to-client bytes rate from the currently tracked
15199 counters, measured in amount of bytes over the period configured in the
15200 table. See also src_bytes_out_rate.
15201
Cyril Bonté62ba8702014-04-22 23:52:25 +020015202sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015203sc0_clr_gpc0([<table>]) : integer
15204sc1_clr_gpc0([<table>]) : integer
15205sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020015206 Clears the first General Purpose Counter associated to the currently tracked
15207 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010015208 stored value is zero, so first invocation will always return zero. This is
15209 typically used as a second ACL in an expression in order to mark a connection
15210 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015211
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015212 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015213 # block if 5 consecutive requests continue to come faster than 10 sess
15214 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015215 acl abuse sc0_http_req_rate gt 10
15216 acl kill sc0_inc_gpc0 gt 5
15217 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015218 tcp-request connection accept if !abuse save
15219 tcp-request connection reject if abuse kill
15220
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015221sc_clr_gpc1(<ctr>[,<table>]) : integer
15222sc0_clr_gpc1([<table>]) : integer
15223sc1_clr_gpc1([<table>]) : integer
15224sc2_clr_gpc1([<table>]) : integer
15225 Clears the second General Purpose Counter associated to the currently tracked
15226 counters, and returns its previous value. Before the first invocation, the
15227 stored value is zero, so first invocation will always return zero. This is
15228 typically used as a second ACL in an expression in order to mark a connection
15229 when a first ACL was verified.
15230
Cyril Bonté62ba8702014-04-22 23:52:25 +020015231sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015232sc0_conn_cnt([<table>]) : integer
15233sc1_conn_cnt([<table>]) : integer
15234sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015235 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015236 counters. See also src_conn_cnt.
15237
Cyril Bonté62ba8702014-04-22 23:52:25 +020015238sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015239sc0_conn_cur([<table>]) : integer
15240sc1_conn_cur([<table>]) : integer
15241sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015242 Returns the current amount of concurrent connections tracking the same
15243 tracked counters. This number is automatically incremented when tracking
15244 begins and decremented when tracking stops. See also src_conn_cur.
15245
Cyril Bonté62ba8702014-04-22 23:52:25 +020015246sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015247sc0_conn_rate([<table>]) : integer
15248sc1_conn_rate([<table>]) : integer
15249sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015250 Returns the average connection rate from the currently tracked counters,
15251 measured in amount of connections over the period configured in the table.
15252 See also src_conn_rate.
15253
Cyril Bonté62ba8702014-04-22 23:52:25 +020015254sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015255sc0_get_gpc0([<table>]) : integer
15256sc1_get_gpc0([<table>]) : integer
15257sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015258 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015259 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015260
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015261sc_get_gpc1(<ctr>[,<table>]) : integer
15262sc0_get_gpc1([<table>]) : integer
15263sc1_get_gpc1([<table>]) : integer
15264sc2_get_gpc1([<table>]) : integer
15265 Returns the value of the second General Purpose Counter associated to the
15266 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
15267
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015268sc_get_gpt0(<ctr>[,<table>]) : integer
15269sc0_get_gpt0([<table>]) : integer
15270sc1_get_gpt0([<table>]) : integer
15271sc2_get_gpt0([<table>]) : integer
15272 Returns the value of the first General Purpose Tag associated to the
15273 currently tracked counters. See also src_get_gpt0.
15274
Cyril Bonté62ba8702014-04-22 23:52:25 +020015275sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015276sc0_gpc0_rate([<table>]) : integer
15277sc1_gpc0_rate([<table>]) : integer
15278sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015279 Returns the average increment rate of the first General Purpose Counter
15280 associated to the currently tracked counters. It reports the frequency
15281 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015282 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15283 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15284 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020015285
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015286sc_gpc1_rate(<ctr>[,<table>]) : integer
15287sc0_gpc1_rate([<table>]) : integer
15288sc1_gpc1_rate([<table>]) : integer
15289sc2_gpc1_rate([<table>]) : integer
15290 Returns the average increment rate of the second General Purpose Counter
15291 associated to the currently tracked counters. It reports the frequency
15292 which the gpc1 counter was incremented over the configured period. See also
15293 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15294 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15295 be returned, as "gpc1" only holds the event count.
15296
Cyril Bonté62ba8702014-04-22 23:52:25 +020015297sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015298sc0_http_err_cnt([<table>]) : integer
15299sc1_http_err_cnt([<table>]) : integer
15300sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015301 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015302 counters. This includes the both request errors and 4xx error responses.
15303 See also src_http_err_cnt.
15304
Cyril Bonté62ba8702014-04-22 23:52:25 +020015305sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015306sc0_http_err_rate([<table>]) : integer
15307sc1_http_err_rate([<table>]) : integer
15308sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015309 Returns the average rate of HTTP errors from the currently tracked counters,
15310 measured in amount of errors over the period configured in the table. This
15311 includes the both request errors and 4xx error responses. See also
15312 src_http_err_rate.
15313
Cyril Bonté62ba8702014-04-22 23:52:25 +020015314sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015315sc0_http_req_cnt([<table>]) : integer
15316sc1_http_req_cnt([<table>]) : integer
15317sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015318 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015319 counters. This includes every started request, valid or not. See also
15320 src_http_req_cnt.
15321
Cyril Bonté62ba8702014-04-22 23:52:25 +020015322sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015323sc0_http_req_rate([<table>]) : integer
15324sc1_http_req_rate([<table>]) : integer
15325sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015326 Returns the average rate of HTTP requests from the currently tracked
15327 counters, measured in amount of requests over the period configured in
15328 the table. This includes every started request, valid or not. See also
15329 src_http_req_rate.
15330
Cyril Bonté62ba8702014-04-22 23:52:25 +020015331sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015332sc0_inc_gpc0([<table>]) : integer
15333sc1_inc_gpc0([<table>]) : integer
15334sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015335 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010015336 tracked counters, and returns its new value. Before the first invocation,
15337 the stored value is zero, so first invocation will increase it to 1 and will
15338 return 1. This is typically used as a second ACL in an expression in order
15339 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020015340
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015341 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015342 acl abuse sc0_http_req_rate gt 10
15343 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015344 tcp-request connection reject if abuse kill
15345
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015346sc_inc_gpc1(<ctr>[,<table>]) : integer
15347sc0_inc_gpc1([<table>]) : integer
15348sc1_inc_gpc1([<table>]) : integer
15349sc2_inc_gpc1([<table>]) : integer
15350 Increments the second General Purpose Counter associated to the currently
15351 tracked counters, and returns its new value. Before the first invocation,
15352 the stored value is zero, so first invocation will increase it to 1 and will
15353 return 1. This is typically used as a second ACL in an expression in order
15354 to mark a connection when a first ACL was verified.
15355
Cyril Bonté62ba8702014-04-22 23:52:25 +020015356sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015357sc0_kbytes_in([<table>]) : integer
15358sc1_kbytes_in([<table>]) : integer
15359sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015360 Returns the total amount of client-to-server data from the currently tracked
15361 counters, measured in kilobytes. The test is currently performed on 32-bit
15362 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020015363
Cyril Bonté62ba8702014-04-22 23:52:25 +020015364sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015365sc0_kbytes_out([<table>]) : integer
15366sc1_kbytes_out([<table>]) : integer
15367sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015368 Returns the total amount of server-to-client data from the currently tracked
15369 counters, measured in kilobytes. The test is currently performed on 32-bit
15370 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020015371
Cyril Bonté62ba8702014-04-22 23:52:25 +020015372sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015373sc0_sess_cnt([<table>]) : integer
15374sc1_sess_cnt([<table>]) : integer
15375sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015376 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020015377 into sessions, which means that they were accepted by a "tcp-request
15378 connection" rule, from the currently tracked counters. A backend may count
15379 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015380 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020015381 with the client. See also src_sess_cnt.
15382
Cyril Bonté62ba8702014-04-22 23:52:25 +020015383sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015384sc0_sess_rate([<table>]) : integer
15385sc1_sess_rate([<table>]) : integer
15386sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015387 Returns the average session rate from the currently tracked counters,
15388 measured in amount of sessions over the period configured in the table. A
15389 session is a connection that got past the early "tcp-request connection"
15390 rules. A backend may count more sessions than connections because each
15391 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015392 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020015393
Cyril Bonté62ba8702014-04-22 23:52:25 +020015394sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020015395sc0_tracked([<table>]) : boolean
15396sc1_tracked([<table>]) : boolean
15397sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020015398 Returns true if the designated session counter is currently being tracked by
15399 the current session. This can be useful when deciding whether or not we want
15400 to set some values in a header passed to the server.
15401
Cyril Bonté62ba8702014-04-22 23:52:25 +020015402sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015403sc0_trackers([<table>]) : integer
15404sc1_trackers([<table>]) : integer
15405sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010015406 Returns the current amount of concurrent connections tracking the same
15407 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015408 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010015409 that it does not rely on any stored information but on the table's reference
15410 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015411 may sometimes be more suited for layer7 tracking. It can be used to tell a
15412 server how many concurrent connections there are from a given address for
15413 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010015414
Willy Tarreau74ca5042013-06-11 23:12:07 +020015415so_id : integer
15416 Returns an integer containing the current listening socket's id. It is useful
15417 in frontends involving many "bind" lines, or to stick all users coming via a
15418 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015419
Willy Tarreau74ca5042013-06-11 23:12:07 +020015420src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010015421 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020015422 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
15423 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
15424 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015425 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
15426 directive is used, it can be the address of a client behind another
15427 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015428 "tcp-request connection" which sees the real address. When the incoming
15429 connection passed through address translation or redirection involving
15430 connection tracking, the original destination address before the redirection
15431 will be reported. On Linux systems, the source and destination may seldom
15432 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
15433 response may reopen a timed out connection and switch what is believed to be
15434 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015435
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010015436 Example:
15437 # add an HTTP header in requests with the originating address' country
15438 http-request set-header X-Country %[src,map_ip(geoip.lst)]
15439
Willy Tarreau74ca5042013-06-11 23:12:07 +020015440src_bytes_in_rate([<table>]) : integer
15441 Returns the average bytes rate from the incoming connection's source address
15442 in the current proxy's stick-table or in the designated stick-table, measured
15443 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015444 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015445
Willy Tarreau74ca5042013-06-11 23:12:07 +020015446src_bytes_out_rate([<table>]) : integer
15447 Returns the average bytes rate to the incoming connection's source address in
15448 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015449 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015450 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015451
Willy Tarreau74ca5042013-06-11 23:12:07 +020015452src_clr_gpc0([<table>]) : integer
15453 Clears the first General Purpose Counter associated to the incoming
15454 connection's source address in the current proxy's stick-table or in the
15455 designated stick-table, and returns its previous value. If the address is not
15456 found, an entry is created and 0 is returned. This is typically used as a
15457 second ACL in an expression in order to mark a connection when a first ACL
15458 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015459
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015460 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015461 # block if 5 consecutive requests continue to come faster than 10 sess
15462 # per second, and reset the counter as soon as the traffic slows down.
15463 acl abuse src_http_req_rate gt 10
15464 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010015465 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015466 tcp-request connection accept if !abuse save
15467 tcp-request connection reject if abuse kill
15468
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015469src_clr_gpc1([<table>]) : integer
15470 Clears the second General Purpose Counter associated to the incoming
15471 connection's source address in the current proxy's stick-table or in the
15472 designated stick-table, and returns its previous value. If the address is not
15473 found, an entry is created and 0 is returned. This is typically used as a
15474 second ACL in an expression in order to mark a connection when a first ACL
15475 was verified.
15476
Willy Tarreau74ca5042013-06-11 23:12:07 +020015477src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015478 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020015479 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015480 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015481 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015482
Willy Tarreau74ca5042013-06-11 23:12:07 +020015483src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015484 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015485 current incoming connection's source address in the current proxy's
15486 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015487 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015488
Willy Tarreau74ca5042013-06-11 23:12:07 +020015489src_conn_rate([<table>]) : integer
15490 Returns the average connection rate from the incoming connection's source
15491 address in the current proxy's stick-table or in the designated stick-table,
15492 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015493 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015494
Willy Tarreau74ca5042013-06-11 23:12:07 +020015495src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015496 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015497 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015498 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015499 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015500
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015501src_get_gpc1([<table>]) : integer
15502 Returns the value of the second General Purpose Counter associated to the
15503 incoming connection's source address in the current proxy's stick-table or in
15504 the designated stick-table. If the address is not found, zero is returned.
15505 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
15506
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015507src_get_gpt0([<table>]) : integer
15508 Returns the value of the first General Purpose Tag associated to the
15509 incoming connection's source address in the current proxy's stick-table or in
15510 the designated stick-table. If the address is not found, zero is returned.
15511 See also sc/sc0/sc1/sc2_get_gpt0.
15512
Willy Tarreau74ca5042013-06-11 23:12:07 +020015513src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015514 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020015515 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015516 stick-table or in the designated stick-table. It reports the frequency
15517 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015518 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15519 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15520 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015521
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015522src_gpc1_rate([<table>]) : integer
15523 Returns the average increment rate of the second General Purpose Counter
15524 associated to the incoming connection's source address in the current proxy's
15525 stick-table or in the designated stick-table. It reports the frequency
15526 which the gpc1 counter was incremented over the configured period. See also
15527 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15528 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15529 be returned, as "gpc1" only holds the event count.
15530
Willy Tarreau74ca5042013-06-11 23:12:07 +020015531src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015532 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015533 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020015534 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015535 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020015536 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015537
Willy Tarreau74ca5042013-06-11 23:12:07 +020015538src_http_err_rate([<table>]) : integer
15539 Returns the average rate of HTTP errors from the incoming connection's source
15540 address in the current proxy's stick-table or in the designated stick-table,
15541 measured in amount of errors over the period configured in the table. This
15542 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015543 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015544
Willy Tarreau74ca5042013-06-11 23:12:07 +020015545src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015546 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015547 source address in the current proxy's stick-table or in the designated stick-
15548 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015549 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015550
Willy Tarreau74ca5042013-06-11 23:12:07 +020015551src_http_req_rate([<table>]) : integer
15552 Returns the average rate of HTTP requests from the incoming connection's
15553 source address in the current proxy's stick-table or in the designated stick-
15554 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020015555 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015556 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015557
Willy Tarreau74ca5042013-06-11 23:12:07 +020015558src_inc_gpc0([<table>]) : integer
15559 Increments the first General Purpose Counter associated to the incoming
15560 connection's source address in the current proxy's stick-table or in the
15561 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015562 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015563 This is typically used as a second ACL in an expression in order to mark a
15564 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020015565
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015566 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020015567 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010015568 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015569 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020015570
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015571src_inc_gpc1([<table>]) : integer
15572 Increments the second General Purpose Counter associated to the incoming
15573 connection's source address in the current proxy's stick-table or in the
15574 designated stick-table, and returns its new value. If the address is not
15575 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
15576 This is typically used as a second ACL in an expression in order to mark a
15577 connection when a first ACL was verified.
15578
Willy Tarreau16e01562016-08-09 16:46:18 +020015579src_is_local : boolean
15580 Returns true if the source address of the incoming connection is local to the
15581 system, or false if the address doesn't exist on the system, meaning that it
15582 comes from a remote machine. Note that UNIX addresses are considered local.
15583 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015584 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020015585 note that the check involves a few system calls, so it's better to do it only
15586 once per connection.
15587
Willy Tarreau74ca5042013-06-11 23:12:07 +020015588src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015589 Returns the total amount of data received from the incoming connection's
15590 source address in the current proxy's stick-table or in the designated
15591 stick-table, measured in kilobytes. If the address is not found, zero is
15592 returned. The test is currently performed on 32-bit integers, which limits
15593 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015594
Willy Tarreau74ca5042013-06-11 23:12:07 +020015595src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015596 Returns the total amount of data sent to the incoming connection's source
15597 address in the current proxy's stick-table or in the designated stick-table,
15598 measured in kilobytes. If the address is not found, zero is returned. The
15599 test is currently performed on 32-bit integers, which limits values to 4
15600 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015601
Willy Tarreau74ca5042013-06-11 23:12:07 +020015602src_port : integer
15603 Returns an integer value corresponding to the TCP source port of the
15604 connection on the client side, which is the port the client connected from.
15605 Usage of this function is very limited as modern protocols do not care much
15606 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010015607
Willy Tarreau74ca5042013-06-11 23:12:07 +020015608src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015609 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020015610 connection's source IPv4 address in the current proxy's stick-table or in the
15611 designated stick-table, that were transformed into sessions, which means that
15612 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015613 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015614
Willy Tarreau74ca5042013-06-11 23:12:07 +020015615src_sess_rate([<table>]) : integer
15616 Returns the average session rate from the incoming connection's source
15617 address in the current proxy's stick-table or in the designated stick-table,
15618 measured in amount of sessions over the period configured in the table. A
15619 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015620 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015621
Willy Tarreau74ca5042013-06-11 23:12:07 +020015622src_updt_conn_cnt([<table>]) : integer
15623 Creates or updates the entry associated to the incoming connection's source
15624 address in the current proxy's stick-table or in the designated stick-table.
15625 This table must be configured to store the "conn_cnt" data type, otherwise
15626 the match will be ignored. The current count is incremented by one, and the
15627 expiration timer refreshed. The updated count is returned, so this match
15628 can't return zero. This was used to reject service abusers based on their
15629 source address. Note: it is recommended to use the more complete "track-sc*"
15630 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015631
15632 Example :
15633 # This frontend limits incoming SSH connections to 3 per 10 second for
15634 # each source address, and rejects excess connections until a 10 second
15635 # silence is observed. At most 20 addresses are tracked.
15636 listen ssh
15637 bind :22
15638 mode tcp
15639 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020015640 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020015641 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015642 server local 127.0.0.1:22
15643
Willy Tarreau74ca5042013-06-11 23:12:07 +020015644srv_id : integer
15645 Returns an integer containing the server's id when processing the response.
15646 While it's almost only used with ACLs, it may be used for logging or
15647 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020015648
vkill1dfd1652019-10-30 16:58:14 +080015649srv_name : string
15650 Returns a string containing the server's name when processing the response.
15651 While it's almost only used with ACLs, it may be used for logging or
15652 debugging.
15653
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200156547.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020015655----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020015656
Willy Tarreau74ca5042013-06-11 23:12:07 +020015657The layer 5 usually describes just the session layer which in haproxy is
15658closest to the session once all the connection handshakes are finished, but
15659when no content is yet made available. The fetch methods described here are
15660usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015661future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020015662
Ben Shillitof25e8e52016-12-02 14:25:37 +00001566351d.all(<prop>[,<prop>*]) : string
15664 Returns values for the properties requested as a string, where values are
15665 separated by the delimiter specified with "51degrees-property-separator".
15666 The device is identified using all the important HTTP headers from the
15667 request. The function can be passed up to five property names, and if a
15668 property name can't be found, the value "NoData" is returned.
15669
15670 Example :
15671 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
15672 # containing the three properties requested using all relevant headers from
15673 # the request.
15674 frontend http-in
15675 bind *:8081
15676 default_backend servers
15677 http-request set-header X-51D-DeviceTypeMobileTablet \
15678 %[51d.all(DeviceType,IsMobile,IsTablet)]
15679
Emeric Brun645ae792014-04-30 14:21:06 +020015680ssl_bc : boolean
15681 Returns true when the back connection was made via an SSL/TLS transport
15682 layer and is locally deciphered. This means the outgoing connection was made
15683 other a server with the "ssl" option.
15684
15685ssl_bc_alg_keysize : integer
15686 Returns the symmetric cipher key size supported in bits when the outgoing
15687 connection was made over an SSL/TLS transport layer.
15688
Olivier Houchard6b77f492018-11-22 18:18:29 +010015689ssl_bc_alpn : string
15690 This extracts the Application Layer Protocol Negotiation field from an
15691 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020015692 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010015693 server. The SSL library must have been built with support for TLS
15694 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15695 not advertised unless the "alpn" keyword on the "server" line specifies a
15696 protocol list. Also, nothing forces the server to pick a protocol from this
15697 list, any other one may be requested. The TLS ALPN extension is meant to
15698 replace the TLS NPN extension. See also "ssl_bc_npn".
15699
Emeric Brun645ae792014-04-30 14:21:06 +020015700ssl_bc_cipher : string
15701 Returns the name of the used cipher when the outgoing connection was made
15702 over an SSL/TLS transport layer.
15703
Patrick Hemmer65674662019-06-04 08:13:03 -040015704ssl_bc_client_random : binary
15705 Returns the client random of the back connection when the incoming connection
15706 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15707 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15708
Emeric Brun74f7ffa2018-02-19 16:14:12 +010015709ssl_bc_is_resumed : boolean
15710 Returns true when the back connection was made over an SSL/TLS transport
15711 layer and the newly created SSL session was resumed using a cached
15712 session or a TLS ticket.
15713
Olivier Houchard6b77f492018-11-22 18:18:29 +010015714ssl_bc_npn : string
15715 This extracts the Next Protocol Negotiation field from an outgoing connection
15716 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020015717 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010015718 built with support for TLS extensions enabled (check haproxy -vv). Note that
15719 the TLS NPN extension is not advertised unless the "npn" keyword on the
15720 "server" line specifies a protocol list. Also, nothing forces the server to
15721 pick a protocol from this list, any other one may be used. Please note that
15722 the TLS NPN extension was replaced with ALPN.
15723
Emeric Brun645ae792014-04-30 14:21:06 +020015724ssl_bc_protocol : string
15725 Returns the name of the used protocol when the outgoing connection was made
15726 over an SSL/TLS transport layer.
15727
Emeric Brunb73a9b02014-04-30 18:49:19 +020015728ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020015729 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015730 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15731 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020015732
Patrick Hemmer65674662019-06-04 08:13:03 -040015733ssl_bc_server_random : binary
15734 Returns the server random of the back connection when the incoming connection
15735 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15736 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15737
Emeric Brun645ae792014-04-30 14:21:06 +020015738ssl_bc_session_id : binary
15739 Returns the SSL ID of the back connection when the outgoing connection was
15740 made over an SSL/TLS transport layer. It is useful to log if we want to know
15741 if session was reused or not.
15742
Patrick Hemmere0275472018-04-28 19:15:51 -040015743ssl_bc_session_key : binary
15744 Returns the SSL session master key of the back connection when the outgoing
15745 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15746 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15747 BoringSSL.
15748
Emeric Brun645ae792014-04-30 14:21:06 +020015749ssl_bc_use_keysize : integer
15750 Returns the symmetric cipher key size used in bits when the outgoing
15751 connection was made over an SSL/TLS transport layer.
15752
Willy Tarreau74ca5042013-06-11 23:12:07 +020015753ssl_c_ca_err : integer
15754 When the incoming connection was made over an SSL/TLS transport layer,
15755 returns the ID of the first error detected during verification of the client
15756 certificate at depth > 0, or 0 if no error was encountered during this
15757 verification process. Please refer to your SSL library's documentation to
15758 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020015759
Willy Tarreau74ca5042013-06-11 23:12:07 +020015760ssl_c_ca_err_depth : integer
15761 When the incoming connection was made over an SSL/TLS transport layer,
15762 returns the depth in the CA chain of the first error detected during the
15763 verification of the client certificate. If no error is encountered, 0 is
15764 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015765
Emeric Brun43e79582014-10-29 19:03:26 +010015766ssl_c_der : binary
15767 Returns the DER formatted certificate presented by the client when the
15768 incoming connection was made over an SSL/TLS transport layer. When used for
15769 an ACL, the value(s) to match against can be passed in hexadecimal form.
15770
Willy Tarreau74ca5042013-06-11 23:12:07 +020015771ssl_c_err : integer
15772 When the incoming connection was made over an SSL/TLS transport layer,
15773 returns the ID of the first error detected during verification at depth 0, or
15774 0 if no error was encountered during this verification process. Please refer
15775 to your SSL library's documentation to find the exhaustive list of error
15776 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020015777
Elliot Otchet71f82972020-01-15 08:12:14 -050015778ssl_c_i_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020015779 When the incoming connection was made over an SSL/TLS transport layer,
15780 returns the full distinguished name of the issuer of the certificate
15781 presented by the client when no <entry> is specified, or the value of the
15782 first given entry found from the beginning of the DN. If a positive/negative
15783 occurrence number is specified as the optional second argument, it returns
15784 the value of the nth given entry value from the beginning/end of the DN.
15785 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
15786 "ssl_c_i_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050015787 The <format> parameter allows you to receive the DN suitable for
15788 consumption by different protocols. Currently supported is rfc2253 for
15789 LDAP v3.
15790 If you'd like to modify the format only you can specify an empty string
15791 and zero for the first two parameters. Example: ssl_c_i_dn(,0,rfc2253)
Willy Tarreau62644772008-07-16 18:36:06 +020015792
Willy Tarreau74ca5042013-06-11 23:12:07 +020015793ssl_c_key_alg : string
15794 Returns the name of the algorithm used to generate the key of the certificate
15795 presented by the client when the incoming connection was made over an SSL/TLS
15796 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020015797
Willy Tarreau74ca5042013-06-11 23:12:07 +020015798ssl_c_notafter : string
15799 Returns the end date presented by the client as a formatted string
15800 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15801 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020015802
Willy Tarreau74ca5042013-06-11 23:12:07 +020015803ssl_c_notbefore : string
15804 Returns the start date presented by the client as a formatted string
15805 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15806 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015807
Elliot Otchet71f82972020-01-15 08:12:14 -050015808ssl_c_s_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020015809 When the incoming connection was made over an SSL/TLS transport layer,
15810 returns the full distinguished name of the subject of the certificate
15811 presented by the client when no <entry> is specified, or the value of the
15812 first given entry found from the beginning of the DN. If a positive/negative
15813 occurrence number is specified as the optional second argument, it returns
15814 the value of the nth given entry value from the beginning/end of the DN.
15815 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
15816 "ssl_c_s_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050015817 The <format> parameter allows you to receive the DN suitable for
15818 consumption by different protocols. Currently supported is rfc2253 for
15819 LDAP v3.
15820 If you'd like to modify the format only you can specify an empty string
15821 and zero for the first two parameters. Example: ssl_c_s_dn(,0,rfc2253)
Willy Tarreaub6672b52011-12-12 17:23:41 +010015822
Willy Tarreau74ca5042013-06-11 23:12:07 +020015823ssl_c_serial : binary
15824 Returns the serial of the certificate presented by the client when the
15825 incoming connection was made over an SSL/TLS transport layer. When used for
15826 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015827
Willy Tarreau74ca5042013-06-11 23:12:07 +020015828ssl_c_sha1 : binary
15829 Returns the SHA-1 fingerprint of the certificate presented by the client when
15830 the incoming connection was made over an SSL/TLS transport layer. This can be
15831 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015832 Note that the output is binary, so if you want to pass that signature to the
15833 server, you need to encode it in hex or base64, such as in the example below:
15834
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015835 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015836 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020015837
Willy Tarreau74ca5042013-06-11 23:12:07 +020015838ssl_c_sig_alg : string
15839 Returns the name of the algorithm used to sign the certificate presented by
15840 the client when the incoming connection was made over an SSL/TLS transport
15841 layer.
Emeric Brun87855892012-10-17 17:39:35 +020015842
Willy Tarreau74ca5042013-06-11 23:12:07 +020015843ssl_c_used : boolean
15844 Returns true if current SSL session uses a client certificate even if current
15845 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020015846
Willy Tarreau74ca5042013-06-11 23:12:07 +020015847ssl_c_verify : integer
15848 Returns the verify result error ID when the incoming connection was made over
15849 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
15850 refer to your SSL library's documentation for an exhaustive list of error
15851 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020015852
Willy Tarreau74ca5042013-06-11 23:12:07 +020015853ssl_c_version : integer
15854 Returns the version of the certificate presented by the client when the
15855 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020015856
Emeric Brun43e79582014-10-29 19:03:26 +010015857ssl_f_der : binary
15858 Returns the DER formatted certificate presented by the frontend when the
15859 incoming connection was made over an SSL/TLS transport layer. When used for
15860 an ACL, the value(s) to match against can be passed in hexadecimal form.
15861
Elliot Otchet71f82972020-01-15 08:12:14 -050015862ssl_f_i_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020015863 When the incoming connection was made over an SSL/TLS transport layer,
15864 returns the full distinguished name of the issuer of the certificate
15865 presented by the frontend when no <entry> is specified, or the value of the
15866 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020015867 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020015868 the value of the nth given entry value from the beginning/end of the DN.
15869 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
15870 "ssl_f_i_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050015871 The <format> parameter allows you to receive the DN suitable for
15872 consumption by different protocols. Currently supported is rfc2253 for
15873 LDAP v3.
15874 If you'd like to modify the format only you can specify an empty string
15875 and zero for the first two parameters. Example: ssl_f_i_dn(,0,rfc2253)
Emeric Brun87855892012-10-17 17:39:35 +020015876
Willy Tarreau74ca5042013-06-11 23:12:07 +020015877ssl_f_key_alg : string
15878 Returns the name of the algorithm used to generate the key of the certificate
15879 presented by the frontend when the incoming connection was made over an
15880 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015881
Willy Tarreau74ca5042013-06-11 23:12:07 +020015882ssl_f_notafter : string
15883 Returns the end date presented by the frontend as a formatted string
15884 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15885 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015886
Willy Tarreau74ca5042013-06-11 23:12:07 +020015887ssl_f_notbefore : string
15888 Returns the start date presented by the frontend as a formatted string
15889 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15890 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020015891
Elliot Otchet71f82972020-01-15 08:12:14 -050015892ssl_f_s_dn([<entry>[,<occ>[,<format>]]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020015893 When the incoming connection was made over an SSL/TLS transport layer,
15894 returns the full distinguished name of the subject of the certificate
15895 presented by the frontend when no <entry> is specified, or the value of the
15896 first given entry found from the beginning of the DN. If a positive/negative
15897 occurrence number is specified as the optional second argument, it returns
15898 the value of the nth given entry value from the beginning/end of the DN.
15899 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
15900 "ssl_f_s_dn(CN)" retrieves the common name.
Elliot Otchet71f82972020-01-15 08:12:14 -050015901 The <format> parameter allows you to receive the DN suitable for
15902 consumption by different protocols. Currently supported is rfc2253 for
15903 LDAP v3.
15904 If you'd like to modify the format only you can specify an empty string
15905 and zero for the first two parameters. Example: ssl_f_s_dn(,0,rfc2253)
Emeric Brunce5ad802012-10-22 14:11:22 +020015906
Willy Tarreau74ca5042013-06-11 23:12:07 +020015907ssl_f_serial : binary
15908 Returns the serial of the certificate presented by the frontend when the
15909 incoming connection was made over an SSL/TLS transport layer. When used for
15910 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020015911
Emeric Brun55f4fa82014-04-30 17:11:25 +020015912ssl_f_sha1 : binary
15913 Returns the SHA-1 fingerprint of the certificate presented by the frontend
15914 when the incoming connection was made over an SSL/TLS transport layer. This
15915 can be used to know which certificate was chosen using SNI.
15916
Willy Tarreau74ca5042013-06-11 23:12:07 +020015917ssl_f_sig_alg : string
15918 Returns the name of the algorithm used to sign the certificate presented by
15919 the frontend when the incoming connection was made over an SSL/TLS transport
15920 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015921
Willy Tarreau74ca5042013-06-11 23:12:07 +020015922ssl_f_version : integer
15923 Returns the version of the certificate presented by the frontend when the
15924 incoming connection was made over an SSL/TLS transport layer.
15925
15926ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015927 Returns true when the front connection was made via an SSL/TLS transport
15928 layer and is locally deciphered. This means it has matched a socket declared
15929 with a "bind" line having the "ssl" option.
15930
Willy Tarreau74ca5042013-06-11 23:12:07 +020015931 Example :
15932 # This passes "X-Proto: https" to servers when client connects over SSL
15933 listen http-https
15934 bind :80
15935 bind :443 ssl crt /etc/haproxy.pem
15936 http-request add-header X-Proto https if { ssl_fc }
15937
15938ssl_fc_alg_keysize : integer
15939 Returns the symmetric cipher key size supported in bits when the incoming
15940 connection was made over an SSL/TLS transport layer.
15941
15942ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015943 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020015944 incoming connection made via a TLS transport layer and locally deciphered by
15945 haproxy. The result is a string containing the protocol name advertised by
15946 the client. The SSL library must have been built with support for TLS
15947 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15948 not advertised unless the "alpn" keyword on the "bind" line specifies a
15949 protocol list. Also, nothing forces the client to pick a protocol from this
15950 list, any other one may be requested. The TLS ALPN extension is meant to
15951 replace the TLS NPN extension. See also "ssl_fc_npn".
15952
Willy Tarreau74ca5042013-06-11 23:12:07 +020015953ssl_fc_cipher : string
15954 Returns the name of the used cipher when the incoming connection was made
15955 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020015956
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015957ssl_fc_cipherlist_bin : binary
15958 Returns the binary form of the client hello cipher list. The maximum returned
15959 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015960 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015961
15962ssl_fc_cipherlist_hex : string
15963 Returns the binary form of the client hello cipher list encoded as
15964 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015965 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015966
15967ssl_fc_cipherlist_str : string
15968 Returns the decoded text form of the client hello cipher list. The maximum
15969 number of ciphers returned is according with the value of
15970 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010015971 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020015972 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015973
15974ssl_fc_cipherlist_xxh : integer
15975 Returns a xxh64 of the cipher list. This hash can be return only is the value
15976 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015977 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015978
Patrick Hemmer65674662019-06-04 08:13:03 -040015979ssl_fc_client_random : binary
15980 Returns the client random of the front connection when the incoming connection
15981 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15982 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15983
Willy Tarreau74ca5042013-06-11 23:12:07 +020015984ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015985 Returns true if a client certificate is present in an incoming connection over
15986 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010015987 Note: on SSL session resumption with Session ID or TLS ticket, client
15988 certificate is not present in the current connection but may be retrieved
15989 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
15990 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015991
Olivier Houchardccaa7de2017-10-02 11:51:03 +020015992ssl_fc_has_early : boolean
15993 Returns true if early data were sent, and the handshake didn't happen yet. As
15994 it has security implications, it is useful to be able to refuse those, or
15995 wait until the handshake happened.
15996
Willy Tarreau74ca5042013-06-11 23:12:07 +020015997ssl_fc_has_sni : boolean
15998 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020015999 in an incoming connection was made over an SSL/TLS transport layer. Returns
16000 true when the incoming connection presents a TLS SNI field. This requires
John Roeslerfb2fce12019-07-10 15:45:51 -050016001 that the SSL library is built with support for TLS extensions enabled (check
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020016002 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020016003
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020016004ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020016005 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010016006 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
16007 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020016008
Willy Tarreau74ca5042013-06-11 23:12:07 +020016009ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016010 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020016011 made via a TLS transport layer and locally deciphered by haproxy. The result
16012 is a string containing the protocol name advertised by the client. The SSL
16013 library must have been built with support for TLS extensions enabled (check
16014 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
16015 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
16016 forces the client to pick a protocol from this list, any other one may be
16017 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020016018
Willy Tarreau74ca5042013-06-11 23:12:07 +020016019ssl_fc_protocol : string
16020 Returns the name of the used protocol when the incoming connection was made
16021 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020016022
Emeric Brunb73a9b02014-04-30 18:49:19 +020016023ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040016024 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020016025 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
16026 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040016027
Patrick Hemmer65674662019-06-04 08:13:03 -040016028ssl_fc_server_random : binary
16029 Returns the server random of the front connection when the incoming connection
16030 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
16031 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
16032
Willy Tarreau74ca5042013-06-11 23:12:07 +020016033ssl_fc_session_id : binary
16034 Returns the SSL ID of the front connection when the incoming connection was
16035 made over an SSL/TLS transport layer. It is useful to stick a given client to
16036 a server. It is important to note that some browsers refresh their session ID
16037 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020016038
Patrick Hemmere0275472018-04-28 19:15:51 -040016039ssl_fc_session_key : binary
16040 Returns the SSL session master key of the front connection when the incoming
16041 connection was made over an SSL/TLS transport layer. It is useful to decrypt
16042 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
16043 BoringSSL.
16044
16045
Willy Tarreau74ca5042013-06-11 23:12:07 +020016046ssl_fc_sni : string
16047 This extracts the Server Name Indication TLS extension (SNI) field from an
16048 incoming connection made via an SSL/TLS transport layer and locally
16049 deciphered by haproxy. The result (when present) typically is a string
16050 matching the HTTPS host name (253 chars or less). The SSL library must have
16051 been built with support for TLS extensions enabled (check haproxy -vv).
16052
16053 This fetch is different from "req_ssl_sni" above in that it applies to the
16054 connection being deciphered by haproxy and not to SSL contents being blindly
16055 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
John Roeslerfb2fce12019-07-10 15:45:51 -050016056 requires that the SSL library is built with support for TLS extensions
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020016057 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020016058
Willy Tarreau74ca5042013-06-11 23:12:07 +020016059 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020016060 ssl_fc_sni_end : suffix match
16061 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020016062
Willy Tarreau74ca5042013-06-11 23:12:07 +020016063ssl_fc_use_keysize : integer
16064 Returns the symmetric cipher key size used in bits when the incoming
16065 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016066
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016067
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200160687.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016069------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016070
Willy Tarreau74ca5042013-06-11 23:12:07 +020016071Fetching samples from buffer contents is a bit different from the previous
16072sample fetches above because the sampled data are ephemeral. These data can
16073only be used when they're available and will be lost when they're forwarded.
16074For this reason, samples fetched from buffer contents during a request cannot
16075be used in a response for example. Even while the data are being fetched, they
16076can change. Sometimes it is necessary to set some delays or combine multiple
16077sample fetch methods to ensure that the expected data are complete and usable,
16078for example through TCP request content inspection. Please see the "tcp-request
16079content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020016080
Willy Tarreau74ca5042013-06-11 23:12:07 +020016081payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010016082 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016083 "stick on", "stick match"), and for "res.payload" when used in the context of
16084 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010016085
Willy Tarreau74ca5042013-06-11 23:12:07 +020016086payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
16087 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010016088 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020016089 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010016090
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020016091req.hdrs : string
16092 Returns the current request headers as string including the last empty line
16093 separating headers from the request body. The last empty line can be used to
16094 detect a truncated header block. This sample fetch is useful for some SPOE
16095 headers analyzers and for advanced logging.
16096
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020016097req.hdrs_bin : binary
16098 Returns the current request headers contained in preparsed binary form. This
16099 is useful for offloading some processing with SPOE. Each string is described
16100 by a length followed by the number of bytes indicated in the length. The
16101 length is represented using the variable integer encoding detailed in the
16102 SPOE documentation. The end of the list is marked by a couple of empty header
16103 names and values (length of 0 for both).
16104
16105 *(<str:header-name><str:header-value>)<empty string><empty string>
16106
16107 int: refer to the SPOE documentation for the encoding
16108 str: <int:length><bytes>
16109
Willy Tarreau74ca5042013-06-11 23:12:07 +020016110req.len : integer
16111req_len : integer (deprecated)
16112 Returns an integer value corresponding to the number of bytes present in the
16113 request buffer. This is mostly used in ACL. It is important to understand
16114 that this test does not return false as long as the buffer is changing. This
16115 means that a check with equality to zero will almost always immediately match
16116 at the beginning of the session, while a test for more data will wait for
16117 that data to come in and return false only when haproxy is certain that no
16118 more data will come in. This test was designed to be used with TCP request
16119 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016120
Willy Tarreau74ca5042013-06-11 23:12:07 +020016121req.payload(<offset>,<length>) : binary
16122 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020016123 in the request buffer. As a special case, if the <length> argument is zero,
16124 the the whole buffer from <offset> to the end is extracted. This can be used
16125 with ACLs in order to check for the presence of some content in a buffer at
16126 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016127
Willy Tarreau74ca5042013-06-11 23:12:07 +020016128 ACL alternatives :
16129 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016130
Willy Tarreau74ca5042013-06-11 23:12:07 +020016131req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16132 This extracts a binary block whose size is specified at <offset1> for <length>
16133 bytes, and which starts at <offset2> if specified or just after the length in
16134 the request buffer. The <offset2> parameter also supports relative offsets if
16135 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016136
Willy Tarreau74ca5042013-06-11 23:12:07 +020016137 ACL alternatives :
16138 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016139
Willy Tarreau74ca5042013-06-11 23:12:07 +020016140 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016141
Willy Tarreau74ca5042013-06-11 23:12:07 +020016142req.proto_http : boolean
16143req_proto_http : boolean (deprecated)
16144 Returns true when data in the request buffer look like HTTP and correctly
16145 parses as such. It is the same parser as the common HTTP request parser which
16146 is used so there should be no surprises. The test does not match until the
16147 request is complete, failed or timed out. This test may be used to report the
16148 protocol in TCP logs, but the biggest use is to block TCP request analysis
16149 until a complete HTTP request is present in the buffer, for example to track
16150 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016151
Willy Tarreau74ca5042013-06-11 23:12:07 +020016152 Example:
16153 # track request counts per "base" (concatenation of Host+URL)
16154 tcp-request inspect-delay 10s
16155 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020016156 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016157
Willy Tarreau74ca5042013-06-11 23:12:07 +020016158req.rdp_cookie([<name>]) : string
16159rdp_cookie([<name>]) : string (deprecated)
16160 When the request buffer looks like the RDP protocol, extracts the RDP cookie
16161 <name>, or any cookie if unspecified. The parser only checks for the first
16162 cookie, as illustrated in the RDP protocol specification. The cookie name is
16163 case insensitive. Generally the "MSTS" cookie name will be used, as it can
16164 contain the user name of the client connecting to the server if properly
16165 configured on the client. The "MSTSHASH" cookie is often used as well for
16166 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016167
Willy Tarreau74ca5042013-06-11 23:12:07 +020016168 This differs from "balance rdp-cookie" in that any balancing algorithm may be
16169 used and thus the distribution of clients to backend servers is not linked to
16170 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
16171 such as "balance roundrobin" or "balance leastconn" will lead to a more even
16172 distribution of clients to backend servers than the hash used by "balance
16173 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016174
Willy Tarreau74ca5042013-06-11 23:12:07 +020016175 ACL derivatives :
16176 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016177
Willy Tarreau74ca5042013-06-11 23:12:07 +020016178 Example :
16179 listen tse-farm
16180 bind 0.0.0.0:3389
16181 # wait up to 5s for an RDP cookie in the request
16182 tcp-request inspect-delay 5s
16183 tcp-request content accept if RDP_COOKIE
16184 # apply RDP cookie persistence
16185 persist rdp-cookie
16186 # Persist based on the mstshash cookie
16187 # This is only useful makes sense if
16188 # balance rdp-cookie is not used
16189 stick-table type string size 204800
16190 stick on req.rdp_cookie(mstshash)
16191 server srv1 1.1.1.1:3389
16192 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016193
Willy Tarreau74ca5042013-06-11 23:12:07 +020016194 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
16195 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016196
Willy Tarreau74ca5042013-06-11 23:12:07 +020016197req.rdp_cookie_cnt([name]) : integer
16198rdp_cookie_cnt([name]) : integer (deprecated)
16199 Tries to parse the request buffer as RDP protocol, then returns an integer
16200 corresponding to the number of RDP cookies found. If an optional cookie name
16201 is passed, only cookies matching this name are considered. This is mostly
16202 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016203
Willy Tarreau74ca5042013-06-11 23:12:07 +020016204 ACL derivatives :
16205 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016206
Alex Zorin4afdd132018-12-30 13:56:28 +110016207req.ssl_alpn : string
16208 Returns a string containing the values of the Application-Layer Protocol
16209 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
16210 ClientHello message. Note that this only applies to raw contents found in the
16211 request buffer and not to the contents deciphered via an SSL data layer, so
16212 this will not work with "bind" lines having the "ssl" option. This is useful
16213 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020016214 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110016215
16216 Examples :
16217 # Wait for a client hello for at most 5 seconds
16218 tcp-request inspect-delay 5s
16219 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020016220 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110016221 default_backend bk_default
16222
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016223req.ssl_ec_ext : boolean
16224 Returns a boolean identifying if client sent the Supported Elliptic Curves
16225 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020016226 message. This can be used to present ECC compatible clients with EC
16227 certificate and to use RSA for all others, on the same IP address. Note that
16228 this only applies to raw contents found in the request buffer and not to
16229 contents deciphered via an SSL data layer, so this will not work with "bind"
16230 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016231
Willy Tarreau74ca5042013-06-11 23:12:07 +020016232req.ssl_hello_type : integer
16233req_ssl_hello_type : integer (deprecated)
16234 Returns an integer value containing the type of the SSL hello message found
16235 in the request buffer if the buffer contains data that parse as a complete
16236 SSL (v3 or superior) client hello message. Note that this only applies to raw
16237 contents found in the request buffer and not to contents deciphered via an
16238 SSL data layer, so this will not work with "bind" lines having the "ssl"
16239 option. This is mostly used in ACL to detect presence of an SSL hello message
16240 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016241
Willy Tarreau74ca5042013-06-11 23:12:07 +020016242req.ssl_sni : string
16243req_ssl_sni : string (deprecated)
16244 Returns a string containing the value of the Server Name TLS extension sent
16245 by a client in a TLS stream passing through the request buffer if the buffer
16246 contains data that parse as a complete SSL (v3 or superior) client hello
16247 message. Note that this only applies to raw contents found in the request
16248 buffer and not to contents deciphered via an SSL data layer, so this will not
16249 work with "bind" lines having the "ssl" option. SNI normally contains the
16250 name of the host the client tries to connect to (for recent browsers). SNI is
16251 useful for allowing or denying access to certain hosts when SSL/TLS is used
16252 by the client. This test was designed to be used with TCP request content
16253 inspection. If content switching is needed, it is recommended to first wait
16254 for a complete client hello (type 1), like in the example below. See also
16255 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016256
Willy Tarreau74ca5042013-06-11 23:12:07 +020016257 ACL derivatives :
16258 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016259
Willy Tarreau74ca5042013-06-11 23:12:07 +020016260 Examples :
16261 # Wait for a client hello for at most 5 seconds
16262 tcp-request inspect-delay 5s
16263 tcp-request content accept if { req_ssl_hello_type 1 }
16264 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
16265 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016266
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053016267req.ssl_st_ext : integer
16268 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
16269 Returns 1 if the client sent SessionTicket TLS Extension
16270 Returns 2 if the client also sent non-zero length TLS SessionTicket
16271 Note that this only applies to raw contents found in the request buffer and
16272 not to contents deciphered via an SSL data layer, so this will not work with
16273 "bind" lines having the "ssl" option. This can for example be used to detect
16274 whether the client sent a SessionTicket or not and stick it accordingly, if
16275 no SessionTicket then stick on SessionID or don't stick as there's no server
16276 side state is there when SessionTickets are in use.
16277
Willy Tarreau74ca5042013-06-11 23:12:07 +020016278req.ssl_ver : integer
16279req_ssl_ver : integer (deprecated)
16280 Returns an integer value containing the version of the SSL/TLS protocol of a
16281 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
16282 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
16283 composed of the major version multiplied by 65536, added to the minor
16284 version. Note that this only applies to raw contents found in the request
16285 buffer and not to contents deciphered via an SSL data layer, so this will not
16286 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010016287 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020016288 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016289
Willy Tarreau74ca5042013-06-11 23:12:07 +020016290 ACL derivatives :
16291 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016292
Willy Tarreau47e8eba2013-09-11 23:28:46 +020016293res.len : integer
16294 Returns an integer value corresponding to the number of bytes present in the
16295 response buffer. This is mostly used in ACL. It is important to understand
16296 that this test does not return false as long as the buffer is changing. This
16297 means that a check with equality to zero will almost always immediately match
16298 at the beginning of the session, while a test for more data will wait for
16299 that data to come in and return false only when haproxy is certain that no
16300 more data will come in. This test was designed to be used with TCP response
16301 content inspection.
16302
Willy Tarreau74ca5042013-06-11 23:12:07 +020016303res.payload(<offset>,<length>) : binary
16304 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020016305 in the response buffer. As a special case, if the <length> argument is zero,
16306 the the whole buffer from <offset> to the end is extracted. This can be used
16307 with ACLs in order to check for the presence of some content in a buffer at
16308 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016309
Willy Tarreau74ca5042013-06-11 23:12:07 +020016310res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16311 This extracts a binary block whose size is specified at <offset1> for <length>
16312 bytes, and which starts at <offset2> if specified or just after the length in
16313 the response buffer. The <offset2> parameter also supports relative offsets
16314 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016315
Willy Tarreau74ca5042013-06-11 23:12:07 +020016316 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016317
Willy Tarreau971f7b62015-09-29 14:06:59 +020016318res.ssl_hello_type : integer
16319rep_ssl_hello_type : integer (deprecated)
16320 Returns an integer value containing the type of the SSL hello message found
16321 in the response buffer if the buffer contains data that parses as a complete
16322 SSL (v3 or superior) hello message. Note that this only applies to raw
16323 contents found in the response buffer and not to contents deciphered via an
16324 SSL data layer, so this will not work with "server" lines having the "ssl"
16325 option. This is mostly used in ACL to detect presence of an SSL hello message
16326 that is supposed to contain an SSL session ID usable for stickiness.
16327
Willy Tarreau74ca5042013-06-11 23:12:07 +020016328wait_end : boolean
16329 This fetch either returns true when the inspection period is over, or does
16330 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016331 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020016332 actions, such as a delayed reject for some special addresses. Since it either
16333 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016334 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020016335 "WAIT_END" is always usable without prior declaration. This test was designed
16336 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016337
Willy Tarreau74ca5042013-06-11 23:12:07 +020016338 Examples :
16339 # delay every incoming request by 2 seconds
16340 tcp-request inspect-delay 2s
16341 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010016342
Willy Tarreau74ca5042013-06-11 23:12:07 +020016343 # don't immediately tell bad guys they are rejected
16344 tcp-request inspect-delay 10s
16345 acl goodguys src 10.0.0.0/24
16346 acl badguys src 10.0.1.0/24
16347 tcp-request content accept if goodguys
16348 tcp-request content reject if badguys WAIT_END
16349 tcp-request content reject
16350
16351
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200163527.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016353--------------------------------------
16354
16355It is possible to fetch samples from HTTP contents, requests and responses.
16356This application layer is also called layer 7. It is only possible to fetch the
16357data in this section when a full HTTP request or response has been parsed from
16358its respective request or response buffer. This is always the case with all
16359HTTP specific rules and for sections running with "mode http". When using TCP
16360content inspection, it may be necessary to support an inspection delay in order
16361to let the request or response come in first. These fetches may require a bit
16362more CPU resources than the layer 4 ones, but not much since the request and
16363response are indexed.
16364
16365base : string
16366 This returns the concatenation of the first Host header and the path part of
16367 the request, which starts at the first slash and ends before the question
16368 mark. It can be useful in virtual hosted environments to detect URL abuses as
16369 well as to improve shared caches efficiency. Using this with a limited size
16370 stick table also allows one to collect statistics about most commonly
16371 requested objects by host/path. With ACLs it can allow simple content
16372 switching rules involving the host and the path at the same time, such as
16373 "www.example.com/favicon.ico". See also "path" and "uri".
16374
16375 ACL derivatives :
16376 base : exact string match
16377 base_beg : prefix match
16378 base_dir : subdir match
16379 base_dom : domain match
16380 base_end : suffix match
16381 base_len : length match
16382 base_reg : regex match
16383 base_sub : substring match
16384
16385base32 : integer
16386 This returns a 32-bit hash of the value returned by the "base" fetch method
16387 above. This is useful to track per-URL activity on high traffic sites without
16388 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016389 memory. The output type is an unsigned integer. The hash function used is
16390 SDBM with full avalanche on the output. Technically, base32 is exactly equal
16391 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020016392
16393base32+src : binary
16394 This returns the concatenation of the base32 fetch above and the src fetch
16395 below. The resulting type is of type binary, with a size of 8 or 20 bytes
16396 depending on the source address family. This can be used to track per-IP,
16397 per-URL counters.
16398
William Lallemand65ad6e12014-01-31 15:08:02 +010016399capture.req.hdr(<idx>) : string
16400 This extracts the content of the header captured by the "capture request
16401 header", idx is the position of the capture keyword in the configuration.
16402 The first entry is an index of 0. See also: "capture request header".
16403
16404capture.req.method : string
16405 This extracts the METHOD of an HTTP request. It can be used in both request
16406 and response. Unlike "method", it can be used in both request and response
16407 because it's allocated.
16408
16409capture.req.uri : string
16410 This extracts the request's URI, which starts at the first slash and ends
16411 before the first space in the request (without the host part). Unlike "path"
16412 and "url", it can be used in both request and response because it's
16413 allocated.
16414
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016415capture.req.ver : string
16416 This extracts the request's HTTP version and returns either "HTTP/1.0" or
16417 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
16418 logs because it relies on a persistent flag.
16419
William Lallemand65ad6e12014-01-31 15:08:02 +010016420capture.res.hdr(<idx>) : string
16421 This extracts the content of the header captured by the "capture response
16422 header", idx is the position of the capture keyword in the configuration.
16423 The first entry is an index of 0.
16424 See also: "capture response header"
16425
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016426capture.res.ver : string
16427 This extracts the response's HTTP version and returns either "HTTP/1.0" or
16428 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
16429 persistent flag.
16430
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016431req.body : binary
16432 This returns the HTTP request's available body as a block of data. It
16433 requires that the request body has been buffered made available using
16434 "option http-buffer-request". In case of chunked-encoded body, currently only
16435 the first chunk is analyzed.
16436
Thierry FOURNIER9826c772015-05-20 15:50:54 +020016437req.body_param([<name>) : string
16438 This fetch assumes that the body of the POST request is url-encoded. The user
16439 can check if the "content-type" contains the value
16440 "application/x-www-form-urlencoded". This extracts the first occurrence of the
16441 parameter <name> in the body, which ends before '&'. The parameter name is
16442 case-sensitive. If no name is given, any parameter will match, and the first
16443 one will be returned. The result is a string corresponding to the value of the
16444 parameter <name> as presented in the request body (no URL decoding is
16445 performed). Note that the ACL version of this fetch iterates over multiple
16446 parameters and will iteratively report all parameters values if no name is
16447 given.
16448
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016449req.body_len : integer
16450 This returns the length of the HTTP request's available body in bytes. It may
16451 be lower than the advertised length if the body is larger than the buffer. It
16452 requires that the request body has been buffered made available using
16453 "option http-buffer-request".
16454
16455req.body_size : integer
16456 This returns the advertised length of the HTTP request's body in bytes. It
16457 will represent the advertised Content-Length header, or the size of the first
16458 chunk in case of chunked encoding. In order to parse the chunks, it requires
16459 that the request body has been buffered made available using
16460 "option http-buffer-request".
16461
Willy Tarreau74ca5042013-06-11 23:12:07 +020016462req.cook([<name>]) : string
16463cook([<name>]) : string (deprecated)
16464 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16465 header line from the request, and returns its value as string. If no name is
16466 specified, the first cookie value is returned. When used with ACLs, all
16467 matching cookies are evaluated. Spaces around the name and the value are
16468 ignored as requested by the Cookie header specification (RFC6265). The cookie
16469 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
16470 well return an empty value if it is present. Use the "found" match to detect
16471 presence. Use the res.cook() variant for response cookies sent by the server.
16472
16473 ACL derivatives :
16474 cook([<name>]) : exact string match
16475 cook_beg([<name>]) : prefix match
16476 cook_dir([<name>]) : subdir match
16477 cook_dom([<name>]) : domain match
16478 cook_end([<name>]) : suffix match
16479 cook_len([<name>]) : length match
16480 cook_reg([<name>]) : regex match
16481 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016482
Willy Tarreau74ca5042013-06-11 23:12:07 +020016483req.cook_cnt([<name>]) : integer
16484cook_cnt([<name>]) : integer (deprecated)
16485 Returns an integer value representing the number of occurrences of the cookie
16486 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016487
Willy Tarreau74ca5042013-06-11 23:12:07 +020016488req.cook_val([<name>]) : integer
16489cook_val([<name>]) : integer (deprecated)
16490 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16491 header line from the request, and converts its value to an integer which is
16492 returned. If no name is specified, the first cookie value is returned. When
16493 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020016494
Willy Tarreau74ca5042013-06-11 23:12:07 +020016495cookie([<name>]) : string (deprecated)
16496 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16497 header line from the request, or a "Set-Cookie" header from the response, and
16498 returns its value as a string. A typical use is to get multiple clients
16499 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020016500 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020016501 multi-peer synchronization and state keeping across restarts. If no name is
16502 specified, the first cookie value is returned. This fetch should not be used
16503 anymore and should be replaced by req.cook() or res.cook() instead as it
16504 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016505
Willy Tarreau74ca5042013-06-11 23:12:07 +020016506hdr([<name>[,<occ>]]) : string
16507 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
16508 used on responses. Please refer to these respective fetches for more details.
16509 In case of doubt about the fetch direction, please use the explicit ones.
16510 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016511 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016512
Willy Tarreau74ca5042013-06-11 23:12:07 +020016513req.fhdr(<name>[,<occ>]) : string
16514 This extracts the last occurrence of header <name> in an HTTP request. When
16515 used from an ACL, all occurrences are iterated over until a match is found.
16516 Optionally, a specific occurrence might be specified as a position number.
16517 Positive values indicate a position from the first occurrence, with 1 being
16518 the first one. Negative values indicate positions relative to the last one,
16519 with -1 being the last one. It differs from req.hdr() in that any commas
16520 present in the value are returned and are not used as delimiters. This is
16521 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016522
Willy Tarreau74ca5042013-06-11 23:12:07 +020016523req.fhdr_cnt([<name>]) : integer
16524 Returns an integer value representing the number of occurrences of request
16525 header field name <name>, or the total number of header fields if <name> is
16526 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
16527 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016528
Willy Tarreau74ca5042013-06-11 23:12:07 +020016529req.hdr([<name>[,<occ>]]) : string
16530 This extracts the last occurrence of header <name> in an HTTP request. When
16531 used from an ACL, all occurrences are iterated over until a match is found.
16532 Optionally, a specific occurrence might be specified as a position number.
16533 Positive values indicate a position from the first occurrence, with 1 being
16534 the first one. Negative values indicate positions relative to the last one,
16535 with -1 being the last one. A typical use is with the X-Forwarded-For header
16536 once converted to IP, associated with an IP stick-table. The function
16537 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000016538 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020016539 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010016540 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010016541
Willy Tarreau74ca5042013-06-11 23:12:07 +020016542 ACL derivatives :
16543 hdr([<name>[,<occ>]]) : exact string match
16544 hdr_beg([<name>[,<occ>]]) : prefix match
16545 hdr_dir([<name>[,<occ>]]) : subdir match
16546 hdr_dom([<name>[,<occ>]]) : domain match
16547 hdr_end([<name>[,<occ>]]) : suffix match
16548 hdr_len([<name>[,<occ>]]) : length match
16549 hdr_reg([<name>[,<occ>]]) : regex match
16550 hdr_sub([<name>[,<occ>]]) : substring match
16551
16552req.hdr_cnt([<name>]) : integer
16553hdr_cnt([<header>]) : integer (deprecated)
16554 Returns an integer value representing the number of occurrences of request
16555 header field name <name>, or the total number of header field values if
16556 <name> is not specified. It is important to remember that one header line may
16557 count as several headers if it has several values. The function considers any
16558 comma as a delimiter for distinct values. If full-line headers are desired
16559 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
16560 detect presence, absence or abuse of a specific header, as well as to block
16561 request smuggling attacks by rejecting requests which contain more than one
16562 of certain headers. See "req.hdr" for more information on header matching.
16563
16564req.hdr_ip([<name>[,<occ>]]) : ip
16565hdr_ip([<name>[,<occ>]]) : ip (deprecated)
16566 This extracts the last occurrence of header <name> in an HTTP request,
16567 converts it to an IPv4 or IPv6 address and returns this address. When used
16568 with ACLs, all occurrences are checked, and if <name> is omitted, every value
16569 of every header is checked. Optionally, a specific occurrence might be
16570 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010016571 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020016572 positions relative to the last one, with -1 being the last one. A typical use
16573 is with the X-Forwarded-For and X-Client-IP headers.
16574
16575req.hdr_val([<name>[,<occ>]]) : integer
16576hdr_val([<name>[,<occ>]]) : integer (deprecated)
16577 This extracts the last occurrence of header <name> in an HTTP request, and
16578 converts it to an integer value. When used with ACLs, all occurrences are
16579 checked, and if <name> is omitted, every value of every header is checked.
16580 Optionally, a specific occurrence might be specified as a position number.
16581 Positive values indicate a position from the first occurrence, with 1 being
16582 the first one. Negative values indicate positions relative to the last one,
16583 with -1 being the last one. A typical use is with the X-Forwarded-For header.
16584
Frédéric Lécailleec891192019-02-26 15:02:35 +010016585
16586
Willy Tarreau74ca5042013-06-11 23:12:07 +020016587http_auth(<userlist>) : boolean
16588 Returns a boolean indicating whether the authentication data received from
16589 the client match a username & password stored in the specified userlist. This
16590 fetch function is not really useful outside of ACLs. Currently only http
16591 basic auth is supported.
16592
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016593http_auth_group(<userlist>) : string
16594 Returns a string corresponding to the user name found in the authentication
16595 data received from the client if both the user name and password are valid
16596 according to the specified userlist. The main purpose is to use it in ACLs
16597 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016598 This fetch function is not really useful outside of ACLs. Currently only http
16599 basic auth is supported.
16600
16601 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016602 http_auth_group(<userlist>) : group ...
16603 Returns true when the user extracted from the request and whose password is
16604 valid according to the specified userlist belongs to at least one of the
16605 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016606
Christopher Fauleta4063562019-08-02 11:51:37 +020016607http_auth_pass : string
16608 Returns the user's password found in the authentication data received from
16609 the client, as supplied in the Authorization header. Not checks are
16610 performed by this sample fetch. Only Basic authentication is supported.
16611
16612http_auth_type : string
16613 Returns the authentication method found in the authentication data received from
16614 the client, as supplied in the Authorization header. Not checks are
16615 performed by this sample fetch. Only Basic authentication is supported.
16616
16617http_auth_user : string
16618 Returns the user name found in the authentication data received from the
16619 client, as supplied in the Authorization header. Not checks are performed by
16620 this sample fetch. Only Basic authentication is supported.
16621
Willy Tarreau74ca5042013-06-11 23:12:07 +020016622http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020016623 Returns true when the request being processed is the first one of the
16624 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020016625 from some requests when a request is not the first one, or to help grouping
16626 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020016627
Willy Tarreau74ca5042013-06-11 23:12:07 +020016628method : integer + string
16629 Returns an integer value corresponding to the method in the HTTP request. For
16630 example, "GET" equals 1 (check sources to establish the matching). Value 9
16631 means "other method" and may be converted to a string extracted from the
16632 stream. This should not be used directly as a sample, this is only meant to
16633 be used from ACLs, which transparently convert methods from patterns to these
16634 integer + string values. Some predefined ACL already check for most common
16635 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016636
Willy Tarreau74ca5042013-06-11 23:12:07 +020016637 ACL derivatives :
16638 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016639
Willy Tarreau74ca5042013-06-11 23:12:07 +020016640 Example :
16641 # only accept GET and HEAD requests
16642 acl valid_method method GET HEAD
16643 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020016644
Willy Tarreau74ca5042013-06-11 23:12:07 +020016645path : string
16646 This extracts the request's URL path, which starts at the first slash and
16647 ends before the question mark (without the host part). A typical use is with
16648 prefetch-capable caches, and with portals which need to aggregate multiple
16649 information from databases and keep them in caches. Note that with outgoing
16650 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010016651 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020016652 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016653
Willy Tarreau74ca5042013-06-11 23:12:07 +020016654 ACL derivatives :
16655 path : exact string match
16656 path_beg : prefix match
16657 path_dir : subdir match
16658 path_dom : domain match
16659 path_end : suffix match
16660 path_len : length match
16661 path_reg : regex match
16662 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016663
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016664query : string
16665 This extracts the request's query string, which starts after the first
16666 question mark. If no question mark is present, this fetch returns nothing. If
16667 a question mark is present but nothing follows, it returns an empty string.
16668 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010016669 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016670 which stops before the question mark.
16671
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016672req.hdr_names([<delim>]) : string
16673 This builds a string made from the concatenation of all header names as they
16674 appear in the request when the rule is evaluated. The default delimiter is
16675 the comma (',') but it may be overridden as an optional argument <delim>. In
16676 this case, only the first character of <delim> is considered.
16677
Willy Tarreau74ca5042013-06-11 23:12:07 +020016678req.ver : string
16679req_ver : string (deprecated)
16680 Returns the version string from the HTTP request, for example "1.1". This can
16681 be useful for logs, but is mostly there for ACL. Some predefined ACL already
16682 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016683
Willy Tarreau74ca5042013-06-11 23:12:07 +020016684 ACL derivatives :
16685 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020016686
Willy Tarreau74ca5042013-06-11 23:12:07 +020016687res.comp : boolean
16688 Returns the boolean "true" value if the response has been compressed by
16689 HAProxy, otherwise returns boolean "false". This may be used to add
16690 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016691
Willy Tarreau74ca5042013-06-11 23:12:07 +020016692res.comp_algo : string
16693 Returns a string containing the name of the algorithm used if the response
16694 was compressed by HAProxy, for example : "deflate". This may be used to add
16695 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016696
Willy Tarreau74ca5042013-06-11 23:12:07 +020016697res.cook([<name>]) : string
16698scook([<name>]) : string (deprecated)
16699 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16700 header line from the response, and returns its value as string. If no name is
16701 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016702
Willy Tarreau74ca5042013-06-11 23:12:07 +020016703 ACL derivatives :
16704 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016705
Willy Tarreau74ca5042013-06-11 23:12:07 +020016706res.cook_cnt([<name>]) : integer
16707scook_cnt([<name>]) : integer (deprecated)
16708 Returns an integer value representing the number of occurrences of the cookie
16709 <name> in the response, or all cookies if <name> is not specified. This is
16710 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016711
Willy Tarreau74ca5042013-06-11 23:12:07 +020016712res.cook_val([<name>]) : integer
16713scook_val([<name>]) : integer (deprecated)
16714 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16715 header line from the response, and converts its value to an integer which is
16716 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016717
Willy Tarreau74ca5042013-06-11 23:12:07 +020016718res.fhdr([<name>[,<occ>]]) : string
16719 This extracts the last occurrence of header <name> in an HTTP response, or of
16720 the last header if no <name> is specified. Optionally, a specific occurrence
16721 might be specified as a position number. Positive values indicate a position
16722 from the first occurrence, with 1 being the first one. Negative values
16723 indicate positions relative to the last one, with -1 being the last one. It
16724 differs from res.hdr() in that any commas present in the value are returned
16725 and are not used as delimiters. If this is not desired, the res.hdr() fetch
16726 should be used instead. This is sometimes useful with headers such as Date or
16727 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016728
Willy Tarreau74ca5042013-06-11 23:12:07 +020016729res.fhdr_cnt([<name>]) : integer
16730 Returns an integer value representing the number of occurrences of response
16731 header field name <name>, or the total number of header fields if <name> is
16732 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
16733 the number of full line headers and does not stop on commas. If this is not
16734 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016735
Willy Tarreau74ca5042013-06-11 23:12:07 +020016736res.hdr([<name>[,<occ>]]) : string
16737shdr([<name>[,<occ>]]) : string (deprecated)
16738 This extracts the last occurrence of header <name> in an HTTP response, or of
16739 the last header if no <name> is specified. Optionally, a specific occurrence
16740 might be specified as a position number. Positive values indicate a position
16741 from the first occurrence, with 1 being the first one. Negative values
16742 indicate positions relative to the last one, with -1 being the last one. This
16743 can be useful to learn some data into a stick-table. The function considers
16744 any comma as a delimiter for distinct values. If this is not desired, the
16745 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016746
Willy Tarreau74ca5042013-06-11 23:12:07 +020016747 ACL derivatives :
16748 shdr([<name>[,<occ>]]) : exact string match
16749 shdr_beg([<name>[,<occ>]]) : prefix match
16750 shdr_dir([<name>[,<occ>]]) : subdir match
16751 shdr_dom([<name>[,<occ>]]) : domain match
16752 shdr_end([<name>[,<occ>]]) : suffix match
16753 shdr_len([<name>[,<occ>]]) : length match
16754 shdr_reg([<name>[,<occ>]]) : regex match
16755 shdr_sub([<name>[,<occ>]]) : substring match
16756
16757res.hdr_cnt([<name>]) : integer
16758shdr_cnt([<name>]) : integer (deprecated)
16759 Returns an integer value representing the number of occurrences of response
16760 header field name <name>, or the total number of header fields if <name> is
16761 not specified. The function considers any comma as a delimiter for distinct
16762 values. If this is not desired, the res.fhdr_cnt() fetch should be used
16763 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016764
Willy Tarreau74ca5042013-06-11 23:12:07 +020016765res.hdr_ip([<name>[,<occ>]]) : ip
16766shdr_ip([<name>[,<occ>]]) : ip (deprecated)
16767 This extracts the last occurrence of header <name> in an HTTP response,
16768 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
16769 specific occurrence might be specified as a position number. Positive values
16770 indicate a position from the first occurrence, with 1 being the first one.
16771 Negative values indicate positions relative to the last one, with -1 being
16772 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016773
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016774res.hdr_names([<delim>]) : string
16775 This builds a string made from the concatenation of all header names as they
16776 appear in the response when the rule is evaluated. The default delimiter is
16777 the comma (',') but it may be overridden as an optional argument <delim>. In
16778 this case, only the first character of <delim> is considered.
16779
Willy Tarreau74ca5042013-06-11 23:12:07 +020016780res.hdr_val([<name>[,<occ>]]) : integer
16781shdr_val([<name>[,<occ>]]) : integer (deprecated)
16782 This extracts the last occurrence of header <name> in an HTTP response, and
16783 converts it to an integer value. Optionally, a specific occurrence might be
16784 specified as a position number. Positive values indicate a position from the
16785 first occurrence, with 1 being the first one. Negative values indicate
16786 positions relative to the last one, with -1 being the last one. This can be
16787 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016788
Willy Tarreau74ca5042013-06-11 23:12:07 +020016789res.ver : string
16790resp_ver : string (deprecated)
16791 Returns the version string from the HTTP response, for example "1.1". This
16792 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020016793
Willy Tarreau74ca5042013-06-11 23:12:07 +020016794 ACL derivatives :
16795 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016796
Willy Tarreau74ca5042013-06-11 23:12:07 +020016797set-cookie([<name>]) : string (deprecated)
16798 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16799 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020016800 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020016801 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016802
Willy Tarreau74ca5042013-06-11 23:12:07 +020016803 This fetch function is deprecated and has been superseded by the "res.cook"
16804 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016805
Willy Tarreau74ca5042013-06-11 23:12:07 +020016806status : integer
16807 Returns an integer containing the HTTP status code in the HTTP response, for
16808 example, 302. It is mostly used within ACLs and integer ranges, for example,
16809 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016810
Thierry Fournier0e00dca2016-04-07 15:47:40 +020016811unique-id : string
16812 Returns the unique-id attached to the request. The directive
16813 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
16814 fails. Note that the unique-id is usually used with HTTP requests, however this
16815 sample fetch can be used with other protocols. Obviously, if it is used with
16816 other protocols than HTTP, the unique-id-format directive must not contain
16817 HTTP parts. See: unique-id-format and unique-id-header
16818
Willy Tarreau74ca5042013-06-11 23:12:07 +020016819url : string
16820 This extracts the request's URL as presented in the request. A typical use is
16821 with prefetch-capable caches, and with portals which need to aggregate
16822 multiple information from databases and keep them in caches. With ACLs, using
16823 "path" is preferred over using "url", because clients may send a full URL as
16824 is normally done with proxies. The only real use is to match "*" which does
16825 not match in "path", and for which there is already a predefined ACL. See
16826 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016827
Willy Tarreau74ca5042013-06-11 23:12:07 +020016828 ACL derivatives :
16829 url : exact string match
16830 url_beg : prefix match
16831 url_dir : subdir match
16832 url_dom : domain match
16833 url_end : suffix match
16834 url_len : length match
16835 url_reg : regex match
16836 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016837
Willy Tarreau74ca5042013-06-11 23:12:07 +020016838url_ip : ip
16839 This extracts the IP address from the request's URL when the host part is
16840 presented as an IP address. Its use is very limited. For instance, a
16841 monitoring system might use this field as an alternative for the source IP in
16842 order to test what path a given source address would follow, or to force an
16843 entry in a table for a given source address. With ACLs it can be used to
16844 restrict access to certain systems through a proxy, for example when combined
16845 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016846
Willy Tarreau74ca5042013-06-11 23:12:07 +020016847url_port : integer
16848 This extracts the port part from the request's URL. Note that if the port is
16849 not specified in the request, port 80 is assumed. With ACLs it can be used to
16850 restrict access to certain systems through a proxy, for example when combined
16851 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016852
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016853urlp([<name>[,<delim>]]) : string
16854url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020016855 This extracts the first occurrence of the parameter <name> in the query
16856 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016857 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
16858 any parameter will match, and the first one will be returned. The result is
16859 a string corresponding to the value of the parameter <name> as presented in
16860 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020016861 stickiness based on a client ID, to extract an application cookie passed as a
16862 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016863 this fetch iterates over multiple parameters and will iteratively report all
16864 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016865
Willy Tarreau74ca5042013-06-11 23:12:07 +020016866 ACL derivatives :
16867 urlp(<name>[,<delim>]) : exact string match
16868 urlp_beg(<name>[,<delim>]) : prefix match
16869 urlp_dir(<name>[,<delim>]) : subdir match
16870 urlp_dom(<name>[,<delim>]) : domain match
16871 urlp_end(<name>[,<delim>]) : suffix match
16872 urlp_len(<name>[,<delim>]) : length match
16873 urlp_reg(<name>[,<delim>]) : regex match
16874 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016875
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016876
Willy Tarreau74ca5042013-06-11 23:12:07 +020016877 Example :
16878 # match http://example.com/foo?PHPSESSIONID=some_id
16879 stick on urlp(PHPSESSIONID)
16880 # match http://example.com/foo;JSESSIONID=some_id
16881 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016882
Jarno Huuskonen676f6222017-03-30 09:19:45 +030016883urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020016884 See "urlp" above. This one extracts the URL parameter <name> in the request
16885 and converts it to an integer value. This can be used for session stickiness
16886 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020016887
Dragan Dosen0070cd52016-06-16 12:19:49 +020016888url32 : integer
16889 This returns a 32-bit hash of the value obtained by concatenating the first
16890 Host header and the whole URL including parameters (not only the path part of
16891 the request, as in the "base32" fetch above). This is useful to track per-URL
16892 activity. A shorter hash is stored, saving a lot of memory. The output type
16893 is an unsigned integer.
16894
16895url32+src : binary
16896 This returns the concatenation of the "url32" fetch and the "src" fetch. The
16897 resulting type is of type binary, with a size of 8 or 20 bytes depending on
16898 the source address family. This can be used to track per-IP, per-URL counters.
16899
Willy Tarreau198a7442008-01-17 12:05:32 +010016900
Christopher Fauletd47941d2020-01-08 14:40:19 +0100169017.3.7. Fetching samples for developers
16902---------------------------------------
16903
16904This set of sample fetch methods is reserved to developers and must never be
16905used on a production environment, except on developer demand, for debugging
16906purposes. Moreover, no special care will be taken on backwards compatibility.
16907There is no warranty the following sample fetches will never change, be renamed
16908or simply removed. So be really careful if you should use one of them. To avoid
16909any ambiguity, these sample fetches are placed in the dedicated scope "internal",
16910for instance "internal.strm.is_htx".
16911
16912internal.htx.data : integer
16913 Returns the size in bytes used by data in the HTX message associated to a
16914 channel. The channel is chosen depending on the sample direction.
16915
16916internal.htx.free : integer
16917 Returns the free space (size - used) in bytes in the HTX message associated
16918 to a channel. The channel is chosen depending on the sample direction.
16919
16920internal.htx.free_data : integer
16921 Returns the free space for the data in bytes in the HTX message associated to
16922 a channel. The channel is chosen depending on the sample direction.
16923
16924internal.htx.has_eom : boolean
16925 Returns true if the HTX message associated to a channel contains an
16926 end-of-message block (EOM). Otherwise, it returns false. The channel is
16927 chosen depending on the sample direction.
16928
16929internal.htx.nbblks : integer
16930 Returns the number of blocks present in the HTX message associated to a
16931 channel. The channel is chosen depending on the sample direction.
16932
16933internal.htx.size : integer
16934 Returns the total size in bytes of the HTX message associated to a
16935 channel. The channel is chosen depending on the sample direction.
16936
16937internal.htx.used : integer
16938 Returns the total size used in bytes (data + metadata) in the HTX message
16939 associated to a channel. The channel is chosen depending on the sample
16940 direction.
16941
16942internal.htx_blk.size(<idx>) : integer
16943 Returns the size of the block at the position <idx> in the HTX message
16944 associated to a channel or 0 if it does not exist. The channel is chosen
16945 depending on the sample direction. <idx> may be any positive integer or one
16946 of the special value :
16947 * head : The oldest inserted block
16948 * tail : The newest inserted block
16949 * fisrt : The first block where to (re)start the analysis
16950
16951internal.htx_blk.type(<idx>) : string
16952 Returns the type of the block at the position <idx> in the HTX message
16953 associated to a channel or "HTX_BLK_UNUSED" if it does not exist. The channel
16954 is chosen depending on the sample direction. <idx> may be any positive
16955 integer or one of the special value :
16956 * head : The oldest inserted block
16957 * tail : The newest inserted block
16958 * fisrt : The first block where to (re)start the analysis
16959
16960internal.htx_blk.data(<idx>) : binary
16961 Returns the value of the DATA block at the position <idx> in the HTX message
16962 associated to a channel or an empty string if it does not exist or if it is
16963 not a DATA block. The channel is chosen depending on the sample direction.
16964 <idx> may be any positive integer or one of the special value :
16965
16966 * head : The oldest inserted block
16967 * tail : The newest inserted block
16968 * fisrt : The first block where to (re)start the analysis
16969
16970internal.htx_blk.hdrname(<idx>) : string
16971 Returns the header name of the HEADER block at the position <idx> in the HTX
16972 message associated to a channel or an empty string if it does not exist or if
16973 it is not an HEADER block. The channel is chosen depending on the sample
16974 direction. <idx> may be any positive integer or one of the special value :
16975
16976 * head : The oldest inserted block
16977 * tail : The newest inserted block
16978 * fisrt : The first block where to (re)start the analysis
16979
16980internal.htx_blk.hdrval(<idx>) : string
16981 Returns the header value of the HEADER block at the position <idx> in the HTX
16982 message associated to a channel or an empty string if it does not exist or if
16983 it is not an HEADER block. The channel is chosen depending on the sample
16984 direction. <idx> may be any positive integer or one of the special value :
16985
16986 * head : The oldest inserted block
16987 * tail : The newest inserted block
16988 * fisrt : The first block where to (re)start the analysis
16989
16990internal.htx_blk.start_line(<idx>) : string
16991 Returns the value of the REQ_SL or RES_SL block at the position <idx> in the
16992 HTX message associated to a channel or an empty string if it does not exist
16993 or if it is not a SL block. The channel is chosen depending on the sample
16994 direction. <idx> may be any positive integer or one of the special value :
16995
16996 * head : The oldest inserted block
16997 * tail : The newest inserted block
16998 * fisrt : The first block where to (re)start the analysis
16999
17000internal.strm.is_htx : boolean
17001 Returns true if the current stream is an HTX stream. It means the data in the
17002 channels buffers are stored using the internal HTX representation. Otherwise,
17003 it returns false.
17004
17005
Willy Tarreau74ca5042013-06-11 23:12:07 +0200170067.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017007---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010017008
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017009Some predefined ACLs are hard-coded so that they do not have to be declared in
17010every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020017011order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010017012
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017013ACL name Equivalent to Usage
17014---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017015FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020017016HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017017HTTP_1.0 req_ver 1.0 match HTTP version 1.0
17018HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010017019HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
17020HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
17021HTTP_URL_SLASH url_beg / match URL beginning with "/"
17022HTTP_URL_STAR url * match URL equal to "*"
17023LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017024METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020017025METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017026METH_GET method GET HEAD match HTTP GET or HEAD method
17027METH_HEAD method HEAD match HTTP HEAD method
17028METH_OPTIONS method OPTIONS match HTTP OPTIONS method
17029METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020017030METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017031METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020017032RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017033REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010017034TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017035WAIT_END wait_end wait for end of content analysis
17036---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010017037
Willy Tarreaub937b7e2010-01-12 15:27:54 +010017038
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170398. Logging
17040----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010017041
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017042One of HAProxy's strong points certainly lies is its precise logs. It probably
17043provides the finest level of information available for such a product, which is
17044very important for troubleshooting complex environments. Standard information
17045provided in logs include client ports, TCP/HTTP state timers, precise session
17046state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017047to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017048headers.
17049
17050In order to improve administrators reactivity, it offers a great transparency
17051about encountered problems, both internal and external, and it is possible to
17052send logs to different sources at the same time with different level filters :
17053
17054 - global process-level logs (system errors, start/stop, etc..)
17055 - per-instance system and internal errors (lack of resource, bugs, ...)
17056 - per-instance external troubles (servers up/down, max connections)
17057 - per-instance activity (client connections), either at the establishment or
17058 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010017059 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060017060 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017061
17062The ability to distribute different levels of logs to different log servers
17063allow several production teams to interact and to fix their problems as soon
17064as possible. For example, the system team might monitor system-wide errors,
17065while the application team might be monitoring the up/down for their servers in
17066real time, and the security team might analyze the activity logs with one hour
17067delay.
17068
17069
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170708.1. Log levels
17071---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017072
Simon Hormandf791f52011-05-29 15:01:10 +090017073TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017074source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090017075HTTP request, HTTP return code, number of bytes transmitted, conditions
17076in which the session ended, and even exchanged cookies values. For example
17077track a particular user's problems. All messages may be sent to up to two
17078syslog servers. Check the "log" keyword in section 4.2 for more information
17079about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017080
17081
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170828.2. Log formats
17083----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017084
William Lallemand48940402012-01-30 16:47:22 +010017085HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090017086and will be detailed in the following sections. A few of them may vary
17087slightly with the configuration, due to indicators specific to certain
17088options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017089
17090 - the default format, which is very basic and very rarely used. It only
17091 provides very basic information about the incoming connection at the moment
17092 it is accepted : source IP:port, destination IP:port, and frontend-name.
17093 This mode will eventually disappear so it will not be described to great
17094 extents.
17095
17096 - the TCP format, which is more advanced. This format is enabled when "option
17097 tcplog" is set on the frontend. HAProxy will then usually wait for the
17098 connection to terminate before logging. This format provides much richer
17099 information, such as timers, connection counts, queue size, etc... This
17100 format is recommended for pure TCP proxies.
17101
17102 - the HTTP format, which is the most advanced for HTTP proxying. This format
17103 is enabled when "option httplog" is set on the frontend. It provides the
17104 same information as the TCP format with some HTTP-specific fields such as
17105 the request, the status code, and captures of headers and cookies. This
17106 format is recommended for HTTP proxies.
17107
Emeric Brun3a058f32009-06-30 18:26:00 +020017108 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
17109 fields arranged in the same order as the CLF format. In this mode, all
17110 timers, captures, flags, etc... appear one per field after the end of the
17111 common fields, in the same order they appear in the standard HTTP format.
17112
William Lallemand48940402012-01-30 16:47:22 +010017113 - the custom log format, allows you to make your own log line.
17114
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017115Next sections will go deeper into details for each of these formats. Format
17116specification will be performed on a "field" basis. Unless stated otherwise, a
17117field is a portion of text delimited by any number of spaces. Since syslog
17118servers are susceptible of inserting fields at the beginning of a line, it is
17119always assumed that the first field is the one containing the process name and
17120identifier.
17121
17122Note : Since log lines may be quite long, the log examples in sections below
17123 might be broken into multiple lines. The example log lines will be
17124 prefixed with 3 closing angle brackets ('>>>') and each time a log is
17125 broken into multiple lines, each non-final line will end with a
17126 backslash ('\') and the next line will start indented by two characters.
17127
17128
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171298.2.1. Default log format
17130-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017131
17132This format is used when no specific option is set. The log is emitted as soon
17133as the connection is accepted. One should note that this currently is the only
17134format which logs the request's destination IP and ports.
17135
17136 Example :
17137 listen www
17138 mode http
17139 log global
17140 server srv1 127.0.0.1:8000
17141
17142 >>> Feb 6 12:12:09 localhost \
17143 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
17144 (www/HTTP)
17145
17146 Field Format Extract from the example above
17147 1 process_name '[' pid ']:' haproxy[14385]:
17148 2 'Connect from' Connect from
17149 3 source_ip ':' source_port 10.0.1.2:33312
17150 4 'to' to
17151 5 destination_ip ':' destination_port 10.0.3.31:8012
17152 6 '(' frontend_name '/' mode ')' (www/HTTP)
17153
17154Detailed fields description :
17155 - "source_ip" is the IP address of the client which initiated the connection.
17156 - "source_port" is the TCP port of the client which initiated the connection.
17157 - "destination_ip" is the IP address the client connected to.
17158 - "destination_port" is the TCP port the client connected to.
17159 - "frontend_name" is the name of the frontend (or listener) which received
17160 and processed the connection.
17161 - "mode is the mode the frontend is operating (TCP or HTTP).
17162
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017163In case of a UNIX socket, the source and destination addresses are marked as
17164"unix:" and the ports reflect the internal ID of the socket which accepted the
17165connection (the same ID as reported in the stats).
17166
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017167It is advised not to use this deprecated format for newer installations as it
17168will eventually disappear.
17169
17170
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171718.2.2. TCP log format
17172---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017173
17174The TCP format is used when "option tcplog" is specified in the frontend, and
17175is the recommended format for pure TCP proxies. It provides a lot of precious
17176information for troubleshooting. Since this format includes timers and byte
17177counts, the log is normally emitted at the end of the session. It can be
17178emitted earlier if "option logasap" is specified, which makes sense in most
17179environments with long sessions such as remote terminals. Sessions which match
17180the "monitor" rules are never logged. It is also possible not to emit logs for
17181sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017182specifying "option dontlognull" in the frontend. Successful connections will
17183not be logged if "option dontlog-normal" is specified in the frontend. A few
17184fields may slightly vary depending on some configuration options, those are
17185marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017186
17187 Example :
17188 frontend fnt
17189 mode tcp
17190 option tcplog
17191 log global
17192 default_backend bck
17193
17194 backend bck
17195 server srv1 127.0.0.1:8000
17196
17197 >>> Feb 6 12:12:56 localhost \
17198 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
17199 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
17200
17201 Field Format Extract from the example above
17202 1 process_name '[' pid ']:' haproxy[14387]:
17203 2 client_ip ':' client_port 10.0.1.2:33313
17204 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
17205 4 frontend_name fnt
17206 5 backend_name '/' server_name bck/srv1
17207 6 Tw '/' Tc '/' Tt* 0/0/5007
17208 7 bytes_read* 212
17209 8 termination_state --
17210 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
17211 10 srv_queue '/' backend_queue 0/0
17212
17213Detailed fields description :
17214 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017215 connection to haproxy. If the connection was accepted on a UNIX socket
17216 instead, the IP address would be replaced with the word "unix". Note that
17217 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017218 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010017219 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017220 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017221
17222 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017223 If the connection was accepted on a UNIX socket instead, the port would be
17224 replaced with the ID of the accepting socket, which is also reported in the
17225 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017226
17227 - "accept_date" is the exact date when the connection was received by haproxy
17228 (which might be very slightly different from the date observed on the
17229 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020017230 the same date which may appear in any upstream firewall's log. When used in
17231 HTTP mode, the accept_date field will be reset to the first moment the
17232 connection is ready to receive a new request (end of previous response for
17233 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017234
17235 - "frontend_name" is the name of the frontend (or listener) which received
17236 and processed the connection.
17237
17238 - "backend_name" is the name of the backend (or listener) which was selected
17239 to manage the connection to the server. This will be the same as the
17240 frontend if no switching rule has been applied, which is common for TCP
17241 applications.
17242
17243 - "server_name" is the name of the last server to which the connection was
17244 sent, which might differ from the first one if there were connection errors
17245 and a redispatch occurred. Note that this server belongs to the backend
17246 which processed the request. If the connection was aborted before reaching
17247 a server, "<NOSRV>" is indicated instead of a server name.
17248
17249 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17250 It can be "-1" if the connection was aborted before reaching the queue.
17251 See "Timers" below for more details.
17252
17253 - "Tc" is the total time in milliseconds spent waiting for the connection to
17254 establish to the final server, including retries. It can be "-1" if the
17255 connection was aborted before a connection could be established. See
17256 "Timers" below for more details.
17257
17258 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017259 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017260 "option logasap" was specified, then the time counting stops at the moment
17261 the log is emitted. In this case, a '+' sign is prepended before the value,
17262 indicating that the final one will be larger. See "Timers" below for more
17263 details.
17264
17265 - "bytes_read" is the total number of bytes transmitted from the server to
17266 the client when the log is emitted. If "option logasap" is specified, the
17267 this value will be prefixed with a '+' sign indicating that the final one
17268 may be larger. Please note that this value is a 64-bit counter, so log
17269 analysis tools must be able to handle it without overflowing.
17270
17271 - "termination_state" is the condition the session was in when the session
17272 ended. This indicates the session state, which side caused the end of
17273 session to happen, and for what reason (timeout, error, ...). The normal
17274 flags should be "--", indicating the session was closed by either end with
17275 no data remaining in buffers. See below "Session state at disconnection"
17276 for more details.
17277
17278 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017279 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017280 limits have been reached. For instance, if actconn is close to 512 when
17281 multiple connection errors occur, chances are high that the system limits
17282 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017283 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017284
17285 - "feconn" is the total number of concurrent connections on the frontend when
17286 the session was logged. It is useful to estimate the amount of resource
17287 required to sustain high loads, and to detect when the frontend's "maxconn"
17288 has been reached. Most often when this value increases by huge jumps, it is
17289 because there is congestion on the backend servers, but sometimes it can be
17290 caused by a denial of service attack.
17291
17292 - "beconn" is the total number of concurrent connections handled by the
17293 backend when the session was logged. It includes the total number of
17294 concurrent connections active on servers as well as the number of
17295 connections pending in queues. It is useful to estimate the amount of
17296 additional servers needed to support high loads for a given application.
17297 Most often when this value increases by huge jumps, it is because there is
17298 congestion on the backend servers, but sometimes it can be caused by a
17299 denial of service attack.
17300
17301 - "srv_conn" is the total number of concurrent connections still active on
17302 the server when the session was logged. It can never exceed the server's
17303 configured "maxconn" parameter. If this value is very often close or equal
17304 to the server's "maxconn", it means that traffic regulation is involved a
17305 lot, meaning that either the server's maxconn value is too low, or that
17306 there aren't enough servers to process the load with an optimal response
17307 time. When only one of the server's "srv_conn" is high, it usually means
17308 that this server has some trouble causing the connections to take longer to
17309 be processed than on other servers.
17310
17311 - "retries" is the number of connection retries experienced by this session
17312 when trying to connect to the server. It must normally be zero, unless a
17313 server is being stopped at the same moment the connection was attempted.
17314 Frequent retries generally indicate either a network problem between
17315 haproxy and the server, or a misconfigured system backlog on the server
17316 preventing new connections from being queued. This field may optionally be
17317 prefixed with a '+' sign, indicating that the session has experienced a
17318 redispatch after the maximal retry count has been reached on the initial
17319 server. In this case, the server name appearing in the log is the one the
17320 connection was redispatched to, and not the first one, though both may
17321 sometimes be the same in case of hashing for instance. So as a general rule
17322 of thumb, when a '+' is present in front of the retry count, this count
17323 should not be attributed to the logged server.
17324
17325 - "srv_queue" is the total number of requests which were processed before
17326 this one in the server queue. It is zero when the request has not gone
17327 through the server queue. It makes it possible to estimate the approximate
17328 server's response time by dividing the time spent in queue by the number of
17329 requests in the queue. It is worth noting that if a session experiences a
17330 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017331 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017332 backend queue unless a redispatch occurs.
17333
17334 - "backend_queue" is the total number of requests which were processed before
17335 this one in the backend's global queue. It is zero when the request has not
17336 gone through the global queue. It makes it possible to estimate the average
17337 queue length, which easily translates into a number of missing servers when
17338 divided by a server's "maxconn" parameter. It is worth noting that if a
17339 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017340 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017341 through both the server queue and the backend queue unless a redispatch
17342 occurs.
17343
17344
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173458.2.3. HTTP log format
17346----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017347
17348The HTTP format is the most complete and the best suited for HTTP proxies. It
17349is enabled by when "option httplog" is specified in the frontend. It provides
17350the same level of information as the TCP format with additional features which
17351are specific to the HTTP protocol. Just like the TCP format, the log is usually
17352emitted at the end of the session, unless "option logasap" is specified, which
17353generally only makes sense for download sites. A session which matches the
17354"monitor" rules will never logged. It is also possible not to log sessions for
17355which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017356frontend. Successful connections will not be logged if "option dontlog-normal"
17357is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017358
17359Most fields are shared with the TCP log, some being different. A few fields may
17360slightly vary depending on some configuration options. Those ones are marked
17361with a star ('*') after the field name below.
17362
17363 Example :
17364 frontend http-in
17365 mode http
17366 option httplog
17367 log global
17368 default_backend bck
17369
17370 backend static
17371 server srv1 127.0.0.1:8000
17372
17373 >>> Feb 6 12:14:14 localhost \
17374 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
17375 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 +010017376 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017377
17378 Field Format Extract from the example above
17379 1 process_name '[' pid ']:' haproxy[14389]:
17380 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017381 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017382 4 frontend_name http-in
17383 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017384 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017385 7 status_code 200
17386 8 bytes_read* 2750
17387 9 captured_request_cookie -
17388 10 captured_response_cookie -
17389 11 termination_state ----
17390 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
17391 13 srv_queue '/' backend_queue 0/0
17392 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
17393 15 '{' captured_response_headers* '}' {}
17394 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010017395
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017396Detailed fields description :
17397 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017398 connection to haproxy. If the connection was accepted on a UNIX socket
17399 instead, the IP address would be replaced with the word "unix". Note that
17400 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017401 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010017402 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017403 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017404
17405 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017406 If the connection was accepted on a UNIX socket instead, the port would be
17407 replaced with the ID of the accepting socket, which is also reported in the
17408 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017409
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017410 - "request_date" is the exact date when the first byte of the HTTP request
17411 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017412
17413 - "frontend_name" is the name of the frontend (or listener) which received
17414 and processed the connection.
17415
17416 - "backend_name" is the name of the backend (or listener) which was selected
17417 to manage the connection to the server. This will be the same as the
17418 frontend if no switching rule has been applied.
17419
17420 - "server_name" is the name of the last server to which the connection was
17421 sent, which might differ from the first one if there were connection errors
17422 and a redispatch occurred. Note that this server belongs to the backend
17423 which processed the request. If the request was aborted before reaching a
17424 server, "<NOSRV>" is indicated instead of a server name. If the request was
17425 intercepted by the stats subsystem, "<STATS>" is indicated instead.
17426
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017427 - "TR" is the total time in milliseconds spent waiting for a full HTTP
17428 request from the client (not counting body) after the first byte was
17429 received. It can be "-1" if the connection was aborted before a complete
John Roeslerfb2fce12019-07-10 15:45:51 -050017430 request could be received or a bad request was received. It should
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017431 always be very small because a request generally fits in one single packet.
17432 Large times here generally indicate network issues between the client and
Willy Tarreau590a0512018-09-05 11:56:48 +020017433 haproxy or requests being typed by hand. See section 8.4 "Timing Events"
17434 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017435
17436 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17437 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020017438 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017439
17440 - "Tc" is the total time in milliseconds spent waiting for the connection to
17441 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020017442 request was aborted before a connection could be established. See section
17443 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017444
17445 - "Tr" is the total time in milliseconds spent waiting for the server to send
17446 a full HTTP response, not counting data. It can be "-1" if the request was
17447 aborted before a complete response could be received. It generally matches
17448 the server's processing time for the request, though it may be altered by
17449 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020017450 "GET" requests generally indicate an overloaded server. See section 8.4
17451 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017452
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017453 - "Ta" is the time the request remained active in haproxy, which is the total
17454 time in milliseconds elapsed between the first byte of the request was
17455 received and the last byte of response was sent. It covers all possible
17456 processing except the handshake (see Th) and idle time (see Ti). There is
17457 one exception, if "option logasap" was specified, then the time counting
17458 stops at the moment the log is emitted. In this case, a '+' sign is
17459 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020017460 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017461
17462 - "status_code" is the HTTP status code returned to the client. This status
17463 is generally set by the server, but it might also be set by haproxy when
17464 the server cannot be reached or when its response is blocked by haproxy.
17465
17466 - "bytes_read" is the total number of bytes transmitted to the client when
17467 the log is emitted. This does include HTTP headers. If "option logasap" is
John Roeslerfb2fce12019-07-10 15:45:51 -050017468 specified, this value will be prefixed with a '+' sign indicating that
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017469 the final one may be larger. Please note that this value is a 64-bit
17470 counter, so log analysis tools must be able to handle it without
17471 overflowing.
17472
17473 - "captured_request_cookie" is an optional "name=value" entry indicating that
17474 the client had this cookie in the request. The cookie name and its maximum
17475 length are defined by the "capture cookie" statement in the frontend
17476 configuration. The field is a single dash ('-') when the option is not
17477 set. Only one cookie may be captured, it is generally used to track session
17478 ID exchanges between a client and a server to detect session crossing
17479 between clients due to application bugs. For more details, please consult
17480 the section "Capturing HTTP headers and cookies" below.
17481
17482 - "captured_response_cookie" is an optional "name=value" entry indicating
17483 that the server has returned a cookie with its response. The cookie name
17484 and its maximum length are defined by the "capture cookie" statement in the
17485 frontend configuration. The field is a single dash ('-') when the option is
17486 not set. Only one cookie may be captured, it is generally used to track
17487 session ID exchanges between a client and a server to detect session
17488 crossing between clients due to application bugs. For more details, please
17489 consult the section "Capturing HTTP headers and cookies" below.
17490
17491 - "termination_state" is the condition the session was in when the session
17492 ended. This indicates the session state, which side caused the end of
17493 session to happen, for what reason (timeout, error, ...), just like in TCP
17494 logs, and information about persistence operations on cookies in the last
17495 two characters. The normal flags should begin with "--", indicating the
17496 session was closed by either end with no data remaining in buffers. See
17497 below "Session state at disconnection" for more details.
17498
17499 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017500 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017501 limits have been reached. For instance, if actconn is close to 512 or 1024
17502 when multiple connection errors occur, chances are high that the system
17503 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017504 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017505 system.
17506
17507 - "feconn" is the total number of concurrent connections on the frontend when
17508 the session was logged. It is useful to estimate the amount of resource
17509 required to sustain high loads, and to detect when the frontend's "maxconn"
17510 has been reached. Most often when this value increases by huge jumps, it is
17511 because there is congestion on the backend servers, but sometimes it can be
17512 caused by a denial of service attack.
17513
17514 - "beconn" is the total number of concurrent connections handled by the
17515 backend when the session was logged. It includes the total number of
17516 concurrent connections active on servers as well as the number of
17517 connections pending in queues. It is useful to estimate the amount of
17518 additional servers needed to support high loads for a given application.
17519 Most often when this value increases by huge jumps, it is because there is
17520 congestion on the backend servers, but sometimes it can be caused by a
17521 denial of service attack.
17522
17523 - "srv_conn" is the total number of concurrent connections still active on
17524 the server when the session was logged. It can never exceed the server's
17525 configured "maxconn" parameter. If this value is very often close or equal
17526 to the server's "maxconn", it means that traffic regulation is involved a
17527 lot, meaning that either the server's maxconn value is too low, or that
17528 there aren't enough servers to process the load with an optimal response
17529 time. When only one of the server's "srv_conn" is high, it usually means
17530 that this server has some trouble causing the requests to take longer to be
17531 processed than on other servers.
17532
17533 - "retries" is the number of connection retries experienced by this session
17534 when trying to connect to the server. It must normally be zero, unless a
17535 server is being stopped at the same moment the connection was attempted.
17536 Frequent retries generally indicate either a network problem between
17537 haproxy and the server, or a misconfigured system backlog on the server
17538 preventing new connections from being queued. This field may optionally be
17539 prefixed with a '+' sign, indicating that the session has experienced a
17540 redispatch after the maximal retry count has been reached on the initial
17541 server. In this case, the server name appearing in the log is the one the
17542 connection was redispatched to, and not the first one, though both may
17543 sometimes be the same in case of hashing for instance. So as a general rule
17544 of thumb, when a '+' is present in front of the retry count, this count
17545 should not be attributed to the logged server.
17546
17547 - "srv_queue" is the total number of requests which were processed before
17548 this one in the server queue. It is zero when the request has not gone
17549 through the server queue. It makes it possible to estimate the approximate
17550 server's response time by dividing the time spent in queue by the number of
17551 requests in the queue. It is worth noting that if a session experiences a
17552 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017553 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017554 backend queue unless a redispatch occurs.
17555
17556 - "backend_queue" is the total number of requests which were processed before
17557 this one in the backend's global queue. It is zero when the request has not
17558 gone through the global queue. It makes it possible to estimate the average
17559 queue length, which easily translates into a number of missing servers when
17560 divided by a server's "maxconn" parameter. It is worth noting that if a
17561 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017562 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017563 through both the server queue and the backend queue unless a redispatch
17564 occurs.
17565
17566 - "captured_request_headers" is a list of headers captured in the request due
17567 to the presence of the "capture request header" statement in the frontend.
17568 Multiple headers can be captured, they will be delimited by a vertical bar
17569 ('|'). When no capture is enabled, the braces do not appear, causing a
17570 shift of remaining fields. It is important to note that this field may
17571 contain spaces, and that using it requires a smarter log parser than when
17572 it's not used. Please consult the section "Capturing HTTP headers and
17573 cookies" below for more details.
17574
17575 - "captured_response_headers" is a list of headers captured in the response
17576 due to the presence of the "capture response header" statement in the
17577 frontend. Multiple headers can be captured, they will be delimited by a
17578 vertical bar ('|'). When no capture is enabled, the braces do not appear,
17579 causing a shift of remaining fields. It is important to note that this
17580 field may contain spaces, and that using it requires a smarter log parser
17581 than when it's not used. Please consult the section "Capturing HTTP headers
17582 and cookies" below for more details.
17583
17584 - "http_request" is the complete HTTP request line, including the method,
17585 request and HTTP version string. Non-printable characters are encoded (see
17586 below the section "Non-printable characters"). This is always the last
17587 field, and it is always delimited by quotes and is the only one which can
17588 contain quotes. If new fields are added to the log format, they will be
17589 added before this field. This field might be truncated if the request is
17590 huge and does not fit in the standard syslog buffer (1024 characters). This
17591 is the reason why this field must always remain the last one.
17592
17593
Cyril Bontédc4d9032012-04-08 21:57:39 +0200175948.2.4. Custom log format
17595------------------------
William Lallemand48940402012-01-30 16:47:22 +010017596
Willy Tarreau2beef582012-12-20 17:22:52 +010017597The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010017598mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010017599
Davor Ocelice9ed2812017-12-25 17:49:28 +010017600HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010017601Variables can take arguments using braces ('{}'), and multiple arguments are
17602separated by commas within the braces. Flags may be added or removed by
17603prefixing them with a '+' or '-' sign.
17604
17605Special variable "%o" may be used to propagate its flags to all other
17606variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010017607("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010017608
Willy Tarreauc8368452012-12-21 00:09:23 +010017609If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020017610as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010017611less common information such as the client's SSL certificate's DN, or to log
17612the key that would be used to store an entry into a stick table.
17613
William Lallemand48940402012-01-30 16:47:22 +010017614Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017615In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010017616in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010017617
Dragan Dosen835b9212016-02-12 13:23:03 +010017618Note: when using the RFC5424 syslog message format, the characters '"',
17619'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
17620https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
17621such cases, the use of the flag "E" should be considered.
17622
William Lallemand48940402012-01-30 16:47:22 +010017623Flags are :
17624 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017625 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010017626 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
17627 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010017628
17629 Example:
17630
17631 log-format %T\ %t\ Some\ Text
17632 log-format %{+Q}o\ %t\ %s\ %{-Q}r
17633
Dragan Dosen835b9212016-02-12 13:23:03 +010017634 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
17635
William Lallemand48940402012-01-30 16:47:22 +010017636At the moment, the default HTTP format is defined this way :
17637
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017638 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
17639 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010017640
William Lallemandbddd4fd2012-02-27 11:23:10 +010017641the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010017642
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017643 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
17644 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
17645 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010017646
William Lallemandbddd4fd2012-02-27 11:23:10 +010017647and the default TCP format is defined this way :
17648
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017649 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
17650 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010017651
William Lallemand48940402012-01-30 16:47:22 +010017652Please refer to the table below for currently defined variables :
17653
William Lallemandbddd4fd2012-02-27 11:23:10 +010017654 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017655 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017656 +---+------+-----------------------------------------------+-------------+
17657 | | %o | special variable, apply flags on all next var | |
17658 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010017659 | | %B | bytes_read (from server to client) | numeric |
17660 | H | %CC | captured_request_cookie | string |
17661 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020017662 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017663 | H | %HM | HTTP method (ex: POST) | string |
17664 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000017665 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017666 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
17667 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010017668 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020017669 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017670 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017671 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017672 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020017673 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080017674 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017675 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
17676 | H | %Ti | idle time before the HTTP request | numeric |
17677 | H | %Tq | Th + Ti + TR | numeric |
17678 | H | %TR | time to receive the full request from 1st byte| numeric |
17679 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017680 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017681 | | %Tt | Tt | numeric |
17682 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017683 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017684 | | %ac | actconn | numeric |
17685 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017686 | | %bc | beconn (backend concurrent connections) | numeric |
17687 | | %bi | backend_source_ip (connecting address) | IP |
17688 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017689 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017690 | | %ci | client_ip (accepted address) | IP |
17691 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017692 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017693 | | %fc | feconn (frontend concurrent connections) | numeric |
17694 | | %fi | frontend_ip (accepting address) | IP |
17695 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020017696 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020017697 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017698 | | %hr | captured_request_headers default style | string |
17699 | | %hrl | captured_request_headers CLF style | string list |
17700 | | %hs | captured_response_headers default style | string |
17701 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020017702 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017703 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017704 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017705 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010017706 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017707 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017708 | | %sc | srv_conn (server concurrent connections) | numeric |
17709 | | %si | server_IP (target address) | IP |
17710 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017711 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017712 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
17713 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017714 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017715 | H | %tr | date_time of HTTP request | date |
17716 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020017717 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017718 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017719 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017720 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010017721
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017722 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010017723
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017724
177258.2.5. Error log format
17726-----------------------
17727
17728When an incoming connection fails due to an SSL handshake or an invalid PROXY
17729protocol header, haproxy will log the event using a shorter, fixed line format.
17730By default, logs are emitted at the LOG_INFO level, unless the option
17731"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017732will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017733logged if the "dontlognull" option is set.
17734
17735The format looks like this :
17736
17737 >>> Dec 3 18:27:14 localhost \
17738 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
17739 Connection error during SSL handshake
17740
17741 Field Format Extract from the example above
17742 1 process_name '[' pid ']:' haproxy[6103]:
17743 2 client_ip ':' client_port 127.0.0.1:56059
17744 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
17745 4 frontend_name "/" bind_name ":" frt/f1:
17746 5 message Connection error during SSL handshake
17747
17748These fields just provide minimal information to help debugging connection
17749failures.
17750
17751
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177528.3. Advanced logging options
17753-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017754
17755Some advanced logging options are often looked for but are not easy to find out
17756just by looking at the various options. Here is an entry point for the few
17757options which can enable better logging. Please refer to the keywords reference
17758for more information about their usage.
17759
17760
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177618.3.1. Disabling logging of external tests
17762------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017763
17764It is quite common to have some monitoring tools perform health checks on
17765haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
17766commercial load-balancer, and sometimes it will simply be a more complete
17767monitoring system such as Nagios. When the tests are very frequent, users often
17768ask how to disable logging for those checks. There are three possibilities :
17769
17770 - if connections come from everywhere and are just TCP probes, it is often
17771 desired to simply disable logging of connections without data exchange, by
17772 setting "option dontlognull" in the frontend. It also disables logging of
17773 port scans, which may or may not be desired.
17774
17775 - if the connection come from a known source network, use "monitor-net" to
17776 declare this network as monitoring only. Any host in this network will then
17777 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017778 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017779 such as other load-balancers.
17780
17781 - if the tests are performed on a known URI, use "monitor-uri" to declare
17782 this URI as dedicated to monitoring. Any host sending this request will
17783 only get the result of a health-check, and the request will not be logged.
17784
17785
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177868.3.2. Logging before waiting for the session to terminate
17787----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017788
17789The problem with logging at end of connection is that you have no clue about
17790what is happening during very long sessions, such as remote terminal sessions
17791or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010017792"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017793just before data transfer begins. This means that in case of TCP, it will still
17794log the connection status to the server, and in case of HTTP, it will log just
17795after processing the server headers. In this case, the number of bytes reported
17796is the number of header bytes sent to the client. In order to avoid confusion
17797with normal logs, the total time field and the number of bytes are prefixed
17798with a '+' sign which means that real numbers are certainly larger.
17799
17800
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200178018.3.3. Raising log level upon errors
17802------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017803
17804Sometimes it is more convenient to separate normal traffic from errors logs,
17805for instance in order to ease error monitoring from log files. When the option
17806"log-separate-errors" is used, connections which experience errors, timeouts,
17807retries, redispatches or HTTP status codes 5xx will see their syslog level
17808raised from "info" to "err". This will help a syslog daemon store the log in
17809a separate file. It is very important to keep the errors in the normal traffic
17810file too, so that log ordering is not altered. You should also be careful if
17811you already have configured your syslog daemon to store all logs higher than
17812"notice" in an "admin" file, because the "err" level is higher than "notice".
17813
17814
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200178158.3.4. Disabling logging of successful connections
17816--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017817
17818Although this may sound strange at first, some large sites have to deal with
17819multiple thousands of logs per second and are experiencing difficulties keeping
17820them intact for a long time or detecting errors within them. If the option
17821"dontlog-normal" is set on the frontend, all normal connections will not be
17822logged. In this regard, a normal connection is defined as one without any
17823error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
17824and a response with a status 5xx is not considered normal and will be logged
17825too. Of course, doing is is really discouraged as it will remove most of the
17826useful information from the logs. Do this only if you have no other
17827alternative.
17828
17829
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200178308.4. Timing events
17831------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017832
17833Timers provide a great help in troubleshooting network problems. All values are
17834reported in milliseconds (ms). These timers should be used in conjunction with
17835the session termination flags. In TCP mode with "option tcplog" set on the
17836frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017837mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
17838addition, three other measures are provided, "Th", "Ti", and "Tq".
17839
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010017840Timings events in HTTP mode:
17841
17842 first request 2nd request
17843 |<-------------------------------->|<-------------- ...
17844 t tr t tr ...
17845 ---|----|----|----|----|----|----|----|----|--
17846 : Th Ti TR Tw Tc Tr Td : Ti ...
17847 :<---- Tq ---->: :
17848 :<-------------- Tt -------------->:
17849 :<--------- Ta --------->:
17850
17851Timings events in TCP mode:
17852
17853 TCP session
17854 |<----------------->|
17855 t t
17856 ---|----|----|----|----|---
17857 | Th Tw Tc Td |
17858 |<------ Tt ------->|
17859
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017860 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017861 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017862 only happen once during the whole connection's lifetime. A large time here
17863 may indicate that the client only pre-established the connection without
17864 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010017865 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020017866 SSL handshake was very expensive to compute. Please note that this time is
17867 reported only before the first request, so it is safe to average it over
17868 all request to calculate the amortized value. The second and subsequent
17869 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017870
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017871 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
17872 counts between the end of the handshakes and the first byte of the HTTP
17873 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020017874 to count after the end of the transmission the previous response. When a
17875 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
17876 after the previous request. Some browsers pre-establish connections to a
17877 server in order to reduce the latency of a future request, and keep them
17878 pending until they need it. This delay will be reported as the idle time. A
17879 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017880
17881 - TR: total time to get the client request (HTTP mode only). It's the time
17882 elapsed between the first bytes received and the moment the proxy received
17883 the empty line marking the end of the HTTP headers. The value "-1"
17884 indicates that the end of headers has never been seen. This happens when
17885 the client closes prematurely or times out. This time is usually very short
17886 since most requests fit in a single packet. A large time may indicate a
17887 request typed by hand during a test.
17888
17889 - Tq: total time to get the client request from the accept date or since the
17890 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010017891 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 +020017892 returns -1 as well. This timer used to be very useful before the arrival of
17893 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
17894 it in favor of TR nowadays, as the idle time adds a lot of noise to the
17895 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017896
17897 - Tw: total time spent in the queues waiting for a connection slot. It
17898 accounts for backend queue as well as the server queues, and depends on the
17899 queue size, and the time needed for the server to complete previous
17900 requests. The value "-1" means that the request was killed before reaching
17901 the queue, which is generally what happens with invalid or denied requests.
17902
17903 - Tc: total time to establish the TCP connection to the server. It's the time
17904 elapsed between the moment the proxy sent the connection request, and the
17905 moment it was acknowledged by the server, or between the TCP SYN packet and
17906 the matching SYN/ACK packet in return. The value "-1" means that the
17907 connection never established.
17908
17909 - Tr: server response time (HTTP mode only). It's the time elapsed between
17910 the moment the TCP connection was established to the server and the moment
17911 the server sent its complete response headers. It purely shows its request
17912 processing time, without the network overhead due to the data transmission.
17913 It is worth noting that when the client has data to send to the server, for
17914 instance during a POST request, the time already runs, and this can distort
17915 apparent response time. For this reason, it's generally wise not to trust
17916 too much this field for POST requests initiated from clients behind an
17917 untrusted network. A value of "-1" here means that the last the response
17918 header (empty line) was never seen, most likely because the server timeout
17919 stroke before the server managed to process the request.
17920
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017921 - Ta: total active time for the HTTP request, between the moment the proxy
17922 received the first byte of the request header and the emission of the last
17923 byte of the response body. The exception is when the "logasap" option is
17924 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
17925 a '+' sign. From this field, we can deduce "Td", the data transmission time,
17926 by subtracting other timers when valid :
17927
17928 Td = Ta - (TR + Tw + Tc + Tr)
17929
17930 Timers with "-1" values have to be excluded from this equation. Note that
17931 "Ta" can never be negative.
17932
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017933 - Tt: total session duration time, between the moment the proxy accepted it
17934 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017935 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
17936 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017937 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017938
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017939 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017940
17941 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017942 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
17943 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017944
17945These timers provide precious indications on trouble causes. Since the TCP
17946protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
17947that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017948due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
17949"Tt" is close to a timeout value specified in the configuration, it often means
17950that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017951
17952Most common cases :
17953
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017954 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
17955 the client and the proxy. This is very rare on local networks but might
17956 happen when clients are on far remote networks and send large requests. It
17957 may happen that values larger than usual appear here without any network
17958 cause. Sometimes, during an attack or just after a resource starvation has
17959 ended, haproxy may accept thousands of connections in a few milliseconds.
17960 The time spent accepting these connections will inevitably slightly delay
17961 processing of other connections, and it can happen that request times in the
17962 order of a few tens of milliseconds are measured after a few thousands of
17963 new connections have been accepted at once. Using one of the keep-alive
17964 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020017965 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017966
17967 - If "Tc" is close to 3000, a packet has probably been lost between the
17968 server and the proxy during the server connection phase. This value should
17969 always be very low, such as 1 ms on local networks and less than a few tens
17970 of ms on remote networks.
17971
Willy Tarreau55165fe2009-05-10 12:02:55 +020017972 - If "Tr" is nearly always lower than 3000 except some rare values which seem
17973 to be the average majored by 3000, there are probably some packets lost
17974 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017975
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017976 - If "Ta" is large even for small byte counts, it generally is because
17977 neither the client nor the server decides to close the connection while
17978 haproxy is running in tunnel mode and both have agreed on a keep-alive
17979 connection mode. In order to solve this issue, it will be needed to specify
17980 one of the HTTP options to manipulate keep-alive or close options on either
17981 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
17982 important when connection regulation is used with the "maxconn" option on
17983 the servers, since no new connection will be sent to the server until
17984 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017985
17986Other noticeable HTTP log cases ('xx' means any value to be ignored) :
17987
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017988 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017989 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017990 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017991
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017992 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017993 or it aborted too early. Check the session termination flags
17994 then "timeout http-request" and "timeout client" settings.
17995
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017996 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017997 servers were out of order, because the request was invalid
17998 or forbidden by ACL rules. Check the session termination
17999 flags.
18000
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020018001 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
18002 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018003 Check the session termination flags, then check the
18004 "timeout connect" setting. Note that the tarpit action might
18005 return similar-looking patterns, with "Tw" equal to the time
18006 the client connection was maintained open.
18007
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020018008 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018009 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020018010 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018011 termination flags, then check the "timeout server" setting.
18012
18013
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180148.5. Session state at disconnection
18015-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018016
18017TCP and HTTP logs provide a session termination indicator in the
18018"termination_state" field, just before the number of active connections. It is
180192-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
18020each of which has a special meaning :
18021
18022 - On the first character, a code reporting the first event which caused the
18023 session to terminate :
18024
18025 C : the TCP session was unexpectedly aborted by the client.
18026
18027 S : the TCP session was unexpectedly aborted by the server, or the
18028 server explicitly refused it.
18029
18030 P : the session was prematurely aborted by the proxy, because of a
18031 connection limit enforcement, because a DENY filter was matched,
18032 because of a security check which detected and blocked a dangerous
18033 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010018034 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020018035
18036 L : the session was locally processed by haproxy and was not passed to
18037 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018038
18039 R : a resource on the proxy has been exhausted (memory, sockets, source
18040 ports, ...). Usually, this appears during the connection phase, and
18041 system logs should contain a copy of the precise error. If this
18042 happens, it must be considered as a very serious anomaly which
18043 should be fixed as soon as possible by any means.
18044
18045 I : an internal error was identified by the proxy during a self-check.
18046 This should NEVER happen, and you are encouraged to report any log
18047 containing this, because this would almost certainly be a bug. It
18048 would be wise to preventively restart the process after such an
18049 event too, in case it would be caused by memory corruption.
18050
Simon Horman752dc4a2011-06-21 14:34:59 +090018051 D : the session was killed by haproxy because the server was detected
18052 as down and was configured to kill all connections when going down.
18053
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070018054 U : the session was killed by haproxy on this backup server because an
18055 active server was detected as up and was configured to kill all
18056 backup connections when going up.
18057
Willy Tarreaua2a64e92011-09-07 23:01:56 +020018058 K : the session was actively killed by an admin operating on haproxy.
18059
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018060 c : the client-side timeout expired while waiting for the client to
18061 send or receive data.
18062
18063 s : the server-side timeout expired while waiting for the server to
18064 send or receive data.
18065
18066 - : normal session completion, both the client and the server closed
18067 with nothing left in the buffers.
18068
18069 - on the second character, the TCP or HTTP session state when it was closed :
18070
Willy Tarreauf7b30a92010-12-06 22:59:17 +010018071 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018072 (HTTP mode only). Nothing was sent to any server.
18073
18074 Q : the proxy was waiting in the QUEUE for a connection slot. This can
18075 only happen when servers have a 'maxconn' parameter set. It can
18076 also happen in the global queue after a redispatch consecutive to
18077 a failed attempt to connect to a dying server. If no redispatch is
18078 reported, then no connection attempt was made to any server.
18079
18080 C : the proxy was waiting for the CONNECTION to establish on the
18081 server. The server might at most have noticed a connection attempt.
18082
18083 H : the proxy was waiting for complete, valid response HEADERS from the
18084 server (HTTP only).
18085
18086 D : the session was in the DATA phase.
18087
18088 L : the proxy was still transmitting LAST data to the client while the
18089 server had already finished. This one is very rare as it can only
18090 happen when the client dies while receiving the last packets.
18091
18092 T : the request was tarpitted. It has been held open with the client
18093 during the whole "timeout tarpit" duration or until the client
18094 closed, both of which will be reported in the "Tw" timer.
18095
18096 - : normal session completion after end of data transfer.
18097
18098 - the third character tells whether the persistence cookie was provided by
18099 the client (only in HTTP mode) :
18100
18101 N : the client provided NO cookie. This is usually the case for new
18102 visitors, so counting the number of occurrences of this flag in the
18103 logs generally indicate a valid trend for the site frequentation.
18104
18105 I : the client provided an INVALID cookie matching no known server.
18106 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020018107 cookies between HTTP/HTTPS sites, persistence conditionally
18108 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018109
18110 D : the client provided a cookie designating a server which was DOWN,
18111 so either "option persist" was used and the client was sent to
18112 this server, or it was not set and the client was redispatched to
18113 another server.
18114
Willy Tarreau996a92c2010-10-13 19:30:47 +020018115 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018116 server.
18117
Willy Tarreau996a92c2010-10-13 19:30:47 +020018118 E : the client provided a valid cookie, but with a last date which was
18119 older than what is allowed by the "maxidle" cookie parameter, so
18120 the cookie is consider EXPIRED and is ignored. The request will be
18121 redispatched just as if there was no cookie.
18122
18123 O : the client provided a valid cookie, but with a first date which was
18124 older than what is allowed by the "maxlife" cookie parameter, so
18125 the cookie is consider too OLD and is ignored. The request will be
18126 redispatched just as if there was no cookie.
18127
Willy Tarreauc89ccb62012-04-05 21:18:22 +020018128 U : a cookie was present but was not used to select the server because
18129 some other server selection mechanism was used instead (typically a
18130 "use-server" rule).
18131
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018132 - : does not apply (no cookie set in configuration).
18133
18134 - the last character reports what operations were performed on the persistence
18135 cookie returned by the server (only in HTTP mode) :
18136
18137 N : NO cookie was provided by the server, and none was inserted either.
18138
18139 I : no cookie was provided by the server, and the proxy INSERTED one.
18140 Note that in "cookie insert" mode, if the server provides a cookie,
18141 it will still be overwritten and reported as "I" here.
18142
Willy Tarreau996a92c2010-10-13 19:30:47 +020018143 U : the proxy UPDATED the last date in the cookie that was presented by
18144 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018145 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020018146 date indicated in the cookie. If any other change happens, such as
18147 a redispatch, then the cookie will be marked as inserted instead.
18148
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018149 P : a cookie was PROVIDED by the server and transmitted as-is.
18150
18151 R : the cookie provided by the server was REWRITTEN by the proxy, which
18152 happens in "cookie rewrite" or "cookie prefix" modes.
18153
18154 D : the cookie provided by the server was DELETED by the proxy.
18155
18156 - : does not apply (no cookie set in configuration).
18157
Willy Tarreau996a92c2010-10-13 19:30:47 +020018158The combination of the two first flags gives a lot of information about what
18159was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018160helpful to detect server saturation, network troubles, local system resource
18161starvation, attacks, etc...
18162
18163The most common termination flags combinations are indicated below. They are
18164alphabetically sorted, with the lowercase set just after the upper case for
18165easier finding and understanding.
18166
18167 Flags Reason
18168
18169 -- Normal termination.
18170
18171 CC The client aborted before the connection could be established to the
18172 server. This can happen when haproxy tries to connect to a recently
18173 dead (or unchecked) server, and the client aborts while haproxy is
18174 waiting for the server to respond or for "timeout connect" to expire.
18175
18176 CD The client unexpectedly aborted during data transfer. This can be
18177 caused by a browser crash, by an intermediate equipment between the
18178 client and haproxy which decided to actively break the connection,
18179 by network routing issues between the client and haproxy, or by a
18180 keep-alive session between the server and the client terminated first
18181 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010018182
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018183 cD The client did not send nor acknowledge any data for as long as the
18184 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020018185 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018186
18187 CH The client aborted while waiting for the server to start responding.
18188 It might be the server taking too long to respond or the client
18189 clicking the 'Stop' button too fast.
18190
18191 cH The "timeout client" stroke while waiting for client data during a
18192 POST request. This is sometimes caused by too large TCP MSS values
18193 for PPPoE networks which cannot transport full-sized packets. It can
18194 also happen when client timeout is smaller than server timeout and
18195 the server takes too long to respond.
18196
18197 CQ The client aborted while its session was queued, waiting for a server
18198 with enough empty slots to accept it. It might be that either all the
18199 servers were saturated or that the assigned server was taking too
18200 long a time to respond.
18201
18202 CR The client aborted before sending a full HTTP request. Most likely
18203 the request was typed by hand using a telnet client, and aborted
18204 too early. The HTTP status code is likely a 400 here. Sometimes this
18205 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020018206 and the client. "option http-ignore-probes" can be used to ignore
18207 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018208
18209 cR The "timeout http-request" stroke before the client sent a full HTTP
18210 request. This is sometimes caused by too large TCP MSS values on the
18211 client side for PPPoE networks which cannot transport full-sized
18212 packets, or by clients sending requests by hand and not typing fast
18213 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020018214 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020018215 some browsers started to implement a "pre-connect" feature consisting
18216 in speculatively connecting to some recently visited web sites just
18217 in case the user would like to visit them. This results in many
18218 connections being established to web sites, which end up in 408
18219 Request Timeout if the timeout strikes first, or 400 Bad Request when
18220 the browser decides to close them first. These ones pollute the log
18221 and feed the error counters. Some versions of some browsers have even
18222 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010018223 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020018224 http-ignore-probes" in the frontend, resulting in connections with
18225 zero data transfer to be totally ignored. This will definitely hide
18226 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018227
18228 CT The client aborted while its session was tarpitted. It is important to
18229 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020018230 wrong tarpit rules have been written. If a lot of them happen, it
18231 might make sense to lower the "timeout tarpit" value to something
18232 closer to the average reported "Tw" timer, in order not to consume
18233 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018234
Willy Tarreau570f2212013-06-10 16:42:09 +020018235 LR The request was intercepted and locally handled by haproxy. Generally
18236 it means that this was a redirect or a stats request.
18237
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018238 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018239 the TCP connection (the proxy received a TCP RST or an ICMP message
18240 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010018241 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018242 or no ARP response on local network). When this happens in HTTP mode,
18243 the status code is likely a 502 or 503 here.
18244
18245 sC The "timeout connect" stroke before a connection to the server could
18246 complete. When this happens in HTTP mode, the status code is likely a
18247 503 or 504 here.
18248
18249 SD The connection to the server died with an error during the data
18250 transfer. This usually means that haproxy has received an RST from
18251 the server or an ICMP message from an intermediate equipment while
18252 exchanging data with the server. This can be caused by a server crash
18253 or by a network issue on an intermediate equipment.
18254
18255 sD The server did not send nor acknowledge any data for as long as the
18256 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010018257 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018258 load-balancers, ...), as well as keep-alive sessions maintained
18259 between the client and the server expiring first on haproxy.
18260
18261 SH The server aborted before sending its full HTTP response headers, or
18262 it crashed while processing the request. Since a server aborting at
18263 this moment is very rare, it would be wise to inspect its logs to
18264 control whether it crashed and why. The logged request may indicate a
18265 small set of faulty requests, demonstrating bugs in the application.
18266 Sometimes this might also be caused by an IDS killing the connection
18267 between haproxy and the server.
18268
18269 sH The "timeout server" stroke before the server could return its
18270 response headers. This is the most common anomaly, indicating too
18271 long transactions, probably caused by server or database saturation.
18272 The immediate workaround consists in increasing the "timeout server"
18273 setting, but it is important to keep in mind that the user experience
18274 will suffer from these long response times. The only long term
18275 solution is to fix the application.
18276
18277 sQ The session spent too much time in queue and has been expired. See
18278 the "timeout queue" and "timeout connect" settings to find out how to
18279 fix this if it happens too often. If it often happens massively in
18280 short periods, it may indicate general problems on the affected
18281 servers due to I/O or database congestion, or saturation caused by
18282 external attacks.
18283
18284 PC The proxy refused to establish a connection to the server because the
18285 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020018286 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018287 so that it does not happen anymore. This status is very rare and
18288 might happen when the global "ulimit-n" parameter is forced by hand.
18289
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018290 PD The proxy blocked an incorrectly formatted chunked encoded message in
18291 a request or a response, after the server has emitted its headers. In
18292 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010018293 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020018294 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018295
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018296 PH The proxy blocked the server's response, because it was invalid,
18297 incomplete, dangerous (cache control), or matched a security filter.
18298 In any case, an HTTP 502 error is sent to the client. One possible
18299 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018300 containing unauthorized characters. It is also possible but quite
18301 rare, that the proxy blocked a chunked-encoding request from the
18302 client due to an invalid syntax, before the server responded. In this
18303 case, an HTTP 400 error is sent to the client and reported in the
18304 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018305
18306 PR The proxy blocked the client's HTTP request, either because of an
18307 invalid HTTP syntax, in which case it returned an HTTP 400 error to
18308 the client, or because a deny filter matched, in which case it
18309 returned an HTTP 403 error.
18310
18311 PT The proxy blocked the client's request and has tarpitted its
18312 connection before returning it a 500 server error. Nothing was sent
18313 to the server. The connection was maintained open for as long as
18314 reported by the "Tw" timer field.
18315
18316 RC A local resource has been exhausted (memory, sockets, source ports)
18317 preventing the connection to the server from establishing. The error
18318 logs will tell precisely what was missing. This is very rare and can
18319 only be solved by proper system tuning.
18320
Willy Tarreau996a92c2010-10-13 19:30:47 +020018321The combination of the two last flags gives a lot of information about how
18322persistence was handled by the client, the server and by haproxy. This is very
18323important to troubleshoot disconnections, when users complain they have to
18324re-authenticate. The commonly encountered flags are :
18325
18326 -- Persistence cookie is not enabled.
18327
18328 NN No cookie was provided by the client, none was inserted in the
18329 response. For instance, this can be in insert mode with "postonly"
18330 set on a GET request.
18331
18332 II A cookie designating an invalid server was provided by the client,
18333 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040018334 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020018335 value can be presented by a client when no other server knows it.
18336
18337 NI No cookie was provided by the client, one was inserted in the
18338 response. This typically happens for first requests from every user
18339 in "insert" mode, which makes it an easy way to count real users.
18340
18341 VN A cookie was provided by the client, none was inserted in the
18342 response. This happens for most responses for which the client has
18343 already got a cookie.
18344
18345 VU A cookie was provided by the client, with a last visit date which is
18346 not completely up-to-date, so an updated cookie was provided in
18347 response. This can also happen if there was no date at all, or if
18348 there was a date but the "maxidle" parameter was not set, so that the
18349 cookie can be switched to unlimited time.
18350
18351 EI A cookie was provided by the client, with a last visit date which is
18352 too old for the "maxidle" parameter, so the cookie was ignored and a
18353 new cookie was inserted in the response.
18354
18355 OI A cookie was provided by the client, with a first visit date which is
18356 too old for the "maxlife" parameter, so the cookie was ignored and a
18357 new cookie was inserted in the response.
18358
18359 DI The server designated by the cookie was down, a new server was
18360 selected and a new cookie was emitted in the response.
18361
18362 VI The server designated by the cookie was not marked dead but could not
18363 be reached. A redispatch happened and selected another one, which was
18364 then advertised in the response.
18365
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018366
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200183678.6. Non-printable characters
18368-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018369
18370In order not to cause trouble to log analysis tools or terminals during log
18371consulting, non-printable characters are not sent as-is into log files, but are
18372converted to the two-digits hexadecimal representation of their ASCII code,
18373prefixed by the character '#'. The only characters that can be logged without
18374being escaped are comprised between 32 and 126 (inclusive). Obviously, the
18375escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
18376is the same for the character '"' which becomes "#22", as well as '{', '|' and
18377'}' when logging headers.
18378
18379Note that the space character (' ') is not encoded in headers, which can cause
18380issues for tools relying on space count to locate fields. A typical header
18381containing spaces is "User-Agent".
18382
18383Last, it has been observed that some syslog daemons such as syslog-ng escape
18384the quote ('"') with a backslash ('\'). The reverse operation can safely be
18385performed since no quote may appear anywhere else in the logs.
18386
18387
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200183888.7. Capturing HTTP cookies
18389---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018390
18391Cookie capture simplifies the tracking a complete user session. This can be
18392achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018393section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018394cookie will simultaneously be checked in the request ("Cookie:" header) and in
18395the response ("Set-Cookie:" header). The respective values will be reported in
18396the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018397locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018398not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
18399user switches to a new session for example, because the server will reassign it
18400a new cookie. It is also possible to detect if a server unexpectedly sets a
18401wrong cookie to a client, leading to session crossing.
18402
18403 Examples :
18404 # capture the first cookie whose name starts with "ASPSESSION"
18405 capture cookie ASPSESSION len 32
18406
18407 # capture the first cookie whose name is exactly "vgnvisitor"
18408 capture cookie vgnvisitor= len 32
18409
18410
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200184118.8. Capturing HTTP headers
18412---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018413
18414Header captures are useful to track unique request identifiers set by an upper
18415proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
18416the response, one can search for information about the response length, how the
18417server asked the cache to behave, or an object location during a redirection.
18418
18419Header captures are performed using the "capture request header" and "capture
18420response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018421section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018422
18423It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018424time. Non-existent headers are logged as empty strings, and if one header
18425appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018426are grouped within braces '{' and '}' in the same order as they were declared,
18427and delimited with a vertical bar '|' without any space. Response headers
18428follow the same representation, but are displayed after a space following the
18429request headers block. These blocks are displayed just before the HTTP request
18430in the logs.
18431
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020018432As a special case, it is possible to specify an HTTP header capture in a TCP
18433frontend. The purpose is to enable logging of headers which will be parsed in
18434an HTTP backend if the request is then switched to this HTTP backend.
18435
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018436 Example :
18437 # This instance chains to the outgoing proxy
18438 listen proxy-out
18439 mode http
18440 option httplog
18441 option logasap
18442 log global
18443 server cache1 192.168.1.1:3128
18444
18445 # log the name of the virtual server
18446 capture request header Host len 20
18447
18448 # log the amount of data uploaded during a POST
18449 capture request header Content-Length len 10
18450
18451 # log the beginning of the referrer
18452 capture request header Referer len 20
18453
18454 # server name (useful for outgoing proxies only)
18455 capture response header Server len 20
18456
18457 # logging the content-length is useful with "option logasap"
18458 capture response header Content-Length len 10
18459
Davor Ocelice9ed2812017-12-25 17:49:28 +010018460 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018461 capture response header Cache-Control len 8
18462
18463 # the Via header will report the next proxy's name
18464 capture response header Via len 20
18465
18466 # log the URL location during a redirection
18467 capture response header Location len 20
18468
18469 >>> Aug 9 20:26:09 localhost \
18470 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
18471 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
18472 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
18473 "GET http://fr.adserver.yahoo.com/"
18474
18475 >>> Aug 9 20:30:46 localhost \
18476 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
18477 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
18478 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018479 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018480
18481 >>> Aug 9 20:30:46 localhost \
18482 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
18483 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
18484 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
18485 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018486 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018487
18488
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200184898.9. Examples of logs
18490---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018491
18492These are real-world examples of logs accompanied with an explanation. Some of
18493them have been made up by hand. The syslog part has been removed for better
18494reading. Their sole purpose is to explain how to decipher them.
18495
18496 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
18497 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
18498 "HEAD / HTTP/1.0"
18499
18500 => long request (6.5s) entered by hand through 'telnet'. The server replied
18501 in 147 ms, and the session ended normally ('----')
18502
18503 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
18504 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
18505 0/9 "HEAD / HTTP/1.0"
18506
18507 => Idem, but the request was queued in the global queue behind 9 other
18508 requests, and waited there for 1230 ms.
18509
18510 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
18511 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
18512 "GET /image.iso HTTP/1.0"
18513
18514 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018515 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018516 14 ms, 243 bytes of headers were sent to the client, and total time from
18517 accept to first data byte is 30 ms.
18518
18519 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
18520 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
18521 "GET /cgi-bin/bug.cgi? HTTP/1.0"
18522
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020018523 => the proxy blocked a server response either because of an "http-response
18524 deny" rule, or because the response was improperly formatted and not
18525 HTTP-compliant, or because it blocked sensitive information which risked
18526 being cached. In this case, the response is replaced with a "502 bad
18527 gateway". The flags ("PH--") tell us that it was haproxy who decided to
18528 return the 502 and not the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018529
18530 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018531 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 +010018532
18533 => the client never completed its request and aborted itself ("C---") after
18534 8.5s, while the proxy was waiting for the request headers ("-R--").
18535 Nothing was sent to any server.
18536
18537 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
18538 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
18539
18540 => The client never completed its request, which was aborted by the
18541 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010018542 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018543 send a 408 return code to the client.
18544
18545 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
18546 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
18547
18548 => This log was produced with "option tcplog". The client timed out after
18549 5 seconds ("c----").
18550
18551 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
18552 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018553 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018554
18555 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018556 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018557 (config says 'retries 3'), and no redispatch (otherwise we would have
18558 seen "/+3"). Status code 503 was returned to the client. There were 115
18559 connections on this server, 202 connections on this proxy, and 205 on
18560 the global process. It is possible that the server refused the
18561 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010018562
Willy Tarreau52b2d222011-09-07 23:48:48 +020018563
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200185649. Supported filters
18565--------------------
18566
18567Here are listed officially supported filters with the list of parameters they
18568accept. Depending on compile options, some of these filters might be
18569unavailable. The list of available filters is reported in haproxy -vv.
18570
18571See also : "filter"
18572
185739.1. Trace
18574----------
18575
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018576filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018577
18578 Arguments:
18579 <name> is an arbitrary name that will be reported in
18580 messages. If no name is provided, "TRACE" is used.
18581
18582 <random-parsing> enables the random parsing of data exchanged between
18583 the client and the server. By default, this filter
18584 parses all available data. With this parameter, it
18585 only parses a random amount of the available data.
18586
Davor Ocelice9ed2812017-12-25 17:49:28 +010018587 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018588 default, this filter forwards all previously parsed
18589 data. With this parameter, it only forwards a random
18590 amount of the parsed data.
18591
Davor Ocelice9ed2812017-12-25 17:49:28 +010018592 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018593
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018594This filter can be used as a base to develop new filters. It defines all
18595callbacks and print a message on the standard error stream (stderr) with useful
18596information for all of them. It may be useful to debug the activity of other
18597filters or, quite simply, HAProxy's activity.
18598
18599Using <random-parsing> and/or <random-forwarding> parameters is a good way to
18600tests the behavior of a filter that parses data exchanged between a client and
18601a server by adding some latencies in the processing.
18602
18603
186049.2. HTTP compression
18605---------------------
18606
18607filter compression
18608
18609The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
18610keyword must still be used to enable and configure the HTTP compression. And
Christopher Fauletb30b3102019-09-12 23:03:09 +020018611when no other filter is used, it is enough. When used with the cache or the
18612fcgi-app enabled, it is also enough. In this case, the compression is always
18613done after the response is stored in the cache. But it is mandatory to
18614explicitly use a filter line to enable the HTTP compression when at least one
18615filter other than the cache or the fcgi-app is used for the same
18616listener/frontend/backend. This is important to know the filters evaluation
18617order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018618
Christopher Fauletb30b3102019-09-12 23:03:09 +020018619See also : "compression", section 9.4 about the cache filter and section 9.5
18620 about the fcgi-app filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018621
18622
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200186239.3. Stream Processing Offload Engine (SPOE)
18624--------------------------------------------
18625
18626filter spoe [engine <name>] config <file>
18627
18628 Arguments :
18629
18630 <name> is the engine name that will be used to find the right scope in
18631 the configuration file. If not provided, all the file will be
18632 parsed.
18633
18634 <file> is the path of the engine configuration file. This file can
18635 contain configuration of several engines. In this case, each
18636 part must be placed in its own scope.
18637
18638The Stream Processing Offload Engine (SPOE) is a filter communicating with
18639external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010018640streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018641exchanged with them are configured in dedicated files, for the main part. It
18642also requires dedicated backends, defined in HAProxy configuration.
18643
18644SPOE communicates with external components using an in-house binary protocol,
18645the Stream Processing Offload Protocol (SPOP).
18646
Tim Düsterhus4896c442016-11-29 02:15:19 +010018647For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018648"doc/SPOE.txt".
18649
18650Important note:
18651 The SPOE filter is highly experimental for now and was not heavily
18652 tested. It is really not production ready. So use it carefully.
18653
Christopher Faulet99a17a22018-12-11 09:18:27 +0100186549.4. Cache
18655----------
18656
18657filter cache <name>
18658
18659 Arguments :
18660
18661 <name> is name of the cache section this filter will use.
18662
18663The cache uses a filter to store cacheable responses. The HTTP rules
18664"cache-store" and "cache-use" must be used to define how and when to use a
John Roeslerfb2fce12019-07-10 15:45:51 -050018665cache. By default the corresponding filter is implicitly defined. And when no
Christopher Fauletb30b3102019-09-12 23:03:09 +020018666other filters than fcgi-app or compression are used, it is enough. In such
18667case, the compression filter is always evaluated after the cache filter. But it
18668is mandatory to explicitly use a filter line to use a cache when at least one
18669filter other than the compression or the fcgi-app is used for the same
Christopher Faulet27d93c32018-12-15 22:32:02 +010018670listener/frontend/backend. This is important to know the filters evaluation
18671order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018672
Christopher Fauletb30b3102019-09-12 23:03:09 +020018673See also : section 9.2 about the compression filter, section 9.5 about the
18674 fcgi-app filter and section 6 about cache.
18675
18676
186779.5. Fcgi-app
18678-------------
18679
18680filter fcg-app <name>
18681
18682 Arguments :
18683
18684 <name> is name of the fcgi-app section this filter will use.
18685
18686The FastCGI application uses a filter to evaluate all custom parameters on the
18687request path, and to process the headers on the response path. the <name> must
18688reference an existing fcgi-app section. The directive "use-fcgi-app" should be
18689used to define the application to use. By default the corresponding filter is
18690implicitly defined. And when no other filters than cache or compression are
18691used, it is enough. But it is mandatory to explicitly use a filter line to a
18692fcgi-app when at least one filter other than the compression or the cache is
18693used for the same backend. This is important to know the filters evaluation
18694order.
18695
18696See also: "use-fcgi-app", section 9.2 about the compression filter, section 9.4
18697 about the cache filter and section 10 about FastCGI application.
18698
18699
1870010. FastCGI applications
18701-------------------------
18702
18703HAProxy is able to send HTTP requests to Responder FastCGI applications. This
18704feature was added in HAProxy 2.1. To do so, servers must be configured to use
18705the FastCGI protocol (using the keyword "proto fcgi" on the server line) and a
18706FastCGI application must be configured and used by the backend managing these
18707servers (using the keyword "use-fcgi-app" into the proxy section). Several
18708FastCGI applications may be defined, but only one can be used at a time by a
18709backend.
18710
18711HAProxy implements all features of the FastCGI specification for Responder
18712application. Especially it is able to multiplex several requests on a simple
18713connection.
18714
1871510.1. Setup
18716-----------
18717
1871810.1.1. Fcgi-app section
18719--------------------------
18720
18721fcgi-app <name>
18722 Declare a FastCGI application named <name>. To be valid, at least the
18723 document root must be defined.
18724
18725acl <aclname> <criterion> [flags] [operator] <value> ...
18726 Declare or complete an access list.
18727
18728 See "acl" keyword in section 4.2 and section 7 about ACL usage for
18729 details. ACLs defined for a FastCGI application are private. They cannot be
18730 used by any other application or by any proxy. In the same way, ACLs defined
18731 in any other section are not usable by a FastCGI application. However,
18732 Pre-defined ACLs are available.
18733
18734docroot <path>
18735 Define the document root on the remote host. <path> will be used to build
18736 the default value of FastCGI parameters SCRIPT_FILENAME and
18737 PATH_TRANSLATED. It is a mandatory setting.
18738
18739index <script-name>
18740 Define the script name that will be appended after an URI that ends with a
18741 slash ("/") to set the default value of the FastCGI parameter SCRIPT_NAME. It
18742 is an optional setting.
18743
18744 Example :
18745 index index.php
18746
18747log-stderr global
18748log-stderr <address> [len <length>] [format <format>]
18749 [sample <ranges>:<smp_size>] <facility> [<level> [<minlevel>]]
18750 Enable logging of STDERR messages reported by the FastCGI application.
18751
18752 See "log" keyword in section 4.2 for details. It is an optional setting. By
18753 default STDERR messages are ignored.
18754
18755pass-header <name> [ { if | unless } <condition> ]
18756 Specify the name of a request header which will be passed to the FastCGI
18757 application. It may optionally be followed by an ACL-based condition, in
18758 which case it will only be evaluated if the condition is true.
18759
18760 Most request headers are already available to the FastCGI application,
18761 prefixed with "HTTP_". Thus, this directive is only required to pass headers
18762 that are purposefully omitted. Currently, the headers "Authorization",
18763 "Proxy-Authorization" and hop-by-hop headers are omitted.
18764
18765 Note that the headers "Content-type" and "Content-length" are never passed to
18766 the FastCGI application because they are already converted into parameters.
18767
18768path-info <regex>
Christopher Faulet28cb3662020-02-14 14:47:37 +010018769 Define a regular expression to extract the script-name and the path-info from
Christopher Faulet6c57f2d2020-02-14 16:55:52 +010018770 the URL-decoded path. Thus, <regex> may have two captures: the first one to
18771 capture the script name and the second one to capture the path-info. The
18772 first one is mandatory, the second one is optional. This way, it is possible
18773 to extract the script-name from the path ignoring the path-info. It is an
18774 optional setting. If it is not defined, no matching is performed on the
18775 path. and the FastCGI parameters PATH_INFO and PATH_TRANSLATED are not
18776 filled.
Christopher Faulet28cb3662020-02-14 14:47:37 +010018777
18778 For security reason, when this regular expression is defined, the newline and
18779 the null characters are forbiden from the path, once URL-decoded. The reason
18780 to such limitation is because otherwise the matching always fails (due to a
18781 limitation one the way regular expression are executed in HAProxy). So if one
18782 of these two characters is found in the URL-decoded path, an error is
18783 returned to the client. The principle of least astonishment is applied here.
Christopher Fauletb30b3102019-09-12 23:03:09 +020018784
18785 Example :
Christopher Faulet6c57f2d2020-02-14 16:55:52 +010018786 path-info ^(/.+\.php)(/.*)?$ # both script-name and path-info may be set
18787 path-info ^(/.+\.php) # the path-info is ignored
Christopher Fauletb30b3102019-09-12 23:03:09 +020018788
18789option get-values
18790no option get-values
18791 Enable or disable the retrieve of variables about connection management.
18792
18793 HAproxy is able to send the record FCGI_GET_VALUES on connection
18794 establishment to retrieve the value for following variables:
18795
18796 * FCGI_MAX_REQS The maximum number of concurrent requests this
18797 application will accept.
18798
William Lallemand93e548e2019-09-30 13:54:02 +020018799 * FCGI_MPXS_CONNS "0" if this application does not multiplex connections,
18800 "1" otherwise.
Christopher Fauletb30b3102019-09-12 23:03:09 +020018801
18802 Some FastCGI applications does not support this feature. Some others close
18803 the connexion immediately after sending their response. So, by default, this
18804 option is disabled.
18805
18806 Note that the maximum number of concurrent requests accepted by a FastCGI
18807 application is a connection variable. It only limits the number of streams
18808 per connection. If the global load must be limited on the application, the
18809 server parameters "maxconn" and "pool-max-conn" must be set. In addition, if
18810 an application does not support connection multiplexing, the maximum number
18811 of concurrent requests is automatically set to 1.
18812
18813option keep-conn
18814no option keep-conn
18815 Instruct the FastCGI application to keep the connection open or not after
18816 sending a response.
18817
18818 If disabled, the FastCGI application closes the connection after responding
18819 to this request. By default, this option is enabled.
18820
18821option max-reqs <reqs>
18822 Define the maximum number of concurrent requests this application will
18823 accept.
18824
18825 This option may be overwritten if the variable FCGI_MAX_REQS is retrieved
18826 during connection establishment. Furthermore, if the application does not
18827 support connection multiplexing, this option will be ignored. By default set
18828 to 1.
18829
18830option mpxs-conns
18831no option mpxs-conns
18832 Enable or disable the support of connection multiplexing.
18833
18834 This option may be overwritten if the variable FCGI_MPXS_CONNS is retrieved
18835 during connection establishment. It is disabled by default.
18836
18837set-param <name> <fmt> [ { if | unless } <condition> ]
18838 Set a FastCGI parameter that should be passed to this application. Its
18839 value, defined by <fmt> must follows the log-format rules (see section 8.2.4
18840 "Custom Log format"). It may optionally be followed by an ACL-based
18841 condition, in which case it will only be evaluated if the condition is true.
18842
18843 With this directive, it is possible to overwrite the value of default FastCGI
18844 parameters. If the value is evaluated to an empty string, the rule is
18845 ignored. These directives are evaluated in their declaration order.
18846
18847 Example :
18848 # PHP only, required if PHP was built with --enable-force-cgi-redirect
18849 set-param REDIRECT_STATUS 200
18850
18851 set-param PHP_AUTH_DIGEST %[req.hdr(Authorization)]
18852
18853
1885410.1.2. Proxy section
18855---------------------
18856
18857use-fcgi-app <name>
18858 Define the FastCGI application to use for the backend.
18859
18860 Arguments :
18861 <name> is the name of the FastCGI application to use.
18862
18863 This keyword is only available for HTTP proxies with the backend capability
18864 and with at least one FastCGI server. However, FastCGI servers can be mixed
18865 with HTTP servers. But except there is a good reason to do so, it is not
18866 recommended (see section 10.3 about the limitations for details). Only one
18867 application may be defined at a time per backend.
18868
18869 Note that, once a FastCGI application is referenced for a backend, depending
18870 on the configuration some processing may be done even if the request is not
18871 sent to a FastCGI server. Rules to set parameters or pass headers to an
18872 application are evaluated.
18873
18874
1887510.1.3. Example
18876---------------
18877
18878 frontend front-http
18879 mode http
18880 bind *:80
18881 bind *:
18882
18883 use_backend back-dynamic if { path_reg ^/.+\.php(/.*)?$ }
18884 default_backend back-static
18885
18886 backend back-static
18887 mode http
18888 server www A.B.C.D:80
18889
18890 backend back-dynamic
18891 mode http
18892 use-fcgi-app php-fpm
18893 server php-fpm A.B.C.D:9000 proto fcgi
18894
18895 fcgi-app php-fpm
18896 log-stderr global
18897 option keep-conn
18898
18899 docroot /var/www/my-app
18900 index index.php
18901 path-info ^(/.+\.php)(/.*)?$
18902
18903
1890410.2. Default parameters
18905------------------------
18906
18907A Responder FastCGI application has the same purpose as a CGI/1.1 program. In
18908the CGI/1.1 specification (RFC3875), several variables must be passed to the
18909scipt. So HAProxy set them and some others commonly used by FastCGI
18910applications. All these variables may be overwritten, with caution though.
18911
18912 +-------------------+-----------------------------------------------------+
18913 | AUTH_TYPE | Identifies the mechanism, if any, used by HAProxy |
18914 | | to authenticate the user. Concretely, only the |
18915 | | BASIC authentication mechanism is supported. |
18916 | | |
18917 +-------------------+-----------------------------------------------------+
18918 | CONTENT_LENGTH | Contains the size of the message-body attached to |
18919 | | the request. It means only requests with a known |
18920 | | size are considered as valid and sent to the |
18921 | | application. |
18922 | | |
18923 +-------------------+-----------------------------------------------------+
18924 | CONTENT_TYPE | Contains the type of the message-body attached to |
18925 | | the request. It may not be set. |
18926 | | |
18927 +-------------------+-----------------------------------------------------+
18928 | DOCUMENT_ROOT | Contains the document root on the remote host under |
18929 | | which the script should be executed, as defined in |
18930 | | the application's configuration. |
18931 | | |
18932 +-------------------+-----------------------------------------------------+
18933 | GATEWAY_INTERFACE | Contains the dialect of CGI being used by HAProxy |
18934 | | to communicate with the FastCGI application. |
18935 | | Concretely, it is set to "CGI/1.1". |
18936 | | |
18937 +-------------------+-----------------------------------------------------+
18938 | PATH_INFO | Contains the portion of the URI path hierarchy |
18939 | | following the part that identifies the script |
18940 | | itself. To be set, the directive "path-info" must |
18941 | | be defined. |
18942 | | |
18943 +-------------------+-----------------------------------------------------+
18944 | PATH_TRANSLATED | If PATH_INFO is set, it is its translated version. |
18945 | | It is the concatenation of DOCUMENT_ROOT and |
18946 | | PATH_INFO. If PATH_INFO is not set, this parameters |
18947 | | is not set too. |
18948 | | |
18949 +-------------------+-----------------------------------------------------+
18950 | QUERY_STRING | Contains the request's query string. It may not be |
18951 | | set. |
18952 | | |
18953 +-------------------+-----------------------------------------------------+
18954 | REMOTE_ADDR | Contains the network address of the client sending |
18955 | | the request. |
18956 | | |
18957 +-------------------+-----------------------------------------------------+
18958 | REMOTE_USER | Contains the user identification string supplied by |
18959 | | client as part of user authentication. |
18960 | | |
18961 +-------------------+-----------------------------------------------------+
18962 | REQUEST_METHOD | Contains the method which should be used by the |
18963 | | script to process the request. |
18964 | | |
18965 +-------------------+-----------------------------------------------------+
18966 | REQUEST_URI | Contains the request's URI. |
18967 | | |
18968 +-------------------+-----------------------------------------------------+
18969 | SCRIPT_FILENAME | Contains the absolute pathname of the script. it is |
18970 | | the concatenation of DOCUMENT_ROOT and SCRIPT_NAME. |
18971 | | |
18972 +-------------------+-----------------------------------------------------+
18973 | SCRIPT_NAME | Contains the name of the script. If the directive |
18974 | | "path-info" is defined, it is the first part of the |
18975 | | URI path hierarchy, ending with the script name. |
18976 | | Otherwise, it is the entire URI path. |
18977 | | |
18978 +-------------------+-----------------------------------------------------+
18979 | SERVER_NAME | Contains the name of the server host to which the |
18980 | | client request is directed. It is the value of the |
18981 | | header "Host", if defined. Otherwise, the |
18982 | | destination address of the connection on the client |
18983 | | side. |
18984 | | |
18985 +-------------------+-----------------------------------------------------+
18986 | SERVER_PORT | Contains the destination TCP port of the connection |
18987 | | on the client side, which is the port the client |
18988 | | connected to. |
18989 | | |
18990 +-------------------+-----------------------------------------------------+
18991 | SERVER_PROTOCOL | Contains the request's protocol. |
18992 | | |
18993 +-------------------+-----------------------------------------------------+
18994 | HTTPS | Set to a non-empty value ("on") if the script was |
18995 | | queried through the HTTPS protocol. |
18996 | | |
18997 +-------------------+-----------------------------------------------------+
18998
18999
1900010.3. Limitations
19001------------------
19002
19003The current implementation have some limitations. The first one is about the
19004way some request headers are hidden to the FastCGI applications. This happens
19005during the headers analysis, on the backend side, before the connection
19006establishment. At this stage, HAProxy know the backend is using a FastCGI
19007application but it don't know if the request will be routed to a FastCGI server
19008or not. But to hide request headers, it simply removes them from the HTX
19009message. So, if the request is finally routed to an HTTP server, it never see
19010these headers. For this reason, it is not recommended to mix FastCGI servers
19011and HTTP servers under the same backend.
19012
19013Similarly, the rules "set-param" and "pass-header" are evaluated during the
19014request headers analysis. So the evaluation is always performed, even if the
19015requests is finally forwarded to an HTTP server.
19016
19017About the rules "set-param", when a rule is applied, a pseudo header is added
19018into the HTX message. So, the same way than for HTTP header rewrites, it may
19019fail if the buffer is full. The rules "set-param" will compete with
19020"http-request" ones.
19021
19022Finally, all FastCGI params and HTTP headers are sent into a unique record
19023FCGI_PARAM. Encoding of this record must be done in one pass, otherwise a
19024processing error is returned. It means the record FCGI_PARAM, once encoded,
19025must not exceeds the size of a buffer. However, there is no reserve to respect
19026here.
William Lallemand86d0df02017-11-24 21:36:45 +010019027
Willy Tarreau0ba27502007-12-24 16:55:16 +010019028/*
19029 * Local variables:
19030 * fill-column: 79
19031 * End:
19032 */