blob: bfc8b9030e4fbe9741db50bb3ce14a67c4417927 [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 Tarreaue54b43a2019-11-25 19:47:40 +01007 2019/11/25
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
Willy Tarreauc57f0e22009-05-10 13:12:33 +020055
564. Proxies
574.1. Proxy keywords matrix
584.2. Alphabetically sorted keywords reference
59
Davor Ocelice9ed2812017-12-25 17:49:28 +0100605. Bind and server options
Willy Tarreau086fbf52012-09-24 20:34:51 +0200615.1. Bind options
625.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200635.3. Server DNS resolution
645.3.1. Global overview
655.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020066
Julien Pivotto6ccee412019-11-27 15:49:54 +0100676. Cache
686.1. Limitation
696.2. Setup
706.2.1. Cache section
716.2.2. Proxy section
72
Willy Tarreau74ca5042013-06-11 23:12:07 +0200737. Using ACLs and fetching samples
747.1. ACL basics
757.1.1. Matching booleans
767.1.2. Matching integers
777.1.3. Matching strings
787.1.4. Matching regular expressions (regexes)
797.1.5. Matching arbitrary data blocks
807.1.6. Matching IPv4 and IPv6 addresses
817.2. Using ACLs to form conditions
827.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200837.3.1. Converters
847.3.2. Fetching samples from internal states
857.3.3. Fetching samples at Layer 4
867.3.4. Fetching samples at Layer 5
877.3.5. Fetching samples from buffer contents (Layer 6)
887.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200897.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020090
918. Logging
928.1. Log levels
938.2. Log formats
948.2.1. Default log format
958.2.2. TCP log format
968.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100978.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100988.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200998.3. Advanced logging options
1008.3.1. Disabling logging of external tests
1018.3.2. Logging before waiting for the session to terminate
1028.3.3. Raising log level upon errors
1038.3.4. Disabling logging of successful connections
1048.4. Timing events
1058.5. Session state at disconnection
1068.6. Non-printable characters
1078.7. Capturing HTTP cookies
1088.8. Capturing HTTP headers
1098.9. Examples of logs
110
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001119. Supported filters
1129.1. Trace
1139.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001149.3. Stream Processing Offload Engine (SPOE)
Christopher Faulet99a17a22018-12-11 09:18:27 +01001159.4. Cache
Christopher Fauletb30b3102019-09-12 23:03:09 +02001169.5. fcgi-app
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200117
Christopher Fauletb30b3102019-09-12 23:03:09 +020011810. FastCGI applications
11910.1. Setup
12010.1.1. Fcgi-app section
12110.1.2. Proxy section
12210.1.3. Example
12310.2. Default parameters
12410.3. Limitations
125
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200126
1271. Quick reminder about HTTP
128----------------------------
129
Davor Ocelice9ed2812017-12-25 17:49:28 +0100130When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200131fully analyzed and indexed, thus it becomes possible to build matching criteria
132on almost anything found in the contents.
133
134However, it is important to understand how HTTP requests and responses are
135formed, and how HAProxy decomposes them. It will then become easier to write
136correct rules and to debug existing configurations.
137
138
1391.1. The HTTP transaction model
140-------------------------------
141
142The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100143to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100144from the client to the server, a request is sent by the client through the
145connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200146will involve a new connection :
147
148 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
149
150In this mode, called the "HTTP close" mode, there are as many connection
151establishments as there are HTTP transactions. Since the connection is closed
152by the server after the response, the client does not need to know the content
153length.
154
155Due to the transactional nature of the protocol, it was possible to improve it
156to avoid closing a connection between two subsequent transactions. In this mode
157however, it is mandatory that the server indicates the content length for each
158response so that the client does not wait indefinitely. For this, a special
159header is used: "Content-length". This mode is called the "keep-alive" mode :
160
161 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
162
163Its advantages are a reduced latency between transactions, and less processing
164power required on the server side. It is generally better than the close mode,
165but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200166a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167
Willy Tarreau95c4e142017-11-26 12:18:55 +0100168Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169keep-alive, but the client does not wait for the first response to send the
170second request. This is useful for fetching large number of images composing a
171page :
172
173 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
174
175This can obviously have a tremendous benefit on performance because the network
176latency is eliminated between subsequent requests. Many HTTP agents do not
177correctly support pipelining since there is no way to associate a response with
178the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100179server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180
Willy Tarreau95c4e142017-11-26 12:18:55 +0100181The next improvement is the multiplexed mode, as implemented in HTTP/2. This
182time, each transaction is assigned a single stream identifier, and all streams
183are multiplexed over an existing connection. Many requests can be sent in
184parallel by the client, and responses can arrive in any order since they also
185carry the stream identifier.
186
Willy Tarreau70dffda2014-01-30 03:07:23 +0100187By default HAProxy operates in keep-alive mode with regards to persistent
188connections: for each connection it processes each request and response, and
189leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100190start of a new request. When it receives HTTP/2 connections from a client, it
191processes all the requests in parallel and leaves the connection idling,
192waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200193
Christopher Faulet315b39c2018-09-21 16:26:19 +0200194HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +0100195 - keep alive : all requests and responses are processed (default)
196 - tunnel : only the first request and response are processed,
Christopher Faulet6c9bbb22019-03-26 21:37:23 +0100197 everything else is forwarded with no analysis (deprecated).
Willy Tarreau70dffda2014-01-30 03:07:23 +0100198 - server close : the server-facing connection is closed after the response.
Christopher Faulet315b39c2018-09-21 16:26:19 +0200199 - close : the connection is actively closed after end of response.
Willy Tarreau70dffda2014-01-30 03:07:23 +0100200
Davor Ocelice9ed2812017-12-25 17:49:28 +0100201For HTTP/2, the connection mode resembles more the "server close" mode : given
202the independence of all streams, there is currently no place to hook the idle
Willy Tarreau95c4e142017-11-26 12:18:55 +0100203server connection after a response, so it is closed after the response. HTTP/2
204is only supported for incoming connections, not on connections going to
205servers.
206
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207
2081.2. HTTP request
209-----------------
210
211First, let's consider this HTTP request :
212
213 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100214 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200215 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
216 2 Host: www.mydomain.com
217 3 User-agent: my small browser
218 4 Accept: image/jpeg, image/gif
219 5 Accept: image/png
220
221
2221.2.1. The Request line
223-----------------------
224
225Line 1 is the "request line". It is always composed of 3 fields :
226
227 - a METHOD : GET
228 - a URI : /serv/login.php?lang=en&profile=2
229 - a version tag : HTTP/1.1
230
231All of them are delimited by what the standard calls LWS (linear white spaces),
232which are commonly spaces, but can also be tabs or line feeds/carriage returns
233followed by spaces/tabs. The method itself cannot contain any colon (':') and
234is limited to alphabetic letters. All those various combinations make it
235desirable that HAProxy performs the splitting itself rather than leaving it to
236the user to write a complex or inaccurate regular expression.
237
238The URI itself can have several forms :
239
240 - A "relative URI" :
241
242 /serv/login.php?lang=en&profile=2
243
244 It is a complete URL without the host part. This is generally what is
245 received by servers, reverse proxies and transparent proxies.
246
247 - An "absolute URI", also called a "URL" :
248
249 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
250
251 It is composed of a "scheme" (the protocol name followed by '://'), a host
252 name or address, optionally a colon (':') followed by a port number, then
253 a relative URI beginning at the first slash ('/') after the address part.
254 This is generally what proxies receive, but a server supporting HTTP/1.1
255 must accept this form too.
256
257 - a star ('*') : this form is only accepted in association with the OPTIONS
258 method and is not relayable. It is used to inquiry a next hop's
259 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100260
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200261 - an address:port combination : 192.168.0.12:80
262 This is used with the CONNECT method, which is used to establish TCP
263 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
264 other protocols too.
265
266In a relative URI, two sub-parts are identified. The part before the question
267mark is called the "path". It is typically the relative path to static objects
268on the server. The part after the question mark is called the "query string".
269It is mostly used with GET requests sent to dynamic scripts and is very
270specific to the language, framework or application in use.
271
Willy Tarreau95c4e142017-11-26 12:18:55 +0100272HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100273assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100274However, haproxy natively processes HTTP/1.x requests and headers, so requests
275received over an HTTP/2 connection are transcoded to HTTP/1.1 before being
276processed. This explains why they still appear as "HTTP/1.1" in haproxy's logs
277as well as in server logs.
278
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200279
2801.2.2. The request headers
281--------------------------
282
283The headers start at the second line. They are composed of a name at the
284beginning of the line, immediately followed by a colon (':'). Traditionally,
285an LWS is added after the colon but that's not required. Then come the values.
286Multiple identical headers may be folded into one single line, delimiting the
287values with commas, provided that their order is respected. This is commonly
288encountered in the "Cookie:" field. A header may span over multiple lines if
289the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
290define a total of 3 values for the "Accept:" header.
291
Davor Ocelice9ed2812017-12-25 17:49:28 +0100292Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200293their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100294"Connection:" header). In HTTP/2, header names are always sent in lower case,
295as can be seen when running in debug mode.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200296
297The end of the headers is indicated by the first empty line. People often say
298that it's a double line feed, which is not exact, even if a double line feed
299is one valid form of empty line.
300
301Fortunately, HAProxy takes care of all these complex combinations when indexing
302headers, checking values and counting them, so there is no reason to worry
303about the way they could be written, but it is important not to accuse an
304application of being buggy if it does unusual, valid things.
305
306Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000307 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200308 in the middle of headers by LWS in order to join multi-line headers. This
309 is necessary for proper analysis and helps less capable HTTP parsers to work
310 correctly and not to be fooled by such complex constructs.
311
312
3131.3. HTTP response
314------------------
315
316An HTTP response looks very much like an HTTP request. Both are called HTTP
317messages. Let's consider this HTTP response :
318
319 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100320 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200321 1 HTTP/1.1 200 OK
322 2 Content-length: 350
323 3 Content-Type: text/html
324
Willy Tarreau816b9792009-09-15 21:25:21 +0200325As a special case, HTTP supports so called "Informational responses" as status
326codes 1xx. These messages are special in that they don't convey any part of the
327response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100328continue to post its request for instance. In the case of a status 100 response
329the requested information will be carried by the next non-100 response message
330following the informational one. This implies that multiple responses may be
331sent to a single request, and that this only works when keep-alive is enabled
332(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
333correctly forward and skip them, and only process the next non-100 response. As
334such, these messages are neither logged nor transformed, unless explicitly
335state otherwise. Status 101 messages indicate that the protocol is changing
336over the same connection and that haproxy must switch to tunnel mode, just as
337if a CONNECT had occurred. Then the Upgrade header would contain additional
338information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200339
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200340
Davor Ocelice9ed2812017-12-25 17:49:28 +01003411.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200342------------------------
343
344Line 1 is the "response line". It is always composed of 3 fields :
345
346 - a version tag : HTTP/1.1
347 - a status code : 200
348 - a reason : OK
349
350The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100351 - 1xx = informational message to be skipped (e.g. 100, 101)
352 - 2xx = OK, content is following (e.g. 200, 206)
353 - 3xx = OK, no content following (e.g. 302, 304)
354 - 4xx = error caused by the client (e.g. 401, 403, 404)
355 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200356
Lukas Tribus23953682017-04-28 13:24:30 +0000357Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100358"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200359found there, but it's a common practice to respect the well-established
360messages. It can be composed of one or multiple words, such as "OK", "Found",
361or "Authentication Required".
362
Davor Ocelice9ed2812017-12-25 17:49:28 +0100363HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200364
365 Code When / reason
366 200 access to stats page, and when replying to monitoring requests
367 301 when performing a redirection, depending on the configured code
368 302 when performing a redirection, depending on the configured code
369 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100370 307 when performing a redirection, depending on the configured code
371 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200372 400 for an invalid or too large request
373 401 when an authentication is required to perform the action (when
374 accessing the stats page)
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200375 403 when a request is forbidden by a "http-request deny" rule
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200376 408 when the request timeout strikes before the request is complete
377 500 when haproxy encounters an unrecoverable internal error, such as a
378 memory allocation failure, which should never happen
379 502 when the server returns an empty, invalid or incomplete response, or
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200380 when an "http-response deny" rule blocks the response.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200381 503 when no server was available to handle the request, or in response to
382 monitoring requests which match the "monitor fail" condition
383 504 when the response timeout strikes before the server responds
384
385The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3864.2).
387
388
3891.3.2. The response headers
390---------------------------
391
392Response headers work exactly like request headers, and as such, HAProxy uses
393the same parsing function for both. Please refer to paragraph 1.2.2 for more
394details.
395
396
3972. Configuring HAProxy
398----------------------
399
4002.1. Configuration file format
401------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200402
403HAProxy's configuration process involves 3 major sources of parameters :
404
405 - the arguments from the command-line, which always take precedence
406 - the "global" section, which sets process-wide parameters
407 - the proxies sections which can take form of "defaults", "listen",
408 "frontend" and "backend".
409
Willy Tarreau0ba27502007-12-24 16:55:16 +0100410The configuration file syntax consists in lines beginning with a keyword
411referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200412delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100413
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200414
William Lallemandf9873ba2015-05-05 17:37:14 +02004152.2. Quoting and escaping
416-------------------------
417
418HAProxy's configuration introduces a quoting and escaping system similar to
419many programming languages. The configuration file supports 3 types: escaping
420with a backslash, weak quoting with double quotes, and strong quoting with
421single quotes.
422
423If spaces have to be entered in strings, then they must be escaped by preceding
424them by a backslash ('\') or by quoting them. Backslashes also have to be
425escaped by doubling or strong quoting them.
426
427Escaping is achieved by preceding a special character by a backslash ('\'):
428
429 \ to mark a space and differentiate it from a delimiter
430 \# to mark a hash and differentiate it from a comment
431 \\ to use a backslash
432 \' to use a single quote and differentiate it from strong quoting
433 \" to use a double quote and differentiate it from weak quoting
434
435Weak quoting is achieved by using double quotes (""). Weak quoting prevents
436the interpretation of:
437
438 space as a parameter separator
439 ' single quote as a strong quoting delimiter
440 # hash as a comment start
441
William Lallemandb2f07452015-05-12 14:27:13 +0200442Weak quoting permits the interpretation of variables, if you want to use a non
443-interpreted dollar within a double quoted string, you should escape it with a
444backslash ("\$"), it does not work outside weak quoting.
445
446Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200447quoting.
448
449Strong quoting is achieved by using single quotes (''). Inside single quotes,
450nothing is interpreted, it's the efficient way to quote regexes.
451
452Quoted and escaped strings are replaced in memory by their interpreted
453equivalent, it allows you to perform concatenation.
454
455 Example:
456 # those are equivalents:
457 log-format %{+Q}o\ %t\ %s\ %{-Q}r
458 log-format "%{+Q}o %t %s %{-Q}r"
459 log-format '%{+Q}o %t %s %{-Q}r'
460 log-format "%{+Q}o %t"' %s %{-Q}r'
461 log-format "%{+Q}o %t"' %s'\ %{-Q}r
462
463 # those are equivalents:
464 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
465 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
466 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
467 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
468
469
William Lallemandb2f07452015-05-12 14:27:13 +02004702.3. Environment variables
471--------------------------
472
473HAProxy's configuration supports environment variables. Those variables are
474interpreted only within double quotes. Variables are expanded during the
475configuration parsing. Variable names must be preceded by a dollar ("$") and
476optionally enclosed with braces ("{}") similarly to what is done in Bourne
477shell. Variable names can contain alphanumerical characters or the character
478underscore ("_") but should not start with a digit.
479
480 Example:
481
482 bind "fd@${FD_APP1}"
483
484 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
485
486 user "$HAPROXY_USER"
487
William Lallemand4d03e432019-06-14 15:35:37 +0200488Some variables are defined by HAProxy, they can be used in the configuration
489file, or could be inherited by a program (See 3.7. Programs):
William Lallemanddaf4cd22018-04-17 16:46:13 +0200490
William Lallemand4d03e432019-06-14 15:35:37 +0200491* HAPROXY_LOCALPEER: defined at the startup of the process which contains the
492 name of the local peer. (See "-L" in the management guide.)
493
494* HAPROXY_CFGFILES: list of the configuration files loaded by HAProxy,
495 separated by semicolons. Can be useful in the case you specified a
496 directory.
497
498* HAPROXY_MWORKER: In master-worker mode, this variable is set to 1.
499
John Roeslerfb2fce12019-07-10 15:45:51 -0500500* HAPROXY_CLI: configured listeners addresses of the stats socket for every
William Lallemand4d03e432019-06-14 15:35:37 +0200501 processes, separated by semicolons.
502
John Roeslerfb2fce12019-07-10 15:45:51 -0500503* HAPROXY_MASTER_CLI: In master-worker mode, listeners addresses of the master
William Lallemand4d03e432019-06-14 15:35:37 +0200504 CLI, separated by semicolons.
505
506See also "external-check command" for other variables.
William Lallemandb2f07452015-05-12 14:27:13 +0200507
5082.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200509----------------
510
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100511Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100512values are generally expressed in milliseconds (unless explicitly stated
513otherwise) but may be expressed in any other unit by suffixing the unit to the
514numeric value. It is important to consider this because it will not be repeated
515for every keyword. Supported units are :
516
517 - us : microseconds. 1 microsecond = 1/1000000 second
518 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
519 - s : seconds. 1s = 1000ms
520 - m : minutes. 1m = 60s = 60000ms
521 - h : hours. 1h = 60m = 3600s = 3600000ms
522 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
523
524
Lukas Tribusaa83a312017-03-21 09:25:09 +00005252.5. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200526-------------
527
528 # Simple configuration for an HTTP proxy listening on port 80 on all
529 # interfaces and forwarding requests to a single backend "servers" with a
530 # single server "server1" listening on 127.0.0.1:8000
531 global
532 daemon
533 maxconn 256
534
535 defaults
536 mode http
537 timeout connect 5000ms
538 timeout client 50000ms
539 timeout server 50000ms
540
541 frontend http-in
542 bind *:80
543 default_backend servers
544
545 backend servers
546 server server1 127.0.0.1:8000 maxconn 32
547
548
549 # The same configuration defined with a single listen block. Shorter but
550 # less expressive, especially in HTTP mode.
551 global
552 daemon
553 maxconn 256
554
555 defaults
556 mode http
557 timeout connect 5000ms
558 timeout client 50000ms
559 timeout server 50000ms
560
561 listen http-in
562 bind *:80
563 server server1 127.0.0.1:8000 maxconn 32
564
565
566Assuming haproxy is in $PATH, test these configurations in a shell with:
567
Willy Tarreauccb289d2010-12-11 20:19:38 +0100568 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200569
570
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005713. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200572--------------------
573
574Parameters in the "global" section are process-wide and often OS-specific. They
575are generally set once for all and do not need being changed once correct. Some
576of them have command-line equivalents.
577
578The following keywords are supported in the "global" section :
579
580 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200581 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200582 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200583 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200584 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200585 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200586 - description
587 - deviceatlas-json-file
588 - deviceatlas-log-level
589 - deviceatlas-separator
590 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900591 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200592 - gid
593 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100594 - hard-stop-after
Christopher Faulet98fbe952019-07-22 16:18:24 +0200595 - h1-case-adjust
596 - h1-case-adjust-file
Willy Tarreaud96f1122019-12-03 07:07:36 +0100597 - insecure-fork-wanted
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100598 - insecure-setuid-wanted
Willy Tarreau6a06a402007-07-15 20:15:28 +0200599 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200600 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100601 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200602 - lua-load
William Lallemand27edc4b2019-05-07 17:49:33 +0200603 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200604 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200605 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200606 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200607 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100608 - presetenv
609 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200610 - uid
611 - ulimit-n
612 - user
Willy Tarreau636848a2019-04-15 19:38:50 +0200613 - set-dumpable
Willy Tarreau1d549722016-02-16 12:41:57 +0100614 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200615 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200616 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200617 - ssl-default-bind-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200618 - ssl-default-bind-options
619 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200620 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200621 - ssl-default-server-options
622 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100623 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100624 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100625 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100626 - 51degrees-data-file
627 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200628 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200629 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200630 - wurfl-data-file
631 - wurfl-information-list
632 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200633 - wurfl-cache-size
William Dauchy0fec3ab2019-10-27 20:08:11 +0100634 - strict-limits
Willy Tarreaud72758d2010-01-12 10:42:19 +0100635
Willy Tarreau6a06a402007-07-15 20:15:28 +0200636 * Performance tuning
William Dauchy0a8824f2019-10-27 20:08:09 +0100637 - busy-polling
Willy Tarreau1746eec2014-04-25 10:46:47 +0200638 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200639 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200640 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100641 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100642 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100643 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200644 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200645 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200646 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200647 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200648 - noepoll
649 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +0000650 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +0200651 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100652 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300653 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000654 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +0100655 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +0200656 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200657 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200658 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000659 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000660 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200661 - tune.buffers.limit
662 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200663 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200664 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100665 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200666 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200667 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200668 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100669 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200670 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200671 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100672 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100673 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100674 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100675 - tune.lua.session-timeout
676 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200677 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100678 - tune.maxaccept
679 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200680 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200681 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200682 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100683 - tune.rcvbuf.client
684 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100685 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200686 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100687 - tune.sndbuf.client
688 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100689 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100690 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200691 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100692 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200693 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200694 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100695 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200696 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100697 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200698 - tune.vars.reqres-max-size
699 - tune.vars.sess-max-size
700 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100701 - tune.zlib.memlevel
702 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100703
Willy Tarreau6a06a402007-07-15 20:15:28 +0200704 * Debugging
705 - debug
706 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200707
708
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007093.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200710------------------------------------
711
Emeric Brunc8e8d122012-10-02 18:42:10 +0200712ca-base <dir>
713 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200714 relative path is used with "ca-file" or "crl-file" directives. Absolute
715 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200716
Willy Tarreau6a06a402007-07-15 20:15:28 +0200717chroot <jail dir>
718 Changes current directory to <jail dir> and performs a chroot() there before
719 dropping privileges. This increases the security level in case an unknown
720 vulnerability would be exploited, since it would make it very hard for the
721 attacker to exploit the system. This only works when the process is started
722 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100723 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100724
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100725cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
726 On Linux 2.6 and above, it is possible to bind a process or a thread to a
727 specific CPU set. This means that the process or the thread will never run on
728 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
729 sets. The first argument is a process set, eventually followed by a thread
730 set. These sets have the format
731
732 all | odd | even | number[-[number]]
733
734 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100735 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100736 ignored. It is possible to specify a range with two such number delimited by
737 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100738 "all", only odd numbers using "odd" or even numbers using "even", just like
739 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100740 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 +0100741 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100742 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100743 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100744 specified. Each "cpu-map" directive will replace the previous ones when they
745 overlap. A thread will be bound on the intersection of its mapping and the
746 one of the process on which it is attached. If the intersection is null, no
747 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100748
Christopher Fauletff4121f2017-11-22 16:38:49 +0100749 Ranges can be partially defined. The higher bound can be omitted. In such
750 case, it is replaced by the corresponding maximum value, 32 or 64 depending
751 on the machine's word size.
752
Christopher Faulet26028f62017-11-22 15:01:51 +0100753 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100754 automatically bind a process or a thread to a CPU by incrementing
755 process/thread and CPU sets. To be valid, both sets must have the same
756 size. No matter the declaration order of the CPU sets, it will be bound from
757 the lowest to the highest bound. Having a process and a thread range with the
758 "auto:" prefix is not supported. Only one range is supported, the other one
759 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100760
761 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100762 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
763
764 cpu-map 1/all 0-3 # bind all threads of the first process on the
765 # first 4 CPUs
766
767 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
768 # or "cpu-map 1-32 0-31" depending on the machine's
769 # word size.
770
Christopher Faulet26028f62017-11-22 15:01:51 +0100771 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100772 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100773 cpu-map auto:1-4 0-3
774 cpu-map auto:1-4 0-1 2-3
775 cpu-map auto:1-4 3 2 1 0
776
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100777 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
778 # and so on.
779 cpu-map auto:1/1-4 0-3
780 cpu-map auto:1/1-4 0-1 2-3
781 cpu-map auto:1/1-4 3 2 1 0
782
Davor Ocelice9ed2812017-12-25 17:49:28 +0100783 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100784 cpu-map auto:all 0-63
785 cpu-map auto:even 0-31
786 cpu-map auto:odd 32-63
787
788 # invalid cpu-map because process and CPU sets have different sizes.
789 cpu-map auto:1-4 0 # invalid
790 cpu-map auto:1 0-3 # invalid
791
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100792 # invalid cpu-map because automatic binding is used with a process range
793 # and a thread range.
794 cpu-map auto:all/all 0 # invalid
795 cpu-map auto:all/1-4 0 # invalid
796 cpu-map auto:1-4/all 0 # invalid
797
Emeric Brunc8e8d122012-10-02 18:42:10 +0200798crt-base <dir>
799 Assigns a default directory to fetch SSL certificates from when a relative
800 path is used with "crtfile" directives. Absolute locations specified after
801 "crtfile" prevail and ignore "crt-base".
802
Willy Tarreau6a06a402007-07-15 20:15:28 +0200803daemon
804 Makes the process fork into background. This is the recommended mode of
805 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100806 disabled by the command line "-db" argument. This option is ignored in
807 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200808
David Carlier8167f302015-06-01 13:50:06 +0200809deviceatlas-json-file <path>
810 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100811 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200812
813deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100814 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200815 optional and set to 0 by default if not set.
816
817deviceatlas-separator <char>
818 Sets the character separator for the API properties results. This directive
819 is optional and set to | by default if not set.
820
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100821deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200822 Sets the client cookie's name used for the detection if the DeviceAtlas
823 Client-side component was used during the request. This directive is optional
824 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100825
Simon Horman98637e52014-06-20 12:30:16 +0900826external-check
Willy Tarreaud96f1122019-12-03 07:07:36 +0100827 Allows the use of an external agent to perform health checks. This is
828 disabled by default as a security precaution, and even when enabled, checks
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100829 may still fail unless "insecure-fork-wanted" is enabled as well. If the
830 program launched makes use of a setuid executable (it should really not),
831 you may also need to set "insecure-setuid-wanted" in the global section.
832 See "option external-check", and "insecure-fork-wanted", and
833 "insecure-setuid-wanted".
Simon Horman98637e52014-06-20 12:30:16 +0900834
Willy Tarreau6a06a402007-07-15 20:15:28 +0200835gid <number>
836 Changes the process' group ID to <number>. It is recommended that the group
837 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
838 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100839 Note that if haproxy is started from a user having supplementary groups, it
840 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200841 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100842
Willy Tarreau11770ce2019-12-03 08:29:22 +0100843group <group name>
844 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
845 See also "gid" and "user".
846
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100847hard-stop-after <time>
848 Defines the maximum time allowed to perform a clean soft-stop.
849
850 Arguments :
851 <time> is the maximum time (by default in milliseconds) for which the
852 instance will remain alive when a soft-stop is received via the
853 SIGUSR1 signal.
854
855 This may be used to ensure that the instance will quit even if connections
856 remain opened during a soft-stop (for example with long timeouts for a proxy
857 in tcp mode). It applies both in TCP and HTTP mode.
858
859 Example:
860 global
861 hard-stop-after 30s
862
Christopher Faulet98fbe952019-07-22 16:18:24 +0200863h1-case-adjust <from> <to>
864 Defines the case adjustment to apply, when enabled, to the header name
865 <from>, to change it to <to> before sending it to HTTP/1 clients or
866 servers. <from> must be in lower case, and <from> and <to> must not differ
867 except for their case. It may be repeated if several header names need to be
868 ajusted. Duplicate entries are not allowed. If a lot of header names have to
869 be adjusted, it might be more convenient to use "h1-case-adjust-file".
870 Please note that no transformation will be applied unless "option
871 h1-case-adjust-bogus-client" or "option h1-case-adjust-bogus-server" is
872 specified in a proxy.
873
874 There is no standard case for header names because, as stated in RFC7230,
875 they are case-insensitive. So applications must handle them in a case-
876 insensitive manner. But some bogus applications violate the standards and
877 erroneously rely on the cases most commonly used by browsers. This problem
878 becomes critical with HTTP/2 because all header names must be exchanged in
879 lower case, and HAProxy follows the same convention. All header names are
880 sent in lower case to clients and servers, regardless of the HTTP version.
881
882 Applications which fail to properly process requests or responses may require
883 to temporarily use such workarounds to adjust header names sent to them for
884 the time it takes the application to be fixed. Please note that an
885 application which requires such workarounds might be vulnerable to content
886 smuggling attacks and must absolutely be fixed.
887
888 Example:
889 global
890 h1-case-adjust content-length Content-Length
891
892 See "h1-case-adjust-file", "option h1-case-adjust-bogus-client" and
893 "option h1-case-adjust-bogus-server".
894
895h1-case-adjust-file <hdrs-file>
896 Defines a file containing a list of key/value pairs used to adjust the case
897 of some header names before sending them to HTTP/1 clients or servers. The
898 file <hdrs-file> must contain 2 header names per line. The first one must be
899 in lower case and both must not differ except for their case. Lines which
900 start with '#' are ignored, just like empty lines. Leading and trailing tabs
901 and spaces are stripped. Duplicate entries are not allowed. Please note that
902 no transformation will be applied unless "option h1-case-adjust-bogus-client"
903 or "option h1-case-adjust-bogus-server" is specified in a proxy.
904
905 If this directive is repeated, only the last one will be processed. It is an
906 alternative to the directive "h1-case-adjust" if a lot of header names need
907 to be adjusted. Please read the risks associated with using this.
908
909 See "h1-case-adjust", "option h1-case-adjust-bogus-client" and
910 "option h1-case-adjust-bogus-server".
911
Willy Tarreaud96f1122019-12-03 07:07:36 +0100912insecure-fork-wanted
913 By default haproxy tries hard to prevent any thread and process creation
914 after it starts. Doing so is particularly important when using Lua files of
915 uncertain origin, and when experimenting with development versions which may
916 still contain bugs whose exploitability is uncertain. And generally speaking
917 it's good hygiene to make sure that no unexpected background activity can be
918 triggered by traffic. But this prevents external checks from working, and may
919 break some very specific Lua scripts which actively rely on the ability to
920 fork. This option is there to disable this protection. Note that it is a bad
921 idea to disable it, as a vulnerability in a library or within haproxy itself
922 will be easier to exploit once disabled. In addition, forking from Lua or
923 anywhere else is not reliable as the forked process may randomly embed a lock
924 set by another thread and never manage to finish an operation. As such it is
925 highly recommended that this option is never used and that any workload
926 requiring such a fork be reconsidered and moved to a safer solution (such as
927 agents instead of external checks). This option supports the "no" prefix to
928 disable it.
929
Willy Tarreaua45a8b52019-12-06 16:31:45 +0100930insecure-setuid-wanted
931 HAProxy doesn't need to call executables at run time (except when using
932 external checks which are strongly recommended against), and is even expected
933 to isolate itself into an empty chroot. As such, there basically is no valid
934 reason to allow a setuid executable to be called without the user being fully
935 aware of the risks. In a situation where haproxy would need to call external
936 checks and/or disable chroot, exploiting a vulnerability in a library or in
937 haproxy itself could lead to the execution of an external program. On Linux
938 it is possible to lock the process so that any setuid bit present on such an
939 executable is ignored. This significantly reduces the risk of privilege
940 escalation in such a situation. This is what haproxy does by default. In case
941 this causes a problem to an external check (for example one which would need
942 the "ping" command), then it is possible to disable this protection by
943 explicitly adding this directive in the global section. If enabled, it is
944 possible to turn it back off by prefixing it with the "no" keyword.
945
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200946log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
947 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100948 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100949 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100950 configured with "log global".
951
952 <address> can be one of:
953
Willy Tarreau2769aa02007-12-27 18:26:09 +0100954 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100955 no port is specified, 514 is used by default (the standard syslog
956 port).
957
David du Colombier24bb5f52011-03-17 10:40:23 +0100958 - An IPv6 address followed by a colon and optionally a UDP port. If
959 no port is specified, 514 is used by default (the standard syslog
960 port).
961
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100962 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +0100963 considerations for chroot (be sure the path is accessible inside
964 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100965 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100966
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100967 - A file descriptor number in the form "fd@<number>", which may point
968 to a pipe, terminal, or socket. In this case unbuffered logs are used
969 and one writev() call per log is performed. This is a bit expensive
970 but acceptable for most workloads. Messages sent this way will not be
971 truncated but may be dropped, in which case the DroppedLogs counter
972 will be incremented. The writev() call is atomic even on pipes for
973 messages up to PIPE_BUF size, which POSIX recommends to be at least
974 512 and which is 4096 bytes on most modern operating systems. Any
975 larger message may be interleaved with messages from other processes.
976 Exceptionally for debugging purposes the file descriptor may also be
977 directed to a file, but doing so will significantly slow haproxy down
978 as non-blocking calls will be ignored. Also there will be no way to
979 purge nor rotate this file without restarting the process. Note that
980 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +0100981 for use with a TCP syslog server. See also the "short" and "raw"
982 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100983
984 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
985 "fd@2", see above.
986
Willy Tarreauc046d162019-08-30 15:24:59 +0200987 - A ring buffer in the form "ring@<name>", which will correspond to an
988 in-memory ring buffer accessible over the CLI using the "show events"
989 command, which will also list existing rings and their sizes. Such
990 buffers are lost on reload or restart but when used as a complement
991 this can help troubleshooting by having the logs instantly available.
992
William Lallemandb2f07452015-05-12 14:27:13 +0200993 You may want to reference some environment variables in the address
994 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100995
Willy Tarreau18324f52014-06-27 18:10:07 +0200996 <length> is an optional maximum line length. Log lines larger than this value
997 will be truncated before being sent. The reason is that syslog
998 servers act differently on log line length. All servers support the
999 default value of 1024, but some servers simply drop larger lines
1000 while others do log them. If a server supports long lines, it may
1001 make sense to set this value here in order to avoid truncating long
1002 lines. Similarly, if a server drops long lines, it is preferable to
1003 truncate them before sending them. Accepted values are 80 to 65535
1004 inclusive. The default value of 1024 is generally fine for all
1005 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +01001006 JSON-formatted logs may require larger values. You may also need to
1007 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +02001008
Dragan Dosen7ad31542015-09-28 17:16:47 +02001009 <format> is the log format used when generating syslog messages. It may be
1010 one of the following :
1011
1012 rfc3164 The RFC3164 syslog message format. This is the default.
1013 (https://tools.ietf.org/html/rfc3164)
1014
1015 rfc5424 The RFC5424 syslog message format.
1016 (https://tools.ietf.org/html/rfc5424)
1017
Willy Tarreaue8746a02018-11-12 08:45:00 +01001018 short A message containing only a level between angle brackets such as
1019 '<3>', followed by the text. The PID, date, time, process name
1020 and system name are omitted. This is designed to be used with a
1021 local log server. This format is compatible with what the systemd
1022 logger consumes.
1023
Willy Tarreauc1b06452018-11-12 11:57:56 +01001024 raw A message containing only the text. The level, PID, date, time,
1025 process name and system name are omitted. This is designed to be
1026 used in containers or during development, where the severity only
1027 depends on the file descriptor used (stdout/stderr).
1028
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02001029 <ranges> A list of comma-separated ranges to identify the logs to sample.
1030 This is used to balance the load of the logs to send to the log
1031 server. The limits of the ranges cannot be null. They are numbered
1032 from 1. The size or period (in number of logs) of the sample must be
1033 set with <sample_size> parameter.
1034
1035 <sample_size>
1036 The size of the sample in number of logs to consider when balancing
1037 their logging loads. It is used to balance the load of the logs to
1038 send to the syslog server. This size must be greater or equal to the
1039 maximum of the high limits of the ranges.
1040 (see also <ranges> parameter).
1041
Robert Tsai81ae1952007-12-05 10:47:29 +01001042 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001043
Willy Tarreaue8746a02018-11-12 08:45:00 +01001044 kern user mail daemon auth syslog lpr news
1045 uucp cron auth2 ftp ntp audit alert cron2
1046 local0 local1 local2 local3 local4 local5 local6 local7
1047
Willy Tarreauc1b06452018-11-12 11:57:56 +01001048 Note that the facility is ignored for the "short" and "raw"
1049 formats, but still required as a positional field. It is
1050 recommended to use "daemon" in this case to make it clear that
1051 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001052
1053 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +02001054 all messages are sent. If a maximum level is specified, only messages with a
1055 severity at least as important as this level will be sent. An optional minimum
1056 level can be specified. If it is set, logs emitted with a more severe level
1057 than this one will be capped to this level. This is used to avoid sending
1058 "emerg" messages on all terminals on some default syslog configurations.
1059 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001060
Cyril Bontédc4d9032012-04-08 21:57:39 +02001061 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +02001062
Joe Williamsdf5b38f2010-12-29 17:05:48 +01001063log-send-hostname [<string>]
1064 Sets the hostname field in the syslog header. If optional "string" parameter
1065 is set the header is set to the string contents, otherwise uses the hostname
1066 of the system. Generally used if one is not relaying logs through an
1067 intermediate syslog server or for simply customizing the hostname printed in
1068 the logs.
1069
Kevinm48936af2010-12-22 16:08:21 +00001070log-tag <string>
1071 Sets the tag field in the syslog header to this string. It defaults to the
1072 program name as launched from the command line, which usually is "haproxy".
1073 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +01001074 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +00001075
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001076lua-load <file>
1077 This global directive loads and executes a Lua file. This directive can be
1078 used multiple times.
1079
William Lallemand4cfede82017-11-24 22:02:34 +01001080master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +02001081 Master-worker mode. It is equivalent to the command line "-W" argument.
1082 This mode will launch a "master" which will monitor the "workers". Using
1083 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001084 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +02001085 or daemon mode. It is recommended to use this mode with multiprocess and
1086 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +01001087 By default, if a worker exits with a bad return code, in the case of a
1088 segfault for example, all workers will be killed, and the master will leave.
1089 It is convenient to combine this behavior with Restart=on-failure in a
1090 systemd unit file in order to relaunch the whole process. If you don't want
1091 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +02001092
William Lallemand4cfede82017-11-24 22:02:34 +01001093 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +02001094
William Lallemand27edc4b2019-05-07 17:49:33 +02001095mworker-max-reloads <number>
1096 In master-worker mode, this option limits the number of time a worker can
John Roeslerfb2fce12019-07-10 15:45:51 -05001097 survive to a reload. If the worker did not leave after a reload, once its
William Lallemand27edc4b2019-05-07 17:49:33 +02001098 number of reloads is greater than this number, the worker will receive a
1099 SIGTERM. This option helps to keep under control the number of workers.
1100 See also "show proc" in the Management Guide.
1101
Willy Tarreau6a06a402007-07-15 20:15:28 +02001102nbproc <number>
1103 Creates <number> processes when going daemon. This requires the "daemon"
1104 mode. By default, only one process is created, which is the recommended mode
1105 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +01001106 process, it may be needed to fork multiple daemons. When set to a value
1107 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreau1f672a82019-01-26 14:20:55 +01001108 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon" and
1109 "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001110
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001111nbthread <number>
1112 This setting is only available when support for threads was built in. It
Willy Tarreau26f6ae12019-02-02 12:56:15 +01001113 makes haproxy run on <number> threads. This is exclusive with "nbproc". While
1114 "nbproc" historically used to be the only way to use multiple processors, it
1115 also involved a number of shortcomings related to the lack of synchronization
1116 between processes (health-checks, peers, stick-tables, stats, ...) which do
1117 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +01001118 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
1119 HAProxy is started in foreground. On some platforms supporting CPU affinity,
1120 when nbproc is not used, the default "nbthread" value is automatically set to
1121 the number of CPUs the process is bound to upon startup. This means that the
1122 thread count can easily be adjusted from the calling process using commands
1123 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1124 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001125
Willy Tarreau6a06a402007-07-15 20:15:28 +02001126pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001127 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +02001128 the "-p" command line argument. The file must be accessible to the user
1129 starting the process. See also "daemon".
1130
Willy Tarreau1d549722016-02-16 12:41:57 +01001131presetenv <name> <value>
1132 Sets environment variable <name> to value <value>. If the variable exists, it
1133 is NOT overwritten. The changes immediately take effect so that the next line
1134 in the configuration file sees the new value. See also "setenv", "resetenv",
1135 and "unsetenv".
1136
1137resetenv [<name> ...]
1138 Removes all environment variables except the ones specified in argument. It
1139 allows to use a clean controlled environment before setting new values with
1140 setenv or unsetenv. Please note that some internal functions may make use of
1141 some environment variables, such as time manipulation functions, but also
1142 OpenSSL or even external checks. This must be used with extreme care and only
1143 after complete validation. The changes immediately take effect so that the
1144 next line in the configuration file sees the new environment. See also
1145 "setenv", "presetenv", and "unsetenv".
1146
Christopher Fauletff4121f2017-11-22 16:38:49 +01001147stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001148 Limits the stats socket to a certain set of processes numbers. By default the
1149 stats socket is bound to all processes, causing a warning to be emitted when
1150 nbproc is greater than 1 because there is no way to select the target process
1151 when connecting. However, by using this setting, it becomes possible to pin
1152 the stats socket to a specific set of processes, typically the first one. The
1153 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001154 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001155 word size (32 or 64). Ranges can be partially defined. The higher bound can
1156 be omitted. In such case, it is replaced by the corresponding maximum
1157 value. A better option consists in using the "process" setting of the "stats
1158 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001159
Baptiste Assmann5626f482015-08-23 10:00:10 +02001160server-state-base <directory>
1161 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001162 file names which do not start with a '/'. See also "server-state-file",
1163 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001164
1165server-state-file <file>
1166 Specifies the path to the file containing state of servers. If the path starts
1167 with a slash ('/'), it is considered absolute, otherwise it is considered
1168 relative to the directory specified using "server-state-base" (if set) or to
1169 the current directory. Before reloading HAProxy, it is possible to save the
1170 servers' current state using the stats command "show servers state". The
1171 output of this command must be written in the file pointed by <file>. When
1172 starting up, before handling traffic, HAProxy will read, load and apply state
1173 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001174 configuration. See also "server-state-base" and "show servers state",
1175 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001176
Willy Tarreau1d549722016-02-16 12:41:57 +01001177setenv <name> <value>
1178 Sets environment variable <name> to value <value>. If the variable exists, it
1179 is overwritten. The changes immediately take effect so that the next line in
1180 the configuration file sees the new value. See also "presetenv", "resetenv",
1181 and "unsetenv".
1182
Willy Tarreau636848a2019-04-15 19:38:50 +02001183set-dumpable
1184 This option is better left disabled by default and enabled only upon a
William Dauchyec730982019-10-27 20:08:10 +01001185 developer's request. If it has been enabled, it may still be forcibly
1186 disabled by prefixing it with the "no" keyword. It has no impact on
1187 performance nor stability but will try hard to re-enable core dumps that were
1188 possibly disabled by file size limitations (ulimit -f), core size limitations
1189 (ulimit -c), or "dumpability" of a process after changing its UID/GID (such
1190 as /proc/sys/fs/suid_dumpable on Linux). Core dumps might still be limited by
1191 the current directory's permissions (check what directory the file is started
1192 from), the chroot directory's permission (it may be needed to temporarily
1193 disable the chroot directive or to move it to a dedicated writable location),
1194 or any other system-specific constraint. For example, some Linux flavours are
1195 notorious for replacing the default core file with a path to an executable
1196 not even installed on the system (check /proc/sys/kernel/core_pattern). Often,
1197 simply writing "core", "core.%p" or "/var/log/core/core.%p" addresses the
1198 issue. When trying to enable this option waiting for a rare issue to
1199 re-appear, it's often a good idea to first try to obtain such a dump by
1200 issuing, for example, "kill -11" to the haproxy process and verify that it
1201 leaves a core where expected when dying.
Willy Tarreau636848a2019-04-15 19:38:50 +02001202
Willy Tarreau610f04b2014-02-13 11:36:41 +01001203ssl-default-bind-ciphers <ciphers>
1204 This setting is only available when support for OpenSSL was built in. It sets
1205 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001206 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001207 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001208 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1209 information and recommendations see e.g.
1210 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1211 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1212 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1213 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001214
1215ssl-default-bind-ciphersuites <ciphersuites>
1216 This setting is only available when support for OpenSSL was built in and
1217 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1218 describing the list of cipher algorithms ("cipher suite") that are negotiated
1219 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1220 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001221 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1222 cipher configuration for TLSv1.2 and earlier, please check the
1223 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001224 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001225
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001226ssl-default-bind-options [<option>]...
1227 This setting is only available when support for OpenSSL was built in. It sets
1228 default ssl-options to force on all "bind" lines. Please check the "bind"
1229 keyword to see available options.
1230
1231 Example:
1232 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001233 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001234
Willy Tarreau610f04b2014-02-13 11:36:41 +01001235ssl-default-server-ciphers <ciphers>
1236 This setting is only available when support for OpenSSL was built in. It
1237 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001238 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001239 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001240 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1241 information and recommendations see e.g.
1242 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1243 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1244 For TLSv1.3 cipher configuration, please check the
1245 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1246 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001247
1248ssl-default-server-ciphersuites <ciphersuites>
1249 This setting is only available when support for OpenSSL was built in and
1250 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1251 string describing the list of cipher algorithms that are negotiated during
1252 the TLSv1.3 handshake with the server, for all "server" lines which do not
1253 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001254 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1255 cipher configuration for TLSv1.2 and earlier, please check the
1256 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1257 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001258
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001259ssl-default-server-options [<option>]...
1260 This setting is only available when support for OpenSSL was built in. It sets
1261 default ssl-options to force on all "server" lines. Please check the "server"
1262 keyword to see available options.
1263
Remi Gacogne47783ef2015-05-29 15:53:22 +02001264ssl-dh-param-file <file>
1265 This setting is only available when support for OpenSSL was built in. It sets
1266 the default DH parameters that are used during the SSL/TLS handshake when
1267 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001268 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001269 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001270 are not specified either by using ssl-dh-param-file or by setting them
1271 directly in the certificate file, pre-generated DH parameters of the size
1272 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1273 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001274 Custom DH parameters may be generated by using the OpenSSL command
1275 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1276 parameters should not be considered secure anymore.
1277
Emeric Brun850efd52014-01-29 12:24:34 +01001278ssl-server-verify [none|required]
1279 The default behavior for SSL verify on servers side. If specified to 'none',
1280 servers certificates are not verified. The default is 'required' except if
1281 forced using cmdline option '-dV'.
1282
Willy Tarreauabb175f2012-09-24 12:43:26 +02001283stats socket [<address:port>|<path>] [param*]
1284 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1285 Connections to this socket will return various statistics outputs and even
1286 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001287 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001288 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001289
Willy Tarreauabb175f2012-09-24 12:43:26 +02001290 All parameters supported by "bind" lines are supported, for instance to
1291 restrict access to some users or their access rights. Please consult
1292 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001293
1294stats timeout <timeout, in milliseconds>
1295 The default timeout on the stats socket is set to 10 seconds. It is possible
1296 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001297 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001298
1299stats maxconn <connections>
1300 By default, the stats socket is limited to 10 concurrent connections. It is
1301 possible to change this value with "stats maxconn".
1302
Willy Tarreau6a06a402007-07-15 20:15:28 +02001303uid <number>
1304 Changes the process' user ID to <number>. It is recommended that the user ID
1305 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1306 be started with superuser privileges in order to be able to switch to another
1307 one. See also "gid" and "user".
1308
1309ulimit-n <number>
1310 Sets the maximum number of per-process file-descriptors to <number>. By
1311 default, it is automatically computed, so it is recommended not to use this
1312 option.
1313
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001314unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1315 [ group <group> ] [ gid <gid> ]
1316
1317 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1318 This is mainly used to simplify declaration of those UNIX sockets and reduce
1319 the risk of errors, since those settings are most commonly required but are
1320 also process-specific. The <prefix> setting can be used to force all socket
1321 path to be relative to that directory. This might be needed to access another
1322 component's chroot. Note that those paths are resolved before haproxy chroots
1323 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1324 all have the same meaning as their homonyms used by the "bind" statement. If
1325 both are specified, the "bind" statement has priority, meaning that the
1326 "unix-bind" settings may be seen as process-wide default settings.
1327
Willy Tarreau1d549722016-02-16 12:41:57 +01001328unsetenv [<name> ...]
1329 Removes environment variables specified in arguments. This can be useful to
1330 hide some sensitive information that are occasionally inherited from the
1331 user's environment during some operations. Variables which did not exist are
1332 silently ignored so that after the operation, it is certain that none of
1333 these variables remain. The changes immediately take effect so that the next
1334 line in the configuration file will not see these variables. See also
1335 "setenv", "presetenv", and "resetenv".
1336
Willy Tarreau6a06a402007-07-15 20:15:28 +02001337user <user name>
1338 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1339 See also "uid" and "group".
1340
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001341node <name>
1342 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1343
1344 This statement is useful in HA configurations where two or more processes or
1345 servers share the same IP address. By setting a different node-name on all
1346 nodes, it becomes easy to immediately spot what server is handling the
1347 traffic.
1348
1349description <text>
1350 Add a text that describes the instance.
1351
1352 Please note that it is required to escape certain characters (# for example)
1353 and this text is inserted into a html page so you should avoid using
1354 "<" and ">" characters.
1355
Thomas Holmesdb04f192015-05-18 13:21:39 +0100135651degrees-data-file <file path>
1357 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001358 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001359
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001360 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001361 compiled with USE_51DEGREES.
1362
Ben Shillitof25e8e52016-12-02 14:25:37 +0000136351degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001364 A list of 51Degrees property names to be load from the dataset. A full list
1365 of names is available on the 51Degrees website:
1366 https://51degrees.com/resources/property-dictionary
1367
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001368 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001369 compiled with USE_51DEGREES.
1370
Dragan Dosen93b38d92015-06-29 16:43:25 +0200137151degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001372 A char that will be appended to every property value in a response header
1373 containing 51Degrees results. If not set that will be set as ','.
1374
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001375 Please note that this option is only available when haproxy has been
1376 compiled with USE_51DEGREES.
1377
137851degrees-cache-size <number>
1379 Sets the size of the 51Degrees converter cache to <number> entries. This
1380 is an LRU cache which reminds previous device detections and their results.
1381 By default, this cache is disabled.
1382
1383 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001384 compiled with USE_51DEGREES.
1385
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001386wurfl-data-file <file path>
1387 The path of the WURFL data file to provide device detection services. The
1388 file should be accessible by HAProxy with relevant permissions.
1389
1390 Please note that this option is only available when haproxy has been compiled
1391 with USE_WURFL=1.
1392
1393wurfl-information-list [<capability>]*
1394 A space-delimited list of WURFL capabilities, virtual capabilities, property
1395 names we plan to use in injected headers. A full list of capability and
1396 virtual capability names is available on the Scientiamobile website :
1397
1398 https://www.scientiamobile.com/wurflCapability
1399
1400 Valid WURFL properties are:
1401 - wurfl_id Contains the device ID of the matched device.
1402
1403 - wurfl_root_id Contains the device root ID of the matched
1404 device.
1405
1406 - wurfl_isdevroot Tells if the matched device is a root device.
1407 Possible values are "TRUE" or "FALSE".
1408
1409 - wurfl_useragent The original useragent coming with this
1410 particular web request.
1411
1412 - wurfl_api_version Contains a string representing the currently
1413 used Libwurfl API version.
1414
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001415 - wurfl_info A string containing information on the parsed
1416 wurfl.xml and its full path.
1417
1418 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1419 WURFL has been loaded successfully.
1420
1421 - wurfl_normalized_useragent The normalized useragent.
1422
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001423 Please note that this option is only available when haproxy has been compiled
1424 with USE_WURFL=1.
1425
1426wurfl-information-list-separator <char>
1427 A char that will be used to separate values in a response header containing
1428 WURFL results. If not set that a comma (',') will be used by default.
1429
1430 Please note that this option is only available when haproxy has been compiled
1431 with USE_WURFL=1.
1432
1433wurfl-patch-file [<file path>]
1434 A list of WURFL patch file paths. Note that patches are loaded during startup
1435 thus before the chroot.
1436
1437 Please note that this option is only available when haproxy has been compiled
1438 with USE_WURFL=1.
1439
paulborilebad132c2019-04-18 11:57:04 +02001440wurfl-cache-size <size>
1441 Sets the WURFL Useragent cache size. For faster lookups, already processed user
1442 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001443 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02001444 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001445
1446 Please note that this option is only available when haproxy has been compiled
1447 with USE_WURFL=1.
1448
William Dauchy0fec3ab2019-10-27 20:08:11 +01001449strict-limits
1450 Makes process fail at startup when a setrlimit fails. Haproxy is tries to set
1451 the best setrlimit according to what has been calculated. If it fails, it
1452 will emit a warning. Use this option if you want an explicit failure of
1453 haproxy when those limits fail. This option is disabled by default. If it has
1454 been enabled, it may still be forcibly disabled by prefixing it with the "no"
1455 keyword.
1456
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014573.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001458-----------------------
1459
Willy Tarreaubeb859a2018-11-22 18:07:59 +01001460busy-polling
1461 In some situations, especially when dealing with low latency on processors
1462 supporting a variable frequency or when running inside virtual machines, each
1463 time the process waits for an I/O using the poller, the processor goes back
1464 to sleep or is offered to another VM for a long time, and it causes
1465 excessively high latencies. This option provides a solution preventing the
1466 processor from sleeping by always using a null timeout on the pollers. This
1467 results in a significant latency reduction (30 to 100 microseconds observed)
1468 at the expense of a risk to overheat the processor. It may even be used with
1469 threads, in which case improperly bound threads may heavily conflict,
1470 resulting in a worse performance and high values for the CPU stolen fields
1471 in "show info" output, indicating which threads are misconfigured. It is
1472 important not to let the process run on the same processor as the network
1473 interrupts when this option is used. It is also better to avoid using it on
1474 multiple CPU threads sharing the same core. This option is disabled by
1475 default. If it has been enabled, it may still be forcibly disabled by
1476 prefixing it with the "no" keyword. It is ignored by the "select" and
1477 "poll" pollers.
1478
Willy Tarreau1746eec2014-04-25 10:46:47 +02001479max-spread-checks <delay in milliseconds>
1480 By default, haproxy tries to spread the start of health checks across the
1481 smallest health check interval of all the servers in a farm. The principle is
1482 to avoid hammering services running on the same server. But when using large
1483 check intervals (10 seconds or more), the last servers in the farm take some
1484 time before starting to be tested, which can be a problem. This parameter is
1485 used to enforce an upper bound on delay between the first and the last check,
1486 even if the servers' check intervals are larger. When servers run with
1487 shorter intervals, their intervals will be respected though.
1488
Willy Tarreau6a06a402007-07-15 20:15:28 +02001489maxconn <number>
1490 Sets the maximum per-process number of concurrent connections to <number>. It
1491 is equivalent to the command-line argument "-n". Proxies will stop accepting
1492 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001493 automatically adjusted according to this value. See also "ulimit-n". Note:
1494 the "select" poller cannot reliably use more than 1024 file descriptors on
1495 some platforms. If your platform only supports select and reports "select
1496 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01001497 below 500 in general). If this value is not set, it will automatically be
1498 calculated based on the current file descriptors limit reported by the
1499 "ulimit -n" command, possibly reduced to a lower value if a memory limit
1500 is enforced, based on the buffer size, memory allocated to compression, SSL
1501 cache size, and use or not of SSL and the associated maxsslconn (which can
1502 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001503
Willy Tarreau81c25d02011-09-07 15:17:21 +02001504maxconnrate <number>
1505 Sets the maximum per-process number of connections per second to <number>.
1506 Proxies will stop accepting connections when this limit is reached. It can be
1507 used to limit the global capacity regardless of each frontend capacity. It is
1508 important to note that this can only be used as a service protection measure,
1509 as there will not necessarily be a fair share between frontends when the
1510 limit is reached, so it's a good idea to also limit each frontend to some
1511 value close to its expected share. Also, lowering tune.maxaccept can improve
1512 fairness.
1513
William Lallemandd85f9172012-11-09 17:05:39 +01001514maxcomprate <number>
1515 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001516 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001517 level will be decreased during the session. If the maximum is reached at the
1518 beginning of a session, the session will not compress at all. If the maximum
1519 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001520 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001521 default value.
1522
William Lallemand072a2bf2012-11-20 17:01:01 +01001523maxcompcpuusage <number>
1524 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1525 for new requests or decreasing the compression level of current requests.
1526 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1527 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1528 case of multiple processes (nbproc > 1), each process manages its individual
1529 usage. A value of 100 disable the limit. The default value is 100. Setting
1530 a lower value will prevent the compression work from slowing the whole
1531 process down and from introducing high latencies.
1532
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001533maxpipes <number>
1534 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1535 are only used by kernel-based tcp splicing. Since a pipe contains two file
1536 descriptors, the "ulimit-n" value will be increased accordingly. The default
1537 value is maxconn/4, which seems to be more than enough for most heavy usages.
1538 The splice code dynamically allocates and releases pipes, and can fall back
1539 to standard copy, so setting this value too low may only impact performance.
1540
Willy Tarreau93e7c002013-10-07 18:51:07 +02001541maxsessrate <number>
1542 Sets the maximum per-process number of sessions per second to <number>.
1543 Proxies will stop accepting connections when this limit is reached. It can be
1544 used to limit the global capacity regardless of each frontend capacity. It is
1545 important to note that this can only be used as a service protection measure,
1546 as there will not necessarily be a fair share between frontends when the
1547 limit is reached, so it's a good idea to also limit each frontend to some
1548 value close to its expected share. Also, lowering tune.maxaccept can improve
1549 fairness.
1550
Willy Tarreau403edff2012-09-06 11:58:37 +02001551maxsslconn <number>
1552 Sets the maximum per-process number of concurrent SSL connections to
1553 <number>. By default there is no SSL-specific limit, which means that the
1554 global maxconn setting will apply to all connections. Setting this limit
1555 avoids having openssl use too much memory and crash when malloc returns NULL
1556 (since it unfortunately does not reliably check for such conditions). Note
1557 that the limit applies both to incoming and outgoing connections, so one
1558 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001559 If this value is not set, but a memory limit is enforced, this value will be
1560 automatically computed based on the memory limit, maxconn, the buffer size,
1561 memory allocated to compression, SSL cache size, and use of SSL in either
1562 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1563 when there is a memory limit, haproxy will automatically adjust these values
1564 so that 100% of the connections can be made over SSL with no risk, and will
1565 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001566
Willy Tarreaue43d5322013-10-07 20:01:52 +02001567maxsslrate <number>
1568 Sets the maximum per-process number of SSL sessions per second to <number>.
1569 SSL listeners will stop accepting connections when this limit is reached. It
1570 can be used to limit the global SSL CPU usage regardless of each frontend
1571 capacity. It is important to note that this can only be used as a service
1572 protection measure, as there will not necessarily be a fair share between
1573 frontends when the limit is reached, so it's a good idea to also limit each
1574 frontend to some value close to its expected share. It is also important to
1575 note that the sessions are accounted before they enter the SSL stack and not
1576 after, which also protects the stack against bad handshakes. Also, lowering
1577 tune.maxaccept can improve fairness.
1578
William Lallemand9d5f5482012-11-07 16:12:57 +01001579maxzlibmem <number>
1580 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1581 When the maximum amount is reached, future sessions will not compress as long
1582 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001583 The default value is 0. The value is available in bytes on the UNIX socket
1584 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1585 "ZlibMemUsage" in bytes.
1586
Willy Tarreau6a06a402007-07-15 20:15:28 +02001587noepoll
1588 Disables the use of the "epoll" event polling system on Linux. It is
1589 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001590 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001591
1592nokqueue
1593 Disables the use of the "kqueue" event polling system on BSD. It is
1594 equivalent to the command-line argument "-dk". The next polling system
1595 used will generally be "poll". See also "nopoll".
1596
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001597noevports
1598 Disables the use of the event ports event polling system on SunOS systems
1599 derived from Solaris 10 and later. It is equivalent to the command-line
1600 argument "-dv". The next polling system used will generally be "poll". See
1601 also "nopoll".
1602
Willy Tarreau6a06a402007-07-15 20:15:28 +02001603nopoll
1604 Disables the use of the "poll" event polling system. It is equivalent to the
1605 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001606 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001607 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
1608 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001609
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001610nosplice
1611 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001612 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001613 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001614 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001615 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1616 be used. This option makes it easier to globally disable kernel splicing in
1617 case of doubt. See also "option splice-auto", "option splice-request" and
1618 "option splice-response".
1619
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001620nogetaddrinfo
1621 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1622 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1623
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001624noreuseport
1625 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1626 command line argument "-dR".
1627
Willy Tarreaud2d33482019-04-25 17:09:07 +02001628profiling.tasks { auto | on | off }
1629 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
1630 the profiling automatically turns on a thread when it starts to suffer from
1631 an average latency of 1000 microseconds or higher as reported in the
1632 "avg_loop_us" activity field, and automatically turns off when the latency
John Roeslerfb2fce12019-07-10 15:45:51 -05001633 returns below 990 microseconds (this value is an average over the last 1024
Willy Tarreaud2d33482019-04-25 17:09:07 +02001634 loops so it does not vary quickly and tends to significantly smooth short
1635 spikes). It may also spontaneously trigger from time to time on overloaded
1636 systems, containers, or virtual machines, or when the system swaps (which
1637 must absolutely never happen on a load balancer).
1638
1639 CPU profiling per task can be very convenient to report where the time is
1640 spent and which requests have what effect on which other request. Enabling
1641 it will typically affect the overall's performance by less than 1%, thus it
1642 is recommended to leave it to the default 'auto' value so that it only
1643 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01001644 supporting the clock_gettime(2) syscall with clock identifiers
1645 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
1646 be zero. This option may be changed at run time using "set profiling" on the
1647 CLI.
1648
Willy Tarreaufe255b72007-10-14 23:09:26 +02001649spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001650 Sometimes it is desirable to avoid sending agent and health checks to
1651 servers at exact intervals, for instance when many logical servers are
1652 located on the same physical server. With the help of this parameter, it
1653 becomes possible to add some randomness in the check interval between 0
1654 and +/- 50%. A value between 2 and 5 seems to show good results. The
1655 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001656
Davor Ocelice9ed2812017-12-25 17:49:28 +01001657ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001658 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001659 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001660 multiple times, it will simply enable multiple crypto engines. Referencing an
1661 unsupported engine will prevent haproxy from starting. Note that many engines
1662 will lead to lower HTTPS performance than pure software with recent
1663 processors. The optional command "algo" sets the default algorithms an ENGINE
1664 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001665 of "ALL" uses the engine for all cryptographic operations. If no list of
1666 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001667 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1668 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1669 openssl configuration file uses:
1670 https://www.openssl.org/docs/man1.0.2/apps/config.html
1671
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001672ssl-mode-async
1673 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001674 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001675 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1676 doesn't support moving read/write buffers and is not compliant with
1677 haproxy's buffer management. So the asynchronous mode is disabled on
John Roeslerfb2fce12019-07-10 15:45:51 -05001678 read/write operations (it is only enabled during initial and renegotiation
Emeric Brunb5e42a82017-06-06 12:35:14 +00001679 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001680
Willy Tarreau33cb0652014-12-23 22:52:37 +01001681tune.buffers.limit <number>
1682 Sets a hard limit on the number of buffers which may be allocated per process.
1683 The default value is zero which means unlimited. The minimum non-zero value
1684 will always be greater than "tune.buffers.reserve" and should ideally always
1685 be about twice as large. Forcing this value can be particularly useful to
1686 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001687 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001688 another one to be released by another session. Since buffers are dynamically
1689 allocated and released, the waiting time is very short and not perceptible
1690 provided that limits remain reasonable. In fact sometimes reducing the limit
1691 may even increase performance by increasing the CPU cache's efficiency. Tests
1692 have shown good results on average HTTP traffic with a limit to 1/10 of the
1693 expected global maxconn setting, which also significantly reduces memory
1694 usage. The memory savings come from the fact that a number of connections
1695 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1696 advised to do so by an haproxy core developer.
1697
Willy Tarreau1058ae72014-12-23 22:40:40 +01001698tune.buffers.reserve <number>
1699 Sets the number of buffers which are pre-allocated and reserved for use only
1700 during memory shortage conditions resulting in failed memory allocations. The
1701 minimum value is 2 and is also the default. There is no reason a user would
1702 want to change this value, it's mostly aimed at haproxy core developers.
1703
Willy Tarreau27a674e2009-08-17 07:23:33 +02001704tune.bufsize <number>
1705 Sets the buffer size to this size (in bytes). Lower values allow more
1706 sessions to coexist in the same amount of RAM, and higher values allow some
1707 applications with very large cookies to work. The default value is 16384 and
1708 can be changed at build time. It is strongly recommended not to change this
1709 from the default value, as very low values will break some services such as
1710 statistics, and values larger than default size will increase memory usage,
1711 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001712 parameter should be decreased by the same factor as this one is increased. In
1713 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1714 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001715 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Willy Tarreauc77d3642018-12-12 06:19:42 +01001716 than this size, haproxy will return HTTP 502 (Bad Gateway). Note that the
1717 value set using this parameter will automatically be rounded up to the next
1718 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02001719
Willy Tarreau43961d52010-10-04 20:39:20 +02001720tune.chksize <number>
1721 Sets the check buffer size to this size (in bytes). Higher values may help
1722 find string or regex patterns in very large pages, though doing so may imply
1723 more memory and CPU usage. The default value is 16384 and can be changed at
1724 build time. It is not recommended to change this value, but to use better
1725 checks whenever possible.
1726
William Lallemandf3747832012-11-09 12:33:10 +01001727tune.comp.maxlevel <number>
1728 Sets the maximum compression level. The compression level affects CPU
1729 usage during compression. This value affects CPU usage during compression.
1730 Each session using compression initializes the compression algorithm with
1731 this value. The default value is 1.
1732
Willy Tarreauc299e1e2019-02-27 11:35:12 +01001733tune.fail-alloc
1734 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
1735 allocation attempt fails. Must be between 0 (no failure) and 100 (no
1736 success). This is useful to debug and make sure memory failures are handled
1737 gracefully.
1738
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001739tune.h2.header-table-size <number>
1740 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1741 cannot be larger than 65536 bytes. A larger value may help certain clients
1742 send more compact requests, depending on their capabilities. This amount of
1743 memory is consumed for each HTTP/2 connection. It is recommended not to
1744 change it.
1745
Willy Tarreaue6baec02017-07-27 11:45:11 +02001746tune.h2.initial-window-size <number>
1747 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001748 can upload before waiting for an acknowledgment from haproxy. This setting
1749 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001750 The default value is 65535, which roughly allows up to 5 Mbps of upload
1751 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1752 over a 1-ms local network. It can make sense to increase this value to allow
1753 faster uploads, or to reduce it to increase fairness when dealing with many
1754 clients. It doesn't affect resource usage.
1755
Willy Tarreau5242ef82017-07-27 11:47:28 +02001756tune.h2.max-concurrent-streams <number>
1757 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1758 number of outstanding requests on a single connection). The default value is
1759 100. A larger one may slightly improve page load time for complex sites when
1760 visited over high latency networks, but increases the amount of resources a
1761 single client may allocate. A value of zero disables the limit so a single
1762 client may create as many streams as allocatable by haproxy. It is highly
1763 recommended not to change this value.
1764
Willy Tarreaua24b35c2019-02-21 13:24:36 +01001765tune.h2.max-frame-size <number>
1766 Sets the HTTP/2 maximum frame size that haproxy announces it is willing to
1767 receive to its peers. The default value is the largest between 16384 and the
1768 buffer size (tune.bufsize). In any case, haproxy will not announce support
1769 for frame sizes larger than buffers. The main purpose of this setting is to
1770 allow to limit the maximum frame size setting when using large buffers. Too
1771 large frame sizes might have performance impact or cause some peers to
1772 misbehave. It is highly recommended not to change this value.
1773
Willy Tarreau193b8c62012-11-22 00:17:38 +01001774tune.http.cookielen <number>
1775 Sets the maximum length of captured cookies. This is the maximum value that
1776 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1777 will automatically be truncated to this one. It is important not to set too
1778 high a value because all cookie captures still allocate this size whatever
1779 their configured value (they share a same pool). This value is per request
1780 per response, so the memory allocated is twice this value per connection.
1781 When not specified, the limit is set to 63 characters. It is recommended not
1782 to change this value.
1783
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001784tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001785 Sets the maximum length of request URI in logs. This prevents truncating long
1786 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001787 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001788 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001789 configuration directives too.
1790 The default value is 1024.
1791
Willy Tarreauac1932d2011-10-24 19:14:41 +02001792tune.http.maxhdr <number>
1793 Sets the maximum number of headers in a request. When a request comes with a
1794 number of headers greater than this value (including the first line), it is
1795 rejected with a "400 Bad Request" status code. Similarly, too large responses
1796 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1797 for all usages, considering that the widely deployed Apache server uses the
1798 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001799 a buggy application to work by the time it gets fixed. The accepted range is
1800 1..32767. Keep in mind that each new header consumes 32bits of memory for
1801 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001802
Willy Tarreau7e312732014-02-12 16:35:14 +01001803tune.idletimer <timeout>
1804 Sets the duration after which haproxy will consider that an empty buffer is
1805 probably associated with an idle stream. This is used to optimally adjust
1806 some packet sizes while forwarding large and small data alternatively. The
1807 decision to use splice() or to send large buffers in SSL is modulated by this
1808 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1809 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001810 which seems to correctly detect end user pauses (e.g. read a page before
John Roeslerfb2fce12019-07-10 15:45:51 -05001811 clicking). There should be no reason for changing this value. Please check
Willy Tarreau7e312732014-02-12 16:35:14 +01001812 tune.ssl.maxrecord below.
1813
Willy Tarreau7ac908b2019-02-27 12:02:18 +01001814tune.listener.multi-queue { on | off }
1815 Enables ('on') or disables ('off') the listener's multi-queue accept which
1816 spreads the incoming traffic to all threads a "bind" line is allowed to run
1817 on instead of taking them for itself. This provides a smoother traffic
1818 distribution and scales much better, especially in environments where threads
1819 may be unevenly loaded due to external activity (network interrupts colliding
1820 with one thread for example). This option is enabled by default, but it may
1821 be forcefully disabled for troubleshooting or for situations where it is
1822 estimated that the operating system already provides a good enough
1823 distribution and connections are extremely short-lived.
1824
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001825tune.lua.forced-yield <number>
1826 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001827 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001828 HAProxy scheduler to process other tasks like accepting connections or
1829 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001830 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001831 lowered. If the Lua code is quite long and its result is absolutely required
1832 to process the data, the <number> can be increased.
1833
Willy Tarreau32f61e22015-03-18 17:54:59 +01001834tune.lua.maxmem
1835 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1836 default it is zero which means unlimited. It is important to set a limit to
1837 ensure that a bug in a script will not result in the system running out of
1838 memory.
1839
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001840tune.lua.session-timeout <timeout>
1841 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001842 preventing infinite loops or spending too much time in Lua. This timeout
1843 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001844 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001845
1846tune.lua.task-timeout <timeout>
1847 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1848 dedicated to the tasks. By default, this timeout isn't set because a task may
1849 remain alive during of the lifetime of HAProxy. For example, a task used to
1850 check servers.
1851
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001852tune.lua.service-timeout <timeout>
1853 This is the execution timeout for the Lua services. This is useful for
1854 preventing infinite loops or spending too much time in Lua. This timeout
1855 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001856 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001857
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001858tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001859 Sets the maximum number of consecutive connections a process may accept in a
1860 row before switching to other work. In single process mode, higher numbers
1861 give better performance at high connection rates. However in multi-process
1862 modes, keeping a bit of fairness between processes generally is better to
1863 increase performance. This value applies individually to each listener, so
1864 that the number of processes a listener is bound to is taken into account.
1865 This value defaults to 64. In multi-process mode, it is divided by twice
1866 the number of processes the listener is bound to. Setting this value to -1
1867 completely disables the limitation. It should normally not be needed to tweak
1868 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001869
1870tune.maxpollevents <number>
1871 Sets the maximum amount of events that can be processed at once in a call to
1872 the polling system. The default value is adapted to the operating system. It
1873 has been noticed that reducing it below 200 tends to slightly decrease
1874 latency at the expense of network bandwidth, and increasing it above 200
1875 tends to trade latency for slightly increased bandwidth.
1876
Willy Tarreau27a674e2009-08-17 07:23:33 +02001877tune.maxrewrite <number>
1878 Sets the reserved buffer space to this size in bytes. The reserved space is
1879 used for header rewriting or appending. The first reads on sockets will never
1880 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1881 bufsize, though that does not make much sense since there are rarely large
1882 numbers of headers to add. Setting it too high prevents processing of large
1883 requests or responses. Setting it too low prevents addition of new headers
1884 to already large requests or to POST requests. It is generally wise to set it
1885 to about 1024. It is automatically readjusted to half of bufsize if it is
1886 larger than that. This means you don't have to worry about it when changing
1887 bufsize.
1888
Willy Tarreauf3045d22015-04-29 16:24:50 +02001889tune.pattern.cache-size <number>
1890 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1891 cache which reminds previous lookups and their results. It is used by ACLs
1892 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1893 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1894 strings. It applies to pattern expressions which means that it will be able
1895 to memorize the result of a lookup among all the patterns specified on a
1896 configuration line (including all those loaded from files). It automatically
1897 invalidates entries which are updated using HTTP actions or on the CLI. The
1898 default cache size is set to 10000 entries, which limits its footprint to
Willy Tarreau403bfbb2019-10-23 06:59:31 +02001899 about 5 MB per process/thread on 32-bit systems and 8 MB per process/thread
1900 on 64-bit systems, as caches are thread/process local. There is a very low
Willy Tarreauf3045d22015-04-29 16:24:50 +02001901 risk of collision in this cache, which is in the order of the size of the
1902 cache divided by 2^64. Typically, at 10000 requests per second with the
1903 default cache size of 10000 entries, there's 1% chance that a brute force
1904 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1905 This is considered much lower than the risk of a memory corruption caused by
1906 aging components. If this is not acceptable, the cache can be disabled by
1907 setting this parameter to 0.
1908
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001909tune.pipesize <number>
1910 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1911 are the default size for the system. But sometimes when using TCP splicing,
1912 it can improve performance to increase pipe sizes, especially if it is
1913 suspected that pipes are not filled and that many calls to splice() are
1914 performed. This has an impact on the kernel's memory footprint, so this must
1915 not be changed if impacts are not understood.
1916
Olivier Houchard88698d92019-04-16 19:07:22 +02001917tune.pool-low-fd-ratio <number>
1918 This setting sets the max number of file descriptors (in percentage) used by
1919 haproxy globally against the maximum number of file descriptors haproxy can
1920 use before we stop putting connection into the idle pool for reuse. The
1921 default is 20.
1922
1923tune.pool-high-fd-ratio <number>
1924 This setting sets the max number of file descriptors (in percentage) used by
1925 haproxy globally against the maximum number of file descriptors haproxy can
1926 use before we start killing idle connections when we can't reuse a connection
1927 and we have to create a new one. The default is 25 (one quarter of the file
1928 descriptor will mean that roughly half of the maximum front connections can
1929 keep an idle connection behind, anything beyond this probably doesn't make
John Roeslerfb2fce12019-07-10 15:45:51 -05001930 much sense in the general case when targeting connection reuse).
Olivier Houchard88698d92019-04-16 19:07:22 +02001931
Willy Tarreaue803de22010-01-21 17:43:04 +01001932tune.rcvbuf.client <number>
1933tune.rcvbuf.server <number>
1934 Forces the kernel socket receive buffer size on the client or the server side
1935 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1936 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05001937 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001938 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001939 order to save kernel memory by preventing it from buffering too large amounts
1940 of received data. Lower values will significantly increase CPU usage though.
1941
Willy Tarreaub22fc302015-12-14 12:04:35 +01001942tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001943 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01001944 socket buffers. One of them is that a read returns more than <recv_enough>
1945 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1946 may be changed by this setting to better deal with workloads involving lots
1947 of short messages such as telnet or SSH sessions.
1948
Olivier Houchard1599b802018-05-24 18:59:04 +02001949tune.runqueue-depth <number>
John Roeslerfb2fce12019-07-10 15:45:51 -05001950 Sets the maximum amount of task that can be processed at once when running
Olivier Houchard1599b802018-05-24 18:59:04 +02001951 tasks. The default value is 200. Increasing it may incur latency when
1952 dealing with I/Os, making it too small can incur extra overhead.
1953
Willy Tarreaue803de22010-01-21 17:43:04 +01001954tune.sndbuf.client <number>
1955tune.sndbuf.server <number>
1956 Forces the kernel socket send buffer size on the client or the server side to
1957 the specified value in bytes. This value applies to all TCP/HTTP frontends
1958 and backends. It should normally never be set, and the default size (0) lets
John Roeslerfb2fce12019-07-10 15:45:51 -05001959 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001960 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001961 order to save kernel memory by preventing it from buffering too large amounts
1962 of received data. Lower values will significantly increase CPU usage though.
1963 Another use case is to prevent write timeouts with extremely slow clients due
1964 to the kernel waiting for a large part of the buffer to be read before
1965 notifying haproxy again.
1966
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001967tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001968 Sets the size of the global SSL session cache, in a number of blocks. A block
1969 is large enough to contain an encoded session without peer certificate.
1970 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001971 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001972 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01001973 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01001974 and reassigned. Higher values reduce the occurrence of such a purge, hence
1975 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1976 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001977 and are shared between all processes if "nbproc" is greater than 1. Setting
1978 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001979
Emeric Brun8dc60392014-05-09 13:52:00 +02001980tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02001981 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02001982 should normally not be used since it will force many renegotiations due to
1983 clients hitting a random process. But it may be required on some operating
1984 systems where none of the SSL cache synchronization method may be used. In
1985 this case, adding a first layer of hash-based load balancing before the SSL
1986 layer might limit the impact of the lack of session sharing.
1987
Emeric Brun4f65bff2012-11-16 15:11:00 +01001988tune.ssl.lifetime <timeout>
1989 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001990 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001991 does not guarantee that sessions will last that long, because if the cache is
1992 full, the longest idle sessions will be purged despite their configured
1993 lifetime. The real usefulness of this setting is to prevent sessions from
1994 being used for too long.
1995
Willy Tarreaubfd59462013-02-21 07:46:09 +01001996tune.ssl.maxrecord <number>
1997 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1998 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1999 data only once it has received a full record. With large records, it means
2000 that clients might have to download up to 16kB of data before starting to
2001 process them. Limiting the value can improve page load times on browsers
2002 located over high latency or low bandwidth networks. It is suggested to find
2003 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
2004 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
2005 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
2006 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01002007 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01002008 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01002009
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002010tune.ssl.default-dh-param <number>
2011 Sets the maximum size of the Diffie-Hellman parameters used for generating
2012 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
2013 final size will try to match the size of the server's RSA (or DSA) key (e.g,
2014 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
2015 this maximum value. Default value if 1024. Only 1024 or higher values are
2016 allowed. Higher values will increase the CPU load, and values greater than
2017 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02002018 used if static Diffie-Hellman parameters are supplied either directly
2019 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02002020
Christopher Faulet31af49d2015-06-09 17:29:50 +02002021tune.ssl.ssl-ctx-cache-size <number>
2022 Sets the size of the cache used to store generated certificates to <number>
2023 entries. This is a LRU cache. Because generating a SSL certificate
2024 dynamically is expensive, they are cached. The default cache size is set to
2025 1000 entries.
2026
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01002027tune.ssl.capture-cipherlist-size <number>
2028 Sets the maximum size of the buffer used for capturing client-hello cipher
2029 list. If the value is 0 (default value) the capture is disabled, otherwise
2030 a buffer is allocated for each SSL/TLS connection.
2031
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002032tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002033tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002034tune.vars.reqres-max-size <size>
2035tune.vars.sess-max-size <size>
2036tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01002037 These five tunes help to manage the maximum amount of memory used by the
2038 variables system. "global" limits the overall amount of memory available for
2039 all scopes. "proc" limits the memory for the process scope, "sess" limits the
2040 memory for the session scope, "txn" for the transaction scope, and "reqres"
2041 limits the memory for each request or response processing.
2042 Memory accounting is hierarchical, meaning more coarse grained limits include
2043 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
2044 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002045
Daniel Schneller0b547052016-03-21 20:46:57 +01002046 For example, when "tune.vars.sess-max-size" is limited to 100,
2047 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
2048 100 either. If we create a variable "txn.var" that contains 100 bytes,
2049 all available space is consumed.
2050 Notice that exceeding the limits at runtime will not result in an error
2051 message, but values might be cut off or corrupted. So make sure to accurately
2052 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02002053
William Lallemanda509e4c2012-11-07 16:54:34 +01002054tune.zlib.memlevel <number>
2055 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002056 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01002057 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01002058 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01002059 between 1 and 9. The default value is 8.
2060
2061tune.zlib.windowsize <number>
2062 Sets the window size (the size of the history buffer) as a parameter of the
2063 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01002064 in better compression at the expense of memory usage. Can be a value between
2065 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002066
Willy Tarreauc57f0e22009-05-10 13:12:33 +020020673.3. Debugging
2068--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02002069
2070debug
2071 Enables debug mode which dumps to stdout all exchanges, and disables forking
2072 into background. It is the equivalent of the command-line argument "-d". It
2073 should never be used in a production configuration since it may prevent full
2074 system startup.
2075
2076quiet
2077 Do not display any message during startup. It is equivalent to the command-
2078 line argument "-q".
2079
Emeric Brunf099e792010-09-27 12:05:28 +02002080
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010020813.4. Userlists
2082--------------
2083It is possible to control access to frontend/backend/listen sections or to
2084http stats by allowing only authenticated and authorized users. To do this,
2085it is required to create at least one userlist and to define users.
2086
2087userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01002088 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002089 used to store authentication & authorization data for independent customers.
2090
2091group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01002092 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002093 attach users to this group by using a comma separated list of names
2094 proceeded by "users" keyword.
2095
Cyril Bontéf0c60612010-02-06 14:44:47 +01002096user <username> [password|insecure-password <password>]
2097 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002098 Adds user <username> to the current userlist. Both secure (encrypted) and
2099 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002100 evaluated using the crypt(3) function, so depending on the system's
2101 capabilities, different algorithms are supported. For example, modern Glibc
2102 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
2103 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002104
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002105 Attention: Be aware that using encrypted passwords might cause significantly
2106 increased CPU usage, depending on the number of requests, and the algorithm
2107 used. For any of the hashed variants, the password for each request must
2108 be processed through the chosen algorithm, before it can be compared to the
2109 value specified in the config file. Most current algorithms are deliberately
2110 designed to be expensive to compute to achieve resistance against brute
2111 force attacks. They do not simply salt/hash the clear text password once,
2112 but thousands of times. This can quickly become a major factor in haproxy's
2113 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002114
2115 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01002116 userlist L1
2117 group G1 users tiger,scott
2118 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002119
Cyril Bontéf0c60612010-02-06 14:44:47 +01002120 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
2121 user scott insecure-password elgato
2122 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002123
Cyril Bontéf0c60612010-02-06 14:44:47 +01002124 userlist L2
2125 group G1
2126 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002127
Cyril Bontéf0c60612010-02-06 14:44:47 +01002128 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
2129 user scott insecure-password elgato groups G1,G2
2130 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002131
2132 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002133
Emeric Brunf099e792010-09-27 12:05:28 +02002134
21353.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002136----------
Emeric Brun94900952015-06-11 18:25:54 +02002137It is possible to propagate entries of any data-types in stick-tables between
2138several haproxy instances over TCP connections in a multi-master fashion. Each
2139instance pushes its local updates and insertions to remote peers. The pushed
2140values overwrite remote ones without aggregation. Interrupted exchanges are
2141automatically detected and recovered from the last known point.
2142In addition, during a soft restart, the old process connects to the new one
2143using such a TCP connection to push all its entries before the new process
2144tries to connect to other peers. That ensures very fast replication during a
2145reload, it typically takes a fraction of a second even for large tables.
2146Note that Server IDs are used to identify servers remotely, so it is important
2147that configurations look similar or at least that the same IDs are forced on
2148each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002149
2150peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002151 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002152 which is referenced by one or more stick-tables.
2153
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002154bind [<address>]:<port_range> [, ...] [param*]
2155 Defines the binding parameters of the local peer of this "peers" section.
2156 Such lines are not supported with "peer" line in the same "peers" section.
2157
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002158disabled
2159 Disables a peers section. It disables both listening and any synchronization
2160 related to this section. This is provided to disable synchronization of stick
2161 tables without having to comment out all "peers" references.
2162
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002163default-bind [param*]
2164 Defines the binding parameters for the local peer, excepted its address.
2165
2166default-server [param*]
2167 Change default options for a server in a "peers" section.
2168
2169 Arguments:
2170 <param*> is a list of parameters for this server. The "default-server"
2171 keyword accepts an important number of options and has a complete
2172 section dedicated to it. Please refer to section 5 for more
2173 details.
2174
2175
2176 See also: "server" and section 5 about server options
2177
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002178enable
2179 This re-enables a disabled peers section which was previously disabled.
2180
Frédéric Lécailleb6f759b2019-11-05 09:57:45 +01002181log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
2182 <facility> [<level> [<minlevel>]]
2183 "peers" sections support the same "log" keyword as for the proxies to
2184 log information about the "peers" listener. See "log" option for proxies for
2185 more details.
2186
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002187peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002188 Defines a peer inside a peers section.
2189 If <peername> is set to the local peer name (by default hostname, or forced
2190 using "-L" command line option), haproxy will listen for incoming remote peer
2191 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
2192 to join the remote peer, and <peername> is used at the protocol level to
2193 identify and validate the remote peer on the server side.
2194
2195 During a soft restart, local peer <ip>:<port> is used by the old instance to
2196 connect the new one and initiate a complete replication (teaching process).
2197
2198 It is strongly recommended to have the exact same peers declaration on all
2199 peers and to only rely on the "-L" command line argument to change the local
2200 peer name. This makes it easier to maintain coherent configuration files
2201 across all peers.
2202
William Lallemandb2f07452015-05-12 14:27:13 +02002203 You may want to reference some environment variables in the address
2204 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01002205
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002206 Note: "peer" keyword may transparently be replaced by "server" keyword (see
2207 "server" keyword explanation below).
2208
2209server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02002210 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002211 with a support for all "server" parameters found in 5.2 paragraph.
2212 If the underlying peer is local, <ip>:<port> parameters must not be present.
2213 These parameters must be provided on a "bind" line (see "bind" keyword
2214 of this "peers" section).
2215 Some of these parameters are irrelevant for "peers" sections.
2216
2217
Cyril Bontédc4d9032012-04-08 21:57:39 +02002218 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002219 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02002220 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002221 peer haproxy1 192.168.0.1:1024
2222 peer haproxy2 192.168.0.2:1024
2223 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02002224
2225 backend mybackend
2226 mode tcp
2227 balance roundrobin
2228 stick-table type ip size 20k peers mypeers
2229 stick on src
2230
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002231 server srv1 192.168.0.30:80
2232 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02002233
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002234 Example:
2235 peers mypeers
2236 bind 127.0.0.11:10001 ssl crt mycerts/pem
2237 default-server ssl verify none
2238 server hostA 127.0.0.10:10000
2239 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02002240
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002241
2242table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
2243 size <size> [expire <expire>] [nopurge] [store <data_type>]*
2244
2245 Configure a stickiness table for the current section. This line is parsed
2246 exactly the same way as the "stick-table" keyword in others section, except
John Roeslerfb2fce12019-07-10 15:45:51 -05002247 for the "peers" argument which is not required here and with an additional
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002248 mandatory first parameter to designate the stick-table. Contrary to others
2249 sections, there may be several "table" lines in "peers" sections (see also
2250 "stick-table" keyword).
2251
2252 Also be aware of the fact that "peers" sections have their own stick-table
2253 namespaces to avoid collisions between stick-table names identical in
2254 different "peers" section. This is internally handled prepending the "peers"
2255 sections names to the name of the stick-tables followed by a '/' character.
2256 If somewhere else in the configuration file you have to refer to such
2257 stick-tables declared in "peers" sections you must use the prefixed version
2258 of the stick-table name as follows:
2259
2260 peers mypeers
2261 peer A ...
2262 peer B ...
2263 table t1 ...
2264
2265 frontend fe1
2266 tcp-request content track-sc0 src table mypeers/t1
2267
2268 This is also this prefixed version of the stick-table names which must be
2269 used to refer to stick-tables through the CLI.
2270
2271 About "peers" protocol, as only "peers" belonging to the same section may
2272 communicate with each others, there is no need to do such a distinction.
2273 Several "peers" sections may declare stick-tables with the same name.
2274 This is shorter version of the stick-table name which is sent over the network.
2275 There is only a '/' character as prefix to avoid stick-table name collisions between
2276 stick-tables declared as backends and stick-table declared in "peers" sections
2277 as follows in this weird but supported configuration:
2278
2279 peers mypeers
2280 peer A ...
2281 peer B ...
2282 table t1 type string size 10m store gpc0
2283
2284 backend t1
2285 stick-table type string size 10m store gpc0 peers mypeers
2286
2287 Here "t1" table declared in "mypeeers" section has "mypeers/t1" as global name.
2288 "t1" table declared as a backend as "t1" as global name. But at peer protocol
2289 level the former table is named "/t1", the latter is again named "t1".
2290
Simon Horman51a1cf62015-02-03 13:00:44 +090022913.6. Mailers
2292------------
2293It is possible to send email alerts when the state of servers changes.
2294If configured email alerts are sent to each mailer that is configured
2295in a mailers section. Email is sent to mailers using SMTP.
2296
Pieter Baauw386a1272015-08-16 15:26:24 +02002297mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09002298 Creates a new mailer list with the name <mailersect>. It is an
2299 independent section which is referenced by one or more proxies.
2300
2301mailer <mailername> <ip>:<port>
2302 Defines a mailer inside a mailers section.
2303
2304 Example:
2305 mailers mymailers
2306 mailer smtp1 192.168.0.1:587
2307 mailer smtp2 192.168.0.2:587
2308
2309 backend mybackend
2310 mode tcp
2311 balance roundrobin
2312
2313 email-alert mailers mymailers
2314 email-alert from test1@horms.org
2315 email-alert to test2@horms.org
2316
2317 server srv1 192.168.0.30:80
2318 server srv2 192.168.0.31:80
2319
Pieter Baauw235fcfc2016-02-13 15:33:40 +01002320timeout mail <time>
2321 Defines the time available for a mail/connection to be made and send to
2322 the mail-server. If not defined the default value is 10 seconds. To allow
2323 for at least two SYN-ACK packets to be send during initial TCP handshake it
2324 is advised to keep this value above 4 seconds.
2325
2326 Example:
2327 mailers mymailers
2328 timeout mail 20s
2329 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09002330
William Lallemandc9515522019-06-12 16:32:11 +020023313.7. Programs
2332-------------
2333In master-worker mode, it is possible to launch external binaries with the
2334master, these processes are called programs. These programs are launched and
2335managed the same way as the workers.
2336
2337During a reload of HAProxy, those processes are dealing with the same
2338sequence as a worker:
2339
2340 - the master is re-executed
2341 - the master sends a SIGUSR1 signal to the program
2342 - if "option start-on-reload" is not disabled, the master launches a new
2343 instance of the program
2344
2345During a stop, or restart, a SIGTERM is sent to the programs.
2346
2347program <name>
2348 This is a new program section, this section will create an instance <name>
2349 which is visible in "show proc" on the master CLI. (See "9.4. Master CLI" in
2350 the management guide).
2351
2352command <command> [arguments*]
2353 Define the command to start with optional arguments. The command is looked
2354 up in the current PATH if it does not include an absolute path. This is a
2355 mandatory option of the program section. Arguments containing spaces must
2356 be enclosed in quotes or double quotes or be prefixed by a backslash.
2357
Andrew Heberle97236962019-07-12 11:50:26 +08002358user <user name>
2359 Changes the executed command user ID to the <user name> from /etc/passwd.
2360 See also "group".
2361
2362group <group name>
2363 Changes the executed command group ID to the <group name> from /etc/group.
2364 See also "user".
2365
William Lallemandc9515522019-06-12 16:32:11 +02002366option start-on-reload
2367no option start-on-reload
2368 Start (or not) a new instance of the program upon a reload of the master.
2369 The default is to start a new instance. This option may only be used in a
2370 program section.
2371
2372
Willy Tarreauc57f0e22009-05-10 13:12:33 +020023734. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02002374----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002375
Willy Tarreau6a06a402007-07-15 20:15:28 +02002376Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02002377 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02002378 - frontend <name>
2379 - backend <name>
2380 - listen <name>
2381
2382A "defaults" section sets default parameters for all other sections following
2383its declaration. Those default parameters are reset by the next "defaults"
2384section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01002385section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002386
2387A "frontend" section describes a set of listening sockets accepting client
2388connections.
2389
2390A "backend" section describes a set of servers to which the proxy will connect
2391to forward incoming connections.
2392
2393A "listen" section defines a complete proxy with its frontend and backend
2394parts combined in one section. It is generally useful for TCP-only traffic.
2395
Willy Tarreau0ba27502007-12-24 16:55:16 +01002396All proxy names must be formed from upper and lower case letters, digits,
2397'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
2398case-sensitive, which means that "www" and "WWW" are two different proxies.
2399
2400Historically, all proxy names could overlap, it just caused troubles in the
2401logs. Since the introduction of content switching, it is mandatory that two
2402proxies with overlapping capabilities (frontend/backend) have different names.
2403However, it is still permitted that a frontend and a backend share the same
2404name, as this configuration seems to be commonly encountered.
2405
2406Right now, two major proxy modes are supported : "tcp", also known as layer 4,
2407and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002408bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002409protocol, and can interact with it by allowing, blocking, switching, adding,
2410modifying, or removing arbitrary contents in requests or responses, based on
2411arbitrary criteria.
2412
Willy Tarreau70dffda2014-01-30 03:07:23 +01002413In HTTP mode, the processing applied to requests and responses flowing over
2414a connection depends in the combination of the frontend's HTTP options and
Julien Pivotto21ad3152019-12-10 13:11:17 +01002415the backend's. HAProxy supports 3 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01002416
2417 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
2418 requests and responses are processed, and connections remain open but idle
2419 between responses and new requests.
2420
Willy Tarreau70dffda2014-01-30 03:07:23 +01002421 - SCL: server close ("option http-server-close") : the server-facing
2422 connection is closed after the end of the response is received, but the
2423 client-facing connection remains open.
2424
Christopher Faulet315b39c2018-09-21 16:26:19 +02002425 - CLO: close ("option httpclose"): the connection is closed after the end of
2426 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002427
2428The effective mode that will be applied to a connection passing through a
2429frontend and a backend can be determined by both proxy modes according to the
2430following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02002431weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002432
Christopher Faulet315b39c2018-09-21 16:26:19 +02002433 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01002434
Christopher Faulet315b39c2018-09-21 16:26:19 +02002435 | KAL | SCL | CLO
2436 ----+-----+-----+----
2437 KAL | KAL | SCL | CLO
2438 ----+-----+-----+----
Christopher Faulet315b39c2018-09-21 16:26:19 +02002439 mode SCL | SCL | SCL | CLO
2440 ----+-----+-----+----
2441 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01002442
Willy Tarreau0ba27502007-12-24 16:55:16 +01002443
Willy Tarreau70dffda2014-01-30 03:07:23 +01002444
Willy Tarreauc57f0e22009-05-10 13:12:33 +020024454.1. Proxy keywords matrix
2446--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002447
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002448The following list of keywords is supported. Most of them may only be used in a
2449limited set of section types. Some of them are marked as "deprecated" because
2450they are inherited from an old syntax which may be confusing or functionally
2451limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002452marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002453option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002454and must be disabled for a specific instance. Such options may also be prefixed
2455with "default" in order to restore default settings regardless of what has been
2456specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002457
Willy Tarreau6a06a402007-07-15 20:15:28 +02002458
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002459 keyword defaults frontend listen backend
2460------------------------------------+----------+----------+---------+---------
2461acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002462backlog X X X -
2463balance X - X X
2464bind - X X -
2465bind-process X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002466capture cookie - X X -
2467capture request header - X X -
2468capture response header - X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002469compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002470cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002471declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002472default-server X - X X
2473default_backend X X X -
2474description - X X X
2475disabled X X X X
2476dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002477email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002478email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002479email-alert mailers X X X X
2480email-alert myhostname X X X X
2481email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002482enabled X X X X
2483errorfile X X X X
2484errorloc X X X X
2485errorloc302 X X X X
2486-- keyword -------------------------- defaults - frontend - listen -- backend -
2487errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002488force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002489filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002490fullconn X - X X
2491grace X X X X
2492hash-type X - X X
2493http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002494http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002495http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002496http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002497http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002498http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002499http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002500id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002501ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002502load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002503log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002504log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002505log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002506log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002507max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002508maxconn X X X -
2509mode X X X X
2510monitor fail - X X -
2511monitor-net X X X -
2512monitor-uri X X X -
2513option abortonclose (*) X - X X
2514option accept-invalid-http-request (*) X X X -
2515option accept-invalid-http-response (*) X - X X
2516option allbackups (*) X - X X
2517option checkcache (*) X - X X
2518option clitcpka (*) X X X -
2519option contstats (*) X X X -
2520option dontlog-normal (*) X X X -
2521option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002522-- keyword -------------------------- defaults - frontend - listen -- backend -
2523option forwardfor X X X X
Christopher Faulet98fbe952019-07-22 16:18:24 +02002524option h1-case-adjust-bogus-client (*) X X X -
2525option h1-case-adjust-bogus-server (*) X - X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002526option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002527option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002528option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002529option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02002530option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002531option http-server-close (*) X X X X
2532option http-use-proxy-header (*) X X X -
2533option httpchk X - X X
2534option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01002535option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002536option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002537option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002538option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002539option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002540option log-health-checks (*) X - X X
2541option log-separate-errors (*) X X X -
2542option logasap (*) X X X -
2543option mysql-check X - X X
2544option nolinger (*) X X X X
2545option originalto X X X X
2546option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002547option pgsql-check X - X X
2548option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002549option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002550option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002551option smtpchk X - X X
2552option socket-stats (*) X X X -
2553option splice-auto (*) X X X X
2554option splice-request (*) X X X X
2555option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002556option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002557option srvtcpka (*) X - X X
2558option ssl-hello-chk X - X X
2559-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002560option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002561option tcp-smart-accept (*) X X X -
2562option tcp-smart-connect (*) X - X X
2563option tcpka X X X X
2564option tcplog X X X X
2565option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002566external-check command X - X X
2567external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002568persist rdp-cookie X - X X
2569rate-limit sessions X X X -
2570redirect - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002571-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002572retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02002573retry-on X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002574server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002575server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002576server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002577source X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002578stats admin - X X X
2579stats auth X X X X
2580stats enable X X X X
2581stats hide-version X X X X
2582stats http-request - X X X
2583stats realm X X X X
2584stats refresh X X X X
2585stats scope X X X X
2586stats show-desc X X X X
2587stats show-legends X X X X
2588stats show-node X X X X
2589stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002590-- keyword -------------------------- defaults - frontend - listen -- backend -
2591stick match - - X X
2592stick on - - X X
2593stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002594stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002595stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002596tcp-check connect - - X X
2597tcp-check expect - - X X
2598tcp-check send - - X X
2599tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002600tcp-request connection - X X -
2601tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002602tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002603tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002604tcp-response content - - X X
2605tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002606timeout check X - X X
2607timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002608timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002609timeout connect X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002610timeout http-keep-alive X X X X
2611timeout http-request X X X X
2612timeout queue X - X X
2613timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002614timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002615timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002616timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002617transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002618unique-id-format X X X -
2619unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002620use_backend - X X -
Christopher Fauletb30b3102019-09-12 23:03:09 +02002621use-fcgi-app - - X X
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002622use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002623------------------------------------+----------+----------+---------+---------
2624 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002625
Willy Tarreau0ba27502007-12-24 16:55:16 +01002626
Willy Tarreauc57f0e22009-05-10 13:12:33 +020026274.2. Alphabetically sorted keywords reference
2628---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002629
2630This section provides a description of each keyword and its usage.
2631
2632
2633acl <aclname> <criterion> [flags] [operator] <value> ...
2634 Declare or complete an access list.
2635 May be used in sections : defaults | frontend | listen | backend
2636 no | yes | yes | yes
2637 Example:
2638 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2639 acl invalid_src src_port 0:1023
2640 acl local_dst hdr(host) -i localhost
2641
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002642 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002643
2644
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002645backlog <conns>
2646 Give hints to the system about the approximate listen backlog desired size
2647 May be used in sections : defaults | frontend | listen | backend
2648 yes | yes | yes | no
2649 Arguments :
2650 <conns> is the number of pending connections. Depending on the operating
2651 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002652 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002653
2654 In order to protect against SYN flood attacks, one solution is to increase
2655 the system's SYN backlog size. Depending on the system, sometimes it is just
2656 tunable via a system parameter, sometimes it is not adjustable at all, and
2657 sometimes the system relies on hints given by the application at the time of
2658 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2659 to the listen() syscall. On systems which can make use of this value, it can
2660 sometimes be useful to be able to specify a different value, hence this
2661 backlog parameter.
2662
2663 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2664 used as a hint and the system accepts up to the smallest greater power of
2665 two, and never more than some limits (usually 32768).
2666
2667 See also : "maxconn" and the target operating system's tuning guide.
2668
2669
Willy Tarreau0ba27502007-12-24 16:55:16 +01002670balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002671balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002672 Define the load balancing algorithm to be used in a backend.
2673 May be used in sections : defaults | frontend | listen | backend
2674 yes | no | yes | yes
2675 Arguments :
2676 <algorithm> is the algorithm used to select a server when doing load
2677 balancing. This only applies when no persistence information
2678 is available, or when a connection is redispatched to another
2679 server. <algorithm> may be one of the following :
2680
2681 roundrobin Each server is used in turns, according to their weights.
2682 This is the smoothest and fairest algorithm when the server's
2683 processing time remains equally distributed. This algorithm
2684 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002685 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002686 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002687 large farms, when a server becomes up after having been down
2688 for a very short time, it may sometimes take a few hundreds
2689 requests for it to be re-integrated into the farm and start
2690 receiving traffic. This is normal, though very rare. It is
2691 indicated here in case you would have the chance to observe
2692 it, so that you don't worry.
2693
2694 static-rr Each server is used in turns, according to their weights.
2695 This algorithm is as similar to roundrobin except that it is
2696 static, which means that changing a server's weight on the
2697 fly will have no effect. On the other hand, it has no design
2698 limitation on the number of servers, and when a server goes
2699 up, it is always immediately reintroduced into the farm, once
2700 the full map is recomputed. It also uses slightly less CPU to
2701 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002702
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002703 leastconn The server with the lowest number of connections receives the
2704 connection. Round-robin is performed within groups of servers
2705 of the same load to ensure that all servers will be used. Use
2706 of this algorithm is recommended where very long sessions are
2707 expected, such as LDAP, SQL, TSE, etc... but is not very well
2708 suited for protocols using short sessions such as HTTP. This
2709 algorithm is dynamic, which means that server weights may be
2710 adjusted on the fly for slow starts for instance.
2711
Willy Tarreauf09c6602012-02-13 17:12:08 +01002712 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002713 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002714 identifier to the highest (see server parameter "id"), which
2715 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002716 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002717 not make sense to use this algorithm without setting maxconn.
2718 The purpose of this algorithm is to always use the smallest
2719 number of servers so that extra servers can be powered off
2720 during non-intensive hours. This algorithm ignores the server
2721 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002722 or IMAP than HTTP, though it can be useful there too. In
2723 order to use this algorithm efficiently, it is recommended
2724 that a cloud controller regularly checks server usage to turn
2725 them off when unused, and regularly checks backend queue to
2726 turn new servers on when the queue inflates. Alternatively,
2727 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002728
Willy Tarreau0ba27502007-12-24 16:55:16 +01002729 source The source IP address is hashed and divided by the total
2730 weight of the running servers to designate which server will
2731 receive the request. This ensures that the same client IP
2732 address will always reach the same server as long as no
2733 server goes down or up. If the hash result changes due to the
2734 number of running servers changing, many clients will be
2735 directed to a different server. This algorithm is generally
2736 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002737 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002738 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002739 static by default, which means that changing a server's
2740 weight on the fly will have no effect, but this can be
2741 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002742
Oskar Stolc8dc41842012-05-19 10:19:54 +01002743 uri This algorithm hashes either the left part of the URI (before
2744 the question mark) or the whole URI (if the "whole" parameter
2745 is present) and divides the hash value by the total weight of
2746 the running servers. The result designates which server will
2747 receive the request. This ensures that the same URI will
2748 always be directed to the same server as long as no server
2749 goes up or down. This is used with proxy caches and
2750 anti-virus proxies in order to maximize the cache hit rate.
2751 Note that this algorithm may only be used in an HTTP backend.
2752 This algorithm is static by default, which means that
2753 changing a server's weight on the fly will have no effect,
2754 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002755
Oskar Stolc8dc41842012-05-19 10:19:54 +01002756 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002757 "depth", both followed by a positive integer number. These
2758 options may be helpful when it is needed to balance servers
2759 based on the beginning of the URI only. The "len" parameter
2760 indicates that the algorithm should only consider that many
2761 characters at the beginning of the URI to compute the hash.
2762 Note that having "len" set to 1 rarely makes sense since most
2763 URIs start with a leading "/".
2764
2765 The "depth" parameter indicates the maximum directory depth
2766 to be used to compute the hash. One level is counted for each
2767 slash in the request. If both parameters are specified, the
2768 evaluation stops when either is reached.
2769
Willy Tarreau0ba27502007-12-24 16:55:16 +01002770 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002771 the query string of each HTTP GET request.
2772
2773 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002774 request entity will be searched for the parameter argument,
2775 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002776 ('?') in the URL. The message body will only start to be
2777 analyzed once either the advertised amount of data has been
2778 received or the request buffer is full. In the unlikely event
2779 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002780 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002781 be randomly balanced if at all. This keyword used to support
2782 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002783
2784 If the parameter is found followed by an equal sign ('=') and
2785 a value, then the value is hashed and divided by the total
2786 weight of the running servers. The result designates which
2787 server will receive the request.
2788
2789 This is used to track user identifiers in requests and ensure
2790 that a same user ID will always be sent to the same server as
2791 long as no server goes up or down. If no value is found or if
2792 the parameter is not found, then a round robin algorithm is
2793 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002794 backend. This algorithm is static by default, which means
2795 that changing a server's weight on the fly will have no
2796 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002797
Cyril Bontédc4d9032012-04-08 21:57:39 +02002798 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2799 request. Just as with the equivalent ACL 'hdr()' function,
2800 the header name in parenthesis is not case sensitive. If the
2801 header is absent or if it does not contain any value, the
2802 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002803
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002804 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002805 reducing the hash algorithm to the main domain part with some
2806 specific headers such as 'Host'. For instance, in the Host
2807 value "haproxy.1wt.eu", only "1wt" will be considered.
2808
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002809 This algorithm is static by default, which means that
2810 changing a server's weight on the fly will have no effect,
2811 but this can be changed using "hash-type".
2812
Willy Tarreau21c741a2019-01-14 18:14:27 +01002813 random
2814 random(<draws>)
2815 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02002816 hashing function. This means that the servers' weights are
2817 respected, dynamic weight changes immediately take effect, as
2818 well as new server additions. Random load balancing can be
2819 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01002820 or removed as it may avoid the hammering effect that could
2821 result from roundrobin or leastconn in this situation. The
2822 hash-balance-factor directive can be used to further improve
2823 fairness of the load balancing, especially in situations
2824 where servers show highly variable response times. When an
2825 argument <draws> is present, it must be an integer value one
2826 or greater, indicating the number of draws before selecting
2827 the least loaded of these servers. It was indeed demonstrated
2828 that picking the least loaded of two servers is enough to
2829 significantly improve the fairness of the algorithm, by
2830 always avoiding to pick the most loaded server within a farm
2831 and getting rid of any bias that could be induced by the
2832 unfair distribution of the consistent list. Higher values N
2833 will take away N-1 of the highest loaded servers at the
2834 expense of performance. With very high values, the algorithm
2835 will converge towards the leastconn's result but much slower.
2836 The default value is 2, which generally shows very good
2837 distribution and performance. This algorithm is also known as
2838 the Power of Two Random Choices and is described here :
2839 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02002840
Emeric Brun736aa232009-06-30 17:56:00 +02002841 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002842 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002843 The RDP cookie <name> (or "mstshash" if omitted) will be
2844 looked up and hashed for each incoming TCP request. Just as
2845 with the equivalent ACL 'req_rdp_cookie()' function, the name
2846 is not case-sensitive. This mechanism is useful as a degraded
2847 persistence mode, as it makes it possible to always send the
2848 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002849 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002850 used instead.
2851
2852 Note that for this to work, the frontend must ensure that an
2853 RDP cookie is already present in the request buffer. For this
2854 you must use 'tcp-request content accept' rule combined with
2855 a 'req_rdp_cookie_cnt' ACL.
2856
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002857 This algorithm is static by default, which means that
2858 changing a server's weight on the fly will have no effect,
2859 but this can be changed using "hash-type".
2860
Cyril Bontédc4d9032012-04-08 21:57:39 +02002861 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002862
Willy Tarreau0ba27502007-12-24 16:55:16 +01002863 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002864 algorithms. Right now, only "url_param" and "uri" support an
2865 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002866
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002867 The load balancing algorithm of a backend is set to roundrobin when no other
2868 algorithm, mode nor option have been set. The algorithm may only be set once
2869 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002870
Lukas Tribus80512b12018-10-27 20:07:40 +02002871 With authentication schemes that require the same connection like NTLM, URI
John Roeslerfb2fce12019-07-10 15:45:51 -05002872 based algorithms must not be used, as they would cause subsequent requests
Lukas Tribus80512b12018-10-27 20:07:40 +02002873 to be routed to different backend servers, breaking the invalid assumptions
2874 NTLM relies on.
2875
Willy Tarreau0ba27502007-12-24 16:55:16 +01002876 Examples :
2877 balance roundrobin
2878 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002879 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002880 balance hdr(User-Agent)
2881 balance hdr(host)
2882 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002883
2884 Note: the following caveats and limitations on using the "check_post"
2885 extension with "url_param" must be considered :
2886
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002887 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002888 to determine if the parameters will be found in the body or entity which
2889 may contain binary data. Therefore another method may be required to
2890 restrict consideration of POST requests that have no URL parameters in
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02002891 the body. (see acl http_end)
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002892
2893 - using a <max_wait> value larger than the request buffer size does not
2894 make sense and is useless. The buffer size is set at build time, and
2895 defaults to 16 kB.
2896
2897 - Content-Encoding is not supported, the parameter search will probably
2898 fail; and load balancing will fall back to Round Robin.
2899
2900 - Expect: 100-continue is not supported, load balancing will fall back to
2901 Round Robin.
2902
Lukas Tribus23953682017-04-28 13:24:30 +00002903 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002904 If the entire parameter value is not present in the first chunk, the
2905 selection of server is undefined (actually, defined by how little
2906 actually appeared in the first chunk).
2907
2908 - This feature does not support generation of a 100, 411 or 501 response.
2909
2910 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002911 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002912 white space or control characters are found, indicating the end of what
2913 might be a URL parameter list. This is probably not a concern with SGML
2914 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002915
Willy Tarreau294d0f02015-08-10 19:40:12 +02002916 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002917
2918
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002919bind [<address>]:<port_range> [, ...] [param*]
2920bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002921 Define one or several listening addresses and/or ports in a frontend.
2922 May be used in sections : defaults | frontend | listen | backend
2923 no | yes | yes | no
2924 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002925 <address> is optional and can be a host name, an IPv4 address, an IPv6
2926 address, or '*'. It designates the address the frontend will
2927 listen on. If unset, all IPv4 addresses of the system will be
2928 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002929 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002930 Optionally, an address family prefix may be used before the
2931 address to force the family regardless of the address format,
2932 which can be useful to specify a path to a unix socket with
2933 no slash ('/'). Currently supported prefixes are :
2934 - 'ipv4@' -> address is always IPv4
2935 - 'ipv6@' -> address is always IPv6
2936 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002937 - 'abns@' -> address is in abstract namespace (Linux only).
2938 Note: since abstract sockets are not "rebindable", they
2939 do not cope well with multi-process mode during
2940 soft-restart, so it is better to avoid them if
2941 nbproc is greater than 1. The effect is that if the
2942 new process fails to start, only one of the old ones
2943 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002944 - 'fd@<n>' -> use file descriptor <n> inherited from the
2945 parent. The fd must be bound and may or may not already
2946 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02002947 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
2948 connected unix socket or of a socketpair. The bind waits
2949 to receive a FD over the unix socket and uses it as if it
2950 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02002951 You may want to reference some environment variables in the
2952 address parameter, see section 2.3 about environment
2953 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002954
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002955 <port_range> is either a unique TCP port, or a port range for which the
2956 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002957 above. The port is mandatory for TCP listeners. Note that in
2958 the case of an IPv6 address, the port is always the number
2959 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002960 - a numerical port (ex: '80')
2961 - a dash-delimited ports range explicitly stating the lower
2962 and upper bounds (ex: '2000-2100') which are included in
2963 the range.
2964
2965 Particular care must be taken against port ranges, because
2966 every <address:port> couple consumes one socket (= a file
2967 descriptor), so it's easy to consume lots of descriptors
2968 with a simple range, and to run out of sockets. Also, each
2969 <address:port> couple must be used only once among all
2970 instances running on a same system. Please note that binding
2971 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002972 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002973 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002974
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002975 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002976 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002977 receive UNIX connections on the socket located at this place.
2978 The path must begin with a slash and by default is absolute.
2979 It can be relative to the prefix defined by "unix-bind" in
2980 the global section. Note that the total length of the prefix
2981 followed by the socket path cannot exceed some system limits
2982 for UNIX sockets, which commonly are set to 107 characters.
2983
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002984 <param*> is a list of parameters common to all sockets declared on the
2985 same line. These numerous parameters depend on OS and build
2986 options and have a complete section dedicated to them. Please
2987 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002988
Willy Tarreau0ba27502007-12-24 16:55:16 +01002989 It is possible to specify a list of address:port combinations delimited by
2990 commas. The frontend will then listen on all of these addresses. There is no
2991 fixed limit to the number of addresses and ports which can be listened on in
2992 a frontend, as well as there is no limit to the number of "bind" statements
2993 in a frontend.
2994
2995 Example :
2996 listen http_proxy
2997 bind :80,:443
2998 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002999 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01003000
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02003001 listen http_https_proxy
3002 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02003003 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02003004
Willy Tarreau24709282013-03-10 21:32:12 +01003005 listen http_https_proxy_explicit
3006 bind ipv6@:80
3007 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
3008 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
3009
Willy Tarreaudad36a32013-03-11 01:20:04 +01003010 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02003011 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01003012
Willy Tarreau55dcaf62015-09-27 15:03:15 +02003013 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
3014 sun_path length is used for the address length. Some other programs
3015 such as socat use the string length only by default. Pass the option
3016 ",unix-tightsocklen=0" to any abstract socket definition in socat to
3017 make it compatible with HAProxy's.
3018
Willy Tarreauceb24bc2010-11-09 12:46:41 +01003019 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02003020 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003021
3022
Christopher Fauletff4121f2017-11-22 16:38:49 +01003023bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003024 Limit visibility of an instance to a certain set of processes numbers.
3025 May be used in sections : defaults | frontend | listen | backend
3026 yes | yes | yes | yes
3027 Arguments :
3028 all All process will see this instance. This is the default. It
3029 may be used to override a default value.
3030
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003031 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003032 option may be combined with other numbers.
3033
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003034 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003035 option may be combined with other numbers. Do not use it
3036 with less than 2 processes otherwise some instances might be
3037 missing from all processes.
3038
Christopher Fauletff4121f2017-11-22 16:38:49 +01003039 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003040 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01003041 the machine's word size. Ranges can be partially defined. The
3042 higher bound can be omitted. In such case, it is replaced by
3043 the corresponding maximum value. If a proxy is bound to
3044 process numbers greater than the configured global.nbproc, it
3045 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02003046 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003047
3048 This keyword limits binding of certain instances to certain processes. This
3049 is useful in order not to have too many processes listening to the same
3050 ports. For instance, on a dual-core machine, it might make sense to set
3051 'nbproc 2' in the global section, then distributes the listeners among 'odd'
3052 and 'even' instances.
3053
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003054 At the moment, it is not possible to reference more than 32 or 64 processes
3055 using this keyword, but this should be more than enough for most setups.
3056 Please note that 'all' really means all processes regardless of the machine's
3057 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003058
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003059 Each "bind" line may further be limited to a subset of the proxy's processes,
3060 please consult the "process" bind keyword in section 5.1.
3061
Willy Tarreaub369a042014-09-16 13:21:03 +02003062 When a frontend has no explicit "bind-process" line, it tries to bind to all
3063 the processes referenced by its "bind" lines. That means that frontends can
3064 easily adapt to their listeners' processes.
3065
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003066 If some backends are referenced by frontends bound to other processes, the
3067 backend automatically inherits the frontend's processes.
3068
3069 Example :
3070 listen app_ip1
3071 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003072 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003073
3074 listen app_ip2
3075 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003076 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003077
3078 listen management
3079 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003080 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003081
Willy Tarreau110ecc12012-11-15 17:50:01 +01003082 listen management
3083 bind 10.0.0.4:80
3084 bind-process 1-4
3085
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003086 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003087
3088
Willy Tarreau0ba27502007-12-24 16:55:16 +01003089capture cookie <name> len <length>
3090 Capture and log a cookie in the request and in the response.
3091 May be used in sections : defaults | frontend | listen | backend
3092 no | yes | yes | no
3093 Arguments :
3094 <name> is the beginning of the name of the cookie to capture. In order
3095 to match the exact name, simply suffix the name with an equal
3096 sign ('='). The full name will appear in the logs, which is
3097 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01003098 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003099
3100 <length> is the maximum number of characters to report in the logs, which
3101 include the cookie name, the equal sign and the value, all in the
3102 standard "name=value" form. The string will be truncated on the
3103 right if it exceeds <length>.
3104
3105 Only the first cookie is captured. Both the "cookie" request headers and the
3106 "set-cookie" response headers are monitored. This is particularly useful to
3107 check for application bugs causing session crossing or stealing between
3108 users, because generally the user's cookies can only change on a login page.
3109
3110 When the cookie was not presented by the client, the associated log column
3111 will report "-". When a request does not cause a cookie to be assigned by the
3112 server, a "-" is reported in the response column.
3113
3114 The capture is performed in the frontend only because it is necessary that
3115 the log format does not change for a given frontend depending on the
3116 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01003117 "capture cookie" statement in a frontend. The maximum capture length is set
3118 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
3119 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003120
3121 Example:
3122 capture cookie ASPSESSION len 32
3123
3124 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003125 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003126
3127
3128capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003129 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003130 May be used in sections : defaults | frontend | listen | backend
3131 no | yes | yes | no
3132 Arguments :
3133 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003134 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003135 appear in the requests, with the first letter of each word in
3136 upper case. The header name will not appear in the logs, only the
3137 value is reported, but the position in the logs is respected.
3138
3139 <length> is the maximum number of characters to extract from the value and
3140 report in the logs. The string will be truncated on the right if
3141 it exceeds <length>.
3142
Willy Tarreau4460d032012-11-21 23:37:37 +01003143 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003144 value will be added to the logs between braces ('{}'). If multiple headers
3145 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003146 in the same order they were declared in the configuration. Non-existent
3147 headers will be logged just as an empty string. Common uses for request
3148 header captures include the "Host" field in virtual hosting environments, the
3149 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003150 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003151 environments to find where the request came from.
3152
3153 Note that when capturing headers such as "User-agent", some spaces may be
3154 logged, making the log analysis more difficult. Thus be careful about what
3155 you log if you know your log parser is not smart enough to rely on the
3156 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003157
Willy Tarreau0900abb2012-11-22 00:21:46 +01003158 There is no limit to the number of captured request headers nor to their
3159 length, though it is wise to keep them low to limit memory usage per session.
3160 In order to keep log format consistent for a same frontend, header captures
3161 can only be declared in a frontend. It is not possible to specify a capture
3162 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003163
3164 Example:
3165 capture request header Host len 15
3166 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01003167 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01003168
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003169 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003170 about logging.
3171
3172
3173capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003174 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003175 May be used in sections : defaults | frontend | listen | backend
3176 no | yes | yes | no
3177 Arguments :
3178 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003179 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003180 appear in the response, with the first letter of each word in
3181 upper case. The header name will not appear in the logs, only the
3182 value is reported, but the position in the logs is respected.
3183
3184 <length> is the maximum number of characters to extract from the value and
3185 report in the logs. The string will be truncated on the right if
3186 it exceeds <length>.
3187
Willy Tarreau4460d032012-11-21 23:37:37 +01003188 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003189 result will be added to the logs between braces ('{}') after the captured
3190 request headers. If multiple headers are captured, they will be delimited by
3191 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003192 the configuration. Non-existent headers will be logged just as an empty
3193 string. Common uses for response header captures include the "Content-length"
3194 header which indicates how many bytes are expected to be returned, the
3195 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003196
Willy Tarreau0900abb2012-11-22 00:21:46 +01003197 There is no limit to the number of captured response headers nor to their
3198 length, though it is wise to keep them low to limit memory usage per session.
3199 In order to keep log format consistent for a same frontend, header captures
3200 can only be declared in a frontend. It is not possible to specify a capture
3201 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003202
3203 Example:
3204 capture response header Content-length len 9
3205 capture response header Location len 15
3206
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003207 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003208 about logging.
3209
3210
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003211compression algo <algorithm> ...
3212compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02003213compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02003214 Enable HTTP compression.
3215 May be used in sections : defaults | frontend | listen | backend
3216 yes | yes | yes | yes
3217 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003218 algo is followed by the list of supported compression algorithms.
3219 type is followed by the list of MIME types that will be compressed.
3220 offload makes haproxy work as a compression offloader only (see notes).
3221
3222 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01003223 identity this is mostly for debugging, and it was useful for developing
3224 the compression feature. Identity does not apply any change on
3225 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003226
Willy Tarreauc91840a2015-03-28 17:00:39 +01003227 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01003228 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01003229
3230 deflate same as "gzip", but with deflate algorithm and zlib format.
3231 Note that this algorithm has ambiguous support on many
3232 browsers and no support at all from recent ones. It is
3233 strongly recommended not to use it for anything else than
3234 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01003235 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003236
Willy Tarreauc91840a2015-03-28 17:00:39 +01003237 raw-deflate same as "deflate" without the zlib wrapper, and used as an
3238 alternative when the browser wants "deflate". All major
3239 browsers understand it and despite violating the standards,
3240 it is known to work better than "deflate", at least on MSIE
3241 and some versions of Safari. Do not use it in conjunction
3242 with "deflate", use either one or the other since both react
3243 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01003244 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003245
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04003246 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003247 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003248 If backend servers support HTTP compression, these directives
3249 will be no-op: haproxy will see the compressed response and will not
3250 compress again. If backend servers do not support HTTP compression and
3251 there is Accept-Encoding header in request, haproxy will compress the
3252 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02003253
3254 The "offload" setting makes haproxy remove the Accept-Encoding header to
3255 prevent backend servers from compressing responses. It is strongly
3256 recommended not to do this because this means that all the compression work
3257 will be done on the single point where haproxy is located. However in some
3258 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003259 with broken HTTP compression implementation which can't be turned off.
3260 In that case haproxy can be used to prevent that gateway from emitting
3261 invalid payloads. In this case, simply removing the header in the
3262 configuration does not work because it applies before the header is parsed,
3263 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02003264 then be used for such scenarios. Note: for now, the "offload" setting is
3265 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02003266
William Lallemand05097442012-11-20 12:14:28 +01003267 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003268 * the request does not advertise a supported compression algorithm in the
3269 "Accept-Encoding" header
3270 * the response message is not HTTP/1.1
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003271 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003272 * response contain neither a "Content-Length" header nor a
3273 "Transfer-Encoding" whose last value is "chunked"
3274 * response contains a "Content-Type" header whose first value starts with
3275 "multipart"
3276 * the response contains the "no-transform" value in the "Cache-control"
3277 header
3278 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
3279 and later
3280 * The response contains a "Content-Encoding" header, indicating that the
3281 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003282 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01003283
Tim Duesterhusb229f012019-01-29 16:38:56 +01003284 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01003285
William Lallemand82fe75c2012-10-23 10:25:10 +02003286 Examples :
3287 compression algo gzip
3288 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01003289
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003290
Willy Tarreau55165fe2009-05-10 12:02:55 +02003291cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003292 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3293 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Olivier Houchard4e694042017-03-14 20:01:29 +01003294 [ dynamic ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003295 Enable cookie-based persistence in a backend.
3296 May be used in sections : defaults | frontend | listen | backend
3297 yes | no | yes | yes
3298 Arguments :
3299 <name> is the name of the cookie which will be monitored, modified or
3300 inserted in order to bring persistence. This cookie is sent to
3301 the client via a "Set-Cookie" header in the response, and is
3302 brought back by the client in a "Cookie" header in all requests.
3303 Special care should be taken to choose a name which does not
3304 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003305 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003306 HTTP/HTTPS), care should be taken to use different cookie names
3307 between all backends if persistence between them is not desired.
3308
3309 rewrite This keyword indicates that the cookie will be provided by the
3310 server and that haproxy will have to modify its value to set the
3311 server's identifier in it. This mode is handy when the management
3312 of complex combinations of "Set-cookie" and "Cache-control"
3313 headers is left to the application. The application can then
3314 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003315 cookie. Since all responses should be monitored, this mode
3316 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003317 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003318 start with this mode for new deployments. This keyword is
3319 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003320
3321 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003322 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003323
Willy Tarreaua79094d2010-08-31 22:54:15 +02003324 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003325 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02003326 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003327 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003328 existing configurations running in the "rewrite" mode. The cookie
3329 will only be a session cookie and will not be stored on the
3330 client's disk. By default, unless the "indirect" option is added,
3331 the server will see the cookies emitted by the client. Due to
3332 caching effects, it is generally wise to add the "nocache" or
3333 "postonly" keywords (see below). The "insert" keyword is not
3334 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003335
3336 prefix This keyword indicates that instead of relying on a dedicated
3337 cookie for the persistence, an existing one will be completed.
3338 This may be needed in some specific environments where the client
3339 does not support more than one single cookie and the application
3340 already needs it. In this case, whenever the server sets a cookie
3341 named <name>, it will be prefixed with the server's identifier
3342 and a delimiter. The prefix will be removed from all client
3343 requests so that the server still finds the cookie it emitted.
3344 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003345 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003346 not compatible with "rewrite" and "insert". Note: it is highly
3347 recommended not to use "indirect" with "prefix", otherwise server
3348 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003349
Willy Tarreaua79094d2010-08-31 22:54:15 +02003350 indirect When this option is specified, no cookie will be emitted to a
3351 client which already has a valid one for the server which has
3352 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003353 it will be removed, unless the "preserve" option is also set. In
3354 "insert" mode, this will additionally remove cookies from the
3355 requests transmitted to the server, making the persistence
3356 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003357 Note: it is highly recommended not to use "indirect" with
3358 "prefix", otherwise server cookie updates would not be sent to
3359 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003360
3361 nocache This option is recommended in conjunction with the insert mode
3362 when there is a cache between the client and HAProxy, as it
3363 ensures that a cacheable response will be tagged non-cacheable if
3364 a cookie needs to be inserted. This is important because if all
3365 persistence cookies are added on a cacheable home page for
3366 instance, then all customers will then fetch the page from an
3367 outer cache and will all share the same persistence cookie,
3368 leading to one server receiving much more traffic than others.
3369 See also the "insert" and "postonly" options.
3370
3371 postonly This option ensures that cookie insertion will only be performed
3372 on responses to POST requests. It is an alternative to the
3373 "nocache" option, because POST responses are not cacheable, so
3374 this ensures that the persistence cookie will never get cached.
3375 Since most sites do not need any sort of persistence before the
3376 first POST which generally is a login request, this is a very
3377 efficient method to optimize caching without risking to find a
3378 persistence cookie in the cache.
3379 See also the "insert" and "nocache" options.
3380
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003381 preserve This option may only be used with "insert" and/or "indirect". It
3382 allows the server to emit the persistence cookie itself. In this
3383 case, if a cookie is found in the response, haproxy will leave it
3384 untouched. This is useful in order to end persistence after a
3385 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003386 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003387 the past. By combining this mechanism with the "disable-on-404"
3388 check option, it is possible to perform a completely graceful
3389 shutdown because users will definitely leave the server after
3390 they logout.
3391
Willy Tarreau4992dd22012-05-31 21:02:17 +02003392 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3393 when a cookie is inserted. This attribute is used so that a
3394 user agent doesn't share the cookie with non-HTTP components.
3395 Please check RFC6265 for more information on this attribute.
3396
3397 secure This option tells haproxy to add a "Secure" cookie attribute when
3398 a cookie is inserted. This attribute is used so that a user agent
3399 never emits this cookie over non-secure channels, which means
3400 that a cookie learned with this flag will be presented only over
3401 SSL/TLS connections. Please check RFC6265 for more information on
3402 this attribute.
3403
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003404 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003405 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003406 name. If the domain begins with a dot, the browser is allowed to
3407 use it for any host ending with that name. It is also possible to
3408 specify several domain names by invoking this option multiple
3409 times. Some browsers might have small limits on the number of
3410 domains, so be careful when doing that. For the record, sending
3411 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003412
Willy Tarreau996a92c2010-10-13 19:30:47 +02003413 maxidle This option allows inserted cookies to be ignored after some idle
3414 time. It only works with insert-mode cookies. When a cookie is
3415 sent to the client, the date this cookie was emitted is sent too.
3416 Upon further presentations of this cookie, if the date is older
3417 than the delay indicated by the parameter (in seconds), it will
3418 be ignored. Otherwise, it will be refreshed if needed when the
3419 response is sent to the client. This is particularly useful to
3420 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003421 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003422 this option is set and a cookie has no date, it is always
3423 accepted, but gets refreshed in the response. This maintains the
3424 ability for admins to access their sites. Cookies that have a
3425 date in the future further than 24 hours are ignored. Doing so
3426 lets admins fix timezone issues without risking kicking users off
3427 the site.
3428
3429 maxlife This option allows inserted cookies to be ignored after some life
3430 time, whether they're in use or not. It only works with insert
3431 mode cookies. When a cookie is first sent to the client, the date
3432 this cookie was emitted is sent too. Upon further presentations
3433 of this cookie, if the date is older than the delay indicated by
3434 the parameter (in seconds), it will be ignored. If the cookie in
3435 the request has no date, it is accepted and a date will be set.
3436 Cookies that have a date in the future further than 24 hours are
3437 ignored. Doing so lets admins fix timezone issues without risking
3438 kicking users off the site. Contrary to maxidle, this value is
3439 not refreshed, only the first visit date counts. Both maxidle and
3440 maxlife may be used at the time. This is particularly useful to
3441 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003442 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003443 is stronger than the maxidle method in that it forces a
3444 redispatch after some absolute delay.
3445
Olivier Houchard4e694042017-03-14 20:01:29 +01003446 dynamic Activate dynamic cookies. When used, a session cookie is
3447 dynamically created for each server, based on the IP and port
3448 of the server, and a secret key, specified in the
3449 "dynamic-cookie-key" backend directive.
3450 The cookie will be regenerated each time the IP address change,
3451 and is only generated for IPv4/IPv6.
3452
Willy Tarreau0ba27502007-12-24 16:55:16 +01003453 There can be only one persistence cookie per HTTP backend, and it can be
3454 declared in a defaults section. The value of the cookie will be the value
3455 indicated after the "cookie" keyword in a "server" statement. If no cookie
3456 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003457
Willy Tarreau0ba27502007-12-24 16:55:16 +01003458 Examples :
3459 cookie JSESSIONID prefix
3460 cookie SRV insert indirect nocache
3461 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003462 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003463
Willy Tarreau294d0f02015-08-10 19:40:12 +02003464 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003465
Willy Tarreau983e01e2010-01-11 18:42:06 +01003466
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003467declare capture [ request | response ] len <length>
3468 Declares a capture slot.
3469 May be used in sections : defaults | frontend | listen | backend
3470 no | yes | yes | no
3471 Arguments:
3472 <length> is the length allowed for the capture.
3473
3474 This declaration is only available in the frontend or listen section, but the
3475 reserved slot can be used in the backends. The "request" keyword allocates a
3476 capture slot for use in the request, and "response" allocates a capture slot
3477 for use in the response.
3478
3479 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003480 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003481 "http-request capture" and "http-response capture".
3482
3483
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003484default-server [param*]
3485 Change default options for a server in a backend
3486 May be used in sections : defaults | frontend | listen | backend
3487 yes | no | yes | yes
3488 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003489 <param*> is a list of parameters for this server. The "default-server"
3490 keyword accepts an important number of options and has a complete
3491 section dedicated to it. Please refer to section 5 for more
3492 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003493
Willy Tarreau983e01e2010-01-11 18:42:06 +01003494 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003495 default-server inter 1000 weight 13
3496
3497 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003498
Willy Tarreau983e01e2010-01-11 18:42:06 +01003499
Willy Tarreau0ba27502007-12-24 16:55:16 +01003500default_backend <backend>
3501 Specify the backend to use when no "use_backend" rule has been matched.
3502 May be used in sections : defaults | frontend | listen | backend
3503 yes | yes | yes | no
3504 Arguments :
3505 <backend> is the name of the backend to use.
3506
3507 When doing content-switching between frontend and backends using the
3508 "use_backend" keyword, it is often useful to indicate which backend will be
3509 used when no rule has matched. It generally is the dynamic backend which
3510 will catch all undetermined requests.
3511
Willy Tarreau0ba27502007-12-24 16:55:16 +01003512 Example :
3513
3514 use_backend dynamic if url_dyn
3515 use_backend static if url_css url_img extension_img
3516 default_backend dynamic
3517
Willy Tarreau98d04852015-05-26 12:18:29 +02003518 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003519
Willy Tarreau0ba27502007-12-24 16:55:16 +01003520
Baptiste Assmann27f51342013-10-09 06:51:49 +02003521description <string>
3522 Describe a listen, frontend or backend.
3523 May be used in sections : defaults | frontend | listen | backend
3524 no | yes | yes | yes
3525 Arguments : string
3526
3527 Allows to add a sentence to describe the related object in the HAProxy HTML
3528 stats page. The description will be printed on the right of the object name
3529 it describes.
3530 No need to backslash spaces in the <string> arguments.
3531
3532
Willy Tarreau0ba27502007-12-24 16:55:16 +01003533disabled
3534 Disable a proxy, frontend or backend.
3535 May be used in sections : defaults | frontend | listen | backend
3536 yes | yes | yes | yes
3537 Arguments : none
3538
3539 The "disabled" keyword is used to disable an instance, mainly in order to
3540 liberate a listening port or to temporarily disable a service. The instance
3541 will still be created and its configuration will be checked, but it will be
3542 created in the "stopped" state and will appear as such in the statistics. It
3543 will not receive any traffic nor will it send any health-checks or logs. It
3544 is possible to disable many instances at once by adding the "disabled"
3545 keyword in a "defaults" section.
3546
3547 See also : "enabled"
3548
3549
Willy Tarreau5ce94572010-06-07 14:35:41 +02003550dispatch <address>:<port>
3551 Set a default server address
3552 May be used in sections : defaults | frontend | listen | backend
3553 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003554 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003555
3556 <address> is the IPv4 address of the default server. Alternatively, a
3557 resolvable hostname is supported, but this name will be resolved
3558 during start-up.
3559
3560 <ports> is a mandatory port specification. All connections will be sent
3561 to this port, and it is not permitted to use port offsets as is
3562 possible with normal servers.
3563
Willy Tarreau787aed52011-04-15 06:45:37 +02003564 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003565 server can take the connection. In the past it was used to forward non
3566 persistent connections to an auxiliary load balancer. Due to its simple
3567 syntax, it has also been used for simple TCP relays. It is recommended not to
3568 use it for more clarity, and to use the "server" directive instead.
3569
3570 See also : "server"
3571
Olivier Houchard4e694042017-03-14 20:01:29 +01003572
3573dynamic-cookie-key <string>
3574 Set the dynamic cookie secret key for a backend.
3575 May be used in sections : defaults | frontend | listen | backend
3576 yes | no | yes | yes
3577 Arguments : The secret key to be used.
3578
3579 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003580 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003581 specified on the "server" line), using a hash of the IP address of the
3582 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003583 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003584 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003585
Willy Tarreau0ba27502007-12-24 16:55:16 +01003586enabled
3587 Enable a proxy, frontend or backend.
3588 May be used in sections : defaults | frontend | listen | backend
3589 yes | yes | yes | yes
3590 Arguments : none
3591
3592 The "enabled" keyword is used to explicitly enable an instance, when the
3593 defaults has been set to "disabled". This is very rarely used.
3594
3595 See also : "disabled"
3596
3597
3598errorfile <code> <file>
3599 Return a file contents instead of errors generated by HAProxy
3600 May be used in sections : defaults | frontend | listen | backend
3601 yes | yes | yes | yes
3602 Arguments :
3603 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003604 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3605 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003606
3607 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003608 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003609 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003610 error pages, and to use absolute paths, since files are read
3611 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003612
3613 It is important to understand that this keyword is not meant to rewrite
3614 errors returned by the server, but errors detected and returned by HAProxy.
3615 This is why the list of supported errors is limited to a small set.
3616
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003617 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3618
Willy Tarreau0ba27502007-12-24 16:55:16 +01003619 The files are returned verbatim on the TCP socket. This allows any trick such
3620 as redirections to another URL or site, as well as tricks to clean cookies,
3621 force enable or disable caching, etc... The package provides default error
3622 files returning the same contents as default errors.
3623
Willy Tarreau59140a22009-02-22 12:02:30 +01003624 The files should not exceed the configured buffer size (BUFSIZE), which
3625 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003626 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003627 loops between the client and HAProxy when all servers are down, causing an
3628 error to be returned instead of an image. For better HTTP compliance, it is
3629 recommended that all header lines end with CR-LF and not LF alone.
3630
Willy Tarreau0ba27502007-12-24 16:55:16 +01003631 The files are read at the same time as the configuration and kept in memory.
3632 For this reason, the errors continue to be returned even when the process is
3633 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003634 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003635 403 status code and interrogating a blocked URL.
3636
3637 See also : "errorloc", "errorloc302", "errorloc303"
3638
Willy Tarreau59140a22009-02-22 12:02:30 +01003639 Example :
3640 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003641 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003642 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3643 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3644
Willy Tarreau2769aa02007-12-27 18:26:09 +01003645
3646errorloc <code> <url>
3647errorloc302 <code> <url>
3648 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3649 May be used in sections : defaults | frontend | listen | backend
3650 yes | yes | yes | yes
3651 Arguments :
3652 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003653 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3654 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003655
3656 <url> it is the exact contents of the "Location" header. It may contain
3657 either a relative URI to an error page hosted on the same site,
3658 or an absolute URI designating an error page on another site.
3659 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003660 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003661
3662 It is important to understand that this keyword is not meant to rewrite
3663 errors returned by the server, but errors detected and returned by HAProxy.
3664 This is why the list of supported errors is limited to a small set.
3665
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003666 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3667
Willy Tarreau2769aa02007-12-27 18:26:09 +01003668 Note that both keyword return the HTTP 302 status code, which tells the
3669 client to fetch the designated URL using the same HTTP method. This can be
3670 quite problematic in case of non-GET methods such as POST, because the URL
3671 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003672 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003673 status code, indicating to the client that the URL must be fetched with a GET
3674 request.
3675
3676 See also : "errorfile", "errorloc303"
3677
3678
3679errorloc303 <code> <url>
3680 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3681 May be used in sections : defaults | frontend | listen | backend
3682 yes | yes | yes | yes
3683 Arguments :
3684 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003685 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3686 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003687
3688 <url> it is the exact contents of the "Location" header. It may contain
3689 either a relative URI to an error page hosted on the same site,
3690 or an absolute URI designating an error page on another site.
3691 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003692 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003693
3694 It is important to understand that this keyword is not meant to rewrite
3695 errors returned by the server, but errors detected and returned by HAProxy.
3696 This is why the list of supported errors is limited to a small set.
3697
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003698 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3699
Willy Tarreau2769aa02007-12-27 18:26:09 +01003700 Note that both keyword return the HTTP 303 status code, which tells the
3701 client to fetch the designated URL using the same HTTP GET method. This
3702 solves the usual problems associated with "errorloc" and the 302 code. It is
3703 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003704 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003705
3706 See also : "errorfile", "errorloc", "errorloc302"
3707
3708
Simon Horman51a1cf62015-02-03 13:00:44 +09003709email-alert from <emailaddr>
3710 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003711 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003712 May be used in sections: defaults | frontend | listen | backend
3713 yes | yes | yes | yes
3714
3715 Arguments :
3716
3717 <emailaddr> is the from email address to use when sending email alerts
3718
3719 Also requires "email-alert mailers" and "email-alert to" to be set
3720 and if so sending email alerts is enabled for the proxy.
3721
Simon Horman64e34162015-02-06 11:11:57 +09003722 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003723 "email-alert myhostname", "email-alert to", section 3.6 about
3724 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003725
3726
3727email-alert level <level>
3728 Declare the maximum log level of messages for which email alerts will be
3729 sent. This acts as a filter on the sending of email alerts.
3730 May be used in sections: defaults | frontend | listen | backend
3731 yes | yes | yes | yes
3732
3733 Arguments :
3734
3735 <level> One of the 8 syslog levels:
3736 emerg alert crit err warning notice info debug
3737 The above syslog levels are ordered from lowest to highest.
3738
3739 By default level is alert
3740
3741 Also requires "email-alert from", "email-alert mailers" and
3742 "email-alert to" to be set and if so sending email alerts is enabled
3743 for the proxy.
3744
Simon Horman1421e212015-04-30 13:10:35 +09003745 Alerts are sent when :
3746
3747 * An un-paused server is marked as down and <level> is alert or lower
3748 * A paused server is marked as down and <level> is notice or lower
3749 * A server is marked as up or enters the drain state and <level>
3750 is notice or lower
3751 * "option log-health-checks" is enabled, <level> is info or lower,
3752 and a health check status update occurs
3753
Simon Horman64e34162015-02-06 11:11:57 +09003754 See also : "email-alert from", "email-alert mailers",
3755 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003756 section 3.6 about mailers.
3757
3758
3759email-alert mailers <mailersect>
3760 Declare the mailers to be used when sending email alerts
3761 May be used in sections: defaults | frontend | listen | backend
3762 yes | yes | yes | yes
3763
3764 Arguments :
3765
3766 <mailersect> is the name of the mailers section to send email alerts.
3767
3768 Also requires "email-alert from" and "email-alert to" to be set
3769 and if so sending email alerts is enabled for the proxy.
3770
Simon Horman64e34162015-02-06 11:11:57 +09003771 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3772 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003773
3774
3775email-alert myhostname <hostname>
3776 Declare the to hostname address to be used when communicating with
3777 mailers.
3778 May be used in sections: defaults | frontend | listen | backend
3779 yes | yes | yes | yes
3780
3781 Arguments :
3782
Baptiste Assmann738bad92015-12-21 15:27:53 +01003783 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003784
3785 By default the systems hostname is used.
3786
3787 Also requires "email-alert from", "email-alert mailers" and
3788 "email-alert to" to be set and if so sending email alerts is enabled
3789 for the proxy.
3790
Simon Horman64e34162015-02-06 11:11:57 +09003791 See also : "email-alert from", "email-alert level", "email-alert mailers",
3792 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003793
3794
3795email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003796 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003797 header of email alerts. This is the address that email alerts are sent to.
3798 May be used in sections: defaults | frontend | listen | backend
3799 yes | yes | yes | yes
3800
3801 Arguments :
3802
3803 <emailaddr> is the to email address to use when sending email alerts
3804
3805 Also requires "email-alert mailers" and "email-alert to" to be set
3806 and if so sending email alerts is enabled for the proxy.
3807
Simon Horman64e34162015-02-06 11:11:57 +09003808 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003809 "email-alert myhostname", section 3.6 about mailers.
3810
3811
Willy Tarreau4de91492010-01-22 19:10:05 +01003812force-persist { if | unless } <condition>
3813 Declare a condition to force persistence on down servers
3814 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003815 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003816
3817 By default, requests are not dispatched to down servers. It is possible to
3818 force this using "option persist", but it is unconditional and redispatches
3819 to a valid server if "option redispatch" is set. That leaves with very little
3820 possibilities to force some requests to reach a server which is artificially
3821 marked down for maintenance operations.
3822
3823 The "force-persist" statement allows one to declare various ACL-based
3824 conditions which, when met, will cause a request to ignore the down status of
3825 a server and still try to connect to it. That makes it possible to start a
3826 server, still replying an error to the health checks, and run a specially
3827 configured browser to test the service. Among the handy methods, one could
3828 use a specific source IP address, or a specific cookie. The cookie also has
3829 the advantage that it can easily be added/removed on the browser from a test
3830 page. Once the service is validated, it is then possible to open the service
3831 to the world by returning a valid response to health checks.
3832
3833 The forced persistence is enabled when an "if" condition is met, or unless an
3834 "unless" condition is met. The final redispatch is always disabled when this
3835 is used.
3836
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003837 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003838 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003839
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003840
3841filter <name> [param*]
3842 Add the filter <name> in the filter list attached to the proxy.
3843 May be used in sections : defaults | frontend | listen | backend
3844 no | yes | yes | yes
3845 Arguments :
3846 <name> is the name of the filter. Officially supported filters are
3847 referenced in section 9.
3848
Tim Düsterhus4896c442016-11-29 02:15:19 +01003849 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003850 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01003851 filter. Please refer to the documentation of the corresponding
3852 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003853
3854 Multiple occurrences of the filter line can be used for the same proxy. The
3855 same filter can be referenced many times if needed.
3856
3857 Example:
3858 listen
3859 bind *:80
3860
3861 filter trace name BEFORE-HTTP-COMP
3862 filter compression
3863 filter trace name AFTER-HTTP-COMP
3864
3865 compression algo gzip
3866 compression offload
3867
3868 server srv1 192.168.0.1:80
3869
3870 See also : section 9.
3871
Willy Tarreau4de91492010-01-22 19:10:05 +01003872
Willy Tarreau2769aa02007-12-27 18:26:09 +01003873fullconn <conns>
3874 Specify at what backend load the servers will reach their maxconn
3875 May be used in sections : defaults | frontend | listen | backend
3876 yes | no | yes | yes
3877 Arguments :
3878 <conns> is the number of connections on the backend which will make the
3879 servers use the maximal number of connections.
3880
Willy Tarreau198a7442008-01-17 12:05:32 +01003881 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003882 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003883 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003884 load. The server will then always accept at least <minconn> connections,
3885 never more than <maxconn>, and the limit will be on the ramp between both
3886 values when the backend has less than <conns> concurrent connections. This
3887 makes it possible to limit the load on the servers during normal loads, but
3888 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003889 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003890
Willy Tarreaufbb78422011-06-05 15:38:35 +02003891 Since it's hard to get this value right, haproxy automatically sets it to
3892 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003893 backend (based on "use_backend" and "default_backend" rules). That way it's
3894 safe to leave it unset. However, "use_backend" involving dynamic names are
3895 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003896
Willy Tarreau2769aa02007-12-27 18:26:09 +01003897 Example :
3898 # The servers will accept between 100 and 1000 concurrent connections each
3899 # and the maximum of 1000 will be reached when the backend reaches 10000
3900 # connections.
3901 backend dynamic
3902 fullconn 10000
3903 server srv1 dyn1:80 minconn 100 maxconn 1000
3904 server srv2 dyn2:80 minconn 100 maxconn 1000
3905
3906 See also : "maxconn", "server"
3907
3908
3909grace <time>
3910 Maintain a proxy operational for some time after a soft stop
3911 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003912 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003913 Arguments :
3914 <time> is the time (by default in milliseconds) for which the instance
3915 will remain operational with the frontend sockets still listening
3916 when a soft-stop is received via the SIGUSR1 signal.
3917
3918 This may be used to ensure that the services disappear in a certain order.
3919 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003920 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003921 needed by the equipment to detect the failure.
3922
3923 Note that currently, there is very little benefit in using this parameter,
3924 and it may in fact complicate the soft-reconfiguration process more than
3925 simplify it.
3926
Willy Tarreau0ba27502007-12-24 16:55:16 +01003927
Andrew Rodland17be45e2016-10-25 17:04:12 -04003928hash-balance-factor <factor>
3929 Specify the balancing factor for bounded-load consistent hashing
3930 May be used in sections : defaults | frontend | listen | backend
3931 yes | no | no | yes
3932 Arguments :
3933 <factor> is the control for the maximum number of concurrent requests to
3934 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01003935 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04003936
3937 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
3938 enables an algorithm that prevents any one server from getting too many
3939 requests at once, even if some hash buckets receive many more requests than
3940 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
3941 <factor> is a percentage greater than 100. For example, if <factor> is 150,
3942 then no server will be allowed to have a load more than 1.5 times the average.
3943 If server weights are used, they will be respected.
3944
3945 If the first-choice server is disqualified, the algorithm will choose another
3946 server based on the request hash, until a server with additional capacity is
3947 found. A higher <factor> allows more imbalance between the servers, while a
3948 lower <factor> means that more servers will be checked on average, affecting
3949 performance. Reasonable values are from 125 to 200.
3950
Willy Tarreau760e81d2018-05-03 07:20:40 +02003951 This setting is also used by "balance random" which internally relies on the
3952 consistent hashing mechanism.
3953
Andrew Rodland17be45e2016-10-25 17:04:12 -04003954 See also : "balance" and "hash-type".
3955
3956
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003957hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003958 Specify a method to use for mapping hashes to servers
3959 May be used in sections : defaults | frontend | listen | backend
3960 yes | no | yes | yes
3961 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003962 <method> is the method used to select a server from the hash computed by
3963 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003964
Bhaskar98634f02013-10-29 23:30:51 -04003965 map-based the hash table is a static array containing all alive servers.
3966 The hashes will be very smooth, will consider weights, but
3967 will be static in that weight changes while a server is up
3968 will be ignored. This means that there will be no slow start.
3969 Also, since a server is selected by its position in the array,
3970 most mappings are changed when the server count changes. This
3971 means that when a server goes up or down, or when a server is
3972 added to a farm, most connections will be redistributed to
3973 different servers. This can be inconvenient with caches for
3974 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003975
Bhaskar98634f02013-10-29 23:30:51 -04003976 consistent the hash table is a tree filled with many occurrences of each
3977 server. The hash key is looked up in the tree and the closest
3978 server is chosen. This hash is dynamic, it supports changing
3979 weights while the servers are up, so it is compatible with the
3980 slow start feature. It has the advantage that when a server
3981 goes up or down, only its associations are moved. When a
3982 server is added to the farm, only a few part of the mappings
3983 are redistributed, making it an ideal method for caches.
3984 However, due to its principle, the distribution will never be
3985 very smooth and it may sometimes be necessary to adjust a
3986 server's weight or its ID to get a more balanced distribution.
3987 In order to get the same distribution on multiple load
3988 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003989 same IDs. Note: consistent hash uses sdbm and avalanche if no
3990 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003991
3992 <function> is the hash function to be used :
3993
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003994 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003995 reimplementation of ndbm) database library. It was found to do
3996 well in scrambling bits, causing better distribution of the keys
3997 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003998 function with good distribution, unless the total server weight
3999 is a multiple of 64, in which case applying the avalanche
4000 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04004001
4002 djb2 this function was first proposed by Dan Bernstein many years ago
4003 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004004 function provides a better distribution than sdbm. It generally
4005 works well with text-based inputs though it can perform extremely
4006 poorly with numeric-only input or when the total server weight is
4007 a multiple of 33, unless the avalanche modifier is also used.
4008
Willy Tarreaua0f42712013-11-14 14:30:35 +01004009 wt6 this function was designed for haproxy while testing other
4010 functions in the past. It is not as smooth as the other ones, but
4011 is much less sensible to the input data set or to the number of
4012 servers. It can make sense as an alternative to sdbm+avalanche or
4013 djb2+avalanche for consistent hashing or when hashing on numeric
4014 data such as a source IP address or a visitor identifier in a URL
4015 parameter.
4016
Willy Tarreau324f07f2015-01-20 19:44:50 +01004017 crc32 this is the most common CRC32 implementation as used in Ethernet,
4018 gzip, PNG, etc. It is slower than the other ones but may provide
4019 a better distribution or less predictable results especially when
4020 used on strings.
4021
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004022 <modifier> indicates an optional method applied after hashing the key :
4023
4024 avalanche This directive indicates that the result from the hash
4025 function above should not be used in its raw form but that
4026 a 4-byte full avalanche hash must be applied first. The
4027 purpose of this step is to mix the resulting bits from the
4028 previous hash in order to avoid any undesired effect when
4029 the input contains some limited values or when the number of
4030 servers is a multiple of one of the hash's components (64
4031 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
4032 result less predictable, but it's also not as smooth as when
4033 using the original function. Some testing might be needed
4034 with some workloads. This hash is one of the many proposed
4035 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004036
Bhaskar98634f02013-10-29 23:30:51 -04004037 The default hash type is "map-based" and is recommended for most usages. The
4038 default function is "sdbm", the selection of a function should be based on
4039 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004040
Andrew Rodland17be45e2016-10-25 17:04:12 -04004041 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004042
4043
Willy Tarreau0ba27502007-12-24 16:55:16 +01004044http-check disable-on-404
4045 Enable a maintenance mode upon HTTP/404 response to health-checks
4046 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01004047 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01004048 Arguments : none
4049
4050 When this option is set, a server which returns an HTTP code 404 will be
4051 excluded from further load-balancing, but will still receive persistent
4052 connections. This provides a very convenient method for Web administrators
4053 to perform a graceful shutdown of their servers. It is also important to note
4054 that a server which is detected as failed while it was in this mode will not
4055 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
4056 will immediately be reinserted into the farm. The status on the stats page
4057 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01004058 option only works in conjunction with the "httpchk" option. If this option
4059 is used with "http-check expect", then it has precedence over it so that 404
4060 responses will still be considered as soft-stop.
4061
4062 See also : "option httpchk", "http-check expect"
4063
4064
4065http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004066 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01004067 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02004068 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01004069 Arguments :
4070 <match> is a keyword indicating how to look for a specific pattern in the
4071 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004072 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01004073 exclamation mark ("!") to negate the match. Spaces are allowed
4074 between the exclamation mark and the keyword. See below for more
4075 details on the supported keywords.
4076
4077 <pattern> is the pattern to look for. It may be a string or a regular
4078 expression. If the pattern contains spaces, they must be escaped
4079 with the usual backslash ('\').
4080
4081 By default, "option httpchk" considers that response statuses 2xx and 3xx
4082 are valid, and that others are invalid. When "http-check expect" is used,
4083 it defines what is considered valid or invalid. Only one "http-check"
4084 statement is supported in a backend. If a server fails to respond or times
4085 out, the check obviously fails. The available matches are :
4086
4087 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004088 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004089 response's status code is exactly this string. If the
4090 "status" keyword is prefixed with "!", then the response
4091 will be considered invalid if the status code matches.
4092
4093 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004094 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004095 response's status code matches the expression. If the
4096 "rstatus" keyword is prefixed with "!", then the response
4097 will be considered invalid if the status code matches.
4098 This is mostly used to check for multiple codes.
4099
4100 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004101 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004102 response's body contains this exact string. If the
4103 "string" keyword is prefixed with "!", then the response
4104 will be considered invalid if the body contains this
4105 string. This can be used to look for a mandatory word at
4106 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004107 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01004108 trace).
4109
4110 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004111 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004112 response's body matches this expression. If the "rstring"
4113 keyword is prefixed with "!", then the response will be
4114 considered invalid if the body matches the expression.
4115 This can be used to look for a mandatory word at the end
4116 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01004117 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01004118
4119 It is important to note that the responses will be limited to a certain size
4120 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
4121 Thus, too large responses may not contain the mandatory pattern when using
4122 "string" or "rstring". If a large response is absolutely required, it is
4123 possible to change the default max size by setting the global variable.
4124 However, it is worth keeping in mind that parsing very large responses can
4125 waste some CPU cycles, especially when regular expressions are used, and that
4126 it is always better to focus the checks on smaller resources.
4127
Cyril Bonté32602d22015-01-30 00:07:07 +01004128 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
4129 will automatically append a "Connection: close" header, meaning that this
4130 header should not be present in the request provided by "option httpchk".
4131
Willy Tarreaubd741542010-03-16 18:46:54 +01004132 Last, if "http-check expect" is combined with "http-check disable-on-404",
4133 then this last one has precedence when the server responds with 404.
4134
4135 Examples :
4136 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004137 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01004138
4139 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004140 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01004141
4142 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004143 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01004144
4145 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004146 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01004147
Willy Tarreaubd741542010-03-16 18:46:54 +01004148 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004149
4150
Willy Tarreauef781042010-01-27 11:53:01 +01004151http-check send-state
4152 Enable emission of a state header with HTTP health checks
4153 May be used in sections : defaults | frontend | listen | backend
4154 yes | no | yes | yes
4155 Arguments : none
4156
4157 When this option is set, haproxy will systematically send a special header
4158 "X-Haproxy-Server-State" with a list of parameters indicating to each server
4159 how they are seen by haproxy. This can be used for instance when a server is
4160 manipulated without access to haproxy and the operator needs to know whether
4161 haproxy still sees it up or not, or if the server is the last one in a farm.
4162
4163 The header is composed of fields delimited by semi-colons, the first of which
4164 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
4165 checks on the total number before transition, just as appears in the stats
4166 interface. Next headers are in the form "<variable>=<value>", indicating in
4167 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08004168 - a variable "address", containing the address of the backend server.
4169 This corresponds to the <address> field in the server declaration. For
4170 unix domain sockets, it will read "unix".
4171
4172 - a variable "port", containing the port of the backend server. This
4173 corresponds to the <port> field in the server declaration. For unix
4174 domain sockets, it will read "unix".
4175
Willy Tarreauef781042010-01-27 11:53:01 +01004176 - a variable "name", containing the name of the backend followed by a slash
4177 ("/") then the name of the server. This can be used when a server is
4178 checked in multiple backends.
4179
4180 - a variable "node" containing the name of the haproxy node, as set in the
4181 global "node" variable, otherwise the system's hostname if unspecified.
4182
4183 - a variable "weight" indicating the weight of the server, a slash ("/")
4184 and the total weight of the farm (just counting usable servers). This
4185 helps to know if other servers are available to handle the load when this
4186 one fails.
4187
4188 - a variable "scur" indicating the current number of concurrent connections
4189 on the server, followed by a slash ("/") then the total number of
4190 connections on all servers of the same backend.
4191
4192 - a variable "qcur" indicating the current number of requests in the
4193 server's queue.
4194
4195 Example of a header received by the application server :
4196 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
4197 scur=13/22; qcur=0
4198
4199 See also : "option httpchk", "http-check disable-on-404"
4200
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004201
4202http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004203 Access control for Layer 7 requests
4204
4205 May be used in sections: defaults | frontend | listen | backend
4206 no | yes | yes | yes
4207
Willy Tarreau20b0de52012-12-24 15:45:22 +01004208 The http-request statement defines a set of rules which apply to layer 7
4209 processing. The rules are evaluated in their declaration order when they are
4210 met in a frontend, listen or backend section. Any rule may optionally be
4211 followed by an ACL-based condition, in which case it will only be evaluated
4212 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004213
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004214 The first keyword is the rule's action. The supported actions are described
4215 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004216
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004217 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004218
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004219 Example:
4220 acl nagios src 192.168.129.3
4221 acl local_net src 192.168.0.0/16
4222 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01004223
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004224 http-request allow if nagios
4225 http-request allow if local_net auth_ok
4226 http-request auth realm Gimme if local_net auth_ok
4227 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01004228
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004229 Example:
4230 acl key req.hdr(X-Add-Acl-Key) -m found
4231 acl add path /addacl
4232 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01004233
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004234 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01004235
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004236 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4237 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004238
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004239 Example:
4240 acl value req.hdr(X-Value) -m found
4241 acl setmap path /setmap
4242 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004243
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004244 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004245
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004246 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4247 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004248
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004249 See also : "stats http-request", section 3.4 about userlists and section 7
4250 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004251
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004252http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004253
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004254 This is used to add a new entry into an ACL. The ACL must be loaded from a
4255 file (even a dummy empty file). The file name of the ACL to be updated is
4256 passed between parentheses. It takes one argument: <key fmt>, which follows
4257 log-format rules, to collect content of the new entry. It performs a lookup
4258 in the ACL before insertion, to avoid duplicated (or more) values. This
4259 lookup is done by a linear search and can be expensive with large lists!
4260 It is the equivalent of the "add acl" command from the stats socket, but can
4261 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06004262
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004263http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004264
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004265 This appends an HTTP header field whose name is specified in <name> and
4266 whose value is defined by <fmt> which follows the log-format rules (see
4267 Custom Log Format in section 8.2.4). This is particularly useful to pass
4268 connection-specific information to the server (e.g. the client's SSL
4269 certificate), or to combine several headers into one. This rule is not
4270 final, so it is possible to add other similar rules. Note that header
4271 addition is performed immediately, so one rule might reuse the resulting
4272 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004273
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004274http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004275
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004276 This stops the evaluation of the rules and lets the request pass the check.
4277 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06004278
Sasha Pachev218f0642014-06-16 12:05:59 -06004279
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004280http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004281
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004282 This stops the evaluation of the rules and immediately responds with an
4283 HTTP 401 or 407 error code to invite the user to present a valid user name
4284 and password. No further "http-request" rules are evaluated. An optional
4285 "realm" parameter is supported, it sets the authentication realm that is
4286 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06004287
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004288 Example:
4289 acl auth_ok http_auth_group(L1) G1
4290 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06004291
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004292http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004293
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02004294 See section 6.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004295
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004296http-request capture <sample> [ len <length> | id <id> ]
4297 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004298
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004299 This captures sample expression <sample> from the request buffer, and
4300 converts it to a string of at most <len> characters. The resulting string is
4301 stored into the next request "capture" slot, so it will possibly appear next
4302 to some captured HTTP headers. It will then automatically appear in the logs,
4303 and it will be possible to extract it using sample fetch rules to feed it
4304 into headers or anything. The length should be limited given that this size
4305 will be allocated for each capture during the whole session life.
4306 Please check section 7.3 (Fetching samples) and "capture request header" for
4307 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004308
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004309 If the keyword "id" is used instead of "len", the action tries to store the
4310 captured string in a previously declared capture slot. This is useful to run
4311 captures in backends. The slot id can be declared by a previous directive
4312 "http-request capture" or with the "declare capture" keyword. If the slot
4313 <id> doesn't exist, then HAProxy fails parsing the configuration to prevent
4314 unexpected behavior at run time.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004315
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004316http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004317
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004318 This is used to delete an entry from an ACL. The ACL must be loaded from a
4319 file (even a dummy empty file). The file name of the ACL to be updated is
4320 passed between parentheses. It takes one argument: <key fmt>, which follows
4321 log-format rules, to collect content of the entry to delete.
4322 It is the equivalent of the "del acl" command from the stats socket, but can
4323 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004324
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004325http-request del-header <name> [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004326
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004327 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004328
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004329http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004330
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004331 This is used to delete an entry from a MAP. The MAP must be loaded from a
4332 file (even a dummy empty file). The file name of the MAP to be updated is
4333 passed between parentheses. It takes one argument: <key fmt>, which follows
4334 log-format rules, to collect content of the entry to delete.
4335 It takes one argument: "file name" It is the equivalent of the "del map"
4336 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02004337
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004338http-request deny [deny_status <status>] [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04004339
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004340 This stops the evaluation of the rules and immediately rejects the request
4341 and emits an HTTP 403 error, or optionally the status code specified as an
4342 argument to "deny_status". The list of permitted status codes is limited to
4343 those that can be overridden by the "errorfile" directive.
4344 No further "http-request" rules are evaluated.
Patrick Hemmer268a7072018-05-11 12:52:31 -04004345
Olivier Houchard602bf7d2019-05-10 13:59:15 +02004346http-request disable-l7-retry [ { if | unless } <condition> ]
4347 This disables any attempt to retry the request if it fails for any other
4348 reason than a connection failure. This can be useful for example to make
4349 sure POST requests aren't retried on failure.
4350
Baptiste Assmann333939c2019-01-21 08:34:50 +01004351http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
4352
4353 This action performs a DNS resolution of the output of <expr> and stores
4354 the result in the variable <var>. It uses the DNS resolvers section
4355 pointed by <resolvers>.
4356 It is possible to choose a resolution preference using the optional
4357 arguments 'ipv4' or 'ipv6'.
4358 When performing the DNS resolution, the client side connection is on
4359 pause waiting till the end of the resolution.
4360 If an IP address can be found, it is stored into <var>. If any kind of
4361 error occurs, then <var> is not set.
4362 One can use this action to discover a server IP address at run time and
4363 based on information found in the request (IE a Host header).
4364 If this action is used to find the server's IP address (using the
4365 "set-dst" action), then the server IP address in the backend must be set
4366 to 0.0.0.0.
4367
4368 Example:
4369 resolvers mydns
4370 nameserver local 127.0.0.53:53
4371 nameserver google 8.8.8.8:53
4372 timeout retry 1s
4373 hold valid 10s
4374 hold nx 3s
4375 hold other 3s
4376 hold obsolete 0s
4377 accepted_payload_size 8192
4378
4379 frontend fe
4380 bind 10.42.0.1:80
4381 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
4382 http-request capture var(txn.myip) len 40
4383
4384 # return 503 when the variable is not set,
4385 # which mean DNS resolution error
4386 use_backend b_503 unless { var(txn.myip) -m found }
4387
4388 default_backend be
4389
4390 backend b_503
4391 # dummy backend used to return 503.
4392 # one can use the errorfile directive to send a nice
4393 # 503 error page to end users
4394
4395 backend be
4396 # rule to prevent HAProxy from reconnecting to services
4397 # on the local network (forged DNS name used to scan the network)
4398 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
4399 http-request set-dst var(txn.myip)
4400 server clear 0.0.0.0:0
4401
4402 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
4403 be used to scan the network or worst won't loop over itself...
4404
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004405http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
4406
4407 This is used to build an HTTP 103 Early Hints response prior to any other one.
4408 This appends an HTTP header field to this response whose name is specified in
4409 <name> and whose value is defined by <fmt> which follows the log-format rules
4410 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01004411 to the client some Link headers to preload resources required to render the
4412 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004413
4414 See RFC 8297 for more information.
4415
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004416http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004417
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004418 This performs an HTTP redirection based on a redirect rule. This is exactly
4419 the same as the "redirect" statement except that it inserts a redirect rule
4420 which can be processed in the middle of other "http-request" rules and that
4421 these rules use the "log-format" strings. See the "redirect" keyword for the
4422 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004423
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004424http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004425
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004426 This stops the evaluation of the rules and immediately closes the connection
4427 without sending any response. It acts similarly to the
4428 "tcp-request content reject" rules. It can be useful to force an immediate
4429 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004430
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004431http-request replace-header <name> <match-regex> <replace-fmt>
4432 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02004433
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004434 This matches the value of all occurences of header field <name> against
4435 <match-regex>. Matching is performed case-sensitively. Matching values are
4436 completely replaced by <replace-fmt>. Format characters are allowed in
4437 <replace-fmt> and work like <fmt> arguments in "http-request add-header".
4438 Standard back-references using the backslash ('\') followed by a number are
4439 supported.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004440
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004441 This action acts on whole header lines, regardless of the number of values
4442 they may contain. Thus it is well-suited to process headers naturally
4443 containing commas in their value, such as If-Modified-Since. Headers that
4444 contain a comma-separated list of values, such as Accept, should be processed
4445 using "http-request replace-value".
William Lallemand86d0df02017-11-24 21:36:45 +01004446
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004447 Example:
4448 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
4449
4450 # applied to:
4451 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4452
4453 # outputs:
4454 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4455
4456 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02004457
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004458 http-request replace-header User-Agent curl foo
4459
4460 # applied to:
4461 User-Agent: curl/7.47.0
Willy Tarreau09448f72014-06-25 18:12:15 +02004462
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004463 # outputs:
4464 User-Agent: foo
Willy Tarreau09448f72014-06-25 18:12:15 +02004465
Willy Tarreau262c3f12019-12-17 06:52:51 +01004466http-request replace-path <match-regex> <replace-fmt>
4467 [ { if | unless } <condition> ]
4468
4469 This works like "replace-header" except that it works on the request's path
4470 component instead of a header. The path component starts at the first '/'
4471 after an optional scheme+authority. It does contain the query string if any
4472 is present. The replacement does not modify the scheme nor authority.
4473
4474 It is worth noting that regular expressions may be more expensive to evaluate
4475 than certain ACLs, so rare replacements may benefit from a condition to avoid
4476 performing the evaluation at all if it does not match.
4477
4478 Example:
4479 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4480 http-request replace-path (.*) /foo\1
4481
4482 # suffix /foo : turn /bar?q=1 into /bar/foo?q=1 :
4483 http-request replace-path ([^?]*)(\?(.*))? \1/foo\2
4484
4485 # strip /foo : turn /foo/bar?q=1 into /bar?q=1
4486 http-request replace-path /foo/(.*) /\1
4487 # or more efficient if only some requests match :
4488 http-request replace-path /foo/(.*) /\1 if { url_beg /foo/ }
4489
Willy Tarreau33810222019-06-12 17:44:02 +02004490http-request replace-uri <match-regex> <replace-fmt>
4491 [ { if | unless } <condition> ]
4492
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004493 This works like "replace-header" except that it works on the request's URI part
4494 instead of a header. The URI part may contain an optional scheme, authority or
4495 query string. These are considered to be part of the value that is matched
4496 against.
4497
4498 It is worth noting that regular expressions may be more expensive to evaluate
4499 than certain ACLs, so rare replacements may benefit from a condition to avoid
4500 performing the evaluation at all if it does not match.
Willy Tarreau33810222019-06-12 17:44:02 +02004501
Willy Tarreau62b59132019-12-17 06:51:20 +01004502 IMPORTANT NOTE: historically in HTTP/1.x, the vast majority of requests sent
4503 by browsers use the "origin form", which differs from the "absolute form" in
4504 that they do not contain a scheme nor authority in the URI portion. Mostly
4505 only requests sent to proxies, those forged by hand and some emitted by
4506 certain applications use the absolute form. As such, "replace-uri" usually
4507 works fine most of the time in HTTP/1.x with rules starting with a "/". But
4508 with HTTP/2, clients are encouraged to send absolute URIs only, which look
4509 like the ones HTTP/1 clients use to talk to proxies. Such partial replace-uri
4510 rules may then fail in HTTP/2 when they work in HTTP/1. Either the rules need
Willy Tarreau262c3f12019-12-17 06:52:51 +01004511 to be adapted to optionally match a scheme and authority, or replace-path
4512 should be used.
Willy Tarreau33810222019-06-12 17:44:02 +02004513
Willy Tarreau62b59132019-12-17 06:51:20 +01004514 Example:
4515 # rewrite all "http" absolute requests to "https":
4516 http-request replace-uri ^http://(.*) https://\1
Willy Tarreau33810222019-06-12 17:44:02 +02004517
Willy Tarreau62b59132019-12-17 06:51:20 +01004518 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4519 http-request replace-uri ([^/:]*://[^/]*)?(.*) \1/foo\2
Willy Tarreau33810222019-06-12 17:44:02 +02004520
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004521http-request replace-value <name> <match-regex> <replace-fmt>
4522 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02004523
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004524 This works like "replace-header" except that it matches the regex against
4525 every comma-delimited value of the header field <name> instead of the
4526 entire header. This is suited for all headers which are allowed to carry
4527 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02004528
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004529 Example:
4530 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004531
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004532 # applied to:
4533 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004534
Tim Duesterhus2252beb2019-10-29 00:05:13 +01004535 # outputs:
4536 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004537
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004538http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4539http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004540
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004541 This actions increments the GPC0 or GPC1 counter according with the sticky
4542 counter designated by <sc-id>. If an error occurs, this action silently fails
4543 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004544
Cédric Dufour0d7712d2019-11-06 18:38:53 +01004545http-request sc-set-gpt0(<sc-id>) { <int> | <expr> }
4546 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004547
Cédric Dufour0d7712d2019-11-06 18:38:53 +01004548 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
4549 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
4550 boolean. If an error occurs, this action silently fails and the actions
4551 evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004552
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004553http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004554
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004555 This is used to set the destination IP address to the value of specified
4556 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
4557 but provides the correct IP in a HTTP header; or you want to mask the IP for
4558 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
4559 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004560
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004561 Arguments:
4562 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4563 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004564
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004565 Example:
4566 http-request set-dst hdr(x-dst)
4567 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01004568
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004569 When possible, set-dst preserves the original destination port as long as the
4570 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004571
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004572http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004573
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004574 This is used to set the destination port address to the value of specified
4575 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
4576 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004577
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004578 Arguments:
4579 <expr> Is a standard HAProxy expression formed by a sample-fetch
4580 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004581
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004582 Example:
4583 http-request set-dst-port hdr(x-port)
4584 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004585
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004586 When possible, set-dst-port preserves the original destination address as
4587 long as the address family supports a port, otherwise it forces the
4588 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004589
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004590http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004591
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004592 This does the same as "http-request add-header" except that the header name
4593 is first removed if it existed. This is useful when passing security
4594 information to the server, where the header must not be manipulated by
4595 external users. Note that the new value is computed before the removal so it
4596 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02004597
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004598 Example:
4599 http-request set-header X-Haproxy-Current-Date %T
4600 http-request set-header X-SSL %[ssl_fc]
4601 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
4602 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4603 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4604 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4605 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4606 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4607 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02004608
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004609http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004610
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004611 This is used to change the log level of the current request when a certain
4612 condition is met. Valid levels are the 8 syslog levels (see the "log"
4613 keyword) plus the special level "silent" which disables logging for this
4614 request. This rule is not final so the last matching rule wins. This rule
4615 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004616
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004617http-request set-map(<file-name>) <key fmt> <value fmt>
4618 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004619
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004620 This is used to add a new entry into a MAP. The MAP must be loaded from a
4621 file (even a dummy empty file). The file name of the MAP to be updated is
4622 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
4623 log-format rules, used to collect MAP key, and <value fmt>, which follows
4624 log-format rules, used to collect content for the new entry.
4625 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4626 more) values. This lookup is done by a linear search and can be expensive
4627 with large lists! It is the equivalent of the "set map" command from the
4628 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004629
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004630http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004631
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004632 This is used to set the Netfilter MARK on all packets sent to the client to
4633 the value passed in <mark> on platforms which support it. This value is an
4634 unsigned 32 bit value which can be matched by netfilter and by the routing
4635 table. It can be expressed both in decimal or hexadecimal format (prefixed by
4636 "0x"). This can be useful to force certain packets to take a different route
4637 (for example a cheaper network path for bulk downloads). This works on Linux
4638 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02004639
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004640http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004641
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004642 This rewrites the request method with the result of the evaluation of format
4643 string <fmt>. There should be very few valid reasons for having to do so as
4644 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004645
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004646http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004647
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004648 This sets the "nice" factor of the current request being processed. It only
4649 has effect against the other requests being processed at the same time.
4650 The default value is 0, unless altered by the "nice" setting on the "bind"
4651 line. The accepted range is -1024..1024. The higher the value, the nicest
4652 the request will be. Lower values will make the request more important than
4653 other ones. This can be useful to improve the speed of some requests, or
4654 lower the priority of non-important requests. Using this setting without
4655 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004656
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004657http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02004658
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004659 This rewrites the request path with the result of the evaluation of format
4660 string <fmt>. The query string, if any, is left intact. If a scheme and
4661 authority is found before the path, they are left intact as well. If the
4662 request doesn't have a path ("*"), this one is replaced with the format.
4663 This can be used to prepend a directory component in front of a path for
4664 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02004665
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004666 Example :
4667 # prepend the host name before the path
4668 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004669
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004670http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02004671
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004672 This is used to set the queue priority class of the current request.
4673 The value must be a sample expression which converts to an integer in the
4674 range -2047..2047. Results outside this range will be truncated.
4675 The priority class determines the order in which queued requests are
4676 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004677
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004678http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004679
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004680 This is used to set the queue priority timestamp offset of the current
4681 request. The value must be a sample expression which converts to an integer
4682 in the range -524287..524287. Results outside this range will be truncated.
4683 When a request is queued, it is ordered first by the priority class, then by
4684 the current timestamp adjusted by the given offset in milliseconds. Lower
4685 values have higher priority.
4686 Note that the resulting timestamp is is only tracked with enough precision
4687 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
4688 adjusted timestamp exceeds this value, it will be misidentified as highest
4689 priority. Thus it is important to set "timeout queue" to a value, where when
4690 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004691
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004692http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004693
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004694 This rewrites the request's query string which appears after the first
4695 question mark ("?") with the result of the evaluation of format string <fmt>.
4696 The part prior to the question mark is left intact. If the request doesn't
4697 contain a question mark and the new value is not empty, then one is added at
4698 the end of the URI, followed by the new value. If a question mark was
4699 present, it will never be removed even if the value is empty. This can be
4700 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004701
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004702 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004703
4704 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004705 # replace "%3D" with "=" in the query string
4706 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004707
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004708http-request set-src <expr> [ { if | unless } <condition> ]
4709 This is used to set the source IP address to the value of specified
4710 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
4711 provides the correct IP in a HTTP header; or you want to mask source IP for
4712 privacy.
4713
4714 Arguments :
4715 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4716 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004717
Cyril Bonté78caf842010-03-10 22:41:43 +01004718 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004719 http-request set-src hdr(x-forwarded-for)
4720 http-request set-src src,ipmask(24)
4721
4722 When possible, set-src preserves the original source port as long as the
4723 address family allows it, otherwise the source port is set to 0.
4724
4725http-request set-src-port <expr> [ { if | unless } <condition> ]
4726
4727 This is used to set the source port address to the value of specified
4728 expression.
4729
4730 Arguments:
4731 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4732 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004733
Willy Tarreau20b0de52012-12-24 15:45:22 +01004734 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004735 http-request set-src-port hdr(x-port)
4736 http-request set-src-port int(4000)
4737
4738 When possible, set-src-port preserves the original source address as long as
4739 the address family supports a port, otherwise it forces the source address to
4740 IPv4 "0.0.0.0" before rewriting the port.
4741
4742http-request set-tos <tos> [ { if | unless } <condition> ]
4743
4744 This is used to set the TOS or DSCP field value of packets sent to the client
4745 to the value passed in <tos> on platforms which support this. This value
4746 represents the whole 8 bits of the IP TOS field, and can be expressed both in
4747 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
4748 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
4749 be used to adjust some routing behavior on border routers based on some
4750 information from the request.
4751
4752 See RFC 2474, 2597, 3260 and 4594 for more information.
4753
4754http-request set-uri <fmt> [ { if | unless } <condition> ]
4755
4756 This rewrites the request URI with the result of the evaluation of format
4757 string <fmt>. The scheme, authority, path and query string are all replaced
4758 at once. This can be used to rewrite hosts in front of proxies, or to
4759 perform complex modifications to the URI such as moving parts between the
4760 path and the query string.
4761 See also "http-request set-path" and "http-request set-query".
4762
4763http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
4764
4765 This is used to set the contents of a variable. The variable is declared
4766 inline.
4767
4768 Arguments:
4769 <var-name> The name of the variable starts with an indication about its
4770 scope. The scopes allowed are:
4771 "proc" : the variable is shared with the whole process
4772 "sess" : the variable is shared with the whole session
4773 "txn" : the variable is shared with the transaction
4774 (request and response)
4775 "req" : the variable is shared only during request
4776 processing
4777 "res" : the variable is shared only during response
4778 processing
4779 This prefix is followed by a name. The separator is a '.'.
4780 The name may only contain characters 'a-z', 'A-Z', '0-9'
4781 and '_'.
4782
4783 <expr> Is a standard HAProxy expression formed by a sample-fetch
4784 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004785
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004786 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004787 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004788
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004789http-request send-spoe-group <engine-name> <group-name>
4790 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004791
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004792 This action is used to trigger sending of a group of SPOE messages. To do so,
4793 the SPOE engine used to send messages must be defined, as well as the SPOE
4794 group to send. Of course, the SPOE engine must refer to an existing SPOE
4795 filter. If not engine name is provided on the SPOE filter line, the SPOE
4796 agent name must be used.
4797
4798 Arguments:
4799 <engine-name> The SPOE engine name.
4800
4801 <group-name> The SPOE group name as specified in the engine
4802 configuration.
4803
4804http-request silent-drop [ { if | unless } <condition> ]
4805
4806 This stops the evaluation of the rules and makes the client-facing connection
4807 suddenly disappear using a system-dependent way that tries to prevent the
4808 client from being notified. The effect it then that the client still sees an
4809 established connection while there's none on HAProxy. The purpose is to
4810 achieve a comparable effect to "tarpit" except that it doesn't use any local
4811 resource at all on the machine running HAProxy. It can resist much higher
4812 loads than "tarpit", and slow down stronger attackers. It is important to
4813 understand the impact of using this mechanism. All stateful equipment placed
4814 between the client and HAProxy (firewalls, proxies, load balancers) will also
4815 keep the established connection for a long time and may suffer from this
4816 action.
4817 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
4818 option is used to block the emission of a TCP reset. On other systems, the
4819 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
4820 router, though it's still delivered to local networks. Do not use it unless
4821 you fully understand how it works.
4822
4823http-request tarpit [deny_status <status>] [ { if | unless } <condition> ]
4824
4825 This stops the evaluation of the rules and immediately blocks the request
4826 without responding for a delay specified by "timeout tarpit" or
4827 "timeout connect" if the former is not set. After that delay, if the client
4828 is still connected, an HTTP error 500 (or optionally the status code
4829 specified as an argument to "deny_status") is returned so that the client
4830 does not suspect it has been tarpitted. Logs will report the flags "PT".
4831 The goal of the tarpit rule is to slow down robots during an attack when
4832 they're limited on the number of concurrent requests. It can be very
4833 efficient against very dumb robots, and will significantly reduce the load
4834 on firewalls compared to a "deny" rule. But when facing "correctly"
4835 developed robots, it can make things worse by forcing haproxy and the front
4836 firewall to support insane number of concurrent connections.
4837 See also the "silent-drop" action.
4838
4839http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
4840http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
4841http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
4842
4843 This enables tracking of sticky counters from current request. These rules do
4844 not stop evaluation and do not change default action. The number of counters
4845 that may be simultaneously tracked by the same connection is set in
4846 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
4847 so the track-sc number is between 0 and (MAX_SESS_STCKTR-1). The first
4848 "track-sc0" rule executed enables tracking of the counters of the specified
4849 table as the first set. The first "track-sc1" rule executed enables tracking
4850 of the counters of the specified table as the second set. The first
4851 "track-sc2" rule executed enables tracking of the counters of the specified
4852 table as the third set. It is a recommended practice to use the first set of
4853 counters for the per-frontend counters and the second set for the per-backend
4854 ones. But this is just a guideline, all may be used everywhere.
4855
4856 Arguments :
4857 <key> is mandatory, and is a sample expression rule as described in
4858 section 7.3. It describes what elements of the incoming request or
4859 connection will be analyzed, extracted, combined, and used to
4860 select which table entry to update the counters.
4861
4862 <table> is an optional table to be used instead of the default one, which
4863 is the stick-table declared in the current proxy. All the counters
4864 for the matches and updates for the key will then be performed in
4865 that table until the session ends.
4866
4867 Once a "track-sc*" rule is executed, the key is looked up in the table and if
4868 it is not found, an entry is allocated for it. Then a pointer to that entry
4869 is kept during all the session's life, and this entry's counters are updated
4870 as often as possible, every time the session's counters are updated, and also
4871 systematically when the session ends. Counters are only updated for events
4872 that happen after the tracking has been started. As an exception, connection
4873 counters and request counters are systematically updated so that they reflect
4874 useful information.
4875
4876 If the entry tracks concurrent connection counters, one connection is counted
4877 for as long as the entry is tracked, and the entry will not expire during
4878 that time. Tracking counters also provides a performance advantage over just
4879 checking the keys, because only one table lookup is performed for all ACL
4880 checks that make use of it.
4881
4882http-request unset-var(<var-name>) [ { if | unless } <condition> ]
4883
4884 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004885
4886 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004887 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004888
Christopher Faulet579d83b2019-11-22 15:34:17 +01004889http-request use-service <service-name> [ { if | unless } <condition> ]
4890
4891 This directive executes the configured HTTP service to reply to the request
4892 and stops the evaluation of the rules. An HTTP service may choose to reply by
4893 sending any valid HTTP response or it may immediately close the connection
4894 without sending any response. Outside natives services, for instance the
4895 Prometheus exporter, it is possible to write your own services in Lua. No
4896 further "http-request" rules are evaluated.
4897
4898 Arguments :
4899 <service-name> is mandatory. It is the service to call
4900
4901 Example:
4902 http-request use-service prometheus-exporter if { path /metrics }
4903
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004904http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004905
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004906 This will delay the processing of the request until the SSL handshake
4907 happened. This is mostly useful to delay processing early data until we're
4908 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004909
Willy Tarreauef781042010-01-27 11:53:01 +01004910
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004911http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004912 Access control for Layer 7 responses
4913
4914 May be used in sections: defaults | frontend | listen | backend
4915 no | yes | yes | yes
4916
4917 The http-response statement defines a set of rules which apply to layer 7
4918 processing. The rules are evaluated in their declaration order when they are
4919 met in a frontend, listen or backend section. Any rule may optionally be
4920 followed by an ACL-based condition, in which case it will only be evaluated
4921 if the condition is true. Since these rules apply on responses, the backend
4922 rules are applied first, followed by the frontend's rules.
4923
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004924 The first keyword is the rule's action. The supported actions are described
4925 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004926
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004927 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004928
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004929 Example:
4930 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004931
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004932 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06004933
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004934 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4935 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06004936
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004937 Example:
4938 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06004939
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004940 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004941
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004942 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4943 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06004944
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004945 See also : "http-request", section 3.4 about userlists and section 7 about
4946 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004947
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004948http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004949
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004950 This is used to add a new entry into an ACL. The ACL must be loaded from a
4951 file (even a dummy empty file). The file name of the ACL to be updated is
4952 passed between parentheses. It takes one argument: <key fmt>, which follows
4953 log-format rules, to collect content of the new entry. It performs a lookup
4954 in the ACL before insertion, to avoid duplicated (or more) values.
4955 This lookup is done by a linear search and can be expensive with large lists!
4956 It is the equivalent of the "add acl" command from the stats socket, but can
4957 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06004958
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004959http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004960
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004961 This appends an HTTP header field whose name is specified in <name> and whose
4962 value is defined by <fmt> which follows the log-format rules (see Custom Log
4963 Format in section 8.2.4). This may be used to send a cookie to a client for
4964 example, or to pass some internal information.
4965 This rule is not final, so it is possible to add other similar rules.
4966 Note that header addition is performed immediately, so one rule might reuse
4967 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004968
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004969http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004970
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004971 This stops the evaluation of the rules and lets the response pass the check.
4972 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06004973
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004974http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004975
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02004976 See section 6.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06004977
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004978http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004979
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004980 This captures sample expression <sample> from the response buffer, and
4981 converts it to a string. The resulting string is stored into the next request
4982 "capture" slot, so it will possibly appear next to some captured HTTP
4983 headers. It will then automatically appear in the logs, and it will be
4984 possible to extract it using sample fetch rules to feed it into headers or
4985 anything. Please check section 7.3 (Fetching samples) and
4986 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004987
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004988 The keyword "id" is the id of the capture slot which is used for storing the
4989 string. The capture slot must be defined in an associated frontend.
4990 This is useful to run captures in backends. The slot id can be declared by a
4991 previous directive "http-response capture" or with the "declare capture"
4992 keyword.
4993 If the slot <id> doesn't exist, then HAProxy fails parsing the configuration
4994 to prevent unexpected behavior at run time.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004995
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004996http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004997
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004998 This is used to delete an entry from an ACL. The ACL must be loaded from a
4999 file (even a dummy empty file). The file name of the ACL to be updated is
5000 passed between parentheses. It takes one argument: <key fmt>, which follows
5001 log-format rules, to collect content of the entry to delete.
5002 It is the equivalent of the "del acl" command from the stats socket, but can
5003 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02005004
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005005http-response del-header <name> [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02005006
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005007 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02005008
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005009http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02005010
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005011 This is used to delete an entry from a MAP. The MAP must be loaded from a
5012 file (even a dummy empty file). The file name of the MAP to be updated is
5013 passed between parentheses. It takes one argument: <key fmt>, which follows
5014 log-format rules, to collect content of the entry to delete.
5015 It takes one argument: "file name" It is the equivalent of the "del map"
5016 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005017
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005018http-response deny [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005019
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005020 This stops the evaluation of the rules and immediately rejects the response
5021 and emits an HTTP 502 error. No further "http-response" rules are evaluated.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005022
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005023http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005024
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005025 This performs an HTTP redirection based on a redirect rule.
5026 This supports a format string similarly to "http-request redirect" rules,
5027 with the exception that only the "location" type of redirect is possible on
5028 the response. See the "redirect" keyword for the rule's syntax. When a
5029 redirect rule is applied during a response, connections to the server are
5030 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005031
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005032http-response replace-header <name> <regex-match> <replace-fmt>
5033 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005034
Tim Duesterhus2252beb2019-10-29 00:05:13 +01005035 This works like "http-request replace-header" except that it works on the
5036 server's response instead of the client's request.
William Lallemand86d0df02017-11-24 21:36:45 +01005037
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005038 Example:
5039 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02005040
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005041 # applied to:
5042 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005043
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005044 # outputs:
5045 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 +02005046
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005047 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005048
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005049http-response replace-value <name> <regex-match> <replace-fmt>
5050 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005051
Tim Duesterhus2252beb2019-10-29 00:05:13 +01005052 This works like "http-response replace-value" except that it works on the
5053 server's response instead of the client's request.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005054
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005055 Example:
5056 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005057
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005058 # applied to:
5059 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01005060
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005061 # outputs:
5062 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005063
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005064http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
5065http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08005066
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005067 This action increments the GPC0 or GPC1 counter according with the sticky
5068 counter designated by <sc-id>. If an error occurs, this action silently fails
5069 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02005070
Cédric Dufour0d7712d2019-11-06 18:38:53 +01005071http-response sc-set-gpt0(<sc-id>) { <int> | <expr> }
5072 [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02005073
Cédric Dufour0d7712d2019-11-06 18:38:53 +01005074 This action sets the 32-bit unsigned GPT0 tag according to the sticky counter
5075 designated by <sc-id> and the value of <int>/<expr>. The expected result is a
5076 boolean. If an error occurs, this action silently fails and the actions
5077 evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01005078
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005079http-response send-spoe-group [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02005080
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005081 This action is used to trigger sending of a group of SPOE messages. To do so,
5082 the SPOE engine used to send messages must be defined, as well as the SPOE
5083 group to send. Of course, the SPOE engine must refer to an existing SPOE
5084 filter. If not engine name is provided on the SPOE filter line, the SPOE
5085 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005086
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005087 Arguments:
5088 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005089
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005090 <group-name> The SPOE group name as specified in the engine
5091 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005092
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005093http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005094
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005095 This does the same as "add-header" except that the header name is first
5096 removed if it existed. This is useful when passing security information to
5097 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005098
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005099http-response set-log-level <level> [ { if | unless } <condition> ]
5100
5101 This is used to change the log level of the current request when a certain
5102 condition is met. Valid levels are the 8 syslog levels (see the "log"
5103 keyword) plus the special level "silent" which disables logging for this
5104 request. This rule is not final so the last matching rule wins. This rule can
5105 be useful to disable health checks coming from another equipment.
5106
5107http-response set-map(<file-name>) <key fmt> <value fmt>
5108
5109 This is used to add a new entry into a MAP. The MAP must be loaded from a
5110 file (even a dummy empty file). The file name of the MAP to be updated is
5111 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5112 log-format rules, used to collect MAP key, and <value fmt>, which follows
5113 log-format rules, used to collect content for the new entry. It performs a
5114 lookup in the MAP before insertion, to avoid duplicated (or more) values.
5115 This lookup is done by a linear search and can be expensive with large lists!
5116 It is the equivalent of the "set map" command from the stats socket, but can
5117 be triggered by an HTTP response.
5118
5119http-response set-mark <mark> [ { if | unless } <condition> ]
5120
5121 This is used to set the Netfilter MARK on all packets sent to the client to
5122 the value passed in <mark> on platforms which support it. This value is an
5123 unsigned 32 bit value which can be matched by netfilter and by the routing
5124 table. It can be expressed both in decimal or hexadecimal format (prefixed
5125 by "0x"). This can be useful to force certain packets to take a different
5126 route (for example a cheaper network path for bulk downloads). This works on
5127 Linux kernels 2.6.32 and above and requires admin privileges.
5128
5129http-response set-nice <nice> [ { if | unless } <condition> ]
5130
5131 This sets the "nice" factor of the current request being processed.
5132 It only has effect against the other requests being processed at the same
5133 time. The default value is 0, unless altered by the "nice" setting on the
5134 "bind" line. The accepted range is -1024..1024. The higher the value, the
5135 nicest the request will be. Lower values will make the request more important
5136 than other ones. This can be useful to improve the speed of some requests, or
5137 lower the priority of non-important requests. Using this setting without
5138 prior experimentation can cause some major slowdown.
5139
5140http-response set-status <status> [reason <str>]
5141 [ { if | unless } <condition> ]
5142
5143 This replaces the response status code with <status> which must be an integer
5144 between 100 and 999. Optionally, a custom reason text can be provided defined
5145 by <str>, or the default reason for the specified code will be used as a
5146 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005147
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005148 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005149 # return "431 Request Header Fields Too Large"
5150 http-response set-status 431
5151 # return "503 Slow Down", custom reason
5152 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005153
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005154http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005155
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005156 This is used to set the TOS or DSCP field value of packets sent to the client
5157 to the value passed in <tos> on platforms which support this.
5158 This value represents the whole 8 bits of the IP TOS field, and can be
5159 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
5160 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
5161 always 0. This can be used to adjust some routing behavior on border routers
5162 based on some information from the request.
5163
5164 See RFC 2474, 2597, 3260 and 4594 for more information.
5165
5166http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5167
5168 This is used to set the contents of a variable. The variable is declared
5169 inline.
5170
5171 Arguments:
5172 <var-name> The name of the variable starts with an indication about its
5173 scope. The scopes allowed are:
5174 "proc" : the variable is shared with the whole process
5175 "sess" : the variable is shared with the whole session
5176 "txn" : the variable is shared with the transaction
5177 (request and response)
5178 "req" : the variable is shared only during request
5179 processing
5180 "res" : the variable is shared only during response
5181 processing
5182 This prefix is followed by a name. The separator is a '.'.
5183 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5184 and '_'.
5185
5186 <expr> Is a standard HAProxy expression formed by a sample-fetch
5187 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005188
5189 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005190 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005191
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005192http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005193
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005194 This stops the evaluation of the rules and makes the client-facing connection
5195 suddenly disappear using a system-dependent way that tries to prevent the
5196 client from being notified. The effect it then that the client still sees an
5197 established connection while there's none on HAProxy. The purpose is to
5198 achieve a comparable effect to "tarpit" except that it doesn't use any local
5199 resource at all on the machine running HAProxy. It can resist much higher
5200 loads than "tarpit", and slow down stronger attackers. It is important to
5201 understand the impact of using this mechanism. All stateful equipment placed
5202 between the client and HAProxy (firewalls, proxies, load balancers) will also
5203 keep the established connection for a long time and may suffer from this
5204 action.
5205 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5206 option is used to block the emission of a TCP reset. On other systems, the
5207 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5208 router, though it's still delivered to local networks. Do not use it unless
5209 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005210
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005211http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5212http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5213http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005214
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005215 This enables tracking of sticky counters from current response. Please refer
5216 to "http-request track-sc" for a complete description. The only difference
5217 from "http-request track-sc" is the <key> sample expression can only make use
5218 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
5219 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
5220 supported, haproxy will fail and warn while parsing the config.
5221
5222http-response unset-var(<var-name>) [ { if | unless } <condition> ]
5223
5224 This is used to unset a variable. See "http-response set-var" for details
5225 about <var-name>.
5226
5227 Example:
5228 http-response unset-var(sess.last_redir)
5229
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02005230
Willy Tarreau30631952015-08-06 15:05:24 +02005231http-reuse { never | safe | aggressive | always }
5232 Declare how idle HTTP connections may be shared between requests
5233
5234 May be used in sections: defaults | frontend | listen | backend
5235 yes | no | yes | yes
5236
5237 By default, a connection established between haproxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01005238 which is considered safe for reuse is moved back to the server's idle
5239 connections pool so that any other request can make use of it. This is the
5240 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02005241
5242 The argument indicates the desired connection reuse strategy :
5243
Olivier Houchard86006a52018-12-14 19:37:49 +01005244 - "never" : idle connections are never shared between sessions. This mode
5245 may be enforced to cancel a different strategy inherited from
5246 a defaults section or for troubleshooting. For example, if an
5247 old bogus application considers that multiple requests over
5248 the same connection come from the same client and it is not
5249 possible to fix the application, it may be desirable to
5250 disable connection sharing in a single backend. An example of
5251 such an application could be an old haproxy using cookie
5252 insertion in tunnel mode and not checking any request past the
5253 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02005254
Olivier Houchard86006a52018-12-14 19:37:49 +01005255 - "safe" : this is the default and the recommended strategy. The first
5256 request of a session is always sent over its own connection,
5257 and only subsequent requests may be dispatched over other
5258 existing connections. This ensures that in case the server
5259 closes the connection when the request is being sent, the
5260 browser can decide to silently retry it. Since it is exactly
5261 equivalent to regular keep-alive, there should be no side
5262 effects.
Willy Tarreau30631952015-08-06 15:05:24 +02005263
5264 - "aggressive" : this mode may be useful in webservices environments where
5265 all servers are not necessarily known and where it would be
5266 appreciable to deliver most first requests over existing
5267 connections. In this case, first requests are only delivered
5268 over existing connections that have been reused at least once,
5269 proving that the server correctly supports connection reuse.
5270 It should only be used when it's sure that the client can
5271 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02005272 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02005273 downsides of rare connection failures.
5274
5275 - "always" : this mode is only recommended when the path to the server is
5276 known for never breaking existing connections quickly after
5277 releasing them. It allows the first request of a session to be
5278 sent to an existing connection. This can provide a significant
5279 performance increase over the "safe" strategy when the backend
5280 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005281 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02005282 sharing. It is recommended that the "http-keep-alive" timeout
5283 remains low in this mode so that no dead connections remain
5284 usable. In most cases, this will lead to the same performance
5285 gains as "aggressive" but with more risks. It should only be
5286 used when it improves the situation over "aggressive".
5287
5288 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005289 connection properties and compatibility. Specifically :
5290 - connections made with "usesrc" followed by a client-dependent value
5291 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005292
5293 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01005294 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005295
Lukas Tribusfd9b68c2018-10-27 20:06:59 +02005296 - connections with certain bogus authentication schemes (relying on the
5297 connection) like NTLM are detected, marked private and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005298
Lukas Tribuse8adfeb2019-11-06 11:50:25 +01005299 A connection pool is involved and configurable with "pool-max-conn".
Willy Tarreau30631952015-08-06 15:05:24 +02005300
5301 Note: connection reuse improves the accuracy of the "server maxconn" setting,
5302 because almost no new connection will be established while idle connections
5303 remain available. This is particularly true with the "always" strategy.
5304
5305 See also : "option http-keep-alive", "server maxconn"
5306
5307
Mark Lamourinec2247f02012-01-04 13:02:01 -05005308http-send-name-header [<header>]
5309 Add the server name to a request. Use the header string given by <header>
Mark Lamourinec2247f02012-01-04 13:02:01 -05005310 May be used in sections: defaults | frontend | listen | backend
5311 yes | no | yes | yes
Mark Lamourinec2247f02012-01-04 13:02:01 -05005312 Arguments :
Mark Lamourinec2247f02012-01-04 13:02:01 -05005313 <header> The header string to use to send the server name
5314
Willy Tarreau81bef7e2019-10-07 14:58:02 +02005315 The "http-send-name-header" statement causes the header field named <header>
5316 to be set to the name of the target server at the moment the request is about
5317 to be sent on the wire. Any existing occurrences of this header are removed.
5318 Upon retries and redispatches, the header field is updated to always reflect
5319 the server being attempted to connect to. Given that this header is modified
5320 very late in the connection setup, it may have unexpected effects on already
5321 modified headers. For example using it with transport-level header such as
5322 connection, content-length, transfer-encoding and so on will likely result in
5323 invalid requests being sent to the server. Additionally it has been reported
5324 that this directive is currently being used as a way to overwrite the Host
5325 header field in outgoing requests; while this trick has been known to work
5326 as a side effect of the feature for some time, it is not officially supported
5327 and might possibly not work anymore in a future version depending on the
5328 technical difficulties this feature induces. A long-term solution instead
5329 consists in fixing the application which required this trick so that it binds
5330 to the correct host name.
Mark Lamourinec2247f02012-01-04 13:02:01 -05005331
5332 See also : "server"
5333
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005334id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02005335 Set a persistent ID to a proxy.
5336 May be used in sections : defaults | frontend | listen | backend
5337 no | yes | yes | yes
5338 Arguments : none
5339
5340 Set a persistent ID for the proxy. This ID must be unique and positive.
5341 An unused ID will automatically be assigned if unset. The first assigned
5342 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005343
5344
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005345ignore-persist { if | unless } <condition>
5346 Declare a condition to ignore persistence
5347 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005348 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005349
5350 By default, when cookie persistence is enabled, every requests containing
5351 the cookie are unconditionally persistent (assuming the target server is up
5352 and running).
5353
5354 The "ignore-persist" statement allows one to declare various ACL-based
5355 conditions which, when met, will cause a request to ignore persistence.
5356 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005357 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005358 persistence for a specific User-Agent (for example, some web crawler bots).
5359
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005360 The persistence is ignored when an "if" condition is met, or unless an
5361 "unless" condition is met.
5362
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005363 Example:
5364 acl url_static path_beg /static /images /img /css
5365 acl url_static path_end .gif .png .jpg .css .js
5366 ignore-persist if url_static
5367
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005368 See also : "force-persist", "cookie", and section 7 about ACL usage.
5369
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005370load-server-state-from-file { global | local | none }
5371 Allow seamless reload of HAProxy
5372 May be used in sections: defaults | frontend | listen | backend
5373 yes | no | yes | yes
5374
5375 This directive points HAProxy to a file where server state from previous
5376 running process has been saved. That way, when starting up, before handling
5377 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01005378 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005379 to tell haproxy which file to use. For now, only 2 arguments to either prevent
5380 loading state or load states from a file containing all backends and servers.
5381 The state file can be generated by running the command "show servers state"
5382 over the stats socket and redirect output.
5383
Davor Ocelice9ed2812017-12-25 17:49:28 +01005384 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005385 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02005386 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005387
5388 Arguments:
5389 global load the content of the file pointed by the global directive
5390 named "server-state-file".
5391
5392 local load the content of the file pointed by the directive
5393 "server-state-file-name" if set. If not set, then the backend
5394 name is used as a file name.
5395
5396 none don't load any stat for this backend
5397
5398 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01005399 - server's IP address is preserved across reloads by default, but the
5400 order can be changed thanks to the server's "init-addr" setting. This
5401 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01005402 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01005403 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005404
5405 - server's weight is applied from previous running process unless it has
5406 has changed between previous and new configuration files.
5407
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005408 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005409
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005410 global
5411 stats socket /tmp/socket
5412 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005413
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005414 defaults
5415 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005416
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005417 backend bk
5418 server s1 127.0.0.1:22 check weight 11
5419 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005420
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005421
5422 Then one can run :
5423
5424 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
5425
5426 Content of the file /tmp/server_state would be like this:
5427
5428 1
5429 # <field names skipped for the doc example>
5430 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5431 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5432
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005433 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005434
5435 global
5436 stats socket /tmp/socket
5437 server-state-base /etc/haproxy/states
5438
5439 defaults
5440 load-server-state-from-file local
5441
5442 backend bk
5443 server s1 127.0.0.1:22 check weight 11
5444 server s2 127.0.0.1:22 check weight 12
5445
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005446
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005447 Then one can run :
5448
5449 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
5450
5451 Content of the file /etc/haproxy/states/bk would be like this:
5452
5453 1
5454 # <field names skipped for the doc example>
5455 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5456 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5457
5458 See also: "server-state-file", "server-state-file-name", and
5459 "show servers state"
5460
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005461
Willy Tarreau2769aa02007-12-27 18:26:09 +01005462log global
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005463log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
5464 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02005465no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01005466 Enable per-instance logging of events and traffic.
5467 May be used in sections : defaults | frontend | listen | backend
5468 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02005469
5470 Prefix :
5471 no should be used when the logger list must be flushed. For example,
5472 if you don't want to inherit from the default logger list. This
5473 prefix does not allow arguments.
5474
Willy Tarreau2769aa02007-12-27 18:26:09 +01005475 Arguments :
5476 global should be used when the instance's logging parameters are the
5477 same as the global ones. This is the most common usage. "global"
5478 replaces <address>, <facility> and <level> with those of the log
5479 entries found in the "global" section. Only one "log global"
5480 statement may be used per instance, and this form takes no other
5481 parameter.
5482
5483 <address> indicates where to send the logs. It takes the same format as
5484 for the "global" section's logs, and can be one of :
5485
5486 - An IPv4 address optionally followed by a colon (':') and a UDP
5487 port. If no port is specified, 514 is used by default (the
5488 standard syslog port).
5489
David du Colombier24bb5f52011-03-17 10:40:23 +01005490 - An IPv6 address followed by a colon (':') and optionally a UDP
5491 port. If no port is specified, 514 is used by default (the
5492 standard syslog port).
5493
Willy Tarreau2769aa02007-12-27 18:26:09 +01005494 - A filesystem path to a UNIX domain socket, keeping in mind
5495 considerations for chroot (be sure the path is accessible
5496 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01005497 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005498
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005499 - A file descriptor number in the form "fd@<number>", which may
5500 point to a pipe, terminal, or socket. In this case unbuffered
5501 logs are used and one writev() call per log is performed. This
5502 is a bit expensive but acceptable for most workloads. Messages
5503 sent this way will not be truncated but may be dropped, in
5504 which case the DroppedLogs counter will be incremented. The
5505 writev() call is atomic even on pipes for messages up to
5506 PIPE_BUF size, which POSIX recommends to be at least 512 and
5507 which is 4096 bytes on most modern operating systems. Any
5508 larger message may be interleaved with messages from other
5509 processes. Exceptionally for debugging purposes the file
5510 descriptor may also be directed to a file, but doing so will
5511 significantly slow haproxy down as non-blocking calls will be
5512 ignored. Also there will be no way to purge nor rotate this
5513 file without restarting the process. Note that the configured
5514 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01005515 with a TCP syslog server. See also the "short" and "raw"
5516 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005517
5518 - "stdout" / "stderr", which are respectively aliases for "fd@1"
5519 and "fd@2", see above.
5520
Willy Tarreauc046d162019-08-30 15:24:59 +02005521 - A ring buffer in the form "ring@<name>", which will correspond
5522 to an in-memory ring buffer accessible over the CLI using the
5523 "show events" command, which will also list existing rings and
5524 their sizes. Such buffers are lost on reload or restart but
5525 when used as a complement this can help troubleshooting by
5526 having the logs instantly available.
5527
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005528 You may want to reference some environment variables in the
5529 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01005530
Willy Tarreau18324f52014-06-27 18:10:07 +02005531 <length> is an optional maximum line length. Log lines larger than this
5532 value will be truncated before being sent. The reason is that
5533 syslog servers act differently on log line length. All servers
5534 support the default value of 1024, but some servers simply drop
5535 larger lines while others do log them. If a server supports long
5536 lines, it may make sense to set this value here in order to avoid
5537 truncating long lines. Similarly, if a server drops long lines,
5538 it is preferable to truncate them before sending them. Accepted
5539 values are 80 to 65535 inclusive. The default value of 1024 is
5540 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01005541 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02005542
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005543 <ranges> A list of comma-separated ranges to identify the logs to sample.
5544 This is used to balance the load of the logs to send to the log
5545 server. The limits of the ranges cannot be null. They are numbered
5546 from 1. The size or period (in number of logs) of the sample must
5547 be set with <sample_size> parameter.
5548
5549 <sample_size>
5550 The size of the sample in number of logs to consider when balancing
5551 their logging loads. It is used to balance the load of the logs to
5552 send to the syslog server. This size must be greater or equal to the
5553 maximum of the high limits of the ranges.
5554 (see also <ranges> parameter).
5555
Willy Tarreauadb345d2018-11-12 07:56:13 +01005556 <format> is the log format used when generating syslog messages. It may be
5557 one of the following :
5558
5559 rfc3164 The RFC3164 syslog message format. This is the default.
5560 (https://tools.ietf.org/html/rfc3164)
5561
5562 rfc5424 The RFC5424 syslog message format.
5563 (https://tools.ietf.org/html/rfc5424)
5564
Willy Tarreaue8746a02018-11-12 08:45:00 +01005565 short A message containing only a level between angle brackets such as
5566 '<3>', followed by the text. The PID, date, time, process name
5567 and system name are omitted. This is designed to be used with a
5568 local log server. This format is compatible with what the
5569 systemd logger consumes.
5570
Willy Tarreauc1b06452018-11-12 11:57:56 +01005571 raw A message containing only the text. The level, PID, date, time,
5572 process name and system name are omitted. This is designed to
5573 be used in containers or during development, where the severity
5574 only depends on the file descriptor used (stdout/stderr).
5575
Willy Tarreau2769aa02007-12-27 18:26:09 +01005576 <facility> must be one of the 24 standard syslog facilities :
5577
Willy Tarreaue8746a02018-11-12 08:45:00 +01005578 kern user mail daemon auth syslog lpr news
5579 uucp cron auth2 ftp ntp audit alert cron2
5580 local0 local1 local2 local3 local4 local5 local6 local7
5581
Willy Tarreauc1b06452018-11-12 11:57:56 +01005582 Note that the facility is ignored for the "short" and "raw"
5583 formats, but still required as a positional field. It is
5584 recommended to use "daemon" in this case to make it clear that
5585 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005586
5587 <level> is optional and can be specified to filter outgoing messages. By
5588 default, all messages are sent. If a level is specified, only
5589 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005590 will be sent. An optional minimum level can be specified. If it
5591 is set, logs emitted with a more severe level than this one will
5592 be capped to this level. This is used to avoid sending "emerg"
5593 messages on all terminals on some default syslog configurations.
5594 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005595
5596 emerg alert crit err warning notice info debug
5597
William Lallemand0f99e342011-10-12 17:50:54 +02005598 It is important to keep in mind that it is the frontend which decides what to
5599 log from a connection, and that in case of content switching, the log entries
5600 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005601
5602 However, backend log declaration define how and where servers status changes
5603 will be logged. Level "notice" will be used to indicate a server going up,
5604 "warning" will be used for termination signals and definitive service
5605 termination, and "alert" will be used for when a server goes down.
5606
5607 Note : According to RFC3164, messages are truncated to 1024 bytes before
5608 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005609
5610 Example :
5611 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01005612 log stdout format short daemon # send log to systemd
5613 log stdout format raw daemon # send everything to stdout
5614 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005615 log 127.0.0.1:514 local0 notice # only send important events
5616 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02005617 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01005618
Willy Tarreau2769aa02007-12-27 18:26:09 +01005619
William Lallemand48940402012-01-30 16:47:22 +01005620log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005621 Specifies the log format string to use for traffic logs
5622 May be used in sections: defaults | frontend | listen | backend
5623 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01005624
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005625 This directive specifies the log format string that will be used for all logs
5626 resulting from traffic passing through the frontend using this line. If the
5627 directive is used in a defaults section, all subsequent frontends will use
5628 the same log format. Please see section 8.2.4 which covers the log format
5629 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01005630
Guillaume de Lafond29f45602017-03-31 19:52:15 +02005631 "log-format" directive overrides previous "option tcplog", "log-format" and
5632 "option httplog" directives.
5633
Dragan Dosen7ad31542015-09-28 17:16:47 +02005634log-format-sd <string>
5635 Specifies the RFC5424 structured-data log format string
5636 May be used in sections: defaults | frontend | listen | backend
5637 yes | yes | yes | no
5638
5639 This directive specifies the RFC5424 structured-data log format string that
5640 will be used for all logs resulting from traffic passing through the frontend
5641 using this line. If the directive is used in a defaults section, all
5642 subsequent frontends will use the same log format. Please see section 8.2.4
5643 which covers the log format string in depth.
5644
5645 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
5646 about the RFC5424 structured-data part.
5647
5648 Note : This log format string will be used only for loggers that have set
5649 log format to "rfc5424".
5650
5651 Example :
5652 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
5653
5654
Willy Tarreau094af4e2015-01-07 15:03:42 +01005655log-tag <string>
5656 Specifies the log tag to use for all outgoing logs
5657 May be used in sections: defaults | frontend | listen | backend
5658 yes | yes | yes | yes
5659
5660 Sets the tag field in the syslog header to this string. It defaults to the
5661 log-tag set in the global section, otherwise the program name as launched
5662 from the command line, which usually is "haproxy". Sometimes it can be useful
5663 to differentiate between multiple processes running on the same host, or to
5664 differentiate customer instances running in the same process. In the backend,
5665 logs about servers up/down will use this tag. As a hint, it can be convenient
5666 to set a log-tag related to a hosted customer in a defaults section then put
5667 all the frontends and backends for that customer, then start another customer
5668 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005669
Willy Tarreauc35362a2014-04-25 13:58:37 +02005670max-keep-alive-queue <value>
5671 Set the maximum server queue size for maintaining keep-alive connections
5672 May be used in sections: defaults | frontend | listen | backend
5673 yes | no | yes | yes
5674
5675 HTTP keep-alive tries to reuse the same server connection whenever possible,
5676 but sometimes it can be counter-productive, for example if a server has a lot
5677 of connections while other ones are idle. This is especially true for static
5678 servers.
5679
5680 The purpose of this setting is to set a threshold on the number of queued
5681 connections at which haproxy stops trying to reuse the same server and prefers
5682 to find another one. The default value, -1, means there is no limit. A value
5683 of zero means that keep-alive requests will never be queued. For very close
5684 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01005685 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02005686 use a value of 10 or less). For remote servers suffering from a high latency,
5687 higher values might be needed to cover for the latency and/or the cost of
5688 picking a different server.
5689
5690 Note that this has no impact on responses which are maintained to the same
5691 server consecutively to a 401 response. They will still go to the same server
5692 even if they have to be queued.
5693
5694 See also : "option http-server-close", "option prefer-last-server", server
5695 "maxconn" and cookie persistence.
5696
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01005697max-session-srv-conns <nb>
5698 Set the maximum number of outgoing connections we can keep idling for a given
5699 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
5700 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02005701
Willy Tarreau2769aa02007-12-27 18:26:09 +01005702maxconn <conns>
5703 Fix the maximum number of concurrent connections on a frontend
5704 May be used in sections : defaults | frontend | listen | backend
5705 yes | yes | yes | no
5706 Arguments :
5707 <conns> is the maximum number of concurrent connections the frontend will
5708 accept to serve. Excess connections will be queued by the system
5709 in the socket's listen queue and will be served once a connection
5710 closes.
5711
5712 If the system supports it, it can be useful on big sites to raise this limit
5713 very high so that haproxy manages connection queues, instead of leaving the
5714 clients with unanswered connection attempts. This value should not exceed the
5715 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01005716 of tune.bufsize (16kB by default) each, as well as some other data resulting
5717 in about 33 kB of RAM being consumed per established connection. That means
5718 that a medium system equipped with 1GB of RAM can withstand around
5719 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005720
5721 Also, when <conns> is set to large values, it is possible that the servers
5722 are not sized to accept such loads, and for this reason it is generally wise
5723 to assign them some reasonable connection limits.
5724
Willy Tarreauc8d5b952019-02-27 17:25:52 +01005725 When this value is set to zero, which is the default, the global "maxconn"
5726 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02005727
Willy Tarreau2769aa02007-12-27 18:26:09 +01005728 See also : "server", global section's "maxconn", "fullconn"
5729
5730
5731mode { tcp|http|health }
5732 Set the running mode or protocol of the instance
5733 May be used in sections : defaults | frontend | listen | backend
5734 yes | yes | yes | yes
5735 Arguments :
5736 tcp The instance will work in pure TCP mode. A full-duplex connection
5737 will be established between clients and servers, and no layer 7
5738 examination will be performed. This is the default mode. It
5739 should be used for SSL, SSH, SMTP, ...
5740
5741 http The instance will work in HTTP mode. The client request will be
5742 analyzed in depth before connecting to any server. Any request
5743 which is not RFC-compliant will be rejected. Layer 7 filtering,
5744 processing and switching will be possible. This is the mode which
5745 brings HAProxy most of its value.
5746
5747 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02005748 to incoming connections and close the connection. Alternatively,
5749 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
5750 instead. Nothing will be logged in either case. This mode is used
5751 to reply to external components health checks. This mode is
5752 deprecated and should not be used anymore as it is possible to do
5753 the same and even better by combining TCP or HTTP modes with the
5754 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005755
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005756 When doing content switching, it is mandatory that the frontend and the
5757 backend are in the same mode (generally HTTP), otherwise the configuration
5758 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005759
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005760 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005761 defaults http_instances
5762 mode http
5763
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005764 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005765
Willy Tarreau0ba27502007-12-24 16:55:16 +01005766
Cyril Bontéf0c60612010-02-06 14:44:47 +01005767monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01005768 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005769 May be used in sections : defaults | frontend | listen | backend
5770 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01005771 Arguments :
5772 if <cond> the monitor request will fail if the condition is satisfied,
5773 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005774 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01005775 are met, for instance a low number of servers both in a
5776 backend and its backup.
5777
5778 unless <cond> the monitor request will succeed only if the condition is
5779 satisfied, and will fail otherwise. Such a condition may be
5780 based on a test on the presence of a minimum number of active
5781 servers in a list of backends.
5782
5783 This statement adds a condition which can force the response to a monitor
5784 request to report a failure. By default, when an external component queries
5785 the URI dedicated to monitoring, a 200 response is returned. When one of the
5786 conditions above is met, haproxy will return 503 instead of 200. This is
5787 very useful to report a site failure to an external component which may base
5788 routing advertisements between multiple sites on the availability reported by
5789 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005790 criterion. Note that "monitor fail" only works in HTTP mode. Both status
5791 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005792
5793 Example:
5794 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01005795 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01005796 acl site_dead nbsrv(dynamic) lt 2
5797 acl site_dead nbsrv(static) lt 2
5798 monitor-uri /site_alive
5799 monitor fail if site_dead
5800
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005801 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005802
5803
5804monitor-net <source>
5805 Declare a source network which is limited to monitor requests
5806 May be used in sections : defaults | frontend | listen | backend
5807 yes | yes | yes | no
5808 Arguments :
5809 <source> is the source IPv4 address or network which will only be able to
5810 get monitor responses to any request. It can be either an IPv4
5811 address, a host name, or an address followed by a slash ('/')
5812 followed by a mask.
5813
5814 In TCP mode, any connection coming from a source matching <source> will cause
5815 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005816 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01005817 forwarding the connection to a remote server.
5818
5819 In HTTP mode, a connection coming from a source matching <source> will be
5820 accepted, the following response will be sent without waiting for a request,
5821 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
5822 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02005823 running without forwarding the request to a backend server. Note that this
5824 response is sent in raw format, without any transformation. This is important
5825 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005826
Willy Tarreau82569f92012-09-27 23:48:56 +02005827 Monitor requests are processed very early, just after tcp-request connection
5828 ACLs which are the only ones able to block them. These connections are short
5829 lived and never wait for any data from the client. They cannot be logged, and
5830 it is the intended purpose. They are only used to report HAProxy's health to
5831 an upper component, nothing more. Please note that "monitor fail" rules do
5832 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005833
Willy Tarreau95cd2832010-03-04 23:36:33 +01005834 Last, please note that only one "monitor-net" statement can be specified in
5835 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005836
Willy Tarreau2769aa02007-12-27 18:26:09 +01005837 Example :
5838 # addresses .252 and .253 are just probing us.
5839 frontend www
5840 monitor-net 192.168.0.252/31
5841
5842 See also : "monitor fail", "monitor-uri"
5843
5844
5845monitor-uri <uri>
5846 Intercept a URI used by external components' monitor requests
5847 May be used in sections : defaults | frontend | listen | backend
5848 yes | yes | yes | no
5849 Arguments :
5850 <uri> is the exact URI which we want to intercept to return HAProxy's
5851 health status instead of forwarding the request.
5852
5853 When an HTTP request referencing <uri> will be received on a frontend,
5854 HAProxy will not forward it nor log it, but instead will return either
5855 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
5856 conditions defined with "monitor fail". This is normally enough for any
5857 front-end HTTP probe to detect that the service is UP and running without
5858 forwarding the request to a backend server. Note that the HTTP method, the
5859 version and all headers are ignored, but the request must at least be valid
5860 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
5861
Willy Tarreau721d8e02017-12-01 18:25:08 +01005862 Monitor requests are processed very early, just after the request is parsed
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02005863 and even before any "http-request". The only rulesets applied before are the
5864 tcp-request ones. They cannot be logged either, and it is the intended
5865 purpose. They are only used to report HAProxy's health to an upper component,
5866 nothing more. However, it is possible to add any number of conditions using
5867 "monitor fail" and ACLs so that the result can be adjusted to whatever check
5868 can be imagined (most often the number of available servers in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005869
5870 Example :
5871 # Use /haproxy_test to report haproxy's status
5872 frontend www
5873 mode http
5874 monitor-uri /haproxy_test
5875
5876 See also : "monitor fail", "monitor-net"
5877
Willy Tarreau0ba27502007-12-24 16:55:16 +01005878
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005879option abortonclose
5880no option abortonclose
5881 Enable or disable early dropping of aborted requests pending in queues.
5882 May be used in sections : defaults | frontend | listen | backend
5883 yes | no | yes | yes
5884 Arguments : none
5885
5886 In presence of very high loads, the servers will take some time to respond.
5887 The per-instance connection queue will inflate, and the response time will
5888 increase respective to the size of the queue times the average per-session
5889 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01005890 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005891 the queue, and slowing down other users, and the servers as well, because the
5892 request will eventually be served, then aborted at the first error
5893 encountered while delivering the response.
5894
5895 As there is no way to distinguish between a full STOP and a simple output
5896 close on the client side, HTTP agents should be conservative and consider
5897 that the client might only have closed its output channel while waiting for
5898 the response. However, this introduces risks of congestion when lots of users
5899 do the same, and is completely useless nowadays because probably no client at
5900 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01005901 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005902 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01005903 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005904 of being the single component to break rare but valid traffic is extremely
5905 low, which adds to the temptation to be able to abort a session early while
5906 still not served and not pollute the servers.
5907
Davor Ocelice9ed2812017-12-25 17:49:28 +01005908 In HAProxy, the user can choose the desired behavior using the option
5909 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005910 compliant and aborted requests will be served. But when the option is
5911 specified, a session with an incoming channel closed will be aborted while
5912 it is still possible, either pending in the queue for a connection slot, or
5913 during the connection establishment if the server has not yet acknowledged
5914 the connection request. This considerably reduces the queue size and the load
5915 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005916 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005917
5918 If this option has been enabled in a "defaults" section, it can be disabled
5919 in a specific instance by prepending the "no" keyword before it.
5920
5921 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
5922
5923
Willy Tarreau4076a152009-04-02 15:18:36 +02005924option accept-invalid-http-request
5925no option accept-invalid-http-request
5926 Enable or disable relaxing of HTTP request parsing
5927 May be used in sections : defaults | frontend | listen | backend
5928 yes | yes | yes | no
5929 Arguments : none
5930
Willy Tarreau91852eb2015-05-01 13:26:00 +02005931 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005932 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005933 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005934 forbidden characters are essentially used to build attacks exploiting server
5935 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5936 server will emit invalid header names for whatever reason (configuration,
5937 implementation) and the issue will not be immediately fixed. In such a case,
5938 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01005939 even if that does not make sense, by specifying this option. Similarly, the
5940 list of characters allowed to appear in a URI is well defined by RFC3986, and
5941 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
5942 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01005943 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02005944 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02005945 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
5946 to pass through (no version specified) and multiple digits for both the major
5947 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005948
5949 This option should never be enabled by default as it hides application bugs
5950 and open security breaches. It should only be deployed after a problem has
5951 been confirmed.
5952
5953 When this option is enabled, erroneous header names will still be accepted in
5954 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01005955 analysis using the "show errors" request on the UNIX stats socket. Similarly,
5956 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02005957 also helps confirming that the issue has been solved.
5958
5959 If this option has been enabled in a "defaults" section, it can be disabled
5960 in a specific instance by prepending the "no" keyword before it.
5961
5962 See also : "option accept-invalid-http-response" and "show errors" on the
5963 stats socket.
5964
5965
5966option accept-invalid-http-response
5967no option accept-invalid-http-response
5968 Enable or disable relaxing of HTTP response parsing
5969 May be used in sections : defaults | frontend | listen | backend
5970 yes | no | yes | yes
5971 Arguments : none
5972
Willy Tarreau91852eb2015-05-01 13:26:00 +02005973 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005974 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005975 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005976 forbidden characters are essentially used to build attacks exploiting server
5977 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5978 server will emit invalid header names for whatever reason (configuration,
5979 implementation) and the issue will not be immediately fixed. In such a case,
5980 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02005981 even if that does not make sense, by specifying this option. This option also
5982 relaxes the test on the HTTP version format, it allows multiple digits for
5983 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005984
5985 This option should never be enabled by default as it hides application bugs
5986 and open security breaches. It should only be deployed after a problem has
5987 been confirmed.
5988
5989 When this option is enabled, erroneous header names will still be accepted in
5990 responses, but the complete response will be captured in order to permit
5991 later analysis using the "show errors" request on the UNIX stats socket.
5992 Doing this also helps confirming that the issue has been solved.
5993
5994 If this option has been enabled in a "defaults" section, it can be disabled
5995 in a specific instance by prepending the "no" keyword before it.
5996
5997 See also : "option accept-invalid-http-request" and "show errors" on the
5998 stats socket.
5999
6000
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006001option allbackups
6002no option allbackups
6003 Use either all backup servers at a time or only the first one
6004 May be used in sections : defaults | frontend | listen | backend
6005 yes | no | yes | yes
6006 Arguments : none
6007
6008 By default, the first operational backup server gets all traffic when normal
6009 servers are all down. Sometimes, it may be preferred to use multiple backups
6010 at once, because one will not be enough. When "option allbackups" is enabled,
6011 the load balancing will be performed among all backup servers when all normal
6012 ones are unavailable. The same load balancing algorithm will be used and the
6013 servers' weights will be respected. Thus, there will not be any priority
6014 order between the backup servers anymore.
6015
6016 This option is mostly used with static server farms dedicated to return a
6017 "sorry" page when an application is completely offline.
6018
6019 If this option has been enabled in a "defaults" section, it can be disabled
6020 in a specific instance by prepending the "no" keyword before it.
6021
6022
6023option checkcache
6024no option checkcache
Godbach7056a352013-12-11 20:01:07 +08006025 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006026 May be used in sections : defaults | frontend | listen | backend
6027 yes | no | yes | yes
6028 Arguments : none
6029
6030 Some high-level frameworks set application cookies everywhere and do not
6031 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006032 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006033 high risk of session crossing or stealing between users traversing the same
6034 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006035 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006036
6037 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006038 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01006039 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006040 response to check if there's a risk of caching a cookie on a client-side
6041 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01006042 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01006043 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01006044 - all those with a return code other than 200, 203, 204, 206, 300, 301,
6045 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01006046 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01006047 - all those that result from a request using a method other than GET, HEAD,
6048 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01006049 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006050 - those with a 'Pragma: no-cache' header
6051 - those with a 'Cache-control: private' header
6052 - those with a 'Cache-control: no-store' header
6053 - those with a 'Cache-control: max-age=0' header
6054 - those with a 'Cache-control: s-maxage=0' header
6055 - those with a 'Cache-control: no-cache' header
6056 - those with a 'Cache-control: no-cache="set-cookie"' header
6057 - those with a 'Cache-control: no-cache="set-cookie,' header
6058 (allowing other fields after set-cookie)
6059
6060 If a response doesn't respect these requirements, then it will be blocked
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02006061 just as if it was from an "http-response deny" rule, with an "HTTP 502 bad
6062 gateway". The session state shows "PH--" meaning that the proxy blocked the
6063 response during headers processing. Additionally, an alert will be sent in
6064 the logs so that admins are informed that there's something to be fixed.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006065
6066 Due to the high impact on the application, the application should be tested
6067 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006068 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01006069 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006070
6071 If this option has been enabled in a "defaults" section, it can be disabled
6072 in a specific instance by prepending the "no" keyword before it.
6073
6074
6075option clitcpka
6076no option clitcpka
6077 Enable or disable the sending of TCP keepalive packets on the client side
6078 May be used in sections : defaults | frontend | listen | backend
6079 yes | yes | yes | no
6080 Arguments : none
6081
6082 When there is a firewall or any session-aware component between a client and
6083 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006084 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006085 components decides to expire a session which has remained idle for too long.
6086
6087 Enabling socket-level TCP keep-alives makes the system regularly send packets
6088 to the other end of the connection, leaving it active. The delay between
6089 keep-alive probes is controlled by the system only and depends both on the
6090 operating system and its tuning parameters.
6091
6092 It is important to understand that keep-alive packets are neither emitted nor
6093 received at the application level. It is only the network stacks which sees
6094 them. For this reason, even if one side of the proxy already uses keep-alives
6095 to maintain its connection alive, those keep-alive packets will not be
6096 forwarded to the other side of the proxy.
6097
6098 Please note that this has nothing to do with HTTP keep-alive.
6099
6100 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
6101 client side of a connection, which should help when session expirations are
6102 noticed between HAProxy and a client.
6103
6104 If this option has been enabled in a "defaults" section, it can be disabled
6105 in a specific instance by prepending the "no" keyword before it.
6106
6107 See also : "option srvtcpka", "option tcpka"
6108
6109
Willy Tarreau0ba27502007-12-24 16:55:16 +01006110option contstats
6111 Enable continuous traffic statistics updates
6112 May be used in sections : defaults | frontend | listen | backend
6113 yes | yes | yes | no
6114 Arguments : none
6115
6116 By default, counters used for statistics calculation are incremented
6117 only when a session finishes. It works quite well when serving small
6118 objects, but with big ones (for example large images or archives) or
6119 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01006120 a hedgehog. With this option enabled counters get incremented frequently
6121 along the session, typically every 5 seconds, which is often enough to
6122 produce clean graphs. Recounting touches a hotpath directly so it is not
6123 not enabled by default, as it can cause a lot of wakeups for very large
6124 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006125
6126
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006127option dontlog-normal
6128no option dontlog-normal
6129 Enable or disable logging of normal, successful connections
6130 May be used in sections : defaults | frontend | listen | backend
6131 yes | yes | yes | no
6132 Arguments : none
6133
6134 There are large sites dealing with several thousand connections per second
6135 and for which logging is a major pain. Some of them are even forced to turn
6136 logs off and cannot debug production issues. Setting this option ensures that
6137 normal connections, those which experience no error, no timeout, no retry nor
6138 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
6139 mode, the response status code is checked and return codes 5xx will still be
6140 logged.
6141
6142 It is strongly discouraged to use this option as most of the time, the key to
6143 complex issues is in the normal logs which will not be logged here. If you
6144 need to separate logs, see the "log-separate-errors" option instead.
6145
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006146 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006147 logging.
6148
6149
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006150option dontlognull
6151no option dontlognull
6152 Enable or disable logging of null connections
6153 May be used in sections : defaults | frontend | listen | backend
6154 yes | yes | yes | no
6155 Arguments : none
6156
6157 In certain environments, there are components which will regularly connect to
6158 various systems to ensure that they are still alive. It can be the case from
6159 another load balancer as well as from monitoring systems. By default, even a
6160 simple port probe or scan will produce a log. If those connections pollute
6161 the logs too much, it is possible to enable option "dontlognull" to indicate
6162 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02006163 which typically corresponds to those probes. Note that errors will still be
6164 returned to the client and accounted for in the stats. If this is not what is
6165 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006166
6167 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01006168 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006169 would not be logged.
6170
6171 If this option has been enabled in a "defaults" section, it can be disabled
6172 in a specific instance by prepending the "no" keyword before it.
6173
Willy Tarreau0f228a02015-05-01 15:37:53 +02006174 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
6175 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006176
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006177
Willy Tarreau87cf5142011-08-19 22:57:24 +02006178option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006179 Enable insertion of the X-Forwarded-For header to requests sent to servers
6180 May be used in sections : defaults | frontend | listen | backend
6181 yes | yes | yes | yes
6182 Arguments :
6183 <network> is an optional argument used to disable this option for sources
6184 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02006185 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01006186 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006187
6188 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
6189 their client address. This is sometimes annoying when the client's IP address
6190 is expected in server logs. To solve this problem, the well-known HTTP header
6191 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
6192 This header contains a value representing the client's IP address. Since this
6193 header is always appended at the end of the existing header list, the server
6194 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02006195 the server's manual to find how to enable use of this standard header. Note
6196 that only the last occurrence of the header must be used, since it is really
6197 possible that the client has already brought one.
6198
Willy Tarreaud72758d2010-01-12 10:42:19 +01006199 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02006200 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01006201 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01006202 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006203 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02006204 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01006205
6206 Sometimes, a same HAProxy instance may be shared between a direct client
6207 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6208 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6209 header for a known source address or network by adding the "except" keyword
6210 followed by the network address. In this case, any source IP matching the
6211 network will not cause an addition of this header. Most common uses are with
6212 private networks or 127.0.0.1.
6213
Willy Tarreau87cf5142011-08-19 22:57:24 +02006214 Alternatively, the keyword "if-none" states that the header will only be
6215 added if it is not present. This should only be used in perfectly trusted
6216 environment, as this might cause a security issue if headers reaching haproxy
6217 are under the control of the end-user.
6218
Willy Tarreauc27debf2008-01-06 08:57:02 +01006219 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02006220 least one of them uses it, the header will be added. Note that the backend's
6221 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02006222 both are defined. In the case of the "if-none" argument, if at least one of
6223 the frontend or the backend does not specify it, it wants the addition to be
6224 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006225
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02006226 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01006227 # Public HTTP address also used by stunnel on the same machine
6228 frontend www
6229 mode http
6230 option forwardfor except 127.0.0.1 # stunnel already adds the header
6231
Ross Westaf72a1d2008-08-03 10:51:45 +02006232 # Those servers want the IP Address in X-Client
6233 backend www
6234 mode http
6235 option forwardfor header X-Client
6236
Willy Tarreau87cf5142011-08-19 22:57:24 +02006237 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006238 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01006239
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006240
Christopher Faulet98fbe952019-07-22 16:18:24 +02006241option h1-case-adjust-bogus-client
6242no option h1-case-adjust-bogus-client
6243 Enable or disable the case adjustment of HTTP/1 headers sent to bogus clients
6244 May be used in sections : defaults | frontend | listen | backend
6245 yes | yes | yes | no
6246 Arguments : none
6247
6248 There is no standard case for header names because, as stated in RFC7230,
6249 they are case-insensitive. So applications must handle them in a case-
6250 insensitive manner. But some bogus applications violate the standards and
6251 erroneously rely on the cases most commonly used by browsers. This problem
6252 becomes critical with HTTP/2 because all header names must be exchanged in
6253 lower case, and HAProxy follows the same convention. All header names are
6254 sent in lower case to clients and servers, regardless of the HTTP version.
6255
6256 When HAProxy receives an HTTP/1 response, its header names are converted to
6257 lower case and manipulated and sent this way to the clients. If a client is
6258 known to violate the HTTP standards and to fail to process a response coming
6259 from HAProxy, it is possible to transform the lower case header names to a
6260 different format when the response is formatted and sent to the client, by
6261 enabling this option and specifying the list of headers to be reformatted
6262 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6263 must only be a temporary workaround for the time it takes the client to be
6264 fixed, because clients which require such workarounds might be vulnerable to
6265 content smuggling attacks and must absolutely be fixed.
6266
6267 Please note that this option will not affect standards-compliant clients.
6268
6269 If this option has been enabled in a "defaults" section, it can be disabled
6270 in a specific instance by prepending the "no" keyword before it.
6271
6272 See also: "option h1-case-adjust-bogus-server", "h1-case-adjust",
6273 "h1-case-adjust-file".
6274
6275
6276option h1-case-adjust-bogus-server
6277no option h1-case-adjust-bogus-server
6278 Enable or disable the case adjustment of HTTP/1 headers sent to bogus servers
6279 May be used in sections : defaults | frontend | listen | backend
6280 yes | no | yes | yes
6281 Arguments : none
6282
6283 There is no standard case for header names because, as stated in RFC7230,
6284 they are case-insensitive. So applications must handle them in a case-
6285 insensitive manner. But some bogus applications violate the standards and
6286 erroneously rely on the cases most commonly used by browsers. This problem
6287 becomes critical with HTTP/2 because all header names must be exchanged in
6288 lower case, and HAProxy follows the same convention. All header names are
6289 sent in lower case to clients and servers, regardless of the HTTP version.
6290
6291 When HAProxy receives an HTTP/1 request, its header names are converted to
6292 lower case and manipulated and sent this way to the servers. If a server is
6293 known to violate the HTTP standards and to fail to process a request coming
6294 from HAProxy, it is possible to transform the lower case header names to a
6295 different format when the request is formatted and sent to the server, by
6296 enabling this option and specifying the list of headers to be reformatted
6297 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6298 must only be a temporary workaround for the time it takes the server to be
6299 fixed, because servers which require such workarounds might be vulnerable to
6300 content smuggling attacks and must absolutely be fixed.
6301
6302 Please note that this option will not affect standards-compliant servers.
6303
6304 If this option has been enabled in a "defaults" section, it can be disabled
6305 in a specific instance by prepending the "no" keyword before it.
6306
6307 See also: "option h1-case-adjust-bogus-client", "h1-case-adjust",
6308 "h1-case-adjust-file".
6309
6310
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006311option http-buffer-request
6312no option http-buffer-request
6313 Enable or disable waiting for whole HTTP request body before proceeding
6314 May be used in sections : defaults | frontend | listen | backend
6315 yes | yes | yes | yes
6316 Arguments : none
6317
6318 It is sometimes desirable to wait for the body of an HTTP request before
6319 taking a decision. This is what is being done by "balance url_param" for
6320 example. The first use case is to buffer requests from slow clients before
6321 connecting to the server. Another use case consists in taking the routing
6322 decision based on the request body's contents. This option placed in a
6323 frontend or backend forces the HTTP processing to wait until either the whole
Christopher Faulet6db8a2e2019-11-19 16:27:25 +01006324 body is received or the request buffer is full. It can have undesired side
6325 effects with some applications abusing HTTP by expecting unbuffered
6326 transmissions between the frontend and the backend, so this should definitely
6327 not be used by default.
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006328
Baptiste Assmanneccdf432015-10-28 13:49:01 +01006329 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006330
6331
Willy Tarreau0f228a02015-05-01 15:37:53 +02006332option http-ignore-probes
6333no option http-ignore-probes
6334 Enable or disable logging of null connections and request timeouts
6335 May be used in sections : defaults | frontend | listen | backend
6336 yes | yes | yes | no
6337 Arguments : none
6338
6339 Recently some browsers started to implement a "pre-connect" feature
6340 consisting in speculatively connecting to some recently visited web sites
6341 just in case the user would like to visit them. This results in many
6342 connections being established to web sites, which end up in 408 Request
6343 Timeout if the timeout strikes first, or 400 Bad Request when the browser
6344 decides to close them first. These ones pollute the log and feed the error
6345 counters. There was already "option dontlognull" but it's insufficient in
6346 this case. Instead, this option does the following things :
6347 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01006348 was received over a connection before it was closed;
6349 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02006350 - prevent any error counter from being incremented
6351
6352 That way the empty connection is silently ignored. Note that it is better
6353 not to use this unless it is clear that it is needed, because it will hide
6354 real problems. The most common reason for not receiving a request and seeing
6355 a 408 is due to an MTU inconsistency between the client and an intermediary
6356 element such as a VPN, which blocks too large packets. These issues are
6357 generally seen with POST requests as well as GET with large cookies. The logs
6358 are often the only way to detect them.
6359
6360 If this option has been enabled in a "defaults" section, it can be disabled
6361 in a specific instance by prepending the "no" keyword before it.
6362
6363 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
6364
6365
Willy Tarreau16bfb022010-01-16 19:48:41 +01006366option http-keep-alive
6367no option http-keep-alive
6368 Enable or disable HTTP keep-alive from client to server
6369 May be used in sections : defaults | frontend | listen | backend
6370 yes | yes | yes | yes
6371 Arguments : none
6372
Willy Tarreau70dffda2014-01-30 03:07:23 +01006373 By default HAProxy operates in keep-alive mode with regards to persistent
6374 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02006375 leaves the connection idle on both sides between the end of a response and
6376 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02006377 as "option http-server-close" or "option httpclose". This option allows to
6378 set back the keep-alive mode, which can be useful when another mode was used
6379 in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01006380
6381 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
6382 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01006383 network) and the fastest session reuse on the server side at the expense
6384 of maintaining idle connections to the servers. In general, it is possible
6385 with this option to achieve approximately twice the request rate that the
6386 "http-server-close" option achieves on small objects. There are mainly two
6387 situations where this option may be useful :
6388
6389 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01006390 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01006391
6392 - when the cost of establishing the connection to the server is significant
6393 compared to the cost of retrieving the associated object from the server.
6394
6395 This last case can happen when the server is a fast static server of cache.
6396 In this case, the server will need to be properly tuned to support high enough
6397 connection counts because connections will last until the client sends another
6398 request.
6399
6400 If the client request has to go to another backend or another server due to
6401 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01006402 immediately be closed and a new one re-opened. Option "prefer-last-server" is
6403 available to try optimize server selection so that if the server currently
6404 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01006405
Willy Tarreau16bfb022010-01-16 19:48:41 +01006406 At the moment, logs will not indicate whether requests came from the same
6407 session or not. The accept date reported in the logs corresponds to the end
6408 of the previous request, and the request time corresponds to the time spent
6409 waiting for a new request. The keep-alive request time is still bound to the
6410 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6411 not set.
6412
Christopher Faulet159e6672019-07-16 15:09:52 +02006413 This option disables and replaces any previous "option httpclose" or "option
6414 http-server-close". When backend and frontend options differ, all of these 4
6415 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006416
Christopher Faulet315b39c2018-09-21 16:26:19 +02006417 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01006418 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01006419 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006420
6421
Willy Tarreau96e31212011-05-30 18:10:30 +02006422option http-no-delay
6423no option http-no-delay
6424 Instruct the system to favor low interactive delays over performance in HTTP
6425 May be used in sections : defaults | frontend | listen | backend
6426 yes | yes | yes | yes
6427 Arguments : none
6428
6429 In HTTP, each payload is unidirectional and has no notion of interactivity.
6430 Any agent is expected to queue data somewhat for a reasonably low delay.
6431 There are some very rare server-to-server applications that abuse the HTTP
6432 protocol and expect the payload phase to be highly interactive, with many
6433 interleaved data chunks in both directions within a single request. This is
6434 absolutely not supported by the HTTP specification and will not work across
6435 most proxies or servers. When such applications attempt to do this through
6436 haproxy, it works but they will experience high delays due to the network
6437 optimizations which favor performance by instructing the system to wait for
6438 enough data to be available in order to only send full packets. Typical
6439 delays are around 200 ms per round trip. Note that this only happens with
6440 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
6441 affected.
6442
6443 When "option http-no-delay" is present in either the frontend or the backend
6444 used by a connection, all such optimizations will be disabled in order to
6445 make the exchanges as fast as possible. Of course this offers no guarantee on
6446 the functionality, as it may break at any other place. But if it works via
6447 HAProxy, it will work as fast as possible. This option should never be used
6448 by default, and should never be used at all unless such a buggy application
6449 is discovered. The impact of using this option is an increase of bandwidth
6450 usage and CPU usage, which may significantly lower performance in high
6451 latency environments.
6452
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006453 See also : "option http-buffer-request"
6454
Willy Tarreau96e31212011-05-30 18:10:30 +02006455
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006456option http-pretend-keepalive
6457no option http-pretend-keepalive
6458 Define whether haproxy will announce keepalive to the server or not
6459 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02006460 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006461 Arguments : none
6462
Christopher Faulet315b39c2018-09-21 16:26:19 +02006463 When running with "option http-server-close" or "option httpclose", haproxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006464 adds a "Connection: close" header to the request forwarded to the server.
6465 Unfortunately, when some servers see this header, they automatically refrain
6466 from using the chunked encoding for responses of unknown length, while this
6467 is totally unrelated. The immediate effect is that this prevents haproxy from
6468 maintaining the client connection alive. A second effect is that a client or
6469 a cache could receive an incomplete response without being aware of it, and
6470 consider the response complete.
6471
6472 By setting "option http-pretend-keepalive", haproxy will make the server
6473 believe it will keep the connection alive. The server will then not fall back
6474 to the abnormal undesired above. When haproxy gets the whole response, it
6475 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02006476 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006477 connection is correctly closed on the server side.
6478
6479 It is recommended not to enable this option by default, because most servers
6480 will more efficiently close the connection themselves after the last packet,
6481 and release its buffers slightly earlier. Also, the added packet on the
6482 network could slightly reduce the overall peak performance. However it is
6483 worth noting that when this option is enabled, haproxy will have slightly
6484 less work to do. So if haproxy is the bottleneck on the whole architecture,
6485 enabling this option might save a few CPU cycles.
6486
Christopher Faulet98db9762018-09-21 10:25:19 +02006487 This option may be set in backend and listen sections. Using it in a frontend
6488 section will be ignored and a warning will be reported during startup. It is
6489 a backend related option, so there is no real reason to set it on a
6490 frontend. This option may be combined with "option httpclose", which will
6491 cause keepalive to be announced to the server and close to be announced to
6492 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006493
6494 If this option has been enabled in a "defaults" section, it can be disabled
6495 in a specific instance by prepending the "no" keyword before it.
6496
Christopher Faulet315b39c2018-09-21 16:26:19 +02006497 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01006498 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006499
Willy Tarreauc27debf2008-01-06 08:57:02 +01006500
Willy Tarreaub608feb2010-01-02 22:47:18 +01006501option http-server-close
6502no option http-server-close
6503 Enable or disable HTTP connection closing on the server side
6504 May be used in sections : defaults | frontend | listen | backend
6505 yes | yes | yes | yes
6506 Arguments : none
6507
Willy Tarreau70dffda2014-01-30 03:07:23 +01006508 By default HAProxy operates in keep-alive mode with regards to persistent
6509 connections: for each connection it processes each request and response, and
6510 leaves the connection idle on both sides between the end of a response and
6511 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02006512 as "option http-server-close" or "option httpclose". Setting "option
6513 http-server-close" enables HTTP connection-close mode on the server side
6514 while keeping the ability to support HTTP keep-alive and pipelining on the
6515 client side. This provides the lowest latency on the client side (slow
6516 network) and the fastest session reuse on the server side to save server
6517 resources, similarly to "option httpclose". It also permits non-keepalive
6518 capable servers to be served in keep-alive mode to the clients if they
6519 conform to the requirements of RFC7230. Please note that some servers do not
6520 always conform to those requirements when they see "Connection: close" in the
6521 request. The effect will be that keep-alive will never be used. A workaround
6522 consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006523
6524 At the moment, logs will not indicate whether requests came from the same
6525 session or not. The accept date reported in the logs corresponds to the end
6526 of the previous request, and the request time corresponds to the time spent
6527 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01006528 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6529 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006530
6531 This option may be set both in a frontend and in a backend. It is enabled if
6532 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02006533 It disables and replaces any previous "option httpclose" or "option
6534 http-keep-alive". Please check section 4 ("Proxies") to see how this option
6535 combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006536
6537 If this option has been enabled in a "defaults" section, it can be disabled
6538 in a specific instance by prepending the "no" keyword before it.
6539
Christopher Faulet315b39c2018-09-21 16:26:19 +02006540 See also : "option httpclose", "option http-pretend-keepalive",
6541 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006542
Willy Tarreau88d349d2010-01-25 12:15:43 +01006543option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01006544no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01006545 Make use of non-standard Proxy-Connection header instead of Connection
6546 May be used in sections : defaults | frontend | listen | backend
6547 yes | yes | yes | no
6548 Arguments : none
6549
Lukas Tribus23953682017-04-28 13:24:30 +00006550 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01006551 Connection header to indicate their wish of persistent or non-persistent
6552 connections, both browsers and proxies ignore this header for proxied
6553 connections and make use of the undocumented, non-standard Proxy-Connection
6554 header instead. The issue begins when trying to put a load balancer between
6555 browsers and such proxies, because there will be a difference between what
6556 haproxy understands and what the client and the proxy agree on.
6557
6558 By setting this option in a frontend, haproxy can automatically switch to use
6559 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006560 defined here as one where the URI begins with neither a '/' nor a '*'. This
6561 is incompatible with the HTTP tunnel mode. Note that this option can only be
6562 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01006563
Willy Tarreau844a7e72010-01-31 21:46:18 +01006564 Also, when this option is set, a request which requires authentication will
6565 automatically switch to use proxy authentication headers if it is itself a
6566 proxied request. That makes it possible to check or enforce authentication in
6567 front of an existing proxy.
6568
Willy Tarreau88d349d2010-01-25 12:15:43 +01006569 This option should normally never be used, except in front of a proxy.
6570
Christopher Faulet315b39c2018-09-21 16:26:19 +02006571 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01006572
Willy Tarreaud63335a2010-02-26 12:56:52 +01006573option httpchk
6574option httpchk <uri>
6575option httpchk <method> <uri>
6576option httpchk <method> <uri> <version>
6577 Enable HTTP protocol to check on the servers health
6578 May be used in sections : defaults | frontend | listen | backend
6579 yes | no | yes | yes
6580 Arguments :
6581 <method> is the optional HTTP method used with the requests. When not set,
6582 the "OPTIONS" method is used, as it generally requires low server
6583 processing and is easy to filter out from the logs. Any method
6584 may be used, though it is not recommended to invent non-standard
6585 ones.
6586
6587 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
6588 which is accessible by default on almost any server, but may be
6589 changed to any other URI. Query strings are permitted.
6590
6591 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
6592 but some servers might behave incorrectly in HTTP 1.0, so turning
6593 it to HTTP/1.1 may sometimes help. Note that the Host field is
6594 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
6595 after "\r\n" following the version string.
6596
6597 By default, server health checks only consist in trying to establish a TCP
6598 connection. When "option httpchk" is specified, a complete HTTP request is
6599 sent once the TCP connection is established, and responses 2xx and 3xx are
6600 considered valid, while all other ones indicate a server failure, including
6601 the lack of any response.
6602
6603 The port and interval are specified in the server configuration.
6604
6605 This option does not necessarily require an HTTP backend, it also works with
6606 plain TCP backends. This is particularly useful to check simple scripts bound
6607 to some dedicated ports using the inetd daemon.
6608
6609 Examples :
6610 # Relay HTTPS traffic to Apache instance and check service availability
6611 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
6612 backend https_relay
6613 mode tcp
6614 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
6615 server apache1 192.168.1.1:443 check port 80
6616
Simon Hormanafc47ee2013-11-25 10:46:35 +09006617 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
6618 "option pgsql-check", "http-check" and the "check", "port" and
6619 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006620
6621
Willy Tarreauc27debf2008-01-06 08:57:02 +01006622option httpclose
6623no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02006624 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01006625 May be used in sections : defaults | frontend | listen | backend
6626 yes | yes | yes | yes
6627 Arguments : none
6628
Willy Tarreau70dffda2014-01-30 03:07:23 +01006629 By default HAProxy operates in keep-alive mode with regards to persistent
6630 connections: for each connection it processes each request and response, and
6631 leaves the connection idle on both sides between the end of a response and
6632 the start of a new request. This mode may be changed by several options such
Christopher Faulet159e6672019-07-16 15:09:52 +02006633 as "option http-server-close" or "option httpclose".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006634
Christopher Faulet315b39c2018-09-21 16:26:19 +02006635 If "option httpclose" is set, HAProxy will close connections with the server
6636 and the client as soon as the request and the response are received. It will
John Roeslerfb2fce12019-07-10 15:45:51 -05006637 also check if a "Connection: close" header is already set in each direction,
Christopher Faulet315b39c2018-09-21 16:26:19 +02006638 and will add one if missing. Any "Connection" header different from "close"
6639 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006640
Christopher Faulet315b39c2018-09-21 16:26:19 +02006641 This option may also be combined with "option http-pretend-keepalive", which
6642 will disable sending of the "Connection: close" header, but will still cause
6643 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006644
6645 This option may be set both in a frontend and in a backend. It is enabled if
6646 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet159e6672019-07-16 15:09:52 +02006647 It disables and replaces any previous "option http-server-close" or "option
6648 http-keep-alive". Please check section 4 ("Proxies") to see how this option
6649 combines with others when frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006650
6651 If this option has been enabled in a "defaults" section, it can be disabled
6652 in a specific instance by prepending the "no" keyword before it.
6653
Christopher Faulet315b39c2018-09-21 16:26:19 +02006654 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01006655
6656
Emeric Brun3a058f32009-06-30 18:26:00 +02006657option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006658 Enable logging of HTTP request, session state and timers
6659 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006660 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02006661 Arguments :
6662 clf if the "clf" argument is added, then the output format will be
6663 the CLF format instead of HAProxy's default HTTP format. You can
6664 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01006665 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02006666 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006667
6668 By default, the log output format is very poor, as it only contains the
6669 source and destination addresses, and the instance name. By specifying
6670 "option httplog", each log line turns into a much richer format including,
6671 but not limited to, the HTTP request, the connection timers, the session
6672 status, the connections numbers, the captured headers and cookies, the
6673 frontend, backend and server name, and of course the source address and
6674 ports.
6675
PiBa-NLbd556bf2014-12-11 21:31:54 +01006676 Specifying only "option httplog" will automatically clear the 'clf' mode
6677 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02006678
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006679 "option httplog" overrides any previous "log-format" directive.
6680
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006681 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006682
Willy Tarreau55165fe2009-05-10 12:02:55 +02006683
6684option http_proxy
6685no option http_proxy
6686 Enable or disable plain HTTP proxy mode
6687 May be used in sections : defaults | frontend | listen | backend
6688 yes | yes | yes | yes
6689 Arguments : none
6690
6691 It sometimes happens that people need a pure HTTP proxy which understands
6692 basic proxy requests without caching nor any fancy feature. In this case,
6693 it may be worth setting up an HAProxy instance with the "option http_proxy"
6694 set. In this mode, no server is declared, and the connection is forwarded to
6695 the IP address and port found in the URL after the "http://" scheme.
6696
6697 No host address resolution is performed, so this only works when pure IP
6698 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006699 it will probably be used only by experts who know they need exactly it. This
6700 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02006701
6702 If this option has been enabled in a "defaults" section, it can be disabled
6703 in a specific instance by prepending the "no" keyword before it.
6704
6705 Example :
6706 # this backend understands HTTP proxy requests and forwards them directly.
6707 backend direct_forward
6708 option httpclose
6709 option http_proxy
6710
6711 See also : "option httpclose"
6712
Willy Tarreau211ad242009-10-03 21:45:07 +02006713
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006714option independent-streams
6715no option independent-streams
6716 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006717 May be used in sections : defaults | frontend | listen | backend
6718 yes | yes | yes | yes
6719 Arguments : none
6720
6721 By default, when data is sent over a socket, both the write timeout and the
6722 read timeout for that socket are refreshed, because we consider that there is
6723 activity on that socket, and we have no other means of guessing if we should
6724 receive data or not.
6725
Davor Ocelice9ed2812017-12-25 17:49:28 +01006726 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006727 exists a situation where it is desirable to disable it, and only refresh the
6728 read timeout if there are incoming data. This happens on sessions with large
6729 timeouts and low amounts of exchanged data such as telnet session. If the
6730 server suddenly disappears, the output data accumulates in the system's
6731 socket buffers, both timeouts are correctly refreshed, and there is no way
6732 to know the server does not receive them, so we don't timeout. However, when
6733 the underlying protocol always echoes sent data, it would be enough by itself
6734 to detect the issue using the read timeout. Note that this problem does not
6735 happen with more verbose protocols because data won't accumulate long in the
6736 socket buffers.
6737
6738 When this option is set on the frontend, it will disable read timeout updates
6739 on data sent to the client. There probably is little use of this case. When
6740 the option is set on the backend, it will disable read timeout updates on
6741 data sent to the server. Doing so will typically break large HTTP posts from
6742 slow lines, so use it with caution.
6743
Willy Tarreauce887fd2012-05-12 12:50:00 +02006744 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006745
6746
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02006747option ldap-check
6748 Use LDAPv3 health checks for server testing
6749 May be used in sections : defaults | frontend | listen | backend
6750 yes | no | yes | yes
6751 Arguments : none
6752
6753 It is possible to test that the server correctly talks LDAPv3 instead of just
6754 testing that it accepts the TCP connection. When this option is set, an
6755 LDAPv3 anonymous simple bind message is sent to the server, and the response
6756 is analyzed to find an LDAPv3 bind response message.
6757
6758 The server is considered valid only when the LDAP response contains success
6759 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
6760
6761 Logging of bind requests is server dependent see your documentation how to
6762 configure it.
6763
6764 Example :
6765 option ldap-check
6766
6767 See also : "option httpchk"
6768
6769
Simon Horman98637e52014-06-20 12:30:16 +09006770option external-check
6771 Use external processes for server health checks
6772 May be used in sections : defaults | frontend | listen | backend
6773 yes | no | yes | yes
6774
6775 It is possible to test the health of a server using an external command.
6776 This is achieved by running the executable set using "external-check
6777 command".
6778
6779 Requires the "external-check" global to be set.
6780
6781 See also : "external-check", "external-check command", "external-check path"
6782
6783
Willy Tarreau211ad242009-10-03 21:45:07 +02006784option log-health-checks
6785no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02006786 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02006787 May be used in sections : defaults | frontend | listen | backend
6788 yes | no | yes | yes
6789 Arguments : none
6790
Willy Tarreaubef1b322014-05-13 21:01:39 +02006791 By default, failed health check are logged if server is UP and successful
6792 health checks are logged if server is DOWN, so the amount of additional
6793 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02006794
Willy Tarreaubef1b322014-05-13 21:01:39 +02006795 When this option is enabled, any change of the health check status or to
6796 the server's health will be logged, so that it becomes possible to know
6797 that a server was failing occasional checks before crashing, or exactly when
6798 it failed to respond a valid HTTP status, then when the port started to
6799 reject connections, then when the server stopped responding at all.
6800
Davor Ocelice9ed2812017-12-25 17:49:28 +01006801 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02006802 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02006803
Willy Tarreaubef1b322014-05-13 21:01:39 +02006804 See also: "option httpchk", "option ldap-check", "option mysql-check",
6805 "option pgsql-check", "option redis-check", "option smtpchk",
6806 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02006807
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006808
6809option log-separate-errors
6810no option log-separate-errors
6811 Change log level for non-completely successful connections
6812 May be used in sections : defaults | frontend | listen | backend
6813 yes | yes | yes | no
6814 Arguments : none
6815
6816 Sometimes looking for errors in logs is not easy. This option makes haproxy
6817 raise the level of logs containing potentially interesting information such
6818 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
6819 level changes from "info" to "err". This makes it possible to log them
6820 separately to a different file with most syslog daemons. Be careful not to
6821 remove them from the original file, otherwise you would lose ordering which
6822 provides very important information.
6823
6824 Using this option, large sites dealing with several thousand connections per
6825 second may log normal traffic to a rotating buffer and only archive smaller
6826 error logs.
6827
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006828 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006829 logging.
6830
Willy Tarreauc27debf2008-01-06 08:57:02 +01006831
6832option logasap
6833no option logasap
6834 Enable or disable early logging of HTTP requests
6835 May be used in sections : defaults | frontend | listen | backend
6836 yes | yes | yes | no
6837 Arguments : none
6838
6839 By default, HTTP requests are logged upon termination so that the total
6840 transfer time and the number of bytes appear in the logs. When large objects
6841 are being transferred, it may take a while before the request appears in the
6842 logs. Using "option logasap", the request gets logged as soon as the server
6843 sends the complete headers. The only missing information in the logs will be
6844 the total number of bytes which will indicate everything except the amount
6845 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006846 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01006847 "Content-Length" response header so that the logs at least indicate how many
6848 bytes are expected to be transferred.
6849
Willy Tarreaucc6c8912009-02-22 10:53:55 +01006850 Examples :
6851 listen http_proxy 0.0.0.0:80
6852 mode http
6853 option httplog
6854 option logasap
6855 log 192.168.2.200 local3
6856
6857 >>> Feb 6 12:14:14 localhost \
6858 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
6859 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
6860 "GET /image.iso HTTP/1.0"
6861
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006862 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01006863 logging.
6864
6865
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006866option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006867 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006868 May be used in sections : defaults | frontend | listen | backend
6869 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006870 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006871 <username> This is the username which will be used when connecting to MySQL
6872 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006873 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006874
6875 If you specify a username, the check consists of sending two MySQL packet,
6876 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01006877 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006878 Error packet. It is a basic but useful test which does not produce error nor
6879 aborted connect on the server. However, it requires adding an authorization
6880 in the MySQL table, like this :
6881
6882 USE mysql;
6883 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
6884 FLUSH PRIVILEGES;
6885
6886 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006887 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006888 Error packet, we don't send anything in this mode. It was reported that it
6889 can generate lockout if check is too frequent and/or if there is not enough
6890 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
6891 value as if a connection is established successfully within fewer than MySQL
6892 "max_connect_errors" attempts after a previous connection was interrupted,
6893 the error count for the host is cleared to zero. If HAProxy's server get
6894 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
6895
6896 Remember that this does not check database presence nor database consistency.
6897 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006898
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02006899 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006900
6901 Most often, an incoming MySQL server needs to see the client's IP address for
6902 various purposes, including IP privilege matching and connection logging.
6903 When possible, it is often wise to masquerade the client's IP address when
6904 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006905 which requires the transparent proxy feature to be compiled in, and the MySQL
6906 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006907
6908 See also: "option httpchk"
6909
6910
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006911option nolinger
6912no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006913 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006914 May be used in sections: defaults | frontend | listen | backend
6915 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006916 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006917
Davor Ocelice9ed2812017-12-25 17:49:28 +01006918 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006919 physically disconnected), the session timeouts triggers and the session is
6920 closed. But it will remain in FIN_WAIT1 state for some time in the system,
6921 using some resources and possibly limiting the ability to establish newer
6922 connections.
6923
6924 When this happens, it is possible to activate "option nolinger" which forces
6925 the system to immediately remove any socket's pending data on close. Thus,
6926 the session is instantly purged from the system's tables. This usually has
6927 side effects such as increased number of TCP resets due to old retransmits
6928 getting immediately rejected. Some firewalls may sometimes complain about
6929 this too.
6930
6931 For this reason, it is not recommended to use this option when not absolutely
6932 needed. You know that you need it when you have thousands of FIN_WAIT1
6933 sessions on your system (TIME_WAIT ones do not count).
6934
6935 This option may be used both on frontends and backends, depending on the side
6936 where it is required. Use it on the frontend for clients, and on the backend
6937 for servers.
6938
6939 If this option has been enabled in a "defaults" section, it can be disabled
6940 in a specific instance by prepending the "no" keyword before it.
6941
6942
Willy Tarreau55165fe2009-05-10 12:02:55 +02006943option originalto [ except <network> ] [ header <name> ]
6944 Enable insertion of the X-Original-To header to requests sent to servers
6945 May be used in sections : defaults | frontend | listen | backend
6946 yes | yes | yes | yes
6947 Arguments :
6948 <network> is an optional argument used to disable this option for sources
6949 matching <network>
6950 <name> an optional argument to specify a different "X-Original-To"
6951 header name.
6952
6953 Since HAProxy can work in transparent mode, every request from a client can
6954 be redirected to the proxy and HAProxy itself can proxy every request to a
6955 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
6956 be lost. This is annoying when you want access rules based on destination ip
6957 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
6958 added by HAProxy to all requests sent to the server. This header contains a
6959 value representing the original destination IP address. Since this must be
6960 configured to always use the last occurrence of this header only. Note that
6961 only the last occurrence of the header must be used, since it is really
6962 possible that the client has already brought one.
6963
6964 The keyword "header" may be used to supply a different header name to replace
6965 the default "X-Original-To". This can be useful where you might already
6966 have a "X-Original-To" header from a different application, and you need
6967 preserve it. Also if your backend server doesn't use the "X-Original-To"
6968 header and requires different one.
6969
6970 Sometimes, a same HAProxy instance may be shared between a direct client
6971 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6972 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6973 header for a known source address or network by adding the "except" keyword
6974 followed by the network address. In this case, any source IP matching the
6975 network will not cause an addition of this header. Most common uses are with
6976 private networks or 127.0.0.1.
6977
6978 This option may be specified either in the frontend or in the backend. If at
6979 least one of them uses it, the header will be added. Note that the backend's
6980 setting of the header subargument takes precedence over the frontend's if
6981 both are defined.
6982
Willy Tarreau55165fe2009-05-10 12:02:55 +02006983 Examples :
6984 # Original Destination address
6985 frontend www
6986 mode http
6987 option originalto except 127.0.0.1
6988
6989 # Those servers want the IP Address in X-Client-Dst
6990 backend www
6991 mode http
6992 option originalto header X-Client-Dst
6993
Christopher Faulet315b39c2018-09-21 16:26:19 +02006994 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02006995
6996
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006997option persist
6998no option persist
6999 Enable or disable forced persistence on down servers
7000 May be used in sections: defaults | frontend | listen | backend
7001 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007002 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007003
7004 When an HTTP request reaches a backend with a cookie which references a dead
7005 server, by default it is redispatched to another server. It is possible to
7006 force the request to be sent to the dead server first using "option persist"
7007 if absolutely needed. A common use case is when servers are under extreme
7008 load and spend their time flapping. In this case, the users would still be
7009 directed to the server they opened the session on, in the hope they would be
7010 correctly served. It is recommended to use "option redispatch" in conjunction
7011 with this option so that in the event it would not be possible to connect to
7012 the server at all (server definitely dead), the client would finally be
7013 redirected to another valid server.
7014
7015 If this option has been enabled in a "defaults" section, it can be disabled
7016 in a specific instance by prepending the "no" keyword before it.
7017
Willy Tarreau4de91492010-01-22 19:10:05 +01007018 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007019
7020
Willy Tarreau0c122822013-12-15 18:49:01 +01007021option pgsql-check [ user <username> ]
7022 Use PostgreSQL health checks for server testing
7023 May be used in sections : defaults | frontend | listen | backend
7024 yes | no | yes | yes
7025 Arguments :
7026 <username> This is the username which will be used when connecting to
7027 PostgreSQL server.
7028
7029 The check sends a PostgreSQL StartupMessage and waits for either
7030 Authentication request or ErrorResponse message. It is a basic but useful
7031 test which does not produce error nor aborted connect on the server.
7032 This check is identical with the "mysql-check".
7033
7034 See also: "option httpchk"
7035
7036
Willy Tarreau9420b122013-12-15 18:58:25 +01007037option prefer-last-server
7038no option prefer-last-server
7039 Allow multiple load balanced requests to remain on the same server
7040 May be used in sections: defaults | frontend | listen | backend
7041 yes | no | yes | yes
7042 Arguments : none
7043
7044 When the load balancing algorithm in use is not deterministic, and a previous
7045 request was sent to a server to which haproxy still holds a connection, it is
7046 sometimes desirable that subsequent requests on a same session go to the same
7047 server as much as possible. Note that this is different from persistence, as
7048 we only indicate a preference which haproxy tries to apply without any form
7049 of warranty. The real use is for keep-alive connections sent to servers. When
7050 this option is used, haproxy will try to reuse the same connection that is
7051 attached to the server instead of rebalancing to another server, causing a
7052 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01007053 not make much sense to use this in combination with hashing algorithms. Note,
7054 haproxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02007055 to a proxy which sends a 407 (authentication required), when the load
7056 balancing algorithm is not deterministic. This is mandatory for use with the
7057 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01007058 troubleshooting some faulty applications. Option prefer-last-server might be
7059 desirable in these environments as well, to avoid redistributing the traffic
7060 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01007061
7062 If this option has been enabled in a "defaults" section, it can be disabled
7063 in a specific instance by prepending the "no" keyword before it.
7064
7065 See also: "option http-keep-alive"
7066
7067
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007068option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07007069option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007070no option redispatch
7071 Enable or disable session redistribution in case of connection failure
7072 May be used in sections: defaults | frontend | listen | backend
7073 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07007074 Arguments :
7075 <interval> The optional integer value that controls how often redispatches
7076 occur when retrying connections. Positive value P indicates a
7077 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01007078 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07007079 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007080 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07007081 positive value of 1 would indicate a redispatch on every retry,
7082 and a positive value of 3 would indicate a redispatch on every
7083 third retry. You can disable redispatches with a value of 0.
7084
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007085
7086 In HTTP mode, if a server designated by a cookie is down, clients may
7087 definitely stick to it because they cannot flush the cookie, so they will not
7088 be able to access the service anymore.
7089
Willy Tarreau59884a62019-01-02 14:48:31 +01007090 Specifying "option redispatch" will allow the proxy to break cookie or
7091 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007092
Joseph Lynch726ab712015-05-11 23:25:34 -07007093 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007094 connection failures. Of course, it requires having "retries" set to a nonzero
7095 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007096
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007097 If this option has been enabled in a "defaults" section, it can be disabled
7098 in a specific instance by prepending the "no" keyword before it.
7099
Christopher Faulet87f1f3d2019-07-18 14:51:20 +02007100 See also : "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007101
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007102
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007103option redis-check
7104 Use redis health checks for server testing
7105 May be used in sections : defaults | frontend | listen | backend
7106 yes | no | yes | yes
7107 Arguments : none
7108
7109 It is possible to test that the server correctly talks REDIS protocol instead
7110 of just testing that it accepts the TCP connection. When this option is set,
7111 a PING redis command is sent to the server, and the response is analyzed to
7112 find the "+PONG" response message.
7113
7114 Example :
7115 option redis-check
7116
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007117 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007118
7119
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007120option smtpchk
7121option smtpchk <hello> <domain>
7122 Use SMTP health checks for server testing
7123 May be used in sections : defaults | frontend | listen | backend
7124 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01007125 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007126 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02007127 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007128 values will be turned into the default command ("HELO").
7129
7130 <domain> is the domain name to present to the server. It may only be
7131 specified (and is mandatory) if the hello command has been
7132 specified. By default, "localhost" is used.
7133
7134 When "option smtpchk" is set, the health checks will consist in TCP
7135 connections followed by an SMTP command. By default, this command is
7136 "HELO localhost". The server's return code is analyzed and only return codes
7137 starting with a "2" will be considered as valid. All other responses,
7138 including a lack of response will constitute an error and will indicate a
7139 dead server.
7140
7141 This test is meant to be used with SMTP servers or relays. Depending on the
7142 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01007143 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007144 25 is often easier than adjusting the configuration.
7145
7146 Most often, an incoming SMTP server needs to see the client's IP address for
7147 various purposes, including spam filtering, anti-spoofing and logging. When
7148 possible, it is often wise to masquerade the client's IP address when
7149 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007150 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007151
7152 Example :
7153 option smtpchk HELO mydomain.org
7154
7155 See also : "option httpchk", "source"
7156
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007157
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02007158option socket-stats
7159no option socket-stats
7160
7161 Enable or disable collecting & providing separate statistics for each socket.
7162 May be used in sections : defaults | frontend | listen | backend
7163 yes | yes | yes | no
7164
7165 Arguments : none
7166
7167
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007168option splice-auto
7169no option splice-auto
7170 Enable or disable automatic kernel acceleration on sockets in both directions
7171 May be used in sections : defaults | frontend | listen | backend
7172 yes | yes | yes | yes
7173 Arguments : none
7174
7175 When this option is enabled either on a frontend or on a backend, haproxy
7176 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007177 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007178 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007179 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007180 are not much aggressive in order to limit excessive use of splicing. This
7181 option requires splicing to be enabled at compile time, and may be globally
7182 disabled with the global option "nosplice". Since splice uses pipes, using it
7183 requires that there are enough spare pipes.
7184
7185 Important note: kernel-based TCP splicing is a Linux-specific feature which
7186 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
7187 transfer data between sockets without copying these data to user-space, thus
7188 providing noticeable performance gains and CPU cycles savings. Since many
7189 early implementations are buggy, corrupt data and/or are inefficient, this
7190 feature is not enabled by default, and it should be used with extreme care.
7191 While it is not possible to detect the correctness of an implementation,
7192 2.6.29 is the first version offering a properly working implementation. In
7193 case of doubt, splicing may be globally disabled using the global "nosplice"
7194 keyword.
7195
7196 Example :
7197 option splice-auto
7198
7199 If this option has been enabled in a "defaults" section, it can be disabled
7200 in a specific instance by prepending the "no" keyword before it.
7201
7202 See also : "option splice-request", "option splice-response", and global
7203 options "nosplice" and "maxpipes"
7204
7205
7206option splice-request
7207no option splice-request
7208 Enable or disable automatic kernel acceleration on sockets for requests
7209 May be used in sections : defaults | frontend | listen | backend
7210 yes | yes | yes | yes
7211 Arguments : none
7212
7213 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007214 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007215 the client to the server. It might still use the recv/send scheme if there
7216 are no spare pipes left. This option requires splicing to be enabled at
7217 compile time, and may be globally disabled with the global option "nosplice".
7218 Since splice uses pipes, using it requires that there are enough spare pipes.
7219
7220 Important note: see "option splice-auto" for usage limitations.
7221
7222 Example :
7223 option splice-request
7224
7225 If this option has been enabled in a "defaults" section, it can be disabled
7226 in a specific instance by prepending the "no" keyword before it.
7227
7228 See also : "option splice-auto", "option splice-response", and global options
7229 "nosplice" and "maxpipes"
7230
7231
7232option splice-response
7233no option splice-response
7234 Enable or disable automatic kernel acceleration on sockets for responses
7235 May be used in sections : defaults | frontend | listen | backend
7236 yes | yes | yes | yes
7237 Arguments : none
7238
7239 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007240 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007241 the server to the client. It might still use the recv/send scheme if there
7242 are no spare pipes left. This option requires splicing to be enabled at
7243 compile time, and may be globally disabled with the global option "nosplice".
7244 Since splice uses pipes, using it requires that there are enough spare pipes.
7245
7246 Important note: see "option splice-auto" for usage limitations.
7247
7248 Example :
7249 option splice-response
7250
7251 If this option has been enabled in a "defaults" section, it can be disabled
7252 in a specific instance by prepending the "no" keyword before it.
7253
7254 See also : "option splice-auto", "option splice-request", and global options
7255 "nosplice" and "maxpipes"
7256
7257
Christopher Fauletba7bc162016-11-07 21:07:38 +01007258option spop-check
7259 Use SPOP health checks for server testing
7260 May be used in sections : defaults | frontend | listen | backend
7261 no | no | no | yes
7262 Arguments : none
7263
7264 It is possible to test that the server correctly talks SPOP protocol instead
7265 of just testing that it accepts the TCP connection. When this option is set,
7266 a HELLO handshake is performed between HAProxy and the server, and the
7267 response is analyzed to check no error is reported.
7268
7269 Example :
7270 option spop-check
7271
7272 See also : "option httpchk"
7273
7274
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007275option srvtcpka
7276no option srvtcpka
7277 Enable or disable the sending of TCP keepalive packets on the server side
7278 May be used in sections : defaults | frontend | listen | backend
7279 yes | no | yes | yes
7280 Arguments : none
7281
7282 When there is a firewall or any session-aware component between a client and
7283 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007284 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007285 components decides to expire a session which has remained idle for too long.
7286
7287 Enabling socket-level TCP keep-alives makes the system regularly send packets
7288 to the other end of the connection, leaving it active. The delay between
7289 keep-alive probes is controlled by the system only and depends both on the
7290 operating system and its tuning parameters.
7291
7292 It is important to understand that keep-alive packets are neither emitted nor
7293 received at the application level. It is only the network stacks which sees
7294 them. For this reason, even if one side of the proxy already uses keep-alives
7295 to maintain its connection alive, those keep-alive packets will not be
7296 forwarded to the other side of the proxy.
7297
7298 Please note that this has nothing to do with HTTP keep-alive.
7299
7300 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
7301 server side of a connection, which should help when session expirations are
7302 noticed between HAProxy and a server.
7303
7304 If this option has been enabled in a "defaults" section, it can be disabled
7305 in a specific instance by prepending the "no" keyword before it.
7306
7307 See also : "option clitcpka", "option tcpka"
7308
7309
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007310option ssl-hello-chk
7311 Use SSLv3 client hello health checks for server testing
7312 May be used in sections : defaults | frontend | listen | backend
7313 yes | no | yes | yes
7314 Arguments : none
7315
7316 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
7317 possible to test that the server correctly talks SSL instead of just testing
7318 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
7319 SSLv3 client hello messages are sent once the connection is established to
7320 the server, and the response is analyzed to find an SSL server hello message.
7321 The server is considered valid only when the response contains this server
7322 hello message.
7323
7324 All servers tested till there correctly reply to SSLv3 client hello messages,
7325 and most servers tested do not even log the requests containing only hello
7326 messages, which is appreciable.
7327
Willy Tarreau763a95b2012-10-04 23:15:39 +02007328 Note that this check works even when SSL support was not built into haproxy
7329 because it forges the SSL message. When SSL support is available, it is best
7330 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007331
Willy Tarreau763a95b2012-10-04 23:15:39 +02007332 See also: "option httpchk", "check-ssl"
7333
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007334
Willy Tarreaued179852013-12-16 01:07:00 +01007335option tcp-check
7336 Perform health checks using tcp-check send/expect sequences
7337 May be used in sections: defaults | frontend | listen | backend
7338 yes | no | yes | yes
7339
7340 This health check method is intended to be combined with "tcp-check" command
7341 lists in order to support send/expect types of health check sequences.
7342
7343 TCP checks currently support 4 modes of operations :
7344 - no "tcp-check" directive : the health check only consists in a connection
7345 attempt, which remains the default mode.
7346
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007347 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01007348 used to send a string along with a connection opening. With some
7349 protocols, it helps sending a "QUIT" message for example that prevents
7350 the server from logging a connection error for each health check. The
7351 check result will still be based on the ability to open the connection
7352 only.
7353
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007354 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01007355 The connection is opened and haproxy waits for the server to present some
7356 contents which must validate some rules. The check result will be based
7357 on the matching between the contents and the rules. This is suited for
7358 POP, IMAP, SMTP, FTP, SSH, TELNET.
7359
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007360 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007361 used to test a hello-type protocol. HAProxy sends a message, the server
7362 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007363 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01007364 suited for protocols which require a binding or a request/response model.
7365 LDAP, MySQL, Redis and SSL are example of such protocols, though they
7366 already all have their dedicated checks with a deeper understanding of
7367 the respective protocols.
7368 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01007369 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01007370
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007371 A fifth mode can be used to insert comments in different steps of the
7372 script.
7373
7374 For each tcp-check rule you create, you can add a "comment" directive,
7375 followed by a string. This string will be reported in the log and stderr
7376 in debug mode. It is useful to make user-friendly error reporting.
7377 The "comment" is of course optional.
7378
7379
Willy Tarreaued179852013-12-16 01:07:00 +01007380 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007381 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007382 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007383 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007384
Davor Ocelice9ed2812017-12-25 17:49:28 +01007385 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007386 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007387 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007388
7389 # look for the redis master server after ensuring it speaks well
7390 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01007391 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01007392 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007393 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007394 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02007395 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007396 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01007397 tcp-check send info\ replication\r\n
7398 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007399 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007400 tcp-check send QUIT\r\n
7401 tcp-check expect string +OK
7402
Davor Ocelice9ed2812017-12-25 17:49:28 +01007403 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01007404 (send many headers before analyzing)
7405 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007406 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01007407 tcp-check send HEAD\ /\ HTTP/1.1\r\n
7408 tcp-check send Host:\ www.mydomain.com\r\n
7409 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
7410 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007411 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01007412
7413
7414 See also : "tcp-check expect", "tcp-check send"
7415
7416
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007417option tcp-smart-accept
7418no option tcp-smart-accept
7419 Enable or disable the saving of one ACK packet during the accept sequence
7420 May be used in sections : defaults | frontend | listen | backend
7421 yes | yes | yes | no
7422 Arguments : none
7423
7424 When an HTTP connection request comes in, the system acknowledges it on
7425 behalf of HAProxy, then the client immediately sends its request, and the
7426 system acknowledges it too while it is notifying HAProxy about the new
7427 connection. HAProxy then reads the request and responds. This means that we
7428 have one TCP ACK sent by the system for nothing, because the request could
7429 very well be acknowledged by HAProxy when it sends its response.
7430
7431 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
7432 sending this useless ACK on platforms which support it (currently at least
7433 Linux). It must not cause any problem, because the system will send it anyway
7434 after 40 ms if the response takes more time than expected to come.
7435
7436 During complex network debugging sessions, it may be desirable to disable
7437 this optimization because delayed ACKs can make troubleshooting more complex
7438 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007439 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007440
7441 It is also possible to force it for non-HTTP proxies by simply specifying
7442 "option tcp-smart-accept". For instance, it can make sense with some services
7443 such as SMTP where the server speaks first.
7444
7445 It is recommended to avoid forcing this option in a defaults section. In case
7446 of doubt, consider setting it back to automatic values by prepending the
7447 "default" keyword before it, or disabling it using the "no" keyword.
7448
Willy Tarreaud88edf22009-06-14 15:48:17 +02007449 See also : "option tcp-smart-connect"
7450
7451
7452option tcp-smart-connect
7453no option tcp-smart-connect
7454 Enable or disable the saving of one ACK packet during the connect sequence
7455 May be used in sections : defaults | frontend | listen | backend
7456 yes | no | yes | yes
7457 Arguments : none
7458
7459 On certain systems (at least Linux), HAProxy can ask the kernel not to
7460 immediately send an empty ACK upon a connection request, but to directly
7461 send the buffer request instead. This saves one packet on the network and
7462 thus boosts performance. It can also be useful for some servers, because they
7463 immediately get the request along with the incoming connection.
7464
7465 This feature is enabled when "option tcp-smart-connect" is set in a backend.
7466 It is not enabled by default because it makes network troubleshooting more
7467 complex.
7468
7469 It only makes sense to enable it with protocols where the client speaks first
7470 such as HTTP. In other situations, if there is no data to send in place of
7471 the ACK, a normal ACK is sent.
7472
7473 If this option has been enabled in a "defaults" section, it can be disabled
7474 in a specific instance by prepending the "no" keyword before it.
7475
7476 See also : "option tcp-smart-accept"
7477
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007478
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007479option tcpka
7480 Enable or disable the sending of TCP keepalive packets on both sides
7481 May be used in sections : defaults | frontend | listen | backend
7482 yes | yes | yes | yes
7483 Arguments : none
7484
7485 When there is a firewall or any session-aware component between a client and
7486 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007487 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007488 components decides to expire a session which has remained idle for too long.
7489
7490 Enabling socket-level TCP keep-alives makes the system regularly send packets
7491 to the other end of the connection, leaving it active. The delay between
7492 keep-alive probes is controlled by the system only and depends both on the
7493 operating system and its tuning parameters.
7494
7495 It is important to understand that keep-alive packets are neither emitted nor
7496 received at the application level. It is only the network stacks which sees
7497 them. For this reason, even if one side of the proxy already uses keep-alives
7498 to maintain its connection alive, those keep-alive packets will not be
7499 forwarded to the other side of the proxy.
7500
7501 Please note that this has nothing to do with HTTP keep-alive.
7502
7503 Using option "tcpka" enables the emission of TCP keep-alive probes on both
7504 the client and server sides of a connection. Note that this is meaningful
7505 only in "defaults" or "listen" sections. If this option is used in a
7506 frontend, only the client side will get keep-alives, and if this option is
7507 used in a backend, only the server side will get keep-alives. For this
7508 reason, it is strongly recommended to explicitly use "option clitcpka" and
7509 "option srvtcpka" when the configuration is split between frontends and
7510 backends.
7511
7512 See also : "option clitcpka", "option srvtcpka"
7513
Willy Tarreau844e3c52008-01-11 16:28:18 +01007514
7515option tcplog
7516 Enable advanced logging of TCP connections with session state and timers
7517 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01007518 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01007519 Arguments : none
7520
7521 By default, the log output format is very poor, as it only contains the
7522 source and destination addresses, and the instance name. By specifying
7523 "option tcplog", each log line turns into a much richer format including, but
7524 not limited to, the connection timers, the session status, the connections
7525 numbers, the frontend, backend and server name, and of course the source
7526 address and ports. This option is useful for pure TCP proxies in order to
7527 find which of the client or server disconnects or times out. For normal HTTP
7528 proxies, it's better to use "option httplog" which is even more complete.
7529
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007530 "option tcplog" overrides any previous "log-format" directive.
7531
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007532 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007533
7534
Willy Tarreau844e3c52008-01-11 16:28:18 +01007535option transparent
7536no option transparent
7537 Enable client-side transparent proxying
7538 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007539 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007540 Arguments : none
7541
7542 This option was introduced in order to provide layer 7 persistence to layer 3
7543 load balancers. The idea is to use the OS's ability to redirect an incoming
7544 connection for a remote address to a local process (here HAProxy), and let
7545 this process know what address was initially requested. When this option is
7546 used, sessions without cookies will be forwarded to the original destination
7547 IP address of the incoming request (which should match that of another
7548 equipment), while requests with cookies will still be forwarded to the
7549 appropriate server.
7550
7551 Note that contrary to a common belief, this option does NOT make HAProxy
7552 present the client's IP to the server when establishing the connection.
7553
Willy Tarreaua1146052011-03-01 09:51:54 +01007554 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007555 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007556
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007557
Simon Horman98637e52014-06-20 12:30:16 +09007558external-check command <command>
7559 Executable to run when performing an external-check
7560 May be used in sections : defaults | frontend | listen | backend
7561 yes | no | yes | yes
7562
7563 Arguments :
7564 <command> is the external command to run
7565
Simon Horman98637e52014-06-20 12:30:16 +09007566 The arguments passed to the to the command are:
7567
Cyril Bonté777be862014-12-02 21:21:35 +01007568 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09007569
Cyril Bonté777be862014-12-02 21:21:35 +01007570 The <proxy_address> and <proxy_port> are derived from the first listener
7571 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
7572 listener the proxy_address will be the path of the socket and the
7573 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
7574 possible to determine a listener, and both <proxy_address> and <proxy_port>
7575 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09007576
Cyril Bonté72cda2a2014-12-27 22:28:39 +01007577 Some values are also provided through environment variables.
7578
7579 Environment variables :
7580 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
7581 applicable, for example in a "backend" section).
7582
7583 HAPROXY_PROXY_ID The backend id.
7584
7585 HAPROXY_PROXY_NAME The backend name.
7586
7587 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
7588 applicable, for example in a "backend" section or
7589 for a UNIX socket).
7590
7591 HAPROXY_SERVER_ADDR The server address.
7592
7593 HAPROXY_SERVER_CURCONN The current number of connections on the server.
7594
7595 HAPROXY_SERVER_ID The server id.
7596
7597 HAPROXY_SERVER_MAXCONN The server max connections.
7598
7599 HAPROXY_SERVER_NAME The server name.
7600
7601 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
7602 socket).
7603
7604 PATH The PATH environment variable used when executing
7605 the command may be set using "external-check path".
7606
William Lallemand4d03e432019-06-14 15:35:37 +02007607 See also "2.3. Environment variables" for other variables.
7608
Simon Horman98637e52014-06-20 12:30:16 +09007609 If the command executed and exits with a zero status then the check is
7610 considered to have passed, otherwise the check is considered to have
7611 failed.
7612
7613 Example :
7614 external-check command /bin/true
7615
7616 See also : "external-check", "option external-check", "external-check path"
7617
7618
7619external-check path <path>
7620 The value of the PATH environment variable used when running an external-check
7621 May be used in sections : defaults | frontend | listen | backend
7622 yes | no | yes | yes
7623
7624 Arguments :
7625 <path> is the path used when executing external command to run
7626
7627 The default path is "".
7628
7629 Example :
7630 external-check path "/usr/bin:/bin"
7631
7632 See also : "external-check", "option external-check",
7633 "external-check command"
7634
7635
Emeric Brun647caf12009-06-30 17:57:00 +02007636persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02007637persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02007638 Enable RDP cookie-based persistence
7639 May be used in sections : defaults | frontend | listen | backend
7640 yes | no | yes | yes
7641 Arguments :
7642 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02007643 default cookie name "msts" will be used. There currently is no
7644 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02007645
7646 This statement enables persistence based on an RDP cookie. The RDP cookie
7647 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01007648 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02007649 and if an RDP cookie is found, it is decoded. If it matches a known server
7650 which is still UP (or if "option persist" is set), then the connection is
7651 forwarded to this server.
7652
7653 Note that this only makes sense in a TCP backend, but for this to work, the
7654 frontend must have waited long enough to ensure that an RDP cookie is present
7655 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007656 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02007657 a single "listen" section.
7658
Willy Tarreau61e28f22010-05-16 22:31:05 +02007659 Also, it is important to understand that the terminal server will emit this
7660 RDP cookie only if it is configured for "token redirection mode", which means
7661 that the "IP address redirection" option is disabled.
7662
Emeric Brun647caf12009-06-30 17:57:00 +02007663 Example :
7664 listen tse-farm
7665 bind :3389
7666 # wait up to 5s for an RDP cookie in the request
7667 tcp-request inspect-delay 5s
7668 tcp-request content accept if RDP_COOKIE
7669 # apply RDP cookie persistence
7670 persist rdp-cookie
7671 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02007672 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02007673 balance rdp-cookie
7674 server srv1 1.1.1.1:3389
7675 server srv2 1.1.1.2:3389
7676
Simon Hormanab814e02011-06-24 14:50:20 +09007677 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
7678 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02007679
7680
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007681rate-limit sessions <rate>
7682 Set a limit on the number of new sessions accepted per second on a frontend
7683 May be used in sections : defaults | frontend | listen | backend
7684 yes | yes | yes | no
7685 Arguments :
7686 <rate> The <rate> parameter is an integer designating the maximum number
7687 of new sessions per second to accept on the frontend.
7688
7689 When the frontend reaches the specified number of new sessions per second, it
7690 stops accepting new connections until the rate drops below the limit again.
7691 During this time, the pending sessions will be kept in the socket's backlog
7692 (in system buffers) and haproxy will not even be aware that sessions are
7693 pending. When applying very low limit on a highly loaded service, it may make
7694 sense to increase the socket's backlog using the "backlog" keyword.
7695
7696 This feature is particularly efficient at blocking connection-based attacks
7697 or service abuse on fragile servers. Since the session rate is measured every
7698 millisecond, it is extremely accurate. Also, the limit applies immediately,
7699 no delay is needed at all to detect the threshold.
7700
7701 Example : limit the connection rate on SMTP to 10 per second max
7702 listen smtp
7703 mode tcp
7704 bind :25
7705 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02007706 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007707
Willy Tarreaua17c2d92011-07-25 08:16:20 +02007708 Note : when the maximum rate is reached, the frontend's status is not changed
7709 but its sockets appear as "WAITING" in the statistics if the
7710 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007711
7712 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
7713
7714
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007715redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
7716redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
7717redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007718 Return an HTTP redirection if/unless a condition is matched
7719 May be used in sections : defaults | frontend | listen | backend
7720 no | yes | yes | yes
7721
7722 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01007723 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007724
Willy Tarreau0140f252008-11-19 21:07:09 +01007725 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007726 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007727 the HTTP "Location" header. When used in an "http-request" rule,
7728 <loc> value follows the log-format rules and can include some
7729 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007730
7731 <pfx> With "redirect prefix", the "Location" header is built from the
7732 concatenation of <pfx> and the complete URI path, including the
7733 query string, unless the "drop-query" option is specified (see
7734 below). As a special case, if <pfx> equals exactly "/", then
7735 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007736 redirect to the same URL (for instance, to insert a cookie). When
7737 used in an "http-request" rule, <pfx> value follows the log-format
7738 rules and can include some dynamic values (see Custom Log Format
7739 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007740
7741 <sch> With "redirect scheme", then the "Location" header is built by
7742 concatenating <sch> with "://" then the first occurrence of the
7743 "Host" header, and then the URI path, including the query string
7744 unless the "drop-query" option is specified (see below). If no
7745 path is found or if the path is "*", then "/" is used instead. If
7746 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007747 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007748 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007749 HTTPS. When used in an "http-request" rule, <sch> value follows
7750 the log-format rules and can include some dynamic values (see
7751 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01007752
7753 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007754 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
7755 with 302 used by default if no code is specified. 301 means
7756 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02007757 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007758 cache the redirection. 303 is equivalent to 302 except that the
7759 browser will fetch the location with a GET method. 307 is just
7760 like 302 but makes it clear that the same method must be reused.
7761 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01007762
7763 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007764 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01007765
7766 - "drop-query"
7767 When this keyword is used in a prefix-based redirection, then the
7768 location will be set without any possible query-string, which is useful
7769 for directing users to a non-secure page for instance. It has no effect
7770 with a location-type redirect.
7771
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007772 - "append-slash"
7773 This keyword may be used in conjunction with "drop-query" to redirect
7774 users who use a URL not ending with a '/' to the same one with the '/'.
7775 It can be useful to ensure that search engines will only see one URL.
7776 For this, a return code 301 is preferred.
7777
Willy Tarreau0140f252008-11-19 21:07:09 +01007778 - "set-cookie NAME[=value]"
7779 A "Set-Cookie" header will be added with NAME (and optionally "=value")
7780 to the response. This is sometimes used to indicate that a user has
7781 been seen, for instance to protect against some types of DoS. No other
7782 cookie option is added, so the cookie will be a session cookie. Note
7783 that for a browser, a sole cookie name without an equal sign is
7784 different from a cookie with an equal sign.
7785
7786 - "clear-cookie NAME[=]"
7787 A "Set-Cookie" header will be added with NAME (and optionally "="), but
7788 with the "Max-Age" attribute set to zero. This will tell the browser to
7789 delete this cookie. It is useful for instance on logout pages. It is
7790 important to note that clearing the cookie "NAME" will not remove a
7791 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
7792 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007793
7794 Example: move the login URL only to HTTPS.
7795 acl clear dst_port 80
7796 acl secure dst_port 8080
7797 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01007798 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01007799 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01007800 acl cookie_set hdr_sub(cookie) SEEN=1
7801
7802 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01007803 redirect prefix https://mysite.com if login_page !secure
7804 redirect prefix http://mysite.com drop-query if login_page !uid_given
7805 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01007806 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007807
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007808 Example: send redirects for request for articles without a '/'.
7809 acl missing_slash path_reg ^/article/[^/]*$
7810 redirect code 301 prefix / drop-query append-slash if missing_slash
7811
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007812 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01007813 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007814
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007815 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02007816 http-request redirect code 301 location \
7817 http://www.%[hdr(host)]%[capture.req.uri] \
7818 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007819
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007820 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007821
Willy Tarreau303c0352008-01-17 19:01:39 +01007822
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007823retries <value>
7824 Set the number of retries to perform on a server after a connection failure
7825 May be used in sections: defaults | frontend | listen | backend
7826 yes | no | yes | yes
7827 Arguments :
7828 <value> is the number of times a connection attempt should be retried on
7829 a server when a connection either is refused or times out. The
7830 default value is 3.
7831
7832 It is important to understand that this value applies to the number of
7833 connection attempts, not full requests. When a connection has effectively
7834 been established to a server, there will be no more retry.
7835
7836 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07007837 a turn-around timer of min("timeout connect", one second) is applied before
7838 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007839
7840 When "option redispatch" is set, the last retry may be performed on another
7841 server even if a cookie references a different server.
7842
7843 See also : "option redispatch"
7844
7845
Olivier Houcharda254a372019-04-05 15:30:12 +02007846retry-on [list of keywords]
7847 Specify when to attempt to automatically retry a failed request
7848 May be used in sections: defaults | frontend | listen | backend
7849 yes | no | yes | yes
7850 Arguments :
7851 <keywords> is a list of keywords or HTTP status codes, each representing a
7852 type of failure event on which an attempt to retry the request
7853 is desired. Please read the notes at the bottom before changing
7854 this setting. The following keywords are supported :
7855
7856 none never retry
7857
7858 conn-failure retry when the connection or the SSL handshake failed
7859 and the request could not be sent. This is the default.
7860
7861 empty-response retry when the server connection was closed after part
7862 of the request was sent, and nothing was received from
7863 the server. This type of failure may be caused by the
7864 request timeout on the server side, poor network
7865 condition, or a server crash or restart while
7866 processing the request.
7867
Olivier Houcharde3249a92019-05-03 23:01:47 +02007868 junk-response retry when the server returned something not looking
7869 like a complete HTTP response. This includes partial
7870 responses headers as well as non-HTTP contents. It
7871 usually is a bad idea to retry on such events, which
7872 may be caused a configuration issue (wrong server port)
7873 or by the request being harmful to the server (buffer
7874 overflow attack for example).
7875
Olivier Houcharda254a372019-04-05 15:30:12 +02007876 response-timeout the server timeout stroke while waiting for the server
7877 to respond to the request. This may be caused by poor
7878 network condition, the reuse of an idle connection
7879 which has expired on the path, or by the request being
7880 extremely expensive to process. It generally is a bad
7881 idea to retry on such events on servers dealing with
7882 heavy database processing (full scans, etc) as it may
7883 amplify denial of service attacks.
7884
Olivier Houchard865d8392019-05-03 22:46:27 +02007885 0rtt-rejected retry requests which were sent over early data and were
7886 rejected by the server. These requests are generally
7887 considered to be safe to retry.
7888
Olivier Houcharda254a372019-04-05 15:30:12 +02007889 <status> any HTTP status code among "404" (Not Found), "408"
7890 (Request Timeout), "425" (Too Early), "500" (Server
7891 Error), "501" (Not Implemented), "502" (Bad Gateway),
7892 "503" (Service Unavailable), "504" (Gateway Timeout).
7893
Olivier Houchardddf0e032019-05-10 18:05:40 +02007894 all-retryable-errors
7895 retry request for any error that are considered
7896 retryable. This currently activates "conn-failure",
7897 "empty-response", "junk-response", "response-timeout",
7898 "0rtt-rejected", "500", "502", "503", and "504".
7899
Olivier Houcharda254a372019-04-05 15:30:12 +02007900 Using this directive replaces any previous settings with the new ones; it is
7901 not cumulative.
7902
7903 Please note that using anything other than "none" and "conn-failure" requires
7904 to allocate a buffer and copy the whole request into it, so it has memory and
7905 performance impacts. Requests not fitting in a single buffer will never be
7906 retried (see the global tune.bufsize setting).
7907
7908 You have to make sure the application has a replay protection mechanism built
7909 in such as a unique transaction IDs passed in requests, or that replaying the
7910 same request has no consequence, or it is very dangerous to use any retry-on
7911 value beside "conn-failure" and "none". Static file servers and caches are
7912 generally considered safe against any type of retry. Using a status code can
7913 be useful to quickly leave a server showing an abnormal behavior (out of
7914 memory, file system issues, etc), but in this case it may be a good idea to
7915 immediately redispatch the connection to another server (please see "option
7916 redispatch" for this). Last, it is important to understand that most causes
7917 of failures are the requests themselves and that retrying a request causing a
7918 server to misbehave will often make the situation even worse for this server,
7919 or for the whole service in case of redispatch.
7920
7921 Unless you know exactly how the application deals with replayed requests, you
7922 should not use this directive.
7923
7924 The default is "conn-failure".
7925
7926 See also: "retries", "option redispatch", "tune.bufsize"
7927
David du Colombier486df472011-03-17 10:40:26 +01007928server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007929 Declare a server in a backend
7930 May be used in sections : defaults | frontend | listen | backend
7931 no | no | yes | yes
7932 Arguments :
7933 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01007934 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05007935 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007936
David du Colombier486df472011-03-17 10:40:26 +01007937 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
7938 resolvable hostname is supported, but this name will be resolved
7939 during start-up. Address "0.0.0.0" or "*" has a special meaning.
7940 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02007941 address as the one from the client connection. This is useful in
7942 transparent proxy architectures where the client's connection is
7943 intercepted and haproxy must forward to the original destination
7944 address. This is more or less what the "transparent" keyword does
7945 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01007946 to report statistics. Optionally, an address family prefix may be
7947 used before the address to force the family regardless of the
7948 address format, which can be useful to specify a path to a unix
7949 socket with no slash ('/'). Currently supported prefixes are :
7950 - 'ipv4@' -> address is always IPv4
7951 - 'ipv6@' -> address is always IPv6
7952 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02007953 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +02007954 - 'sockpair@' -> address is the FD of a connected unix
7955 socket or of a socketpair. During a connection, the
7956 backend creates a pair of connected sockets, and passes
7957 one of them over the FD. The bind part will use the
7958 received socket as the client FD. Should be used
7959 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02007960 You may want to reference some environment variables in the
7961 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01007962 variables. The "init-addr" setting can be used to modify the way
7963 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007964
Willy Tarreaub6205fd2012-09-24 12:27:33 +02007965 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007966 be sent to this port. If unset, the same port the client
7967 connected to will be used. The port may also be prefixed by a "+"
7968 or a "-". In this case, the server's port will be determined by
7969 adding this value to the client's port.
7970
7971 <param*> is a list of parameters for this server. The "server" keywords
7972 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007973 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007974
7975 Examples :
7976 server first 10.1.1.1:1080 cookie first check inter 1000
7977 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01007978 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02007979 server backup "${SRV_BACKUP}:1080" backup
7980 server www1_dc1 "${LAN_DC1}.101:80"
7981 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007982
Willy Tarreau55dcaf62015-09-27 15:03:15 +02007983 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
7984 sun_path length is used for the address length. Some other programs
7985 such as socat use the string length only by default. Pass the option
7986 ",unix-tightsocklen=0" to any abstract socket definition in socat to
7987 make it compatible with HAProxy's.
7988
Mark Lamourinec2247f02012-01-04 13:02:01 -05007989 See also: "default-server", "http-send-name-header" and section 5 about
7990 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007991
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02007992server-state-file-name [<file>]
7993 Set the server state file to read, load and apply to servers available in
7994 this backend. It only applies when the directive "load-server-state-from-file"
7995 is set to "local". When <file> is not provided or if this directive is not
7996 set, then backend name is used. If <file> starts with a slash '/', then it is
7997 considered as an absolute path. Otherwise, <file> is concatenated to the
7998 global directive "server-state-file-base".
7999
8000 Example: the minimal configuration below would make HAProxy look for the
8001 state server file '/etc/haproxy/states/bk':
8002
8003 global
8004 server-state-file-base /etc/haproxy/states
8005
8006 backend bk
8007 load-server-state-from-file
8008
8009 See also: "server-state-file-base", "load-server-state-from-file", and
8010 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008011
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02008012server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
8013 Set a template to initialize servers with shared parameters.
8014 The names of these servers are built from <prefix> and <num | range> parameters.
8015 May be used in sections : defaults | frontend | listen | backend
8016 no | no | yes | yes
8017
8018 Arguments:
8019 <prefix> A prefix for the server names to be built.
8020
8021 <num | range>
8022 If <num> is provided, this template initializes <num> servers
8023 with 1 up to <num> as server name suffixes. A range of numbers
8024 <num_low>-<num_high> may also be used to use <num_low> up to
8025 <num_high> as server name suffixes.
8026
8027 <fqdn> A FQDN for all the servers this template initializes.
8028
8029 <port> Same meaning as "server" <port> argument (see "server" keyword).
8030
8031 <params*>
8032 Remaining server parameters among all those supported by "server"
8033 keyword.
8034
8035 Examples:
8036 # Initializes 3 servers with srv1, srv2 and srv3 as names,
8037 # google.com as FQDN, and health-check enabled.
8038 server-template srv 1-3 google.com:80 check
8039
8040 # or
8041 server-template srv 3 google.com:80 check
8042
8043 # would be equivalent to:
8044 server srv1 google.com:80 check
8045 server srv2 google.com:80 check
8046 server srv3 google.com:80 check
8047
8048
8049
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008050source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008051source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008052source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008053 Set the source address for outgoing connections
8054 May be used in sections : defaults | frontend | listen | backend
8055 yes | no | yes | yes
8056 Arguments :
8057 <addr> is the IPv4 address HAProxy will bind to before connecting to a
8058 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01008059
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008060 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01008061 the most appropriate address to reach its destination. Optionally
8062 an address family prefix may be used before the address to force
8063 the family regardless of the address format, which can be useful
8064 to specify a path to a unix socket with no slash ('/'). Currently
8065 supported prefixes are :
8066 - 'ipv4@' -> address is always IPv4
8067 - 'ipv6@' -> address is always IPv6
8068 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008069 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02008070 You may want to reference some environment variables in the
8071 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008072
8073 <port> is an optional port. It is normally not needed but may be useful
8074 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008075 the system will select a free port. Note that port ranges are not
8076 supported in the backend. If you want to force port ranges, you
8077 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008078
8079 <addr2> is the IP address to present to the server when connections are
8080 forwarded in full transparent proxy mode. This is currently only
8081 supported on some patched Linux kernels. When this address is
8082 specified, clients connecting to the server will be presented
8083 with this address, while health checks will still use the address
8084 <addr>.
8085
8086 <port2> is the optional port to present to the server when connections
8087 are forwarded in full transparent proxy mode (see <addr2> above).
8088 The default value of zero means the system will select a free
8089 port.
8090
Willy Tarreaubce70882009-09-07 11:51:47 +02008091 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
8092 This is the name of a comma-separated header list which can
8093 contain multiple IP addresses. By default, the last occurrence is
8094 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01008095 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02008096 by previous proxy, typically Stunnel. In order to use another
8097 occurrence from the last one, please see the <occ> parameter
8098 below. When the header (or occurrence) is not found, no binding
8099 is performed so that the proxy's default IP address is used. Also
8100 keep in mind that the header name is case insensitive, as for any
8101 HTTP header.
8102
8103 <occ> is the occurrence number of a value to be used in a multi-value
8104 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008105 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02008106 address. Positive values indicate a position from the first
8107 occurrence, 1 being the first one. Negative values indicate
8108 positions relative to the last one, -1 being the last one. This
8109 is helpful for situations where an X-Forwarded-For header is set
8110 at the entry point of an infrastructure and must be used several
8111 proxy layers away. When this value is not specified, -1 is
8112 assumed. Passing a zero here disables the feature.
8113
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008114 <name> is an optional interface name to which to bind to for outgoing
8115 traffic. On systems supporting this features (currently, only
8116 Linux), this allows one to bind all traffic to the server to
8117 this interface even if it is not the one the system would select
8118 based on routing tables. This should be used with extreme care.
8119 Note that using this option requires root privileges.
8120
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008121 The "source" keyword is useful in complex environments where a specific
8122 address only is allowed to connect to the servers. It may be needed when a
8123 private address must be used through a public gateway for instance, and it is
8124 known that the system cannot determine the adequate source address by itself.
8125
8126 An extension which is available on certain patched Linux kernels may be used
8127 through the "usesrc" optional keyword. It makes it possible to connect to the
8128 servers with an IP address which does not belong to the system itself. This
8129 is called "full transparent proxy mode". For this to work, the destination
8130 servers have to route their traffic back to this address through the machine
8131 running HAProxy, and IP forwarding must generally be enabled on this machine.
8132
8133 In this "full transparent proxy" mode, it is possible to force a specific IP
8134 address to be presented to the servers. This is not much used in fact. A more
8135 common use is to tell HAProxy to present the client's IP address. For this,
8136 there are two methods :
8137
8138 - present the client's IP and port addresses. This is the most transparent
8139 mode, but it can cause problems when IP connection tracking is enabled on
8140 the machine, because a same connection may be seen twice with different
8141 states. However, this solution presents the huge advantage of not
8142 limiting the system to the 64k outgoing address+port couples, because all
8143 of the client ranges may be used.
8144
8145 - present only the client's IP address and select a spare port. This
8146 solution is still quite elegant but slightly less transparent (downstream
8147 firewalls logs will not match upstream's). It also presents the downside
8148 of limiting the number of concurrent connections to the usual 64k ports.
8149 However, since the upstream and downstream ports are different, local IP
8150 connection tracking on the machine will not be upset by the reuse of the
8151 same session.
8152
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008153 This option sets the default source for all servers in the backend. It may
8154 also be specified in a "defaults" section. Finer source address specification
8155 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008156 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008157
Baptiste Assmann91bd3372015-07-17 21:59:42 +02008158 In order to work, "usesrc" requires root privileges.
8159
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008160 Examples :
8161 backend private
8162 # Connect to the servers using our 192.168.1.200 source address
8163 source 192.168.1.200
8164
8165 backend transparent_ssl1
8166 # Connect to the SSL farm from the client's source address
8167 source 192.168.1.200 usesrc clientip
8168
8169 backend transparent_ssl2
8170 # Connect to the SSL farm from the client's source address and port
8171 # not recommended if IP conntrack is present on the local machine.
8172 source 192.168.1.200 usesrc client
8173
8174 backend transparent_ssl3
8175 # Connect to the SSL farm from the client's source address. It
8176 # is more conntrack-friendly.
8177 source 192.168.1.200 usesrc clientip
8178
8179 backend transparent_smtp
8180 # Connect to the SMTP farm from the client's source address/port
8181 # with Tproxy version 4.
8182 source 0.0.0.0 usesrc clientip
8183
Willy Tarreaubce70882009-09-07 11:51:47 +02008184 backend transparent_http
8185 # Connect to the servers using the client's IP as seen by previous
8186 # proxy.
8187 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
8188
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008189 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008190 the Linux kernel on www.balabit.com, the "bind" keyword.
8191
Willy Tarreau844e3c52008-01-11 16:28:18 +01008192
Cyril Bonté66c327d2010-10-12 00:14:37 +02008193stats admin { if | unless } <cond>
8194 Enable statistics admin level if/unless a condition is matched
8195 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008196 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008197
8198 This statement enables the statistics admin level if/unless a condition is
8199 matched.
8200
8201 The admin level allows to enable/disable servers from the web interface. By
8202 default, statistics page is read-only for security reasons.
8203
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008204 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8205 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008206 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008207
Cyril Bonté23b39d92011-02-10 22:54:44 +01008208 Currently, the POST request is limited to the buffer size minus the reserved
8209 buffer space, which means that if the list of servers is too long, the
8210 request won't be processed. It is recommended to alter few servers at a
8211 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008212
8213 Example :
8214 # statistics admin level only for localhost
8215 backend stats_localhost
8216 stats enable
8217 stats admin if LOCALHOST
8218
8219 Example :
8220 # statistics admin level always enabled because of the authentication
8221 backend stats_auth
8222 stats enable
8223 stats auth admin:AdMiN123
8224 stats admin if TRUE
8225
8226 Example :
8227 # statistics admin level depends on the authenticated user
8228 userlist stats-auth
8229 group admin users admin
8230 user admin insecure-password AdMiN123
8231 group readonly users haproxy
8232 user haproxy insecure-password haproxy
8233
8234 backend stats_auth
8235 stats enable
8236 acl AUTH http_auth(stats-auth)
8237 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8238 stats http-request auth unless AUTH
8239 stats admin if AUTH_ADMIN
8240
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008241 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8242 "bind-process", section 3.4 about userlists and section 7 about
8243 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008244
8245
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008246stats auth <user>:<passwd>
8247 Enable statistics with authentication and grant access to an account
8248 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008249 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008250 Arguments :
8251 <user> is a user name to grant access to
8252
8253 <passwd> is the cleartext password associated to this user
8254
8255 This statement enables statistics with default settings, and restricts access
8256 to declared users only. It may be repeated as many times as necessary to
8257 allow as many users as desired. When a user tries to access the statistics
8258 without a valid account, a "401 Forbidden" response will be returned so that
8259 the browser asks the user to provide a valid user and password. The real
8260 which will be returned to the browser is configurable using "stats realm".
8261
8262 Since the authentication method is HTTP Basic Authentication, the passwords
8263 circulate in cleartext on the network. Thus, it was decided that the
8264 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008265 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008266
8267 It is also possible to reduce the scope of the proxies which appear in the
8268 report using "stats scope".
8269
8270 Though this statement alone is enough to enable statistics reporting, it is
8271 recommended to set all other settings in order to avoid relying on default
8272 unobvious parameters.
8273
8274 Example :
8275 # public access (limited to this backend only)
8276 backend public_www
8277 server srv1 192.168.0.1:80
8278 stats enable
8279 stats hide-version
8280 stats scope .
8281 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008282 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008283 stats auth admin1:AdMiN123
8284 stats auth admin2:AdMiN321
8285
8286 # internal monitoring access (unlimited)
8287 backend private_monitoring
8288 stats enable
8289 stats uri /admin?stats
8290 stats refresh 5s
8291
8292 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8293
8294
8295stats enable
8296 Enable statistics reporting with default settings
8297 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008298 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008299 Arguments : none
8300
8301 This statement enables statistics reporting with default settings defined
8302 at build time. Unless stated otherwise, these settings are used :
8303 - stats uri : /haproxy?stats
8304 - stats realm : "HAProxy Statistics"
8305 - stats auth : no authentication
8306 - stats scope : no restriction
8307
8308 Though this statement alone is enough to enable statistics reporting, it is
8309 recommended to set all other settings in order to avoid relying on default
8310 unobvious parameters.
8311
8312 Example :
8313 # public access (limited to this backend only)
8314 backend public_www
8315 server srv1 192.168.0.1:80
8316 stats enable
8317 stats hide-version
8318 stats scope .
8319 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008320 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008321 stats auth admin1:AdMiN123
8322 stats auth admin2:AdMiN321
8323
8324 # internal monitoring access (unlimited)
8325 backend private_monitoring
8326 stats enable
8327 stats uri /admin?stats
8328 stats refresh 5s
8329
8330 See also : "stats auth", "stats realm", "stats uri"
8331
8332
Willy Tarreaud63335a2010-02-26 12:56:52 +01008333stats hide-version
8334 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008335 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008336 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008337 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008338
Willy Tarreaud63335a2010-02-26 12:56:52 +01008339 By default, the stats page reports some useful status information along with
8340 the statistics. Among them is HAProxy's version. However, it is generally
8341 considered dangerous to report precise version to anyone, as it can help them
8342 target known weaknesses with specific attacks. The "stats hide-version"
8343 statement removes the version from the statistics report. This is recommended
8344 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008345
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008346 Though this statement alone is enough to enable statistics reporting, it is
8347 recommended to set all other settings in order to avoid relying on default
8348 unobvious parameters.
8349
Willy Tarreaud63335a2010-02-26 12:56:52 +01008350 Example :
8351 # public access (limited to this backend only)
8352 backend public_www
8353 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008354 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008355 stats hide-version
8356 stats scope .
8357 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008358 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008359 stats auth admin1:AdMiN123
8360 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008361
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008362 # internal monitoring access (unlimited)
8363 backend private_monitoring
8364 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008365 stats uri /admin?stats
8366 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008367
Willy Tarreaud63335a2010-02-26 12:56:52 +01008368 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008369
Willy Tarreau983e01e2010-01-11 18:42:06 +01008370
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008371stats http-request { allow | deny | auth [realm <realm>] }
8372 [ { if | unless } <condition> ]
8373 Access control for statistics
8374
8375 May be used in sections: defaults | frontend | listen | backend
8376 no | no | yes | yes
8377
8378 As "http-request", these set of options allow to fine control access to
8379 statistics. Each option may be followed by if/unless and acl.
8380 First option with matched condition (or option without condition) is final.
8381 For "deny" a 403 error will be returned, for "allow" normal processing is
8382 performed, for "auth" a 401/407 error code is returned so the client
8383 should be asked to enter a username and password.
8384
8385 There is no fixed limit to the number of http-request statements per
8386 instance.
8387
8388 See also : "http-request", section 3.4 about userlists and section 7
8389 about ACL usage.
8390
8391
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008392stats realm <realm>
8393 Enable statistics and set authentication realm
8394 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008395 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008396 Arguments :
8397 <realm> is the name of the HTTP Basic Authentication realm reported to
8398 the browser. The browser uses it to display it in the pop-up
8399 inviting the user to enter a valid username and password.
8400
8401 The realm is read as a single word, so any spaces in it should be escaped
8402 using a backslash ('\').
8403
8404 This statement is useful only in conjunction with "stats auth" since it is
8405 only related to authentication.
8406
8407 Though this statement alone is enough to enable statistics reporting, it is
8408 recommended to set all other settings in order to avoid relying on default
8409 unobvious parameters.
8410
8411 Example :
8412 # public access (limited to this backend only)
8413 backend public_www
8414 server srv1 192.168.0.1:80
8415 stats enable
8416 stats hide-version
8417 stats scope .
8418 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008419 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008420 stats auth admin1:AdMiN123
8421 stats auth admin2:AdMiN321
8422
8423 # internal monitoring access (unlimited)
8424 backend private_monitoring
8425 stats enable
8426 stats uri /admin?stats
8427 stats refresh 5s
8428
8429 See also : "stats auth", "stats enable", "stats uri"
8430
8431
8432stats refresh <delay>
8433 Enable statistics with automatic refresh
8434 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008435 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008436 Arguments :
8437 <delay> is the suggested refresh delay, specified in seconds, which will
8438 be returned to the browser consulting the report page. While the
8439 browser is free to apply any delay, it will generally respect it
8440 and refresh the page this every seconds. The refresh interval may
8441 be specified in any other non-default time unit, by suffixing the
8442 unit after the value, as explained at the top of this document.
8443
8444 This statement is useful on monitoring displays with a permanent page
8445 reporting the load balancer's activity. When set, the HTML report page will
8446 include a link "refresh"/"stop refresh" so that the user can select whether
8447 he wants automatic refresh of the page or not.
8448
8449 Though this statement alone is enough to enable statistics reporting, it is
8450 recommended to set all other settings in order to avoid relying on default
8451 unobvious parameters.
8452
8453 Example :
8454 # public access (limited to this backend only)
8455 backend public_www
8456 server srv1 192.168.0.1:80
8457 stats enable
8458 stats hide-version
8459 stats scope .
8460 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008461 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008462 stats auth admin1:AdMiN123
8463 stats auth admin2:AdMiN321
8464
8465 # internal monitoring access (unlimited)
8466 backend private_monitoring
8467 stats enable
8468 stats uri /admin?stats
8469 stats refresh 5s
8470
8471 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8472
8473
8474stats scope { <name> | "." }
8475 Enable statistics and limit access scope
8476 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008477 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008478 Arguments :
8479 <name> is the name of a listen, frontend or backend section to be
8480 reported. The special name "." (a single dot) designates the
8481 section in which the statement appears.
8482
8483 When this statement is specified, only the sections enumerated with this
8484 statement will appear in the report. All other ones will be hidden. This
8485 statement may appear as many times as needed if multiple sections need to be
8486 reported. Please note that the name checking is performed as simple string
8487 comparisons, and that it is never checked that a give section name really
8488 exists.
8489
8490 Though this statement alone is enough to enable statistics reporting, it is
8491 recommended to set all other settings in order to avoid relying on default
8492 unobvious parameters.
8493
8494 Example :
8495 # public access (limited to this backend only)
8496 backend public_www
8497 server srv1 192.168.0.1:80
8498 stats enable
8499 stats hide-version
8500 stats scope .
8501 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008502 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008503 stats auth admin1:AdMiN123
8504 stats auth admin2:AdMiN321
8505
8506 # internal monitoring access (unlimited)
8507 backend private_monitoring
8508 stats enable
8509 stats uri /admin?stats
8510 stats refresh 5s
8511
8512 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8513
Willy Tarreaud63335a2010-02-26 12:56:52 +01008514
Willy Tarreauc9705a12010-07-27 20:05:50 +02008515stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01008516 Enable reporting of a description on the statistics page.
8517 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008518 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008519
Willy Tarreauc9705a12010-07-27 20:05:50 +02008520 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01008521 description from global section is automatically used instead.
8522
8523 This statement is useful for users that offer shared services to their
8524 customers, where node or description should be different for each customer.
8525
8526 Though this statement alone is enough to enable statistics reporting, it is
8527 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008528 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008529
8530 Example :
8531 # internal monitoring access (unlimited)
8532 backend private_monitoring
8533 stats enable
8534 stats show-desc Master node for Europe, Asia, Africa
8535 stats uri /admin?stats
8536 stats refresh 5s
8537
8538 See also: "show-node", "stats enable", "stats uri" and "description" in
8539 global section.
8540
8541
8542stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02008543 Enable reporting additional information on the statistics page
8544 May be used in sections : defaults | frontend | listen | backend
8545 yes | yes | yes | yes
8546 Arguments : none
8547
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008548 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01008549 - cap: capabilities (proxy)
8550 - mode: one of tcp, http or health (proxy)
8551 - id: SNMP ID (proxy, socket, server)
8552 - IP (socket, server)
8553 - cookie (backend, server)
8554
8555 Though this statement alone is enough to enable statistics reporting, it is
8556 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008557 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008558
8559 See also: "stats enable", "stats uri".
8560
8561
8562stats show-node [ <name> ]
8563 Enable reporting of a host name on the statistics page.
8564 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008565 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008566 Arguments:
8567 <name> is an optional name to be reported. If unspecified, the
8568 node name from global section is automatically used instead.
8569
8570 This statement is useful for users that offer shared services to their
8571 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01008572 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008573
8574 Though this statement alone is enough to enable statistics reporting, it is
8575 recommended to set all other settings in order to avoid relying on default
8576 unobvious parameters.
8577
8578 Example:
8579 # internal monitoring access (unlimited)
8580 backend private_monitoring
8581 stats enable
8582 stats show-node Europe-1
8583 stats uri /admin?stats
8584 stats refresh 5s
8585
8586 See also: "show-desc", "stats enable", "stats uri", and "node" in global
8587 section.
8588
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008589
8590stats uri <prefix>
8591 Enable statistics and define the URI prefix to access them
8592 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008593 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008594 Arguments :
8595 <prefix> is the prefix of any URI which will be redirected to stats. This
8596 prefix may contain a question mark ('?') to indicate part of a
8597 query string.
8598
8599 The statistics URI is intercepted on the relayed traffic, so it appears as a
8600 page within the normal application. It is strongly advised to ensure that the
8601 selected URI will never appear in the application, otherwise it will never be
8602 possible to reach it in the application.
8603
8604 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008605 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008606 It is generally a good idea to include a question mark in the URI so that
8607 intermediate proxies refrain from caching the results. Also, since any string
8608 beginning with the prefix will be accepted as a stats request, the question
8609 mark helps ensuring that no valid URI will begin with the same words.
8610
8611 It is sometimes very convenient to use "/" as the URI prefix, and put that
8612 statement in a "listen" instance of its own. That makes it easy to dedicate
8613 an address or a port to statistics only.
8614
8615 Though this statement alone is enough to enable statistics reporting, it is
8616 recommended to set all other settings in order to avoid relying on default
8617 unobvious parameters.
8618
8619 Example :
8620 # public access (limited to this backend only)
8621 backend public_www
8622 server srv1 192.168.0.1:80
8623 stats enable
8624 stats hide-version
8625 stats scope .
8626 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008627 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008628 stats auth admin1:AdMiN123
8629 stats auth admin2:AdMiN321
8630
8631 # internal monitoring access (unlimited)
8632 backend private_monitoring
8633 stats enable
8634 stats uri /admin?stats
8635 stats refresh 5s
8636
8637 See also : "stats auth", "stats enable", "stats realm"
8638
8639
Willy Tarreaud63335a2010-02-26 12:56:52 +01008640stick match <pattern> [table <table>] [{if | unless} <cond>]
8641 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008642 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01008643 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008644
8645 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008646 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008647 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008648 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008649 stickiness table. This rule is mandatory.
8650
8651 <table> is an optional stickiness table name. If unspecified, the same
8652 backend's table is used. A stickiness table is declared using
8653 the "stick-table" statement.
8654
8655 <cond> is an optional matching condition. It makes it possible to match
8656 on a certain criterion only when other conditions are met (or
8657 not met). For instance, it could be used to match on a source IP
8658 address except when a request passes through a known proxy, in
8659 which case we'd match on a header containing that IP address.
8660
8661 Some protocols or applications require complex stickiness rules and cannot
8662 always simply rely on cookies nor hashing. The "stick match" statement
8663 describes a rule to extract the stickiness criterion from an incoming request
8664 or connection. See section 7 for a complete list of possible patterns and
8665 transformation rules.
8666
8667 The table has to be declared using the "stick-table" statement. It must be of
8668 a type compatible with the pattern. By default it is the one which is present
8669 in the same backend. It is possible to share a table with other backends by
8670 referencing it using the "table" keyword. If another table is referenced,
8671 the server's ID inside the backends are used. By default, all server IDs
8672 start at 1 in each backend, so the server ordering is enough. But in case of
8673 doubt, it is highly recommended to force server IDs using their "id" setting.
8674
8675 It is possible to restrict the conditions where a "stick match" statement
8676 will apply, using "if" or "unless" followed by a condition. See section 7 for
8677 ACL based conditions.
8678
8679 There is no limit on the number of "stick match" statements. The first that
8680 applies and matches will cause the request to be directed to the same server
8681 as was used for the request which created the entry. That way, multiple
8682 matches can be used as fallbacks.
8683
8684 The stick rules are checked after the persistence cookies, so they will not
8685 affect stickiness if a cookie has already been used to select a server. That
8686 way, it becomes very easy to insert cookies and match on IP addresses in
8687 order to maintain stickiness between HTTP and HTTPS.
8688
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008689 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8690 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008691 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008692
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008693 Example :
8694 # forward SMTP users to the same server they just used for POP in the
8695 # last 30 minutes
8696 backend pop
8697 mode tcp
8698 balance roundrobin
8699 stick store-request src
8700 stick-table type ip size 200k expire 30m
8701 server s1 192.168.1.1:110
8702 server s2 192.168.1.1:110
8703
8704 backend smtp
8705 mode tcp
8706 balance roundrobin
8707 stick match src table pop
8708 server s1 192.168.1.1:25
8709 server s2 192.168.1.1:25
8710
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008711 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008712 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008713
8714
8715stick on <pattern> [table <table>] [{if | unless} <condition>]
8716 Define a request pattern to associate a user to a server
8717 May be used in sections : defaults | frontend | listen | backend
8718 no | no | yes | yes
8719
8720 Note : This form is exactly equivalent to "stick match" followed by
8721 "stick store-request", all with the same arguments. Please refer
8722 to both keywords for details. It is only provided as a convenience
8723 for writing more maintainable configurations.
8724
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008725 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8726 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008727 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008728
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008729 Examples :
8730 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01008731 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008732
8733 # ...is strictly equivalent to this one :
8734 stick match src table pop if !localhost
8735 stick store-request src table pop if !localhost
8736
8737
8738 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
8739 # well as HTTP without cookie. Share the same table between both accesses.
8740 backend http
8741 mode http
8742 balance roundrobin
8743 stick on src table https
8744 cookie SRV insert indirect nocache
8745 server s1 192.168.1.1:80 cookie s1
8746 server s2 192.168.1.1:80 cookie s2
8747
8748 backend https
8749 mode tcp
8750 balance roundrobin
8751 stick-table type ip size 200k expire 30m
8752 stick on src
8753 server s1 192.168.1.1:443
8754 server s2 192.168.1.1:443
8755
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008756 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008757
8758
8759stick store-request <pattern> [table <table>] [{if | unless} <condition>]
8760 Define a request pattern used to create an entry in a stickiness table
8761 May be used in sections : defaults | frontend | listen | backend
8762 no | no | yes | yes
8763
8764 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008765 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008766 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008767 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008768 server is selected.
8769
8770 <table> is an optional stickiness table name. If unspecified, the same
8771 backend's table is used. A stickiness table is declared using
8772 the "stick-table" statement.
8773
8774 <cond> is an optional storage condition. It makes it possible to store
8775 certain criteria only when some conditions are met (or not met).
8776 For instance, it could be used to store the source IP address
8777 except when the request passes through a known proxy, in which
8778 case we'd store a converted form of a header containing that IP
8779 address.
8780
8781 Some protocols or applications require complex stickiness rules and cannot
8782 always simply rely on cookies nor hashing. The "stick store-request" statement
8783 describes a rule to decide what to extract from the request and when to do
8784 it, in order to store it into a stickiness table for further requests to
8785 match it using the "stick match" statement. Obviously the extracted part must
8786 make sense and have a chance to be matched in a further request. Storing a
8787 client's IP address for instance often makes sense. Storing an ID found in a
8788 URL parameter also makes sense. Storing a source port will almost never make
8789 any sense because it will be randomly matched. See section 7 for a complete
8790 list of possible patterns and transformation rules.
8791
8792 The table has to be declared using the "stick-table" statement. It must be of
8793 a type compatible with the pattern. By default it is the one which is present
8794 in the same backend. It is possible to share a table with other backends by
8795 referencing it using the "table" keyword. If another table is referenced,
8796 the server's ID inside the backends are used. By default, all server IDs
8797 start at 1 in each backend, so the server ordering is enough. But in case of
8798 doubt, it is highly recommended to force server IDs using their "id" setting.
8799
8800 It is possible to restrict the conditions where a "stick store-request"
8801 statement will apply, using "if" or "unless" followed by a condition. This
8802 condition will be evaluated while parsing the request, so any criteria can be
8803 used. See section 7 for ACL based conditions.
8804
8805 There is no limit on the number of "stick store-request" statements, but
8806 there is a limit of 8 simultaneous stores per request or response. This
8807 makes it possible to store up to 8 criteria, all extracted from either the
8808 request or the response, regardless of the number of rules. Only the 8 first
8809 ones which match will be kept. Using this, it is possible to feed multiple
8810 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01008811 another protocol or access method. Using multiple store-request rules with
8812 the same table is possible and may be used to find the best criterion to rely
8813 on, by arranging the rules by decreasing preference order. Only the first
8814 extracted criterion for a given table will be stored. All subsequent store-
8815 request rules referencing the same table will be skipped and their ACLs will
8816 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008817
8818 The "store-request" rules are evaluated once the server connection has been
8819 established, so that the table will contain the real server that processed
8820 the request.
8821
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008822 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8823 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008824 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008825
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008826 Example :
8827 # forward SMTP users to the same server they just used for POP in the
8828 # last 30 minutes
8829 backend pop
8830 mode tcp
8831 balance roundrobin
8832 stick store-request src
8833 stick-table type ip size 200k expire 30m
8834 server s1 192.168.1.1:110
8835 server s2 192.168.1.1:110
8836
8837 backend smtp
8838 mode tcp
8839 balance roundrobin
8840 stick match src table pop
8841 server s1 192.168.1.1:25
8842 server s2 192.168.1.1:25
8843
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008844 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02008845 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008846
8847
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008848stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02008849 size <size> [expire <expire>] [nopurge] [peers <peersect>]
8850 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08008851 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008852 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02008853 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008854
8855 Arguments :
8856 ip a table declared with "type ip" will only store IPv4 addresses.
8857 This form is very compact (about 50 bytes per entry) and allows
8858 very fast entry lookup and stores with almost no overhead. This
8859 is mainly used to store client source IP addresses.
8860
David du Colombier9a6d3c92011-03-17 10:40:24 +01008861 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
8862 This form is very compact (about 60 bytes per entry) and allows
8863 very fast entry lookup and stores with almost no overhead. This
8864 is mainly used to store client source IP addresses.
8865
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008866 integer a table declared with "type integer" will store 32bit integers
8867 which can represent a client identifier found in a request for
8868 instance.
8869
8870 string a table declared with "type string" will store substrings of up
8871 to <len> characters. If the string provided by the pattern
8872 extractor is larger than <len>, it will be truncated before
8873 being stored. During matching, at most <len> characters will be
8874 compared between the string in the table and the extracted
8875 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008876 to 32 characters.
8877
8878 binary a table declared with "type binary" will store binary blocks
8879 of <len> bytes. If the block provided by the pattern
8880 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02008881 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008882 is shorter than <len>, it will be padded by 0. When not
8883 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008884
8885 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02008886 "string" type table (See type "string" above). Or the number
8887 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008888 changing this parameter as memory usage will proportionally
8889 increase.
8890
8891 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01008892 value directly impacts memory usage. Count approximately
8893 50 bytes per entry, plus the size of a string if any. The size
8894 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008895
8896 [nopurge] indicates that we refuse to purge older entries when the table
8897 is full. When not specified and the table is full when haproxy
8898 wants to store an entry in it, it will flush a few of the oldest
8899 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01008900 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008901 be desirable to refuse new entries instead of purging the older
8902 ones. That may be the case when the amount of data to store is
8903 far above the hardware limits and we prefer not to offer access
8904 to new clients than to reject the ones already connected. When
8905 using this parameter, be sure to properly set the "expire"
8906 parameter (see below).
8907
Emeric Brunf099e792010-09-27 12:05:28 +02008908 <peersect> is the name of the peers section to use for replication. Entries
8909 which associate keys to server IDs are kept synchronized with
8910 the remote peers declared in this section. All entries are also
8911 automatically learned from the local peer (old process) during a
8912 soft restart.
8913
Willy Tarreau1abc6732015-05-01 19:21:02 +02008914 NOTE : each peers section may be referenced only by tables
8915 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008916
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008917 <expire> defines the maximum duration of an entry in the table since it
8918 was last created, refreshed or matched. The expiration delay is
8919 defined using the standard time format, similarly as the various
8920 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03008921 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02008922 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008923 be removed once full. Be sure not to use the "nopurge" parameter
8924 if not expiration delay is specified.
8925
Willy Tarreau08d5f982010-06-06 13:34:54 +02008926 <data_type> is used to store additional information in the stick-table. This
8927 may be used by ACLs in order to control various criteria related
8928 to the activity of the client matching the stick-table. For each
8929 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02008930 that the additional data can fit. Several data types may be
8931 stored with an entry. Multiple data types may be specified after
8932 the "store" keyword, as a comma-separated list. Alternatively,
8933 it is possible to repeat the "store" keyword followed by one or
8934 several data types. Except for the "server_id" type which is
8935 automatically detected and enabled, all data types must be
8936 explicitly declared to be stored. If an ACL references a data
8937 type which is not stored, the ACL will simply not match. Some
8938 data types require an argument which must be passed just after
8939 the type between parenthesis. See below for the supported data
8940 types and their arguments.
8941
8942 The data types that can be stored with an entry are the following :
8943 - server_id : this is an integer which holds the numeric ID of the server a
8944 request was assigned to. It is used by the "stick match", "stick store",
8945 and "stick on" rules. It is automatically enabled when referenced.
8946
8947 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
8948 integer which may be used for anything. Most of the time it will be used
8949 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01008950 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02008951
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008952 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
8953 over a period. It is a positive 32-bit integer integer which may be used
8954 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01008955 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008956 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01008957 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02008958
Frédéric Lécaille6778b272018-01-29 15:22:53 +01008959 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
8960 integer which may be used for anything. Most of the time it will be used
8961 to put a special tag on some entries, for instance to note that a
8962 specific behavior was detected and must be known for future matches.
8963
8964 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
8965 over a period. It is a positive 32-bit integer integer which may be used
8966 for anything. Just like <gpc1>, it counts events, but instead of keeping
8967 a cumulative number, it maintains the rate at which the counter is
8968 incremented. Most of the time it will be used to measure the frequency of
8969 occurrence of certain events (e.g. requests to a specific URL).
8970
Willy Tarreauc9705a12010-07-27 20:05:50 +02008971 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
8972 the absolute number of connections received from clients which matched
8973 this entry. It does not mean the connections were accepted, just that
8974 they were received.
8975
8976 - conn_cur : Current Connections. It is a positive 32-bit integer which
8977 stores the concurrent connection counts for the entry. It is incremented
8978 once an incoming connection matches the entry, and decremented once the
8979 connection leaves. That way it is possible to know at any time the exact
8980 number of concurrent connections for an entry.
8981
8982 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8983 integer parameter <period> which indicates in milliseconds the length
8984 of the period over which the average is measured. It reports the average
8985 incoming connection rate over that period, in connections per period. The
8986 result is an integer which can be matched using ACLs.
8987
8988 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
8989 the absolute number of sessions received from clients which matched this
8990 entry. A session is a connection that was accepted by the layer 4 rules.
8991
8992 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
8993 integer parameter <period> which indicates in milliseconds the length
8994 of the period over which the average is measured. It reports the average
8995 incoming session rate over that period, in sessions per period. The
8996 result is an integer which can be matched using ACLs.
8997
8998 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
8999 counts the absolute number of HTTP requests received from clients which
9000 matched this entry. It does not matter whether they are valid requests or
9001 not. Note that this is different from sessions when keep-alive is used on
9002 the client side.
9003
9004 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9005 integer parameter <period> which indicates in milliseconds the length
9006 of the period over which the average is measured. It reports the average
9007 HTTP request rate over that period, in requests per period. The result is
9008 an integer which can be matched using ACLs. It does not matter whether
9009 they are valid requests or not. Note that this is different from sessions
9010 when keep-alive is used on the client side.
9011
9012 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
9013 counts the absolute number of HTTP requests errors induced by clients
9014 which matched this entry. Errors are counted on invalid and truncated
9015 requests, as well as on denied or tarpitted requests, and on failed
9016 authentications. If the server responds with 4xx, then the request is
9017 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01009018 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02009019
9020 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9021 integer parameter <period> which indicates in milliseconds the length
9022 of the period over which the average is measured. It reports the average
9023 HTTP request error rate over that period, in requests per period (see
9024 http_err_cnt above for what is accounted as an error). The result is an
9025 integer which can be matched using ACLs.
9026
9027 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009028 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02009029 which matched this entry. Headers are included in the count. This may be
9030 used to limit abuse of upload features on photo or video servers.
9031
9032 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9033 integer parameter <period> which indicates in milliseconds the length
9034 of the period over which the average is measured. It reports the average
9035 incoming bytes rate over that period, in bytes per period. It may be used
9036 to detect users which upload too much and too fast. Warning: with large
9037 uploads, it is possible that the amount of uploaded data will be counted
9038 once upon termination, thus causing spikes in the average transfer speed
9039 instead of having a smooth one. This may partially be smoothed with
9040 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
9041 recommended for better fairness.
9042
9043 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009044 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02009045 matched this entry. Headers are included in the count. This may be used
9046 to limit abuse of bots sucking the whole site.
9047
9048 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
9049 an integer parameter <period> which indicates in milliseconds the length
9050 of the period over which the average is measured. It reports the average
9051 outgoing bytes rate over that period, in bytes per period. It may be used
9052 to detect users which download too much and too fast. Warning: with large
9053 transfers, it is possible that the amount of transferred data will be
9054 counted once upon termination, thus causing spikes in the average
9055 transfer speed instead of having a smooth one. This may partially be
9056 smoothed with "option contstats" though this is not perfect yet. Use of
9057 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02009058
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009059 There is only one stick-table per proxy. At the moment of writing this doc,
9060 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009061 to be required, simply create a dummy backend with a stick-table in it and
9062 reference it.
9063
9064 It is important to understand that stickiness based on learning information
9065 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01009066 lost upon restart unless peers are properly configured to transfer such
9067 information upon restart (recommended). In general it can be good as a
9068 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009069
Willy Tarreauc9705a12010-07-27 20:05:50 +02009070 Last, memory requirements may be important when storing many data types.
9071 Indeed, storing all indicators above at once in each entry requires 116 bytes
9072 per entry, or 116 MB for a 1-million entries table. This is definitely not
9073 something that can be ignored.
9074
9075 Example:
9076 # Keep track of counters of up to 1 million IP addresses over 5 minutes
9077 # and store a general purpose counter and the average connection rate
9078 # computed over a sliding window of 30 seconds.
9079 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
9080
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009081 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01009082 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009083
9084
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009085stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01009086 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009087 May be used in sections : defaults | frontend | listen | backend
9088 no | no | yes | yes
9089
9090 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009091 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009092 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01009093 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009094 server is selected.
9095
9096 <table> is an optional stickiness table name. If unspecified, the same
9097 backend's table is used. A stickiness table is declared using
9098 the "stick-table" statement.
9099
9100 <cond> is an optional storage condition. It makes it possible to store
9101 certain criteria only when some conditions are met (or not met).
9102 For instance, it could be used to store the SSL session ID only
9103 when the response is a SSL server hello.
9104
9105 Some protocols or applications require complex stickiness rules and cannot
9106 always simply rely on cookies nor hashing. The "stick store-response"
9107 statement describes a rule to decide what to extract from the response and
9108 when to do it, in order to store it into a stickiness table for further
9109 requests to match it using the "stick match" statement. Obviously the
9110 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009111 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009112 See section 7 for a complete list of possible patterns and transformation
9113 rules.
9114
9115 The table has to be declared using the "stick-table" statement. It must be of
9116 a type compatible with the pattern. By default it is the one which is present
9117 in the same backend. It is possible to share a table with other backends by
9118 referencing it using the "table" keyword. If another table is referenced,
9119 the server's ID inside the backends are used. By default, all server IDs
9120 start at 1 in each backend, so the server ordering is enough. But in case of
9121 doubt, it is highly recommended to force server IDs using their "id" setting.
9122
9123 It is possible to restrict the conditions where a "stick store-response"
9124 statement will apply, using "if" or "unless" followed by a condition. This
9125 condition will be evaluated while parsing the response, so any criteria can
9126 be used. See section 7 for ACL based conditions.
9127
9128 There is no limit on the number of "stick store-response" statements, but
9129 there is a limit of 8 simultaneous stores per request or response. This
9130 makes it possible to store up to 8 criteria, all extracted from either the
9131 request or the response, regardless of the number of rules. Only the 8 first
9132 ones which match will be kept. Using this, it is possible to feed multiple
9133 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009134 another protocol or access method. Using multiple store-response rules with
9135 the same table is possible and may be used to find the best criterion to rely
9136 on, by arranging the rules by decreasing preference order. Only the first
9137 extracted criterion for a given table will be stored. All subsequent store-
9138 response rules referencing the same table will be skipped and their ACLs will
9139 not be evaluated. However, even if a store-request rule references a table, a
9140 store-response rule may also use the same table. This means that each table
9141 may learn exactly one element from the request and one element from the
9142 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009143
9144 The table will contain the real server that processed the request.
9145
9146 Example :
9147 # Learn SSL session ID from both request and response and create affinity.
9148 backend https
9149 mode tcp
9150 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02009151 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009152 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009153
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009154 acl clienthello req_ssl_hello_type 1
9155 acl serverhello rep_ssl_hello_type 2
9156
9157 # use tcp content accepts to detects ssl client and server hello.
9158 tcp-request inspect-delay 5s
9159 tcp-request content accept if clienthello
9160
9161 # no timeout on response inspect delay by default.
9162 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009163
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009164 # SSL session ID (SSLID) may be present on a client or server hello.
9165 # Its length is coded on 1 byte at offset 43 and its value starts
9166 # at offset 44.
9167
9168 # Match and learn on request if client hello.
9169 stick on payload_lv(43,1) if clienthello
9170
9171 # Learn on response if server hello.
9172 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02009173
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009174 server s1 192.168.1.1:443
9175 server s2 192.168.1.1:443
9176
9177 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
9178 extraction.
9179
9180
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009181tcp-check connect [params*]
9182 Opens a new connection
9183 May be used in sections: defaults | frontend | listen | backend
9184 no | no | yes | yes
9185
9186 When an application lies on more than a single TCP port or when HAProxy
9187 load-balance many services in a single backend, it makes sense to probe all
9188 the services individually before considering a server as operational.
9189
9190 When there are no TCP port configured on the server line neither server port
9191 directive, then the 'tcp-check connect port <port>' must be the first step
9192 of the sequence.
9193
9194 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9195 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9196 do.
9197
9198 Parameters :
9199 They are optional and can be used to describe how HAProxy should open and
9200 use the TCP connection.
9201
9202 port if not set, check port or server port is used.
9203 It tells HAProxy where to open the connection to.
9204 <port> must be a valid TCP port source integer, from 1 to 65535.
9205
9206 send-proxy send a PROXY protocol string
9207
9208 ssl opens a ciphered connection
9209
9210 Examples:
9211 # check HTTP and HTTPs services on a server.
9212 # first open port 80 thanks to server line port directive, then
9213 # tcp-check opens port 443, ciphered and run a request on it:
9214 option tcp-check
9215 tcp-check connect
9216 tcp-check send GET\ /\ HTTP/1.0\r\n
9217 tcp-check send Host:\ haproxy.1wt.eu\r\n
9218 tcp-check send \r\n
9219 tcp-check expect rstring (2..|3..)
9220 tcp-check connect port 443 ssl
9221 tcp-check send GET\ /\ HTTP/1.0\r\n
9222 tcp-check send Host:\ haproxy.1wt.eu\r\n
9223 tcp-check send \r\n
9224 tcp-check expect rstring (2..|3..)
9225 server www 10.0.0.1 check port 80
9226
9227 # check both POP and IMAP from a single server:
9228 option tcp-check
9229 tcp-check connect port 110
9230 tcp-check expect string +OK\ POP3\ ready
9231 tcp-check connect port 143
9232 tcp-check expect string *\ OK\ IMAP4\ ready
9233 server mail 10.0.0.1 check
9234
9235 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9236
9237
9238tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009239 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009240 May be used in sections: defaults | frontend | listen | backend
9241 no | no | yes | yes
9242
9243 Arguments :
9244 <match> is a keyword indicating how to look for a specific pattern in the
9245 response. The keyword may be one of "string", "rstring" or
9246 binary.
9247 The keyword may be preceded by an exclamation mark ("!") to negate
9248 the match. Spaces are allowed between the exclamation mark and the
9249 keyword. See below for more details on the supported keywords.
9250
9251 <pattern> is the pattern to look for. It may be a string or a regular
9252 expression. If the pattern contains spaces, they must be escaped
9253 with the usual backslash ('\').
9254 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009255 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009256 two digits will represent a byte. The hexadecimal digits may be
9257 used upper or lower case.
9258
9259
9260 The available matches are intentionally similar to their http-check cousins :
9261
9262 string <string> : test the exact string matches in the response buffer.
9263 A health check response will be considered valid if the
9264 response's buffer contains this exact string. If the
9265 "string" keyword is prefixed with "!", then the response
9266 will be considered invalid if the body contains this
9267 string. This can be used to look for a mandatory pattern
9268 in a protocol response, or to detect a failure when a
9269 specific error appears in a protocol banner.
9270
9271 rstring <regex> : test a regular expression on the response buffer.
9272 A health check response will be considered valid if the
9273 response's buffer matches this expression. If the
9274 "rstring" keyword is prefixed with "!", then the response
9275 will be considered invalid if the body matches the
9276 expression.
9277
9278 binary <hexstring> : test the exact string in its hexadecimal form matches
9279 in the response buffer. A health check response will
9280 be considered valid if the response's buffer contains
9281 this exact hexadecimal string.
9282 Purpose is to match data on binary protocols.
9283
9284 It is important to note that the responses will be limited to a certain size
9285 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9286 Thus, too large responses may not contain the mandatory pattern when using
9287 "string", "rstring" or binary. If a large response is absolutely required, it
9288 is possible to change the default max size by setting the global variable.
9289 However, it is worth keeping in mind that parsing very large responses can
9290 waste some CPU cycles, especially when regular expressions are used, and that
9291 it is always better to focus the checks on smaller resources. Also, in its
9292 current state, the check will not find any string nor regex past a null
9293 character in the response. Similarly it is not possible to request matching
9294 the null character.
9295
9296 Examples :
9297 # perform a POP check
9298 option tcp-check
9299 tcp-check expect string +OK\ POP3\ ready
9300
9301 # perform an IMAP check
9302 option tcp-check
9303 tcp-check expect string *\ OK\ IMAP4\ ready
9304
9305 # look for the redis master server
9306 option tcp-check
9307 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009308 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009309 tcp-check send info\ replication\r\n
9310 tcp-check expect string role:master
9311 tcp-check send QUIT\r\n
9312 tcp-check expect string +OK
9313
9314
9315 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9316 "tcp-check send-binary", "http-check expect", tune.chksize
9317
9318
9319tcp-check send <data>
9320 Specify a string to be sent as a question during a generic health check
9321 May be used in sections: defaults | frontend | listen | backend
9322 no | no | yes | yes
9323
9324 <data> : the data to be sent as a question during a generic health check
9325 session. For now, <data> must be a string.
9326
9327 Examples :
9328 # look for the redis master server
9329 option tcp-check
9330 tcp-check send info\ replication\r\n
9331 tcp-check expect string role:master
9332
9333 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9334 "tcp-check send-binary", tune.chksize
9335
9336
Davor Ocelice9ed2812017-12-25 17:49:28 +01009337tcp-check send-binary <hexstring>
9338 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009339 tcp health check
9340 May be used in sections: defaults | frontend | listen | backend
9341 no | no | yes | yes
9342
9343 <data> : the data to be sent as a question during a generic health check
9344 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009345 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009346 response buffer. A health check response will be considered
9347 valid if the response's buffer contains this exact
9348 hexadecimal string.
9349 Purpose is to send binary data to ask on binary protocols.
9350
9351 Examples :
9352 # redis check in binary
9353 option tcp-check
9354 tcp-check send-binary 50494e470d0a # PING\r\n
9355 tcp-check expect binary 2b504F4e47 # +PONG
9356
9357
9358 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9359 "tcp-check send", tune.chksize
9360
9361
Willy Tarreaue9656522010-08-17 15:40:09 +02009362tcp-request connection <action> [{if | unless} <condition>]
9363 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009364 May be used in sections : defaults | frontend | listen | backend
9365 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009366 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009367 <action> defines the action to perform if the condition applies. See
9368 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009369
Willy Tarreaue9656522010-08-17 15:40:09 +02009370 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009371
9372 Immediately after acceptance of a new incoming connection, it is possible to
9373 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009374 or dropped or have its counters tracked. Those conditions cannot make use of
9375 any data contents because the connection has not been read from yet, and the
9376 buffers are not yet allocated. This is used to selectively and very quickly
9377 accept or drop connections from various sources with a very low overhead. If
9378 some contents need to be inspected in order to take the decision, the
9379 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009380
Willy Tarreaue9656522010-08-17 15:40:09 +02009381 The "tcp-request connection" rules are evaluated in their exact declaration
9382 order. If no rule matches or if there is no rule, the default action is to
9383 accept the incoming connection. There is no specific limit to the number of
9384 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009385
Willy Tarreaua9083d02015-05-08 15:27:59 +02009386 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02009387 - accept :
9388 accepts the connection if the condition is true (when used with "if")
9389 or false (when used with "unless"). The first such rule executed ends
9390 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009391
Willy Tarreaue9656522010-08-17 15:40:09 +02009392 - reject :
9393 rejects the connection if the condition is true (when used with "if")
9394 or false (when used with "unless"). The first such rule executed ends
9395 the rules evaluation. Rejected connections do not even become a
9396 session, which is why they are accounted separately for in the stats,
9397 as "denied connections". They are not considered for the session
9398 rate-limit and are not logged either. The reason is that these rules
9399 should only be used to filter extremely high connection rates such as
9400 the ones encountered during a massive DDoS attack. Under these extreme
9401 conditions, the simple action of logging each event would make the
9402 system collapse and would considerably lower the filtering capacity. If
9403 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02009404 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009405
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009406 - expect-proxy layer4 :
9407 configures the client-facing connection to receive a PROXY protocol
9408 header before any byte is read from the socket. This is equivalent to
9409 having the "accept-proxy" keyword on the "bind" line, except that using
9410 the TCP rule allows the PROXY protocol to be accepted only for certain
9411 IP address ranges using an ACL. This is convenient when multiple layers
9412 of load balancers are passed through by traffic coming from public
9413 hosts.
9414
Bertrand Jacquin90759682016-06-06 15:35:39 +01009415 - expect-netscaler-cip layer4 :
9416 configures the client-facing connection to receive a NetScaler Client
9417 IP insertion protocol header before any byte is read from the socket.
9418 This is equivalent to having the "accept-netscaler-cip" keyword on the
9419 "bind" line, except that using the TCP rule allows the PROXY protocol
9420 to be accepted only for certain IP address ranges using an ACL. This
9421 is convenient when multiple layers of load balancers are passed
9422 through by traffic coming from public hosts.
9423
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009424 - capture <sample> len <length> :
9425 This only applies to "tcp-request content" rules. It captures sample
9426 expression <sample> from the request buffer, and converts it to a
9427 string of at most <len> characters. The resulting string is stored into
9428 the next request "capture" slot, so it will possibly appear next to
9429 some captured HTTP headers. It will then automatically appear in the
9430 logs, and it will be possible to extract it using sample fetch rules to
9431 feed it into headers or anything. The length should be limited given
9432 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02009433 session life. Please check section 7.3 (Fetching samples) and "capture
9434 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009435
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009436 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02009437 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009438 rules do not stop evaluation and do not change default action. The
9439 number of counters that may be simultaneously tracked by the same
9440 connection is set in MAX_SESS_STKCTR at build time (reported in
John Roeslerfb2fce12019-07-10 15:45:51 -05009441 haproxy -vv) which defaults to 3, so the track-sc number is between 0
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009442 and (MAX_SESS_STCKTR-1). The first "track-sc0" rule executed enables
9443 tracking of the counters of the specified table as the first set. The
9444 first "track-sc1" rule executed enables tracking of the counters of the
9445 specified table as the second set. The first "track-sc2" rule executed
9446 enables tracking of the counters of the specified table as the third
9447 set. It is a recommended practice to use the first set of counters for
9448 the per-frontend counters and the second set for the per-backend ones.
9449 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009450
Willy Tarreaue9656522010-08-17 15:40:09 +02009451 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009452 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02009453 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01009454 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009455 and used to select which table entry to update the counters.
9456 Note that "tcp-request connection" cannot use content-based
9457 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009458
Willy Tarreaue9656522010-08-17 15:40:09 +02009459 <table> is an optional table to be used instead of the default one,
9460 which is the stick-table declared in the current proxy. All
9461 the counters for the matches and updates for the key will
9462 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009463
Willy Tarreaue9656522010-08-17 15:40:09 +02009464 Once a "track-sc*" rule is executed, the key is looked up in the table
9465 and if it is not found, an entry is allocated for it. Then a pointer to
9466 that entry is kept during all the session's life, and this entry's
9467 counters are updated as often as possible, every time the session's
9468 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009469 Counters are only updated for events that happen after the tracking has
9470 been started. For example, connection counters will not be updated when
9471 tracking layer 7 information, since the connection event happens before
9472 layer7 information is extracted.
9473
Willy Tarreaue9656522010-08-17 15:40:09 +02009474 If the entry tracks concurrent connection counters, one connection is
9475 counted for as long as the entry is tracked, and the entry will not
9476 expire during that time. Tracking counters also provides a performance
9477 advantage over just checking the keys, because only one table lookup is
9478 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009479
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009480 - sc-inc-gpc0(<sc-id>):
9481 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
9482 counter designated by <sc-id>. If an error occurs, this action silently
9483 fails and the actions evaluation continues.
9484
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009485 - sc-inc-gpc1(<sc-id>):
9486 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
9487 counter designated by <sc-id>. If an error occurs, this action silently
9488 fails and the actions evaluation continues.
9489
Cédric Dufour0d7712d2019-11-06 18:38:53 +01009490 - sc-set-gpt0(<sc-id>) { <int> | <expr> }:
9491 This action sets the 32-bit unsigned GPT0 tag according to the sticky
9492 counter designated by <sc-id> and the value of <int>/<expr>. The
9493 expected result is a boolean. If an error occurs, this action silently
9494 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009495
William Lallemand2e785f22016-05-25 01:48:42 +02009496 - set-src <expr> :
9497 Is used to set the source IP address to the value of specified
9498 expression. Useful if you want to mask source IP for privacy.
9499 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009500 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +02009501
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009502 Arguments:
9503 <expr> Is a standard HAProxy expression formed by a sample-fetch
9504 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +02009505
9506 Example:
William Lallemand2e785f22016-05-25 01:48:42 +02009507 tcp-request connection set-src src,ipmask(24)
9508
Willy Tarreau0c630532016-10-21 17:52:58 +02009509 When possible, set-src preserves the original source port as long as the
9510 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +02009511
William Lallemand44be6402016-05-25 01:51:35 +02009512 - set-src-port <expr> :
9513 Is used to set the source port address to the value of specified
9514 expression.
9515
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009516 Arguments:
9517 <expr> Is a standard HAProxy expression formed by a sample-fetch
9518 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +02009519
9520 Example:
William Lallemand44be6402016-05-25 01:51:35 +02009521 tcp-request connection set-src-port int(4000)
9522
Willy Tarreau0c630532016-10-21 17:52:58 +02009523 When possible, set-src-port preserves the original source address as long
9524 as the address family supports a port, otherwise it forces the source
9525 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02009526
William Lallemand13e9b0c2016-05-25 02:34:07 +02009527 - set-dst <expr> :
9528 Is used to set the destination IP address to the value of specified
9529 expression. Useful if you want to mask IP for privacy in log.
9530 If you want to provide an IP from a HTTP header use "http-request
9531 set-dst". If you want to connect to the new address/port, use
9532 '0.0.0.0:0' as a server address in the backend.
9533
9534 <expr> Is a standard HAProxy expression formed by a sample-fetch
9535 followed by some converters.
9536
9537 Example:
9538
9539 tcp-request connection set-dst dst,ipmask(24)
9540 tcp-request connection set-dst ipv4(10.0.0.1)
9541
Willy Tarreau0c630532016-10-21 17:52:58 +02009542 When possible, set-dst preserves the original destination port as long as
9543 the address family allows it, otherwise the destination port is set to 0.
9544
William Lallemand13e9b0c2016-05-25 02:34:07 +02009545 - set-dst-port <expr> :
9546 Is used to set the destination port address to the value of specified
9547 expression. If you want to connect to the new address/port, use
9548 '0.0.0.0:0' as a server address in the backend.
9549
9550
9551 <expr> Is a standard HAProxy expression formed by a sample-fetch
9552 followed by some converters.
9553
9554 Example:
9555
9556 tcp-request connection set-dst-port int(4000)
9557
Willy Tarreau0c630532016-10-21 17:52:58 +02009558 When possible, set-dst-port preserves the original destination address as
9559 long as the address family supports a port, otherwise it forces the
9560 destination address to IPv4 "0.0.0.0" before rewriting the port.
9561
Willy Tarreau2d392c22015-08-24 01:43:45 +02009562 - "silent-drop" :
9563 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009564 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009565 to prevent the client from being notified. The effect it then that the
9566 client still sees an established connection while there's none on
9567 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9568 except that it doesn't use any local resource at all on the machine
9569 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009570 slow down stronger attackers. It is important to understand the impact
9571 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009572 client and HAProxy (firewalls, proxies, load balancers) will also keep
9573 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +01009574 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009575 TCP_REPAIR socket option is used to block the emission of a TCP
9576 reset. On other systems, the socket's TTL is reduced to 1 so that the
9577 TCP reset doesn't pass the first router, though it's still delivered to
9578 local networks. Do not use it unless you fully understand how it works.
9579
Willy Tarreaue9656522010-08-17 15:40:09 +02009580 Note that the "if/unless" condition is optional. If no condition is set on
9581 the action, it is simply performed unconditionally. That can be useful for
9582 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009583
Willy Tarreaue9656522010-08-17 15:40:09 +02009584 Example: accept all connections from white-listed hosts, reject too fast
9585 connection without counting them, and track accepted connections.
9586 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009587
Willy Tarreaue9656522010-08-17 15:40:09 +02009588 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009589 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009590 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009591
Willy Tarreaue9656522010-08-17 15:40:09 +02009592 Example: accept all connections from white-listed hosts, count all other
9593 connections and reject too fast ones. This results in abusive ones
9594 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009595
Willy Tarreaue9656522010-08-17 15:40:09 +02009596 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009597 tcp-request connection track-sc0 src
9598 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009599
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009600 Example: enable the PROXY protocol for traffic coming from all known proxies.
9601
9602 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9603
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009604 See section 7 about ACL usage.
9605
Willy Tarreau4f614292016-10-21 17:49:36 +02009606 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009607
9608
Willy Tarreaue9656522010-08-17 15:40:09 +02009609tcp-request content <action> [{if | unless} <condition>]
9610 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02009611 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009612 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02009613 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009614 <action> defines the action to perform if the condition applies. See
9615 below.
Willy Tarreau62644772008-07-16 18:36:06 +02009616
Willy Tarreaue9656522010-08-17 15:40:09 +02009617 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02009618
Davor Ocelice9ed2812017-12-25 17:49:28 +01009619 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +02009620 called "TCP content inspection". During this stage, ACL-based rules are
9621 evaluated every time the request contents are updated, until either an
9622 "accept" or a "reject" rule matches, or the TCP request inspection delay
9623 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02009624
Willy Tarreaue9656522010-08-17 15:40:09 +02009625 The first difference between these rules and "tcp-request connection" rules
9626 is that "tcp-request content" rules can make use of contents to take a
9627 decision. Most often, these decisions will consider a protocol recognition or
9628 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01009629 both frontends and backends. In case of HTTP keep-alive with the client, all
9630 tcp-request content rules are evaluated again, so haproxy keeps a record of
9631 what sticky counters were assigned by a "tcp-request connection" versus a
9632 "tcp-request content" rule, and flushes all the content-related ones after
9633 processing an HTTP request, so that they may be evaluated again by the rules
9634 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009635 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01009636 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02009637
Willy Tarreaue9656522010-08-17 15:40:09 +02009638 Content-based rules are evaluated in their exact declaration order. If no
9639 rule matches or if there is no rule, the default action is to accept the
9640 contents. There is no specific limit to the number of rules which may be
9641 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02009642
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009643 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009644 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +01009645 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009646 - reject : the request is rejected and the connection is closed
9647 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -04009648 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009649 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009650 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009651 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +01009652 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +02009653 - set-dst <expr>
9654 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009655 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +01009656 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +02009657 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +01009658 - send-spoe-group <engine-name> <group-name>
Christopher Faulet579d83b2019-11-22 15:34:17 +01009659 - use-service <service-name>
Willy Tarreau62644772008-07-16 18:36:06 +02009660
Willy Tarreaue9656522010-08-17 15:40:09 +02009661 They have the same meaning as their counter-parts in "tcp-request connection"
9662 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +01009663 For "do-resolve" action, please check the "http-request do-resolve"
9664 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +02009665
Willy Tarreauf3338342014-01-28 21:40:28 +01009666 While there is nothing mandatory about it, it is recommended to use the
9667 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
9668 content" rules in the frontend, and track-sc2 for "tcp-request content"
9669 rules in the backend, because that makes the configuration more readable
9670 and easier to troubleshoot, but this is just a guideline and all counters
9671 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02009672
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009673 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02009674 the action, it is simply performed unconditionally. That can be useful for
9675 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02009676
Willy Tarreaue9656522010-08-17 15:40:09 +02009677 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02009678 rules, since HTTP-specific ACL matches are able to preliminarily parse the
9679 contents of a buffer before extracting the required data. If the buffered
9680 contents do not parse as a valid HTTP message, then the ACL does not match.
9681 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01009682 processing, so there is no risk of parsing something differently. In an HTTP
9683 backend connected to from an HTTP frontend, it is guaranteed that HTTP
9684 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02009685
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009686 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009687 are present when the rule is processed. The rule processing engine is able to
9688 wait until the inspect delay expires when the data to be tracked is not yet
9689 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009690
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +02009691 The "set-dst" and "set-dst-port" are used to set respectively the destination
9692 IP and port. More information on how to use it at "http-request set-dst".
9693
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009694 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +02009695 declared inline. For "tcp-request session" rules, only session-level
9696 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009697
Daniel Schneller0b547052016-03-21 20:46:57 +01009698 <var-name> The name of the variable starts with an indication about
9699 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009700 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009701 "sess" : the variable is shared with the whole session
9702 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009703 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009704 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009705 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009706 "res" : the variable is shared only during response
9707 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009708 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009709 The name may only contain characters 'a-z', 'A-Z', '0-9',
9710 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009711
9712 <expr> Is a standard HAProxy expression formed by a sample-fetch
9713 followed by some converters.
9714
Christopher Faulet85d79c92016-11-09 16:54:56 +01009715 The "unset-var" is used to unset a variable. See above for details about
9716 <var-name>.
9717
Patrick Hemmer268a7072018-05-11 12:52:31 -04009718 The "set-priority-class" is used to set the queue priority class of the
9719 current request. The value must be a sample expression which converts to an
9720 integer in the range -2047..2047. Results outside this range will be
9721 truncated. The priority class determines the order in which queued requests
9722 are processed. Lower values have higher priority.
9723
9724 The "set-priority-offset" is used to set the queue priority timestamp offset
9725 of the current request. The value must be a sample expression which converts
9726 to an integer in the range -524287..524287. Results outside this range will be
9727 truncated. When a request is queued, it is ordered first by the priority
9728 class, then by the current timestamp adjusted by the given offset in
9729 milliseconds. Lower values have higher priority.
9730 Note that the resulting timestamp is is only tracked with enough precision for
9731 524,287ms (8m44s287ms). If the request is queued long enough to where the
9732 adjusted timestamp exceeds this value, it will be misidentified as highest
9733 priority. Thus it is important to set "timeout queue" to a value, where when
9734 combined with the offset, does not exceed this limit.
9735
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009736 The "send-spoe-group" is used to trigger sending of a group of SPOE
9737 messages. To do so, the SPOE engine used to send messages must be defined, as
9738 well as the SPOE group to send. Of course, the SPOE engine must refer to an
9739 existing SPOE filter. If not engine name is provided on the SPOE filter line,
9740 the SPOE agent name must be used.
9741
9742 <engine-name> The SPOE engine name.
9743
9744 <group-name> The SPOE group name as specified in the engine configuration.
9745
Christopher Faulet579d83b2019-11-22 15:34:17 +01009746 The "use-service" is used to executes a TCP service which will reply to the
9747 request and stop the evaluation of the rules. This service may choose to
9748 reply by sending any valid response or it may immediately close the
9749 connection without sending anything. Outside natives services, it is possible
9750 to write your own services in Lua. No further "tcp-request" rules are
9751 evaluated.
9752
9753 Example:
9754 tcp-request content use-service lua.deny { src -f /etc/haproxy/blacklist.lst }
9755
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009756 Example:
9757
9758 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +01009759 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009760
Willy Tarreau62644772008-07-16 18:36:06 +02009761 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02009762 # Accept HTTP requests containing a Host header saying "example.com"
9763 # and reject everything else.
9764 acl is_host_com hdr(Host) -i example.com
9765 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02009766 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02009767 tcp-request content reject
9768
9769 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02009770 # reject SMTP connection if client speaks first
9771 tcp-request inspect-delay 30s
9772 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009773 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02009774
9775 # Forward HTTPS connection only if client speaks
9776 tcp-request inspect-delay 30s
9777 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009778 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02009779 tcp-request content reject
9780
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009781 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009782 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009783 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009784 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009785 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
9786 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009787
9788 Example:
9789 # track request counts per "base" (concatenation of Host+URL)
9790 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02009791 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009792
Willy Tarreaue9656522010-08-17 15:40:09 +02009793 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009794 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +02009795
9796 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +01009797 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02009798 # protecting all our sites
9799 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009800 tcp-request connection track-sc0 src
9801 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02009802 ...
9803 use_backend http_dynamic if { path_end .php }
9804
9805 backend http_dynamic
9806 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009807 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02009808 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009809 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03009810 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009811 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02009812 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02009813
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009814 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02009815
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03009816 See also : "tcp-request connection", "tcp-request session",
9817 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +02009818
9819
9820tcp-request inspect-delay <timeout>
9821 Set the maximum allowed time to wait for data during content inspection
9822 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009823 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02009824 Arguments :
9825 <timeout> is the timeout value specified in milliseconds by default, but
9826 can be in any other unit if the number is suffixed by the unit,
9827 as explained at the top of this document.
9828
9829 People using haproxy primarily as a TCP relay are often worried about the
9830 risk of passing any type of protocol to a server without any analysis. In
9831 order to be able to analyze the request contents, we must first withhold
9832 the data then analyze them. This statement simply enables withholding of
9833 data for at most the specified amount of time.
9834
Willy Tarreaufb356202010-08-03 14:02:05 +02009835 TCP content inspection applies very early when a connection reaches a
9836 frontend, then very early when the connection is forwarded to a backend. This
9837 means that a connection may experience a first delay in the frontend and a
9838 second delay in the backend if both have tcp-request rules.
9839
Willy Tarreau62644772008-07-16 18:36:06 +02009840 Note that when performing content inspection, haproxy will evaluate the whole
9841 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009842 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02009843 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01009844 contents are definitive. If no delay is set, haproxy will not wait at all
9845 and will immediately apply a verdict based on the available information.
9846 Obviously this is unlikely to be very useful and might even be racy, so such
9847 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02009848
9849 As soon as a rule matches, the request is released and continues as usual. If
9850 the timeout is reached and no rule matches, the default policy will be to let
9851 it pass through unaffected.
9852
9853 For most protocols, it is enough to set it to a few seconds, as most clients
9854 send the full request immediately upon connection. Add 3 or more seconds to
9855 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01009856 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +01009857 before the server (e.g. SMTP), or to wait for a client to talk before passing
9858 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02009859 least the inspection delay, otherwise it will expire first. If the client
9860 closes the connection or if the buffer is full, the delay immediately expires
9861 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02009862
Willy Tarreau55165fe2009-05-10 12:02:55 +02009863 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02009864 "timeout client".
9865
9866
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009867tcp-response content <action> [{if | unless} <condition>]
9868 Perform an action on a session response depending on a layer 4-7 condition
9869 May be used in sections : defaults | frontend | listen | backend
9870 no | no | yes | yes
9871 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009872 <action> defines the action to perform if the condition applies. See
9873 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009874
9875 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
9876
Davor Ocelice9ed2812017-12-25 17:49:28 +01009877 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009878 called "TCP content inspection". During this stage, ACL-based rules are
9879 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009880 "accept", "close" or a "reject" rule matches, or a TCP response inspection
9881 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009882
9883 Most often, these decisions will consider a protocol recognition or validity.
9884
9885 Content-based rules are evaluated in their exact declaration order. If no
9886 rule matches or if there is no rule, the default action is to accept the
9887 contents. There is no specific limit to the number of rules which may be
9888 inserted.
9889
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009890 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009891 - accept :
9892 accepts the response if the condition is true (when used with "if")
9893 or false (when used with "unless"). The first such rule executed ends
9894 the rules evaluation.
9895
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009896 - close :
9897 immediately closes the connection with the server if the condition is
9898 true (when used with "if"), or false (when used with "unless"). The
9899 first such rule executed ends the rules evaluation. The main purpose of
9900 this action is to force a connection to be finished between a client
9901 and a server after an exchange when the application protocol expects
9902 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009903 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02009904 protocols.
9905
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009906 - reject :
9907 rejects the response if the condition is true (when used with "if")
9908 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009909 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009910
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009911 - set-var(<var-name>) <expr>
9912 Sets a variable.
9913
Christopher Faulet85d79c92016-11-09 16:54:56 +01009914 - unset-var(<var-name>)
9915 Unsets a variable.
9916
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009917 - sc-inc-gpc0(<sc-id>):
9918 This action increments the GPC0 counter according to the sticky
9919 counter designated by <sc-id>. If an error occurs, this action fails
9920 silently and the actions evaluation continues.
9921
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009922 - sc-inc-gpc1(<sc-id>):
9923 This action increments the GPC1 counter according to the sticky
9924 counter designated by <sc-id>. If an error occurs, this action fails
9925 silently and the actions evaluation continues.
9926
Cédric Dufour0d7712d2019-11-06 18:38:53 +01009927 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
9928 This action sets the 32-bit unsigned GPT0 tag according to the sticky
9929 counter designated by <sc-id> and the value of <int>/<expr>. The
9930 expected result is a boolean. If an error occurs, this action silently
9931 fails and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009932
Willy Tarreau2d392c22015-08-24 01:43:45 +02009933 - "silent-drop" :
9934 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009935 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009936 to prevent the client from being notified. The effect it then that the
9937 client still sees an established connection while there's none on
9938 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9939 except that it doesn't use any local resource at all on the machine
9940 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009941 slow down stronger attackers. It is important to understand the impact
9942 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009943 client and HAProxy (firewalls, proxies, load balancers) will also keep
9944 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +01009945 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009946 TCP_REPAIR socket option is used to block the emission of a TCP
9947 reset. On other systems, the socket's TTL is reduced to 1 so that the
9948 TCP reset doesn't pass the first router, though it's still delivered to
9949 local networks. Do not use it unless you fully understand how it works.
9950
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009951 - send-spoe-group <engine-name> <group-name>
9952 Send a group of SPOE messages.
9953
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009954 Note that the "if/unless" condition is optional. If no condition is set on
9955 the action, it is simply performed unconditionally. That can be useful for
9956 for changing the default action to a reject.
9957
Jamie Gloudonaaa21002012-08-25 00:18:33 -04009958 It is perfectly possible to match layer 7 contents with "tcp-response
9959 content" rules, but then it is important to ensure that a full response has
9960 been buffered, otherwise no contents will match. In order to achieve this,
9961 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02009962 period.
9963
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009964 The "set-var" is used to set the content of a variable. The variable is
9965 declared inline.
9966
Daniel Schneller0b547052016-03-21 20:46:57 +01009967 <var-name> The name of the variable starts with an indication about
9968 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +01009969 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +01009970 "sess" : the variable is shared with the whole session
9971 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009972 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +01009973 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009974 processing
Daniel Schneller0b547052016-03-21 20:46:57 +01009975 "res" : the variable is shared only during response
9976 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009977 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +01009978 The name may only contain characters 'a-z', 'A-Z', '0-9',
9979 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02009980
9981 <expr> Is a standard HAProxy expression formed by a sample-fetch
9982 followed by some converters.
9983
9984 Example:
9985
9986 tcp-request content set-var(sess.my_var) src
9987
Christopher Faulet85d79c92016-11-09 16:54:56 +01009988 The "unset-var" is used to unset a variable. See above for details about
9989 <var-name>.
9990
9991 Example:
9992
9993 tcp-request content unset-var(sess.my_var)
9994
Christopher Faulet76c09ef2017-09-21 11:03:52 +02009995 The "send-spoe-group" is used to trigger sending of a group of SPOE
9996 messages. To do so, the SPOE engine used to send messages must be defined, as
9997 well as the SPOE group to send. Of course, the SPOE engine must refer to an
9998 existing SPOE filter. If not engine name is provided on the SPOE filter line,
9999 the SPOE agent name must be used.
10000
10001 <engine-name> The SPOE engine name.
10002
10003 <group-name> The SPOE group name as specified in the engine configuration.
10004
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010005 See section 7 about ACL usage.
10006
10007 See also : "tcp-request content", "tcp-response inspect-delay"
10008
10009
Willy Tarreau4f614292016-10-21 17:49:36 +020010010tcp-request session <action> [{if | unless} <condition>]
10011 Perform an action on a validated session depending on a layer 5 condition
10012 May be used in sections : defaults | frontend | listen | backend
10013 no | yes | yes | no
10014 Arguments :
10015 <action> defines the action to perform if the condition applies. See
10016 below.
10017
10018 <condition> is a standard layer5-only ACL-based condition (see section 7).
10019
Davor Ocelice9ed2812017-12-25 17:49:28 +010010020 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020010021 it is possible to evaluate some conditions to decide whether this session
10022 must be accepted or dropped or have its counters tracked. Those conditions
10023 cannot make use of any data contents because no buffers are allocated yet and
10024 the processing cannot wait at this stage. The main use case it to copy some
10025 early information into variables (since variables are accessible in the
10026 session), or to keep track of some information collected after the handshake,
10027 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010010028 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020010029 extracted information can thus be copied to a variable or tracked using
10030 "track-sc" rules. Of course it is also possible to decide to accept/reject as
10031 with other rulesets. Most operations performed here could also be performed
10032 in "tcp-request content" rules, except that in HTTP these rules are evaluated
10033 for each new request, and that might not always be acceptable. For example a
10034 rule might increment a counter on each evaluation. It would also be possible
10035 that a country is resolved by geolocation from the source IP address,
10036 assigned to a session-wide variable, then the source address rewritten from
10037 an HTTP header for all requests. If some contents need to be inspected in
10038 order to take the decision, the "tcp-request content" statements must be used
10039 instead.
10040
10041 The "tcp-request session" rules are evaluated in their exact declaration
10042 order. If no rule matches or if there is no rule, the default action is to
10043 accept the incoming session. There is no specific limit to the number of
10044 rules which may be inserted.
10045
10046 Several types of actions are supported :
10047 - accept : the request is accepted
10048 - reject : the request is rejected and the connection is closed
10049 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
10050 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010051 - sc-inc-gpc1(<sc-id>)
Cédric Dufour0d7712d2019-11-06 18:38:53 +010010052 - sc-set-gpt0(<sc-id>) { <int> | <expr> }
Willy Tarreau4f614292016-10-21 17:49:36 +020010053 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010054 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010055 - silent-drop
10056
10057 These actions have the same meaning as their respective counter-parts in
10058 "tcp-request connection" and "tcp-request content", so please refer to these
10059 sections for a complete description.
10060
10061 Note that the "if/unless" condition is optional. If no condition is set on
10062 the action, it is simply performed unconditionally. That can be useful for
10063 "track-sc*" actions as well as for changing the default action to a reject.
10064
10065 Example: track the original source address by default, or the one advertised
10066 in the PROXY protocol header for connection coming from the local
10067 proxies. The first connection-level rule enables receipt of the
10068 PROXY protocol for these ones, the second rule tracks whatever
10069 address we decide to keep after optional decoding.
10070
10071 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10072 tcp-request session track-sc0 src
10073
10074 Example: accept all sessions from white-listed hosts, reject too fast
10075 sessions without counting them, and track accepted sessions.
10076 This results in session rate being capped from abusive sources.
10077
10078 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10079 tcp-request session reject if { src_sess_rate gt 10 }
10080 tcp-request session track-sc0 src
10081
10082 Example: accept all sessions from white-listed hosts, count all other
10083 sessions and reject too fast ones. This results in abusive ones
10084 being blocked as long as they don't slow down.
10085
10086 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10087 tcp-request session track-sc0 src
10088 tcp-request session reject if { sc0_sess_rate gt 10 }
10089
10090 See section 7 about ACL usage.
10091
10092 See also : "tcp-request connection", "tcp-request content", "stick-table"
10093
10094
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010095tcp-response inspect-delay <timeout>
10096 Set the maximum allowed time to wait for a response during content inspection
10097 May be used in sections : defaults | frontend | listen | backend
10098 no | no | yes | yes
10099 Arguments :
10100 <timeout> is the timeout value specified in milliseconds by default, but
10101 can be in any other unit if the number is suffixed by the unit,
10102 as explained at the top of this document.
10103
10104 See also : "tcp-response content", "tcp-request inspect-delay".
10105
10106
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010107timeout check <timeout>
10108 Set additional check timeout, but only after a connection has been already
10109 established.
10110
10111 May be used in sections: defaults | frontend | listen | backend
10112 yes | no | yes | yes
10113 Arguments:
10114 <timeout> is the timeout value specified in milliseconds by default, but
10115 can be in any other unit if the number is suffixed by the unit,
10116 as explained at the top of this document.
10117
10118 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
10119 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010010120 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010121 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010010122 (Please also note that there is no valid reason to have such long connect
10123 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
10124 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010125
10126 If "timeout check" is not set haproxy uses "inter" for complete check
10127 timeout (connect + read) exactly like all <1.3.15 version.
10128
10129 In most cases check request is much simpler and faster to handle than normal
10130 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010010131 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010132
10133 This parameter is specific to backends, but can be specified once for all in
10134 "defaults" sections. This is in fact one of the easiest solutions not to
10135 forget about it.
10136
Willy Tarreau41a340d2008-01-22 12:25:31 +010010137 See also: "timeout connect", "timeout queue", "timeout server",
10138 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010139
10140
Willy Tarreau0ba27502007-12-24 16:55:16 +010010141timeout client <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010010142 Set the maximum inactivity time on the client side.
10143 May be used in sections : defaults | frontend | listen | backend
10144 yes | yes | yes | no
10145 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010146 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010147 can be in any other unit if the number is suffixed by the unit,
10148 as explained at the top of this document.
10149
10150 The inactivity timeout applies when the client is expected to acknowledge or
10151 send data. In HTTP mode, this timeout is particularly important to consider
10152 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010153 response while it is reading data sent by the server. That said, for the
10154 first phase, it is preferable to set the "timeout http-request" to better
10155 protect HAProxy from Slowloris like attacks. The value is specified in
10156 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010010157 suffixed by the unit, as specified at the top of this document. In TCP mode
10158 (and to a lesser extent, in HTTP mode), it is highly recommended that the
10159 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010160 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010010161 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010010162 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
10163 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020010164 which overrides "timeout client" and "timeout server" for tunnels, as well as
10165 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010166
10167 This parameter is specific to frontends, but can be specified once for all in
10168 "defaults" sections. This is in fact one of the easiest solutions not to
10169 forget about it. An unspecified timeout results in an infinite timeout, which
10170 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010171 during startup because it may result in accumulation of expired sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010172 the system if the system's timeouts are not configured either.
10173
Willy Tarreau95c4e142017-11-26 12:18:55 +010010174 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010010175
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010176 See also : "timeout server", "timeout tunnel", "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010177
Willy Tarreau0ba27502007-12-24 16:55:16 +010010178
Willy Tarreau05cdd962014-05-10 14:30:07 +020010179timeout client-fin <timeout>
10180 Set the inactivity timeout on the client side for half-closed connections.
10181 May be used in sections : defaults | frontend | listen | backend
10182 yes | yes | yes | no
10183 Arguments :
10184 <timeout> is the timeout value specified in milliseconds by default, but
10185 can be in any other unit if the number is suffixed by the unit,
10186 as explained at the top of this document.
10187
10188 The inactivity timeout applies when the client is expected to acknowledge or
10189 send data while one direction is already shut down. This timeout is different
10190 from "timeout client" in that it only applies to connections which are closed
10191 in one direction. This is particularly useful to avoid keeping connections in
10192 FIN_WAIT state for too long when clients do not disconnect cleanly. This
10193 problem is particularly common long connections such as RDP or WebSocket.
10194 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010010195 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
10196 frame was sent, often indicating an expectation that the connection quickly
10197 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020010198
10199 This parameter is specific to frontends, but can be specified once for all in
10200 "defaults" sections. By default it is not set, so half-closed connections
10201 will use the other timeouts (timeout.client or timeout.tunnel).
10202
10203 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
10204
10205
Willy Tarreau0ba27502007-12-24 16:55:16 +010010206timeout connect <timeout>
Willy Tarreau0ba27502007-12-24 16:55:16 +010010207 Set the maximum time to wait for a connection attempt to a server to succeed.
10208 May be used in sections : defaults | frontend | listen | backend
10209 yes | no | yes | yes
10210 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010211 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010212 can be in any other unit if the number is suffixed by the unit,
10213 as explained at the top of this document.
10214
10215 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010216 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010010217 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010010218 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010219 connect timeout also presets both queue and tarpit timeouts to the same value
10220 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010221
10222 This parameter is specific to backends, but can be specified once for all in
10223 "defaults" sections. This is in fact one of the easiest solutions not to
10224 forget about it. An unspecified timeout results in an infinite timeout, which
10225 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010226 during startup because it may result in accumulation of failed sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010227 the system if the system's timeouts are not configured either.
10228
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010229 See also: "timeout check", "timeout queue", "timeout server", "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010230
Willy Tarreau0ba27502007-12-24 16:55:16 +010010231
Willy Tarreaub16a5742010-01-10 14:46:16 +010010232timeout http-keep-alive <timeout>
10233 Set the maximum allowed time to wait for a new HTTP request to appear
10234 May be used in sections : defaults | frontend | listen | backend
10235 yes | yes | yes | yes
10236 Arguments :
10237 <timeout> is the timeout value specified in milliseconds by default, but
10238 can be in any other unit if the number is suffixed by the unit,
10239 as explained at the top of this document.
10240
10241 By default, the time to wait for a new request in case of keep-alive is set
10242 by "timeout http-request". However this is not always convenient because some
10243 people want very short keep-alive timeouts in order to release connections
10244 faster, and others prefer to have larger ones but still have short timeouts
10245 once the request has started to present itself.
10246
10247 The "http-keep-alive" timeout covers these needs. It will define how long to
10248 wait for a new HTTP request to start coming after a response was sent. Once
10249 the first byte of request has been seen, the "http-request" timeout is used
10250 to wait for the complete request to come. Note that empty lines prior to a
10251 new request do not refresh the timeout and are not counted as a new request.
10252
10253 There is also another difference between the two timeouts : when a connection
10254 expires during timeout http-keep-alive, no error is returned, the connection
10255 just closes. If the connection expires in "http-request" while waiting for a
10256 connection to complete, a HTTP 408 error is returned.
10257
10258 In general it is optimal to set this value to a few tens to hundreds of
10259 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010260 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010261 1 millisecond) it will probably only accept pipelined requests but not the
10262 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010263 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010264
10265 If this parameter is not set, the "http-request" timeout applies, and if both
10266 are not set, "timeout client" still applies at the lower level. It should be
10267 set in the frontend to take effect, unless the frontend is in TCP mode, in
10268 which case the HTTP backend's timeout will be used.
10269
Willy Tarreau95c4e142017-11-26 12:18:55 +010010270 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10271 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010272 (where we only have one connection per client and a connection setup).
10273
Willy Tarreaub16a5742010-01-10 14:46:16 +010010274 See also : "timeout http-request", "timeout client".
10275
10276
Willy Tarreau036fae02008-01-06 13:24:40 +010010277timeout http-request <timeout>
10278 Set the maximum allowed time to wait for a complete HTTP request
10279 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010280 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010281 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010282 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010283 can be in any other unit if the number is suffixed by the unit,
10284 as explained at the top of this document.
10285
10286 In order to offer DoS protection, it may be required to lower the maximum
10287 accepted time to receive a complete HTTP request without affecting the client
10288 timeout. This helps protecting against established connections on which
10289 nothing is sent. The client timeout cannot offer a good protection against
10290 this abuse because it is an inactivity timeout, which means that if the
10291 attacker sends one character every now and then, the timeout will not
10292 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010293 types, the request will be aborted if it does not complete in time. When the
10294 timeout expires, an HTTP 408 response is sent to the client to inform it
10295 about the problem, and the connection is closed. The logs will report
10296 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010297 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010298 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10299 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010300
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010301 By default, this timeout only applies to the header part of the request,
10302 and not to any data. As soon as the empty line is received, this timeout is
10303 not used anymore. When combined with "option http-buffer-request", this
10304 timeout also applies to the body of the request..
10305 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010306 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010307
10308 Generally it is enough to set it to a few seconds, as most clients send the
10309 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010310 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010311 generally work on local networks as long as there are no packet losses. This
10312 will prevent people from sending bare HTTP requests using telnet.
10313
10314 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010315 chunk of the incoming request. It should be set in the frontend to take
10316 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10317 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010318
Willy Tarreau0f228a02015-05-01 15:37:53 +020010319 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010320 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010321
Willy Tarreau844e3c52008-01-11 16:28:18 +010010322
10323timeout queue <timeout>
10324 Set the maximum time to wait in the queue for a connection slot to be free
10325 May be used in sections : defaults | frontend | listen | backend
10326 yes | no | yes | yes
10327 Arguments :
10328 <timeout> is the timeout value specified in milliseconds by default, but
10329 can be in any other unit if the number is suffixed by the unit,
10330 as explained at the top of this document.
10331
10332 When a server's maxconn is reached, connections are left pending in a queue
10333 which may be server-specific or global to the backend. In order not to wait
10334 indefinitely, a timeout is applied to requests pending in the queue. If the
10335 timeout is reached, it is considered that the request will almost never be
10336 served, so it is dropped and a 503 error is returned to the client.
10337
10338 The "timeout queue" statement allows to fix the maximum time for a request to
10339 be left pending in a queue. If unspecified, the same value as the backend's
10340 connection timeout ("timeout connect") is used, for backwards compatibility
10341 with older versions with no "timeout queue" parameter.
10342
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010343 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010344
10345
10346timeout server <timeout>
Willy Tarreau844e3c52008-01-11 16:28:18 +010010347 Set the maximum inactivity time on the server side.
10348 May be used in sections : defaults | frontend | listen | backend
10349 yes | no | yes | yes
10350 Arguments :
10351 <timeout> is the timeout value specified in milliseconds by default, but
10352 can be in any other unit if the number is suffixed by the unit,
10353 as explained at the top of this document.
10354
10355 The inactivity timeout applies when the server is expected to acknowledge or
10356 send data. In HTTP mode, this timeout is particularly important to consider
10357 during the first phase of the server's response, when it has to send the
10358 headers, as it directly represents the server's processing time for the
10359 request. To find out what value to put there, it's often good to start with
10360 what would be considered as unacceptable response times, then check the logs
10361 to observe the response time distribution, and adjust the value accordingly.
10362
10363 The value is specified in milliseconds by default, but can be in any other
10364 unit if the number is suffixed by the unit, as specified at the top of this
10365 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
10366 recommended that the client timeout remains equal to the server timeout in
10367 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010368 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010010369 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010010370 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
10371 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020010372 "timeout tunnel", which overrides "timeout client" and "timeout server" for
10373 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010374
10375 This parameter is specific to backends, but can be specified once for all in
10376 "defaults" sections. This is in fact one of the easiest solutions not to
10377 forget about it. An unspecified timeout results in an infinite timeout, which
10378 is not recommended. Such a usage is accepted and works but reports a warning
John Roeslerfb2fce12019-07-10 15:45:51 -050010379 during startup because it may result in accumulation of expired sessions in
Willy Tarreau844e3c52008-01-11 16:28:18 +010010380 the system if the system's timeouts are not configured either.
10381
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010382 See also : "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010383
Willy Tarreau05cdd962014-05-10 14:30:07 +020010384
10385timeout server-fin <timeout>
10386 Set the inactivity timeout on the server side for half-closed connections.
10387 May be used in sections : defaults | frontend | listen | backend
10388 yes | no | yes | yes
10389 Arguments :
10390 <timeout> is the timeout value specified in milliseconds by default, but
10391 can be in any other unit if the number is suffixed by the unit,
10392 as explained at the top of this document.
10393
10394 The inactivity timeout applies when the server is expected to acknowledge or
10395 send data while one direction is already shut down. This timeout is different
10396 from "timeout server" in that it only applies to connections which are closed
10397 in one direction. This is particularly useful to avoid keeping connections in
10398 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
10399 This problem is particularly common long connections such as RDP or WebSocket.
10400 Note that this timeout can override "timeout tunnel" when a connection shuts
10401 down in one direction. This setting was provided for completeness, but in most
10402 situations, it should not be needed.
10403
10404 This parameter is specific to backends, but can be specified once for all in
10405 "defaults" sections. By default it is not set, so half-closed connections
10406 will use the other timeouts (timeout.server or timeout.tunnel).
10407
10408 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
10409
Willy Tarreau844e3c52008-01-11 16:28:18 +010010410
10411timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010010412 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010010413 May be used in sections : defaults | frontend | listen | backend
10414 yes | yes | yes | yes
10415 Arguments :
10416 <timeout> is the tarpit duration specified in milliseconds by default, but
10417 can be in any other unit if the number is suffixed by the unit,
10418 as explained at the top of this document.
10419
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020010420 When a connection is tarpitted using "http-request tarpit", it is maintained
10421 open with no activity for a certain amount of time, then closed. "timeout
10422 tarpit" defines how long it will be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010423
10424 The value is specified in milliseconds by default, but can be in any other
10425 unit if the number is suffixed by the unit, as specified at the top of this
10426 document. If unspecified, the same value as the backend's connection timeout
10427 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010010428 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010429
Tim Duesterhus86e6b6e2019-05-14 20:57:59 +020010430 See also : "timeout connect".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010431
10432
Willy Tarreauce887fd2012-05-12 12:50:00 +020010433timeout tunnel <timeout>
10434 Set the maximum inactivity time on the client and server side for tunnels.
10435 May be used in sections : defaults | frontend | listen | backend
10436 yes | no | yes | yes
10437 Arguments :
10438 <timeout> is the timeout value specified in milliseconds by default, but
10439 can be in any other unit if the number is suffixed by the unit,
10440 as explained at the top of this document.
10441
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010442 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020010443 between a client and a server, and the connection remains inactive in both
10444 directions. This timeout supersedes both the client and server timeouts once
10445 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010010446 analyzer remains attached to either connection (e.g. tcp content rules are
10447 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020010448 when switching to the WebSocket protocol, or forwarding a CONNECT request
10449 to a proxy), or after the first response when no keepalive/close option is
10450 specified.
10451
Willy Tarreau05cdd962014-05-10 14:30:07 +020010452 Since this timeout is usually used in conjunction with long-lived connections,
10453 it usually is a good idea to also set "timeout client-fin" to handle the
10454 situation where a client suddenly disappears from the net and does not
10455 acknowledge a close, or sends a shutdown and does not acknowledge pending
10456 data anymore. This can happen in lossy networks where firewalls are present,
10457 and is detected by the presence of large amounts of sessions in a FIN_WAIT
10458 state.
10459
Willy Tarreauce887fd2012-05-12 12:50:00 +020010460 The value is specified in milliseconds by default, but can be in any other
10461 unit if the number is suffixed by the unit, as specified at the top of this
10462 document. Whatever the expected normal idle time, it is a good practice to
10463 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010010464 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020010465
10466 This parameter is specific to backends, but can be specified once for all in
10467 "defaults" sections. This is in fact one of the easiest solutions not to
10468 forget about it.
10469
10470 Example :
10471 defaults http
10472 option http-server-close
10473 timeout connect 5s
10474 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020010475 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020010476 timeout server 30s
10477 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
10478
Willy Tarreau05cdd962014-05-10 14:30:07 +020010479 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020010480
10481
Willy Tarreau844e3c52008-01-11 16:28:18 +010010482transparent (deprecated)
10483 Enable client-side transparent proxying
10484 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010010485 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010010486 Arguments : none
10487
10488 This keyword was introduced in order to provide layer 7 persistence to layer
10489 3 load balancers. The idea is to use the OS's ability to redirect an incoming
10490 connection for a remote address to a local process (here HAProxy), and let
10491 this process know what address was initially requested. When this option is
10492 used, sessions without cookies will be forwarded to the original destination
10493 IP address of the incoming request (which should match that of another
10494 equipment), while requests with cookies will still be forwarded to the
10495 appropriate server.
10496
10497 The "transparent" keyword is deprecated, use "option transparent" instead.
10498
10499 Note that contrary to a common belief, this option does NOT make HAProxy
10500 present the client's IP to the server when establishing the connection.
10501
Willy Tarreau844e3c52008-01-11 16:28:18 +010010502 See also: "option transparent"
10503
William Lallemanda73203e2012-03-12 12:48:57 +010010504unique-id-format <string>
10505 Generate a unique ID for each request.
10506 May be used in sections : defaults | frontend | listen | backend
10507 yes | yes | yes | no
10508 Arguments :
10509 <string> is a log-format string.
10510
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010511 This keyword creates a ID for each request using the custom log format. A
10512 unique ID is useful to trace a request passing through many components of
10513 a complex infrastructure. The newly created ID may also be logged using the
10514 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010010515
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010516 The format should be composed from elements that are guaranteed to be
10517 unique when combined together. For instance, if multiple haproxy instances
10518 are involved, it might be important to include the node name. It is often
10519 needed to log the incoming connection's source and destination addresses
10520 and ports. Note that since multiple requests may be performed over the same
10521 connection, including a request counter may help differentiate them.
10522 Similarly, a timestamp may protect against a rollover of the counter.
10523 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010010524
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010525 It is recommended to use hexadecimal notation for many fields since it
10526 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010010527
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010528 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010529
Julien Vehentf21be322014-03-07 08:27:34 -050010530 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010531
10532 will generate:
10533
10534 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10535
10536 See also: "unique-id-header"
10537
10538unique-id-header <name>
10539 Add a unique ID header in the HTTP request.
10540 May be used in sections : defaults | frontend | listen | backend
10541 yes | yes | yes | no
10542 Arguments :
10543 <name> is the name of the header.
10544
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010545 Add a unique-id header in the HTTP request sent to the server, using the
10546 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010010547
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010548 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010549
Julien Vehentf21be322014-03-07 08:27:34 -050010550 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010551 unique-id-header X-Unique-ID
10552
10553 will generate:
10554
10555 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10556
10557 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010010558
Willy Tarreauf51658d2014-04-23 01:21:56 +020010559use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010560 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010561 May be used in sections : defaults | frontend | listen | backend
10562 no | yes | yes | no
10563 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010564 <backend> is the name of a valid backend or "listen" section, or a
10565 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010566
Willy Tarreauf51658d2014-04-23 01:21:56 +020010567 <condition> is a condition composed of ACLs, as described in section 7. If
10568 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010569
10570 When doing content-switching, connections arrive on a frontend and are then
10571 dispatched to various backends depending on a number of conditions. The
10572 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010573 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010010574 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010575 source address validation) or combined with a "tcp-request" rule to wait for
10576 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010577
10578 There may be as many "use_backend" rules as desired. All of these rules are
10579 evaluated in their declaration order, and the first one which matches will
10580 assign the backend.
10581
10582 In the first form, the backend will be used if the condition is met. In the
10583 second form, the backend will be used if the condition is not met. If no
10584 condition is valid, the backend defined with "default_backend" will be used.
10585 If no default backend is defined, either the servers in the same section are
10586 used (in case of a "listen" section) or, in case of a frontend, no server is
10587 used and a 503 service unavailable response is returned.
10588
Willy Tarreau51aecc72009-07-12 09:47:04 +020010589 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010590 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020010591 and backend processing will immediately follow, or the backend will wait for
10592 a complete HTTP request to get in. This feature is useful when a frontend
10593 must decode several protocols on a unique port, one of them being HTTP.
10594
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010595 When <backend> is a simple name, it is resolved at configuration time, and an
10596 error is reported if the specified backend does not exist. If <backend> is
10597 a log-format string instead, no check may be done at configuration time, so
10598 the backend name is resolved dynamically at run time. If the resulting
10599 backend name does not correspond to any valid backend, no other rule is
10600 evaluated, and the default_backend directive is applied instead. Note that
10601 when using dynamic backend names, it is highly recommended to use a prefix
10602 that no other backend uses in order to ensure that an unauthorized backend
10603 cannot be forced from the request.
10604
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010605 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010606 used to detect the association between frontends and backends to compute the
10607 backend's "fullconn" setting. This cannot be done for dynamic names.
10608
10609 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
10610 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010010611
Christopher Fauletb30b3102019-09-12 23:03:09 +020010612use-fcgi-app <name>
10613 Defines the FastCGI application to use for the backend.
10614 May be used in sections : defaults | frontend | listen | backend
10615 no | no | yes | yes
10616 Arguments :
10617 <name> is the name of the FastCGI application to use.
10618
10619 See section 10.1 about FastCGI application setup for details.
Willy Tarreau036fae02008-01-06 13:24:40 +010010620
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010621use-server <server> if <condition>
10622use-server <server> unless <condition>
10623 Only use a specific server if/unless an ACL-based condition is matched.
10624 May be used in sections : defaults | frontend | listen | backend
10625 no | no | yes | yes
10626 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010627 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010628
10629 <condition> is a condition composed of ACLs, as described in section 7.
10630
10631 By default, connections which arrive to a backend are load-balanced across
10632 the available servers according to the configured algorithm, unless a
10633 persistence mechanism such as a cookie is used and found in the request.
10634
10635 Sometimes it is desirable to forward a particular request to a specific
10636 server without having to declare a dedicated backend for this server. This
10637 can be achieved using the "use-server" rules. These rules are evaluated after
10638 the "redirect" rules and before evaluating cookies, and they have precedence
10639 on them. There may be as many "use-server" rules as desired. All of these
10640 rules are evaluated in their declaration order, and the first one which
10641 matches will assign the server.
10642
10643 If a rule designates a server which is down, and "option persist" is not used
10644 and no force-persist rule was validated, it is ignored and evaluation goes on
10645 with the next rules until one matches.
10646
10647 In the first form, the server will be used if the condition is met. In the
10648 second form, the server will be used if the condition is not met. If no
10649 condition is valid, the processing continues and the server will be assigned
10650 according to other persistence mechanisms.
10651
10652 Note that even if a rule is matched, cookie processing is still performed but
10653 does not assign the server. This allows prefixed cookies to have their prefix
10654 stripped.
10655
10656 The "use-server" statement works both in HTTP and TCP mode. This makes it
10657 suitable for use with content-based inspection. For instance, a server could
10658 be selected in a farm according to the TLS SNI field. And if these servers
10659 have their weight set to zero, they will not be used for other traffic.
10660
10661 Example :
10662 # intercept incoming TLS requests based on the SNI field
10663 use-server www if { req_ssl_sni -i www.example.com }
10664 server www 192.168.0.1:443 weight 0
10665 use-server mail if { req_ssl_sni -i mail.example.com }
10666 server mail 192.168.0.1:587 weight 0
10667 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000010668 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010669 # all the rest is forwarded to this server
10670 server default 192.168.0.2:443 check
10671
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010672 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010673
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010674
Davor Ocelice9ed2812017-12-25 17:49:28 +0100106755. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010676--------------------------
10677
10678The "bind", "server" and "default-server" keywords support a number of settings
10679depending on some build options and on the system HAProxy was built on. These
10680settings generally each consist in one word sometimes followed by a value,
10681written on the same line as the "bind" or "server" line. All these options are
10682described in this section.
10683
10684
106855.1. Bind options
10686-----------------
10687
10688The "bind" keyword supports a certain number of settings which are all passed
10689as arguments on the same line. The order in which those arguments appear makes
10690no importance, provided that they appear after the bind address. All of these
10691parameters are optional. Some of them consist in a single words (booleans),
10692while other ones expect a value after them. In this case, the value must be
10693provided immediately after the setting name.
10694
10695The currently supported settings are the following ones.
10696
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010697accept-netscaler-cip <magic number>
10698 Enforces the use of the NetScaler Client IP insertion protocol over any
10699 connection accepted by any of the TCP sockets declared on the same line. The
10700 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
10701 the incoming connection to be used everywhere an address is used, with the
10702 only exception of "tcp-request connection" rules which will only see the
10703 real connection address. Logs will reflect the addresses indicated in the
10704 protocol, unless it is violated, in which case the real address will still
10705 be used. This keyword combined with support from external components can be
10706 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010010707 mechanism which is not always reliable and not even always usable. See also
10708 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
10709 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010010710
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010711accept-proxy
10712 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020010713 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
10714 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010715 3/4 addresses of the incoming connection to be used everywhere an address is
10716 used, with the only exception of "tcp-request connection" rules which will
10717 only see the real connection address. Logs will reflect the addresses
10718 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010010719 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010720 components can be used as an efficient and reliable alternative to the
10721 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010722 usable. See also "tcp-request connection expect-proxy" for a finer-grained
10723 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010724
Olivier Houchardc2aae742017-09-22 18:26:28 +020010725allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010010726 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010010727 due to security considerations. Because it is vulnerable to replay attacks,
John Roeslerfb2fce12019-07-10 15:45:51 -050010728 you should only allow if for requests that are safe to replay, i.e. requests
Olivier Houchard69752962019-01-08 15:35:32 +010010729 that are idempotent. You can use the "wait-for-handshake" action for any
10730 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020010731
Willy Tarreauab861d32013-04-02 02:30:41 +020010732alpn <protocols>
10733 This enables the TLS ALPN extension and advertises the specified protocol
10734 list as supported on top of ALPN. The protocol list consists in a comma-
10735 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050010736 quotes). This requires that the SSL library is built with support for TLS
Willy Tarreauab861d32013-04-02 02:30:41 +020010737 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010010738 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
10739 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
10740 now obsolete NPN extension. At the time of writing this, most browsers still
10741 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
10742 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
10743 are expected to be supported, both versions can be advertised, in order of
10744 preference, like below :
10745
10746 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020010747
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010748backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010010749 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010750 backlog is used instead, which generally defaults to the maxconn value.
10751
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010010752curves <curves>
10753 This setting is only available when support for OpenSSL was built in. It sets
10754 the string describing the list of elliptic curves algorithms ("curve suite")
10755 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
10756 string is a colon-delimited list of curve name.
10757 Example: "X25519:P-256" (without quote)
10758 When "curves" is set, "ecdhe" parameter is ignored.
10759
Emeric Brun7fb34422012-09-28 15:26:15 +020010760ecdhe <named curve>
10761 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010010762 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
10763 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020010764
Emeric Brunfd33a262012-10-11 16:28:27 +020010765ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010766 This setting is only available when support for OpenSSL was built in. It
10767 designates a PEM file from which to load CA certificates used to verify
10768 client's certificate.
10769
Emeric Brunb6dc9342012-09-28 17:55:37 +020010770ca-ignore-err [all|<errorID>,...]
10771 This setting is only available when support for OpenSSL was built in.
10772 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
10773 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
10774 error is ignored.
10775
Christopher Faulet31af49d2015-06-09 17:29:50 +020010776ca-sign-file <cafile>
10777 This setting is only available when support for OpenSSL was built in. It
10778 designates a PEM file containing both the CA certificate and the CA private
10779 key used to create and sign server's certificates. This is a mandatory
10780 setting when the dynamic generation of certificates is enabled. See
10781 'generate-certificates' for details.
10782
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000010783ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020010784 This setting is only available when support for OpenSSL was built in. It is
10785 the CA private key passphrase. This setting is optional and used only when
10786 the dynamic generation of certificates is enabled. See
10787 'generate-certificates' for details.
10788
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010789ciphers <ciphers>
10790 This setting is only available when support for OpenSSL was built in. It sets
10791 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000010792 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000010793 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020010794 information and recommendations see e.g.
10795 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
10796 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
10797 cipher configuration, please check the "ciphersuites" keyword.
10798
10799ciphersuites <ciphersuites>
10800 This setting is only available when support for OpenSSL was built in and
10801 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
10802 the list of cipher algorithms ("cipher suite") that are negotiated during the
10803 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000010804 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
10805 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010806
Emeric Brunfd33a262012-10-11 16:28:27 +020010807crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020010808 This setting is only available when support for OpenSSL was built in. It
10809 designates a PEM file from which to load certificate revocation list used
10810 to verify client's certificate.
10811
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010812crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000010813 This setting is only available when support for OpenSSL was built in. It
10814 designates a PEM file containing both the required certificates and any
10815 associated private keys. This file can be built by concatenating multiple
10816 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
10817 requires an intermediate certificate, this can also be concatenated into this
10818 file.
10819
10820 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
10821 are loaded.
10822
10823 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010824 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010825 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
10826 specified multiple times in order to load certificates from multiple files or
10827 directories. The certificates will be presented to clients who provide a
10828 valid TLS Server Name Indication field matching one of their CN or alt
Davor Ocelice9ed2812017-12-25 17:49:28 +010010829 subjects. Wildcards are supported, where a wildcard character '*' is used
10830 instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010831 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000010832
10833 If no SNI is provided by the client or if the SSL library does not support
10834 TLS extensions, or if the client provides an SNI hostname which does not
10835 match any certificate, then the first loaded certificate will be presented.
10836 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010010837 recommended to load the default one first as a file or to ensure that it will
10838 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000010839
Emeric Brune032bfa2012-09-28 13:01:45 +020010840 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010841
Davor Ocelice9ed2812017-12-25 17:49:28 +010010842 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000010843 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010010844 choose a web server that the CA believes requires an intermediate CA (for
10845 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000010846 others, e.g. nginx, result in a wrong bundle that will not work for some
10847 clients).
10848
Emeric Brun4147b2e2014-06-16 18:36:30 +020010849 For each PEM file, haproxy checks for the presence of file at the same path
10850 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
10851 Status Request extension (also known as "OCSP stapling") is automatically
10852 enabled. The content of this file is optional. If not empty, it must contain
10853 a valid OCSP Response in DER format. In order to be valid an OCSP Response
10854 must comply with the following rules: it has to indicate a good status,
10855 it has to be a single response for the certificate of the PEM file, and it
10856 has to be valid at the moment of addition. If these rules are not respected
10857 the OCSP Response is ignored and a warning is emitted. In order to identify
10858 which certificate an OCSP Response applies to, the issuer's certificate is
10859 necessary. If the issuer's certificate is not found in the PEM file, it will
10860 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
10861 if it exists otherwise it will fail with an error.
10862
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010010863 For each PEM file, haproxy also checks for the presence of file at the same
10864 path suffixed by ".sctl". If such file is found, support for Certificate
10865 Transparency (RFC6962) TLS extension is enabled. The file must contain a
10866 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
10867 to check basic syntax, but no signatures are verified.
10868
yanbzhu6c25e9e2016-01-05 12:52:02 -050010869 There are cases where it is desirable to support multiple key types, e.g. RSA
10870 and ECDSA in the cipher suites offered to the clients. This allows clients
10871 that support EC certificates to be able to use EC ciphers, while
10872 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050010873
10874 In order to provide this functionality, multiple PEM files, each with a
10875 different key type, are required. To associate these PEM files into a
10876 "cert bundle" that is recognized by haproxy, they must be named in the
10877 following way: All PEM files that are to be bundled must have the same base
10878 name, with a suffix indicating the key type. Currently, three suffixes are
10879 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
10880 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
10881 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
10882 suffix matters. To load this bundle into haproxy, specify the base name only:
10883
10884 Example : bind :8443 ssl crt example.pem
10885
yanbzhu6c25e9e2016-01-05 12:52:02 -050010886 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050010887 a cert bundle.
10888
Davor Ocelice9ed2812017-12-25 17:49:28 +010010889 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050010890 support multiple key types. PEM files are combined based on Common Name
10891 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
10892 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
10893 entries in the certificates in that bundle, haproxy will not be able to
10894 provide multi-cert support.
10895
10896 Assuming bundle in the example above contained the following:
10897
10898 Filename | CN | SAN
10899 -------------------+-----------------+-------------------
10900 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050010901 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050010902 example.pem.ecdsa | www.example.com | ecdsa.example.com
10903 -------------------+-----------------+-------------------
10904
10905 Users connecting with an SNI of "www.example.com" will be able
10906 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
10907 "rsa.example.com" will only be able to use RSA cipher suites, and users
10908 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010909 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
10910 no need to bundle certificates. ECDSA certificate will be preferred if client
10911 support it.
yanbzhud19630c2015-12-14 15:10:25 -050010912
10913 If a directory name is given as the <cert> argument, haproxy will
10914 automatically search and load bundled files in that directory.
10915
10916 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
10917 bundling. Each certificate can have its own .ocsp and .issuer file. At this
10918 time, sctl is not supported in multi-certificate bundling.
10919
Emeric Brunb6dc9342012-09-28 17:55:37 +020010920crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000010921 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010010922 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010923 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000010924 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020010925
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010926crt-list <file>
10927 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010928 designates a list of PEM file with an optional ssl configuration and a SNI
10929 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010930
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010931 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
10932
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020010933 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
10934 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010935 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010936 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010937
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020010938 Wildcards are supported in the SNI filter. Negative filter are also supported,
10939 only useful in combination with a wildcard filter to exclude a particular SNI.
10940 The certificates will be presented to clients who provide a valid TLS Server
10941 Name Indication field matching one of the SNI filters. If no SNI filter is
10942 specified, the CN and alt subjects are used. This directive may be specified
10943 multiple times. See the "crt" option for more information. The default
10944 certificate is still needed to meet OpenSSL expectations. If it is not used,
10945 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010010946
yanbzhu6c25e9e2016-01-05 12:52:02 -050010947 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020010948 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020010949 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
10950 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
10951 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050010952
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010953 crt-list file example:
10954 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010010955 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010956 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010010957 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010010958
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010959defer-accept
10960 Is an optional keyword which is supported only on certain Linux kernels. It
10961 states that a connection will only be accepted once some data arrive on it,
10962 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010010963 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020010964 performance by ensuring that most of the request is already available when
10965 the connection is accepted. On the other hand, it will not be able to detect
10966 connections which don't talk. It is important to note that this option is
10967 broken in all kernels up to 2.6.31, as the connection is never accepted until
10968 the client talks. This can cause issues with front firewalls which would see
10969 an established connection while the proxy will only see it in SYN_RECV. This
10970 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
10971
William Lallemandf6975e92017-05-26 17:42:10 +020010972expose-fd listeners
10973 This option is only usable with the stats socket. It gives your stats socket
10974 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020010975 During a reload with the master-worker mode, the process is automatically
10976 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010010977 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020010978
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010979force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010980 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010981 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010982 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010983 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010984
10985force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010986 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010987 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010988 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010989
10990force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010991 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010992 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010993 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010994
10995force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010996 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010010997 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020010998 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020010999
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011000force-tlsv13
11001 This option enforces use of TLSv1.3 only on SSL connections instantiated from
11002 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011003 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011004
Christopher Faulet31af49d2015-06-09 17:29:50 +020011005generate-certificates
11006 This setting is only available when support for OpenSSL was built in. It
11007 enables the dynamic SSL certificates generation. A CA certificate and its
11008 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
11009 a transparent forward proxy, SSL requests generate errors because of a common
11010 name mismatch on the certificate presented to the client. With this option
11011 enabled, HAProxy will try to forge a certificate using the SNI hostname
11012 indicated by the client. This is done only if no certificate matches the SNI
11013 hostname (see 'crt-list'). If an error occurs, the default certificate is
11014 used, else the 'strict-sni' option is set.
11015 It can also be used when HAProxy is configured as a reverse proxy to ease the
11016 deployment of an architecture with many backends.
11017
11018 Creating a SSL certificate is an expensive operation, so a LRU cache is used
11019 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010011020 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020011021 certificate is used many times.
11022
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011023gid <gid>
11024 Sets the group of the UNIX sockets to the designated system gid. It can also
11025 be set by default in the global section's "unix-bind" statement. Note that
11026 some platforms simply ignore this. This setting is equivalent to the "group"
11027 setting except that the group ID is used instead of its name. This setting is
11028 ignored by non UNIX sockets.
11029
11030group <group>
11031 Sets the group of the UNIX sockets to the designated system group. It can
11032 also be set by default in the global section's "unix-bind" statement. Note
11033 that some platforms simply ignore this. This setting is equivalent to the
11034 "gid" setting except that the group name is used instead of its gid. This
11035 setting is ignored by non UNIX sockets.
11036
11037id <id>
11038 Fixes the socket ID. By default, socket IDs are automatically assigned, but
11039 sometimes it is more convenient to fix them to ease monitoring. This value
11040 must be strictly positive and unique within the listener/frontend. This
11041 option can only be used when defining only a single socket.
11042
11043interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010011044 Restricts the socket to a specific interface. When specified, only packets
11045 received from that particular interface are processed by the socket. This is
11046 currently only supported on Linux. The interface must be a primary system
11047 interface, not an aliased interface. It is also possible to bind multiple
11048 frontends to the same address if they are bound to different interfaces. Note
11049 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010011050 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
11051 uses the same interface as inbound traffic, and its associated routing table,
11052 even if there are explicit routes through different interfaces configured.
11053 This can prove useful to address asymmetric routing issues when the same
11054 client IP addresses need to be able to reach frontends hosted on different
11055 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011056
Willy Tarreauabb175f2012-09-24 12:43:26 +020011057level <level>
11058 This setting is used with the stats sockets only to restrict the nature of
11059 the commands that can be issued on the socket. It is ignored by other
11060 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010011061 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020011062 read, and no change is allowed. It would make sense on systems where it
11063 is not easy to restrict access to the socket.
11064 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011065 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020011066 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010011067 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020011068 all counters).
11069
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020011070severity-output <format>
11071 This setting is used with the stats sockets only to configure severity
11072 level output prepended to informational feedback messages. Severity
11073 level of messages can range between 0 and 7, conforming to syslog
11074 rfc5424. Valid and successful socket commands requesting data
11075 (i.e. "show map", "get acl foo" etc.) will never have a severity level
11076 prepended. It is ignored by other sockets. <format> can be one of :
11077 - "none" (default) no severity level is prepended to feedback messages.
11078 - "number" severity level is prepended as a number.
11079 - "string" severity level is prepended as a string following the
11080 rfc5424 convention.
11081
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011082maxconn <maxconn>
11083 Limits the sockets to this number of concurrent connections. Extraneous
11084 connections will remain in the system's backlog until a connection is
11085 released. If unspecified, the limit will be the same as the frontend's
11086 maxconn. Note that in case of port ranges or multiple addresses, the same
11087 value will be applied to each socket. This setting enables different
11088 limitations on expensive sockets, for instance SSL entries which may easily
11089 eat all memory.
11090
11091mode <mode>
11092 Sets the octal mode used to define access permissions on the UNIX socket. It
11093 can also be set by default in the global section's "unix-bind" statement.
11094 Note that some platforms simply ignore this. This setting is ignored by non
11095 UNIX sockets.
11096
11097mss <maxseg>
11098 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
11099 connections. This can be used to force a lower MSS for certain specific
11100 ports, for instance for connections passing through a VPN. Note that this
11101 relies on a kernel feature which is theoretically supported under Linux but
11102 was buggy in all versions prior to 2.6.28. It may or may not work on other
11103 operating systems. It may also not change the advertised value but change the
11104 effective size of outgoing segments. The commonly advertised value for TCPv4
11105 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
11106 positive, it will be used as the advertised MSS. If it is negative, it will
11107 indicate by how much to reduce the incoming connection's advertised MSS for
11108 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
11109
11110name <name>
11111 Sets an optional name for these sockets, which will be reported on the stats
11112 page.
11113
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011114namespace <name>
11115 On Linux, it is possible to specify which network namespace a socket will
11116 belong to. This directive makes it possible to explicitly bind a listener to
11117 a namespace different from the default one. Please refer to your operating
11118 system's documentation to find more details about network namespaces.
11119
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011120nice <nice>
11121 Sets the 'niceness' of connections initiated from the socket. Value must be
11122 in the range -1024..1024 inclusive, and defaults to zero. Positive values
11123 means that such connections are more friendly to others and easily offer
11124 their place in the scheduler. On the opposite, negative values mean that
11125 connections want to run with a higher priority than others. The difference
11126 only happens under high loads when the system is close to saturation.
11127 Negative values are appropriate for low-latency or administration services,
11128 and high values are generally recommended for CPU intensive tasks such as SSL
11129 processing or bulk transfers which are less sensible to latency. For example,
11130 it may make sense to use a positive value for an SMTP socket and a negative
11131 one for an RDP socket.
11132
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011133no-ca-names
11134 This setting is only available when support for OpenSSL was built in. It
11135 prevents from send CA names in server hello message when ca-file is used.
11136
Emeric Brun9b3009b2012-10-05 11:55:06 +020011137no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011138 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011139 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011140 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011141 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011142 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
11143 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011144
Emeric Brun90ad8722012-10-02 14:00:59 +020011145no-tls-tickets
11146 This setting is only available when support for OpenSSL was built in. It
11147 disables the stateless session resumption (RFC 5077 TLS Ticket
11148 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011149 session resumption is more expensive in CPU usage. This option is also
11150 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020011151
Emeric Brun9b3009b2012-10-05 11:55:06 +020011152no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011153 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011154 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011155 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011156 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011157 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11158 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011159
Emeric Brun9b3009b2012-10-05 11:55:06 +020011160no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020011161 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011162 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011163 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011164 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011165 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11166 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011167
Emeric Brun9b3009b2012-10-05 11:55:06 +020011168no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020011169 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011170 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011171 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011172 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011173 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11174 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011175
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011176no-tlsv13
11177 This setting is only available when support for OpenSSL was built in. It
11178 disables support for TLSv1.3 on any sockets instantiated from the listener
11179 when SSL is supported. Note that SSLv2 is forced disabled in the code and
11180 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011181 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11182 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011183
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011184npn <protocols>
11185 This enables the NPN TLS extension and advertises the specified protocol list
11186 as supported on top of NPN. The protocol list consists in a comma-delimited
11187 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050011188 This requires that the SSL library is built with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020011189 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010011190 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11191 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
11192 version of OpenSSL, NPN might still be used as most clients still support it
11193 at the time of writing this. It is possible to enable both NPN and ALPN
11194 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011195
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011196prefer-client-ciphers
11197 Use the client's preference when selecting the cipher suite, by default
11198 the server's preference is enforced. This option is also available on
11199 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020011200 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
11201 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
11202 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011203
Christopher Fauletc644fa92017-11-23 22:44:11 +010011204process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010011205 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010011206 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010011207 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010011208 between the two is applied. If in the end the listener is not allowed to run
11209 on any remaining process, a warning is emitted, and the listener will either
11210 run on the first process of the listener if a single process was specified,
11211 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010011212 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010011213 connections for this listener, for the the process set. If multiple processes
11214 and threads are configured, a warning is emitted, as it either results from a
11215 configuration error or a misunderstanding of these models. For the unlikely
11216 case where several ranges are needed, this directive may be repeated.
11217 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010011218
11219 all | odd | even | number[-[number]]
11220
11221 Ranges can be partially defined. The higher bound can be omitted. In such
11222 case, it is replaced by the corresponding maximum value. The main purpose of
11223 this directive is to be used with the stats sockets and have one different
11224 socket per process. The second purpose is to have multiple bind lines sharing
11225 the same IP:port but not the same process in a listener, so that the system
11226 can distribute the incoming connections into multiple queues and allow a
11227 smoother inter-process load balancing. Currently Linux 3.9 and above is known
11228 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020011229
Christopher Fauleta717b992018-04-10 14:43:00 +020011230proto <name>
11231 Forces the multiplexer's protocol to use for the incoming connections. It
11232 must be compatible with the mode of the frontend (TCP or HTTP). It must also
11233 be usable on the frontend side. The list of available protocols is reported
11234 in haproxy -vv.
11235 Idea behind this optipon is to bypass the selection of the best multiplexer's
11236 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080011237 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020011238 h2" on the bind line.
11239
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011240ssl
11241 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011242 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011243 certificate is necessary (see "crt" above). All contents in the buffers will
11244 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011245 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11246 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011247
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011248ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11249 This option enforces use of <version> or lower on SSL connections instantiated
11250 from this listener. This option is also available on global statement
11251 "ssl-default-bind-options". See also "ssl-min-ver".
11252
11253ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11254 This option enforces use of <version> or upper on SSL connections instantiated
11255 from this listener. This option is also available on global statement
11256 "ssl-default-bind-options". See also "ssl-max-ver".
11257
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011258strict-sni
11259 This setting is only available when support for OpenSSL was built in. The
11260 SSL/TLS negotiation is allow only if the client provided an SNI which match
11261 a certificate. The default certificate is not used.
11262 See the "crt" option for more information.
11263
Willy Tarreau2af207a2015-02-04 00:45:58 +010011264tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011265 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011266 listening socket. This option is available on Linux since version 2.6.37. It
11267 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011268 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011269 useful on long-lived connections experiencing long idle periods such as
11270 remote terminals or database connection pools, where the client and server
11271 timeouts must remain high to allow a long period of idle, but where it is
11272 important to detect that the client has disappeared in order to release all
11273 resources associated with its connection (and the server's session). The
11274 argument is a delay expressed in milliseconds by default. This only works
11275 for regular TCP connections, and is ignored for other protocols.
11276
Willy Tarreau1c862c52012-10-05 16:21:00 +020011277tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011278 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011279 enables TCP Fast Open on the listening socket, which means that clients which
11280 support this feature will be able to send a request and receive a response
11281 during the 3-way handshake starting from second connection, thus saving one
11282 round-trip after the first connection. This only makes sense with protocols
11283 that use high connection rates and where each round trip matters. This can
11284 possibly cause issues with many firewalls which do not accept data on SYN
11285 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011286 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11287 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11288 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011289
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011290tls-ticket-keys <keyfile>
11291 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010011292 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
11293 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
11294 The first key determines the key length used for next keys: you can't mix
11295 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
11296 build option (default 3) and at least as many keys need to be present in
11297 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
11298 penultimate one for encryption. This enables easy key rotation by just
11299 appending new key to the file and reloading the process. Keys must be
11300 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
11301 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011302 storage such as hard drives (hint: use tmpfs and don't swap those files).
11303 Lifetime hint can be changed using tune.ssl.timeout.
11304
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011305transparent
11306 Is an optional keyword which is supported only on certain Linux kernels. It
11307 indicates that the addresses will be bound even if they do not belong to the
11308 local machine, and that packets targeting any of these addresses will be
11309 intercepted just as if the addresses were locally configured. This normally
11310 requires that IP forwarding is enabled. Caution! do not use this with the
11311 default address '*', as it would redirect any traffic for the specified port.
11312 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11313 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11314 kernel version. Some distribution kernels include backports of the feature,
11315 so check for support with your vendor.
11316
Willy Tarreau77e3af92012-11-24 15:07:23 +010011317v4v6
11318 Is an optional keyword which is supported only on most recent systems
11319 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11320 and IPv6 when it uses the default address. Doing so is sometimes necessary
11321 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011322 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011323
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011324v6only
11325 Is an optional keyword which is supported only on most recent systems
11326 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11327 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011328 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11329 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011330
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011331uid <uid>
11332 Sets the owner of the UNIX sockets to the designated system uid. It can also
11333 be set by default in the global section's "unix-bind" statement. Note that
11334 some platforms simply ignore this. This setting is equivalent to the "user"
11335 setting except that the user numeric ID is used instead of its name. This
11336 setting is ignored by non UNIX sockets.
11337
11338user <user>
11339 Sets the owner of the UNIX sockets to the designated system user. It can also
11340 be set by default in the global section's "unix-bind" statement. Note that
11341 some platforms simply ignore this. This setting is equivalent to the "uid"
11342 setting except that the user name is used instead of its uid. This setting is
11343 ignored by non UNIX sockets.
11344
Emeric Brun1a073b42012-09-28 17:07:34 +020011345verify [none|optional|required]
11346 This setting is only available when support for OpenSSL was built in. If set
11347 to 'none', client certificate is not requested. This is the default. In other
11348 cases, a client certificate is requested. If the client does not provide a
11349 certificate after the request and if 'verify' is set to 'required', then the
11350 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011351 certificate provided by the client is always verified using CAs from
11352 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11353 is aborted, regardless of the 'verify' option, unless the error code exactly
11354 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011355
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200113565.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011357------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011358
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011359The "server" and "default-server" keywords support a certain number of settings
11360which are all passed as arguments on the server line. The order in which those
11361arguments appear does not count, and they are all optional. Some of those
11362settings are single words (booleans) while others expect one or several values
11363after them. In this case, the values must immediately follow the setting name.
11364Except default-server, all those settings must be specified after the server's
11365address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011366
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011367 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011368 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020011369
Frédéric Lécailled2376272017-03-21 18:52:12 +010011370Note that all these settings are supported both by "server" and "default-server"
11371keywords, except "id" which is only supported by "server".
11372
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011373The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011374
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011375addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011376 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010011377 to send health-checks or to probe the agent-check. On some servers, it may be
11378 desirable to dedicate an IP address to specific component able to perform
11379 complex tests which are more suitable to health-checks than the application.
11380 This parameter is ignored if the "check" parameter is not set. See also the
11381 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011382
Simon Hormand60d6912013-11-25 10:46:36 +090011383agent-check
11384 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010011385 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010011386 to the port set by the "agent-port" parameter and reading an ASCII string
11387 terminated by the first '\r' or '\n' met. The string is made of a series of
11388 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090011389
Willy Tarreau81f5d942013-12-09 20:51:51 +010011390 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090011391 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020011392 weight of a server as configured when haproxy starts. Note that a zero
11393 weight is reported on the stats page as "DRAIN" since it has the same
11394 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090011395
Davor Ocelice9ed2812017-12-25 17:49:28 +010011396 - The string "maxconn:" followed by an integer (no space between). Values
11397 in this format will set the maxconn of a server. The maximum number of
11398 connections advertised needs to be multiplied by the number of load
11399 balancers and different backends that use this health check to get the
11400 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020011401
Willy Tarreau81f5d942013-12-09 20:51:51 +010011402 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011403 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090011404
Willy Tarreau81f5d942013-12-09 20:51:51 +010011405 - The word "drain". This will turn the server's administrative state to the
11406 DRAIN mode, thus it will not accept any new connections other than those
11407 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090011408
Willy Tarreau81f5d942013-12-09 20:51:51 +010011409 - The word "maint". This will turn the server's administrative state to the
11410 MAINT mode, thus it will not accept any new connections at all, and health
11411 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090011412
Willy Tarreau81f5d942013-12-09 20:51:51 +010011413 - The words "down", "failed", or "stopped", optionally followed by a
11414 description string after a sharp ('#'). All of these mark the server's
11415 operating state as DOWN, but since the word itself is reported on the stats
11416 page, the difference allows an administrator to know if the situation was
11417 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010011418 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010011419 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090011420
Willy Tarreau81f5d942013-12-09 20:51:51 +010011421 - The word "up" sets back the server's operating state as UP if health checks
11422 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090011423
Willy Tarreau81f5d942013-12-09 20:51:51 +010011424 Parameters which are not advertised by the agent are not changed. For
11425 example, an agent might be designed to monitor CPU usage and only report a
11426 relative weight and never interact with the operating status. Similarly, an
11427 agent could be designed as an end-user interface with 3 radio buttons
11428 allowing an administrator to change only the administrative state. However,
11429 it is important to consider that only the agent may revert its own actions,
11430 so if a server is set to DRAIN mode or to DOWN state using the agent, the
11431 agent must implement the other equivalent actions to bring the service into
11432 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090011433
Simon Horman2f1f9552013-11-25 10:46:37 +090011434 Failure to connect to the agent is not considered an error as connectivity
11435 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010011436 parameter. Warning though, it is not a good idea to stop an agent after it
11437 reports "down", since only an agent reporting "up" will be able to turn the
11438 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010011439 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090011440
Willy Tarreau81f5d942013-12-09 20:51:51 +010011441 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010011442 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090011443
James Brown55f9ff12015-10-21 18:19:05 -070011444agent-send <string>
11445 If this option is specified, haproxy will send the given string (verbatim)
11446 to the agent server upon connection. You could, for example, encode
11447 the backend name into this string, which would enable your agent to send
11448 different responses based on the backend. Make sure to include a '\n' if
11449 you want to terminate your request with a newline.
11450
Simon Hormand60d6912013-11-25 10:46:36 +090011451agent-inter <delay>
11452 The "agent-inter" parameter sets the interval between two agent checks
11453 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11454
11455 Just as with every other time-based parameter, it may be entered in any
11456 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
11457 parameter also serves as a timeout for agent checks "timeout check" is
11458 not set. In order to reduce "resonance" effects when multiple servers are
11459 hosted on the same hardware, the agent and health checks of all servers
11460 are started with a small time offset between them. It is also possible to
11461 add some random noise in the agent and health checks interval using the
11462 global "spread-checks" keyword. This makes sense for instance when a lot
11463 of backends use the same servers.
11464
11465 See also the "agent-check" and "agent-port" parameters.
11466
Misiek768d8602017-01-09 09:52:43 +010011467agent-addr <addr>
11468 The "agent-addr" parameter sets address for agent check.
11469
11470 You can offload agent-check to another target, so you can make single place
11471 managing status and weights of servers defined in haproxy in case you can't
11472 make self-aware and self-managing services. You can specify both IP or
11473 hostname, it will be resolved.
11474
Simon Hormand60d6912013-11-25 10:46:36 +090011475agent-port <port>
11476 The "agent-port" parameter sets the TCP port used for agent checks.
11477
11478 See also the "agent-check" and "agent-inter" parameters.
11479
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011480allow-0rtt
11481 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020011482 Note that early data will be sent only if the client used early data, or
11483 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011484
Olivier Houchardc7566002018-11-20 23:33:50 +010011485alpn <protocols>
11486 This enables the TLS ALPN extension and advertises the specified protocol
11487 list as supported on top of ALPN. The protocol list consists in a comma-
11488 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roeslerfb2fce12019-07-10 15:45:51 -050011489 quotes). This requires that the SSL library is built with support for TLS
Olivier Houchardc7566002018-11-20 23:33:50 +010011490 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
11491 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
11492 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11493 now obsolete NPN extension.
11494 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
11495 be advertised, in order of preference, like below :
11496
11497 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
11498
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011499backup
11500 When "backup" is present on a server line, the server is only used in load
11501 balancing when all other non-backup servers are unavailable. Requests coming
11502 with a persistence cookie referencing the server will always be served
11503 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010011504 the "allbackups" option is set in the backend. See also the "no-backup" and
11505 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011506
Emeric Brunef42d922012-10-11 16:11:36 +020011507ca-file <cafile>
11508 This setting is only available when support for OpenSSL was built in. It
11509 designates a PEM file from which to load CA certificates used to verify
11510 server's certificate.
11511
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011512check
11513 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010011514 always considered available. If "check" is set, the server is available when
11515 accepting periodic TCP connections, to ensure that it is really able to serve
11516 requests. The default address and port to send the tests to are those of the
11517 server, and the default source is the same as the one defined in the
11518 backend. It is possible to change the address using the "addr" parameter, the
11519 port using the "port" parameter, the source address using the "source"
11520 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090011521 parameters. The request method is define in the backend using the "httpchk",
11522 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010011523 refer to those options and parameters for more information. See also
11524 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011525
Willy Tarreau6c16adc2012-10-05 00:04:16 +020011526check-send-proxy
11527 This option forces emission of a PROXY protocol line with outgoing health
11528 checks, regardless of whether the server uses send-proxy or not for the
11529 normal traffic. By default, the PROXY protocol is enabled for health checks
11530 if it is already enabled for normal traffic and if no "port" nor "addr"
11531 directive is present. However, if such a directive is present, the
11532 "check-send-proxy" option needs to be used to force the use of the
11533 protocol. See also the "send-proxy" option for more information.
11534
Olivier Houchard92150142018-12-21 19:47:01 +010011535check-alpn <protocols>
11536 Defines which protocols to advertise with ALPN. The protocol list consists in
11537 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
11538 (without quotes). If it is not set, the server ALPN is used.
11539
Jérôme Magninae9bb762018-12-09 16:08:26 +010011540check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020011541 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010011542 over SSL. It is only possible to use a string to set <sni>. If you want to
11543 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020011544
Willy Tarreau763a95b2012-10-04 23:15:39 +020011545check-ssl
11546 This option forces encryption of all health checks over SSL, regardless of
11547 whether the server uses SSL or not for the normal traffic. This is generally
11548 used when an explicit "port" or "addr" directive is specified and SSL health
11549 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011550 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020011551 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
11552 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011553 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010011554 See the "ssl" option for more information and "no-check-ssl" to disable
11555 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011556
Alexander Liu2a54bb72019-05-22 19:44:48 +080011557check-via-socks4
John Roeslerfb2fce12019-07-10 15:45:51 -050011558 This option enables outgoing health checks using upstream socks4 proxy. By
Alexander Liu2a54bb72019-05-22 19:44:48 +080011559 default, the health checks won't go through socks tunnel even it was enabled
11560 for normal traffic.
11561
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011562ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011563 This setting is only available when support for OpenSSL was built in. This
11564 option sets the string describing the list of cipher algorithms that is
11565 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011566 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
11567 information and recommendations see e.g.
11568 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11569 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11570 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011571
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011572ciphersuites <ciphersuites>
11573 This setting is only available when support for OpenSSL was built in and
11574 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
11575 describing the list of cipher algorithms that is negotiated during the TLS
11576 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011577 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
11578 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
11579 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011580
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011581cookie <value>
11582 The "cookie" parameter sets the cookie value assigned to the server to
11583 <value>. This value will be checked in incoming requests, and the first
11584 operational server possessing the same value will be selected. In return, in
11585 cookie insertion or rewrite modes, this value will be assigned to the cookie
11586 sent to the client. There is nothing wrong in having several servers sharing
11587 the same cookie value, and it is in fact somewhat common between normal and
11588 backup servers. See also the "cookie" keyword in backend section.
11589
Emeric Brunef42d922012-10-11 16:11:36 +020011590crl-file <crlfile>
11591 This setting is only available when support for OpenSSL was built in. It
11592 designates a PEM file from which to load certificate revocation list used
11593 to verify server's certificate.
11594
Emeric Bruna7aa3092012-10-26 12:58:00 +020011595crt <cert>
11596 This setting is only available when support for OpenSSL was built in.
11597 It designates a PEM file from which to load both a certificate and the
11598 associated private key. This file can be built by concatenating both PEM
11599 files into one. This certificate will be sent if the server send a client
11600 certificate request.
11601
Willy Tarreau96839092010-03-29 10:02:24 +020011602disabled
11603 The "disabled" keyword starts the server in the "disabled" state. That means
11604 that it is marked down in maintenance mode, and no connection other than the
11605 ones allowed by persist mode will reach it. It is very well suited to setup
11606 new servers, because normal traffic will never reach them, while it is still
11607 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011608 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011609
Frédéric Lécailled2376272017-03-21 18:52:12 +010011610enabled
11611 This option may be used as 'server' setting to reset any 'disabled'
11612 setting which would have been inherited from 'default-server' directive as
11613 default value.
11614 It may also be used as 'default-server' setting to reset any previous
11615 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011616
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011617error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010011618 If health observing is enabled, the "error-limit" parameter specifies the
11619 number of consecutive errors that triggers event selected by the "on-error"
11620 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011621
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011622 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011623
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011624fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011625 The "fall" parameter states that a server will be considered as dead after
11626 <count> consecutive unsuccessful health checks. This value defaults to 3 if
11627 unspecified. See also the "check", "inter" and "rise" parameters.
11628
Emeric Brun8694b9a2012-10-05 14:39:07 +020011629force-sslv3
11630 This option enforces use of SSLv3 only when SSL is used to communicate with
11631 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011632 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011633 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011634
11635force-tlsv10
11636 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011637 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011638 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011639
11640force-tlsv11
11641 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011642 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011643 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011644
11645force-tlsv12
11646 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011647 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011648 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011649
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011650force-tlsv13
11651 This option enforces use of TLSv1.3 only when SSL is used to communicate with
11652 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011653 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011654
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011655id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020011656 Set a persistent ID for the server. This ID must be positive and unique for
11657 the proxy. An unused ID will automatically be assigned if unset. The first
11658 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011659
Willy Tarreau6a031d12016-11-07 19:42:35 +010011660init-addr {last | libc | none | <ip>},[...]*
11661 Indicate in what order the server's address should be resolved upon startup
11662 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010011663 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010011664 method which succeeds is used. If the end of the list is reached without
11665 finding a working method, an error is thrown. Method "last" suggests to pick
11666 the address which appears in the state file (see "server-state-file"). Method
11667 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
11668 depending on the operating system and build options). Method "none"
11669 specifically indicates that the server should start without any valid IP
11670 address in a down state. It can be useful to ignore some DNS issues upon
11671 startup, waiting for the situation to get fixed later. Finally, an IP address
11672 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011673 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010011674 server used to catch old sessions and present them with a decent error
11675 message for example. When the "first" load balancing algorithm is used, this
11676 IP address could point to a fake server used to trigger the creation of new
11677 instances on the fly. This option defaults to "last,libc" indicating that the
11678 previous address found in the state file (if any) is used first, otherwise
11679 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011680 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010011681
11682 Example:
11683 defaults
11684 # never fail on address resolution
11685 default-server init-addr last,libc,none
11686
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011687inter <delay>
11688fastinter <delay>
11689downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011690 The "inter" parameter sets the interval between two consecutive health checks
11691 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11692 It is also possible to use "fastinter" and "downinter" to optimize delays
11693 between checks depending on the server state :
11694
Pieter Baauw44fc9df2015-09-17 21:30:46 +020011695 Server state | Interval used
11696 ----------------------------------------+----------------------------------
11697 UP 100% (non-transitional) | "inter"
11698 ----------------------------------------+----------------------------------
11699 Transitionally UP (going down "fall"), | "fastinter" if set,
11700 Transitionally DOWN (going up "rise"), | "inter" otherwise.
11701 or yet unchecked. |
11702 ----------------------------------------+----------------------------------
11703 DOWN 100% (non-transitional) | "downinter" if set,
11704 | "inter" otherwise.
11705 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010011706
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011707 Just as with every other time-based parameter, they can be entered in any
11708 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
11709 serves as a timeout for health checks sent to servers if "timeout check" is
11710 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090011711 hosted on the same hardware, the agent and health checks of all servers
11712 are started with a small time offset between them. It is also possible to
11713 add some random noise in the agent and health checks interval using the
11714 global "spread-checks" keyword. This makes sense for instance when a lot
11715 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011716
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011717maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011718 The "maxconn" parameter specifies the maximal number of concurrent
11719 connections that will be sent to this server. If the number of incoming
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010011720 concurrent connections goes higher than this value, they will be queued,
11721 waiting for a slot to be released. This parameter is very important as it can
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011722 save fragile servers from going down under extreme loads. If a "minconn"
11723 parameter is specified, the limit becomes dynamic. The default value is "0"
11724 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
11725 the backend's "fullconn" keyword.
11726
Tim Duesterhuscefbbd92019-11-27 22:35:27 +010011727 In HTTP mode this parameter limits the number of concurrent requests instead
11728 of the number of connections. Multiple requests might be multiplexed over a
11729 single TCP connection to the server. As an example if you specify a maxconn
11730 of 50 you might see between 1 and 50 actual server connections, but no more
11731 than 50 concurrent requests.
11732
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011733maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011734 The "maxqueue" parameter specifies the maximal number of connections which
11735 will wait in the queue for this server. If this limit is reached, next
11736 requests will be redispatched to other servers instead of indefinitely
11737 waiting to be served. This will break persistence but may allow people to
11738 quickly re-log in when the server they try to connect to is dying. The
11739 default value is "0" which means the queue is unlimited. See also the
11740 "maxconn" and "minconn" parameters.
11741
Willy Tarreau9c538e02019-01-23 10:21:49 +010011742max-reuse <count>
11743 The "max-reuse" argument indicates the HTTP connection processors that they
11744 should not reuse a server connection more than this number of times to send
11745 new requests. Permitted values are -1 (the default), which disables this
11746 limit, or any positive value. Value zero will effectively disable keep-alive.
11747 This is only used to work around certain server bugs which cause them to leak
11748 resources over time. The argument is not necessarily respected by the lower
11749 layers as there might be technical limitations making it impossible to
11750 enforce. At least HTTP/2 connections to servers will respect it.
11751
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011752minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011753 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
11754 limit following the backend's load. The server will always accept at least
11755 <minconn> connections, never more than <maxconn>, and the limit will be on
11756 the ramp between both values when the backend has less than <fullconn>
11757 concurrent connections. This makes it possible to limit the load on the
11758 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011759 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011760 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011761
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011762namespace <name>
11763 On Linux, it is possible to specify which network namespace a socket will
11764 belong to. This directive makes it possible to explicitly bind a server to
11765 a namespace different from the default one. Please refer to your operating
11766 system's documentation to find more details about network namespaces.
11767
Frédéric Lécailled2376272017-03-21 18:52:12 +010011768no-agent-check
11769 This option may be used as "server" setting to reset any "agent-check"
11770 setting which would have been inherited from "default-server" directive as
11771 default value.
11772 It may also be used as "default-server" setting to reset any previous
11773 "default-server" "agent-check" setting.
11774
11775no-backup
11776 This option may be used as "server" setting to reset any "backup"
11777 setting which would have been inherited from "default-server" directive as
11778 default value.
11779 It may also be used as "default-server" setting to reset any previous
11780 "default-server" "backup" setting.
11781
11782no-check
11783 This option may be used as "server" setting to reset any "check"
11784 setting which would have been inherited from "default-server" directive as
11785 default value.
11786 It may also be used as "default-server" setting to reset any previous
11787 "default-server" "check" setting.
11788
11789no-check-ssl
11790 This option may be used as "server" setting to reset any "check-ssl"
11791 setting which would have been inherited from "default-server" directive as
11792 default value.
11793 It may also be used as "default-server" setting to reset any previous
11794 "default-server" "check-ssl" setting.
11795
Frédéric Lécailled2376272017-03-21 18:52:12 +010011796no-send-proxy
11797 This option may be used as "server" setting to reset any "send-proxy"
11798 setting which would have been inherited from "default-server" directive as
11799 default value.
11800 It may also be used as "default-server" setting to reset any previous
11801 "default-server" "send-proxy" setting.
11802
11803no-send-proxy-v2
11804 This option may be used as "server" setting to reset any "send-proxy-v2"
11805 setting which would have been inherited from "default-server" directive as
11806 default value.
11807 It may also be used as "default-server" setting to reset any previous
11808 "default-server" "send-proxy-v2" setting.
11809
11810no-send-proxy-v2-ssl
11811 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
11812 setting which would have been inherited from "default-server" directive as
11813 default value.
11814 It may also be used as "default-server" setting to reset any previous
11815 "default-server" "send-proxy-v2-ssl" setting.
11816
11817no-send-proxy-v2-ssl-cn
11818 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
11819 setting which would have been inherited from "default-server" directive as
11820 default value.
11821 It may also be used as "default-server" setting to reset any previous
11822 "default-server" "send-proxy-v2-ssl-cn" setting.
11823
11824no-ssl
11825 This option may be used as "server" setting to reset any "ssl"
11826 setting which would have been inherited from "default-server" directive as
11827 default value.
11828 It may also be used as "default-server" setting to reset any previous
11829 "default-server" "ssl" setting.
11830
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010011831no-ssl-reuse
11832 This option disables SSL session reuse when SSL is used to communicate with
11833 the server. It will force the server to perform a full handshake for every
11834 new connection. It's probably only useful for benchmarking, troubleshooting,
11835 and for paranoid users.
11836
Emeric Brun9b3009b2012-10-05 11:55:06 +020011837no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011838 This option disables support for SSLv3 when SSL is used to communicate with
11839 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011840 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011841
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011842 Supported in default-server: No
11843
Emeric Brunf9c5c472012-10-11 15:28:34 +020011844no-tls-tickets
11845 This setting is only available when support for OpenSSL was built in. It
11846 disables the stateless session resumption (RFC 5077 TLS Ticket
11847 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011848 session resumption is more expensive in CPU usage for servers. This option
11849 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010011850 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020011851
Emeric Brun9b3009b2012-10-05 11:55:06 +020011852no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020011853 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020011854 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11855 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011856 often makes sense to disable it when communicating with local servers. This
11857 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011858 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011859
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011860 Supported in default-server: No
11861
Emeric Brun9b3009b2012-10-05 11:55:06 +020011862no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020011863 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020011864 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11865 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011866 often makes sense to disable it when communicating with local servers. This
11867 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011868 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011869
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011870 Supported in default-server: No
11871
Emeric Brun9b3009b2012-10-05 11:55:06 +020011872no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020011873 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011874 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11875 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011876 often makes sense to disable it when communicating with local servers. This
11877 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011878 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011879
11880 Supported in default-server: No
11881
11882no-tlsv13
11883 This option disables support for TLSv1.3 when SSL is used to communicate with
11884 the server. Note that SSLv2 is disabled in the code and cannot be enabled
11885 using any configuration option. TLSv1 is more expensive than SSLv3 so it
11886 often makes sense to disable it when communicating with local servers. This
11887 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011888 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011889
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020011890 Supported in default-server: No
11891
Frédéric Lécailled2376272017-03-21 18:52:12 +010011892no-verifyhost
11893 This option may be used as "server" setting to reset any "verifyhost"
11894 setting which would have been inherited from "default-server" directive as
11895 default value.
11896 It may also be used as "default-server" setting to reset any previous
11897 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011898
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020011899no-tfo
11900 This option may be used as "server" setting to reset any "tfo"
11901 setting which would have been inherited from "default-server" directive as
11902 default value.
11903 It may also be used as "default-server" setting to reset any previous
11904 "default-server" "tfo" setting.
11905
Simon Hormanfa461682011-06-25 09:39:49 +090011906non-stick
11907 Never add connections allocated to this sever to a stick-table.
11908 This may be used in conjunction with backup to ensure that
11909 stick-table persistence is disabled for backup servers.
11910
Olivier Houchardc7566002018-11-20 23:33:50 +010011911npn <protocols>
11912 This enables the NPN TLS extension and advertises the specified protocol list
11913 as supported on top of NPN. The protocol list consists in a comma-delimited
11914 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roeslerfb2fce12019-07-10 15:45:51 -050011915 This requires that the SSL library is built with support for TLS extensions
Olivier Houchardc7566002018-11-20 23:33:50 +010011916 enabled (check with haproxy -vv). Note that the NPN extension has been
11917 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11918 only available starting with OpenSSL 1.0.2.
11919
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011920observe <mode>
11921 This option enables health adjusting based on observing communication with
11922 the server. By default this functionality is disabled and enabling it also
11923 requires to enable health checks. There are two supported modes: "layer4" and
11924 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
11925 significant. In layer7, which is only allowed for http proxies, responses
11926 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010011927 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011928
11929 See also the "check", "on-error" and "error-limit".
11930
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011931on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011932 Select what should happen when enough consecutive errors are detected.
11933 Currently, four modes are available:
11934 - fastinter: force fastinter
11935 - fail-check: simulate a failed check, also forces fastinter (default)
11936 - sudden-death: simulate a pre-fatal failed health check, one more failed
11937 check will mark a server down, forces fastinter
11938 - mark-down: mark the server immediately down and force fastinter
11939
11940 See also the "check", "observe" and "error-limit".
11941
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011942on-marked-down <action>
11943 Modify what occurs when a server is marked down.
11944 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011945 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
11946 all connections to the server are immediately terminated when the server
11947 goes down. It might be used if the health check detects more complex cases
11948 than a simple connection status, and long timeouts would cause the service
11949 to remain unresponsive for too long a time. For instance, a health check
11950 might detect that a database is stuck and that there's no chance to reuse
11951 existing connections anymore. Connections killed this way are logged with
11952 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090011953
11954 Actions are disabled by default
11955
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011956on-marked-up <action>
11957 Modify what occurs when a server is marked up.
11958 Currently one action is available:
11959 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
11960 done only if the server is not in backup state and if it is not disabled
11961 (it must have an effective weight > 0). This can be used sometimes to force
11962 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010011963 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
11964 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070011965 with extreme care. Sessions killed because a server comes up are logged
11966 with an 'U' termination code (for "Up").
11967
11968 Actions are disabled by default
11969
Olivier Houchard006e3102018-12-10 18:30:32 +010011970pool-max-conn <max>
11971 Set the maximum number of idling connections for a server. -1 means unlimited
11972 connections, 0 means no idle connections. The default is -1. When idle
11973 connections are enabled, orphaned idle connections which do not belong to any
11974 client session anymore are moved to a dedicated pool so that they remain
11975 usable by future clients. This only applies to connections that can be shared
11976 according to the same principles as those applying to "http-reuse".
11977
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010011978pool-purge-delay <delay>
11979 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010011980 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020011981 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010011982
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011983port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011984 Using the "port" parameter, it becomes possible to use a different port to
11985 send health-checks. On some servers, it may be desirable to dedicate a port
11986 to a specific component able to perform complex tests which are more suitable
11987 to health-checks than the application. It is common to run a simple script in
11988 inetd for instance. This parameter is ignored if the "check" parameter is not
11989 set. See also the "addr" parameter.
11990
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020011991proto <name>
11992
11993 Forces the multiplexer's protocol to use for the outgoing connections to this
11994 server. It must be compatible with the mode of the backend (TCP or HTTP). It
11995 must also be usable on the backend side. The list of available protocols is
11996 reported in haproxy -vv.
11997 Idea behind this optipon is to bypass the selection of the best multiplexer's
11998 protocol for all connections established to this server.
11999
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012000redir <prefix>
12001 The "redir" parameter enables the redirection mode for all GET and HEAD
12002 requests addressing this server. This means that instead of having HAProxy
12003 forward the request to the server, it will send an "HTTP 302" response with
12004 the "Location" header composed of this prefix immediately followed by the
12005 requested URI beginning at the leading '/' of the path component. That means
12006 that no trailing slash should be used after <prefix>. All invalid requests
12007 will be rejected, and all non-GET or HEAD requests will be normally served by
12008 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012009 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012010 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012011 users to a remote location in case of local disaster. Main use consists in
12012 increasing bandwidth for static servers by having the clients directly
12013 connect to them. Note: never use a relative location here, it would cause a
12014 loop between the client and HAProxy!
12015
12016 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
12017
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012018rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012019 The "rise" parameter states that a server will be considered as operational
12020 after <count> consecutive successful health checks. This value defaults to 2
12021 if unspecified. See also the "check", "inter" and "fall" parameters.
12022
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012023resolve-opts <option>,<option>,...
12024 Comma separated list of options to apply to DNS resolution linked to this
12025 server.
12026
12027 Available options:
12028
12029 * allow-dup-ip
12030 By default, HAProxy prevents IP address duplication in a backend when DNS
12031 resolution at runtime is in operation.
12032 That said, for some cases, it makes sense that two servers (in the same
12033 backend, being resolved by the same FQDN) have the same IP address.
12034 For such case, simply enable this option.
12035 This is the opposite of prevent-dup-ip.
12036
Daniel Corbettf8716912019-11-17 09:48:56 -050012037 * ignore-weight
12038 Ignore any weight that is set within an SRV record. This is useful when
12039 you would like to control the weights using an alternate method, such as
12040 using an "agent-check" or through the runtime api.
12041
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012042 * prevent-dup-ip
12043 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
12044 an IP address already set to a server in the same backend and sharing the
12045 same fqdn.
12046 This is the opposite of allow-dup-ip.
12047
12048 Example:
12049 backend b_myapp
12050 default-server init-addr none resolvers dns
12051 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
12052 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
12053
12054 With the option allow-dup-ip set:
12055 * if the nameserver returns a single IP address, then both servers will use
12056 it
12057 * If the nameserver returns 2 IP addresses, then each server will pick up a
12058 different address
12059
12060 Default value: not set
12061
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012062resolve-prefer <family>
12063 When DNS resolution is enabled for a server and multiple IP addresses from
12064 different families are returned, HAProxy will prefer using an IP address
12065 from the family mentioned in the "resolve-prefer" parameter.
12066 Available families: "ipv4" and "ipv6"
12067
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020012068 Default value: ipv6
12069
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012070 Example:
12071
12072 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012073
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012074resolve-net <network>[,<network[,...]]
John Roeslerfb2fce12019-07-10 15:45:51 -050012075 This option prioritizes the choice of an ip address matching a network. This is
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012076 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010012077 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010012078 different datacenters. The latency between datacenter is not negligible, so
12079 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012080 configured network, another address is selected.
12081
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012082 Example:
12083
12084 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012085
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012086resolvers <id>
12087 Points to an existing "resolvers" section to resolve current server's
12088 hostname.
12089
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012090 Example:
12091
12092 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012093
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012094 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012095
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012096send-proxy
12097 The "send-proxy" parameter enforces use of the PROXY protocol over any
12098 connection established to this server. The PROXY protocol informs the other
12099 end about the layer 3/4 addresses of the incoming connection, so that it can
12100 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010012101 upper layer protocol. For connections accepted by an "accept-proxy" or
12102 "accept-netscaler-cip" listener, the advertised address will be used. Only
12103 TCPv4 and TCPv6 address families are supported. Other families such as
12104 Unix sockets, will report an UNKNOWN family. Servers using this option can
12105 fully be chained to another instance of haproxy listening with an
12106 "accept-proxy" setting. This setting must not be used if the server isn't
12107 aware of the protocol. When health checks are sent to the server, the PROXY
12108 protocol is automatically used when this option is set, unless there is an
12109 explicit "port" or "addr" directive, in which case an explicit
12110 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012111 See also the "no-send-proxy" option of this section and "accept-proxy" and
12112 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012113
David Safb76832014-05-08 23:42:08 -040012114send-proxy-v2
12115 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
12116 over any connection established to this server. The PROXY protocol informs
12117 the other end about the layer 3/4 addresses of the incoming connection, so
12118 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020012119 whatever the upper layer protocol. It also send ALPN information if an alpn
12120 have been negotiated. This setting must not be used if the server isn't aware
12121 of this version of the protocol. See also the "no-send-proxy-v2" option of
12122 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012123
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012124proxy-v2-options <option>[,<option>]*
12125 The "proxy-v2-options" parameter add option to send in PROXY protocol version
12126 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010012127 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
12128 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010012129 certificate, "cert-key": key algorithm of the used certificate), "authority":
12130 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010012131 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012132
David Safb76832014-05-08 23:42:08 -040012133send-proxy-v2-ssl
12134 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12135 2 over any connection established to this server. The PROXY protocol informs
12136 the other end about the layer 3/4 addresses of the incoming connection, so
12137 that it can know the client's address or the public address it accessed to,
12138 whatever the upper layer protocol. In addition, the SSL information extension
12139 of the PROXY protocol is added to the PROXY protocol header. This setting
12140 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012141 See also the "no-send-proxy-v2-ssl" option of this section and the
12142 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012143
12144send-proxy-v2-ssl-cn
12145 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12146 2 over any connection established to this server. The PROXY protocol informs
12147 the other end about the layer 3/4 addresses of the incoming connection, so
12148 that it can know the client's address or the public address it accessed to,
12149 whatever the upper layer protocol. In addition, the SSL information extension
12150 of the PROXY protocol, along along with the Common Name from the subject of
12151 the client certificate (if any), is added to the PROXY protocol header. This
12152 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012153 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
12154 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012155
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012156slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012157 The "slowstart" parameter for a server accepts a value in milliseconds which
12158 indicates after how long a server which has just come back up will run at
12159 full speed. Just as with every other time-based parameter, it can be entered
12160 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
12161 linearly from 0 to 100% during this time. The limitation applies to two
12162 parameters :
12163
12164 - maxconn: the number of connections accepted by the server will grow from 1
12165 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
12166
12167 - weight: when the backend uses a dynamic weighted algorithm, the weight
12168 grows linearly from 1 to 100%. In this case, the weight is updated at every
12169 health-check. For this reason, it is important that the "inter" parameter
12170 is smaller than the "slowstart", in order to maximize the number of steps.
12171
12172 The slowstart never applies when haproxy starts, otherwise it would cause
12173 trouble to running servers. It only applies when a server has been previously
12174 seen as failed.
12175
Willy Tarreau732eac42015-07-09 11:40:25 +020012176sni <expression>
12177 The "sni" parameter evaluates the sample fetch expression, converts it to a
12178 string and uses the result as the host name sent in the SNI TLS extension to
12179 the server. A typical use case is to send the SNI received from the client in
12180 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020012181 expression, though alternatives such as req.hdr(host) can also make sense. If
12182 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012183 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010012184 "verify" directive for more details. If you want to set a SNI for health
12185 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020012186
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012187source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020012188source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012189source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012190 The "source" parameter sets the source address which will be used when
12191 connecting to the server. It follows the exact same parameters and principle
12192 as the backend "source" keyword, except that it only applies to the server
12193 referencing it. Please consult the "source" keyword for details.
12194
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012195 Additionally, the "source" statement on a server line allows one to specify a
12196 source port range by indicating the lower and higher bounds delimited by a
12197 dash ('-'). Some operating systems might require a valid IP address when a
12198 source port range is specified. It is permitted to have the same IP/range for
12199 several servers. Doing so makes it possible to bypass the maximum of 64k
12200 total concurrent connections. The limit will then reach 64k connections per
12201 server.
12202
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000012203 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
12204 specifying the source address without port(s).
12205
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012206ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020012207 This option enables SSL ciphering on outgoing connections to the server. It
12208 is critical to verify server certificates using "verify" when using SSL to
12209 connect to servers, otherwise the communication is prone to trivial man in
12210 the-middle attacks rendering SSL useless. When this option is used, health
12211 checks are automatically sent in SSL too unless there is a "port" or an
12212 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012213 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
12214 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012215
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012216ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12217 This option enforces use of <version> or lower when SSL is used to communicate
12218 with the server. This option is also available on global statement
12219 "ssl-default-server-options". See also "ssl-min-ver".
12220
12221ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12222 This option enforces use of <version> or upper when SSL is used to communicate
12223 with the server. This option is also available on global statement
12224 "ssl-default-server-options". See also "ssl-max-ver".
12225
Frédéric Lécailled2376272017-03-21 18:52:12 +010012226ssl-reuse
12227 This option may be used as "server" setting to reset any "no-ssl-reuse"
12228 setting which would have been inherited from "default-server" directive as
12229 default value.
12230 It may also be used as "default-server" setting to reset any previous
12231 "default-server" "no-ssl-reuse" setting.
12232
12233stick
12234 This option may be used as "server" setting to reset any "non-stick"
12235 setting which would have been inherited from "default-server" directive as
12236 default value.
12237 It may also be used as "default-server" setting to reset any previous
12238 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012239
Alexander Liu2a54bb72019-05-22 19:44:48 +080012240socks4 <addr>:<port>
John Roeslerfb2fce12019-07-10 15:45:51 -050012241 This option enables upstream socks4 tunnel for outgoing connections to the
Alexander Liu2a54bb72019-05-22 19:44:48 +080012242 server. Using this option won't force the health check to go via socks4 by
12243 default. You will have to use the keyword "check-via-socks4" to enable it.
12244
Willy Tarreau163d4622015-10-13 16:16:41 +020012245tcp-ut <delay>
12246 Sets the TCP User Timeout for all outgoing connections to this server. This
12247 option is available on Linux since version 2.6.37. It allows haproxy to
12248 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012249 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020012250 long-lived connections experiencing long idle periods such as remote
12251 terminals or database connection pools, where the client and server timeouts
12252 must remain high to allow a long period of idle, but where it is important to
12253 detect that the server has disappeared in order to release all resources
12254 associated with its connection (and the client's session). One typical use
12255 case is also to force dead server connections to die when health checks are
12256 too slow or during a soft reload since health checks are then disabled. The
12257 argument is a delay expressed in milliseconds by default. This only works for
12258 regular TCP connections, and is ignored for other protocols.
12259
Willy Tarreau034c88c2017-01-23 23:36:45 +010012260tfo
12261 This option enables using TCP fast open when connecting to servers, on
12262 systems that support it (currently only the Linux kernel >= 4.11).
12263 See the "tfo" bind option for more information about TCP fast open.
12264 Please note that when using tfo, you should also use the "conn-failure",
12265 "empty-response" and "response-timeout" keywords for "retry-on", or haproxy
Frédéric Lécaille1b9423d2019-07-04 14:19:06 +020012266 won't be able to retry the connection on failure. See also "no-tfo".
Willy Tarreau034c88c2017-01-23 23:36:45 +010012267
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012268track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020012269 This option enables ability to set the current state of the server by tracking
12270 another one. It is possible to track a server which itself tracks another
12271 server, provided that at the end of the chain, a server has health checks
12272 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012273 used, it has to be enabled on both proxies.
12274
Frédéric Lécailled2376272017-03-21 18:52:12 +010012275tls-tickets
12276 This option may be used as "server" setting to reset any "no-tls-tickets"
12277 setting which would have been inherited from "default-server" directive as
12278 default value.
12279 It may also be used as "default-server" setting to reset any previous
12280 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012281
Emeric Brunef42d922012-10-11 16:11:36 +020012282verify [none|required]
12283 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010012284 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012285 certificate provided by the server is verified using CAs from 'ca-file' and
12286 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012287 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012288 the name passed using the "sni" directive, or if not provided, the static
12289 host name passed using the "verifyhost" directive. When no name is found, the
12290 certificate's names are ignored. For this reason, without SNI it's important
12291 to use "verifyhost". On verification failure the handshake is aborted. It is
12292 critically important to verify server certificates when using SSL to connect
12293 to servers, otherwise the communication is prone to trivial man-in-the-middle
12294 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
12295 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020012296
Evan Broderbe554312013-06-27 00:05:25 -070012297verifyhost <hostname>
12298 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012299 only takes effect if 'verify required' is also specified. This directive sets
12300 a default static hostname to check the server's certificate against when no
12301 SNI was used to connect to the server. If SNI is not used, this is the only
12302 way to enable hostname verification. This static hostname, when set, will
12303 also be used for health checks (which cannot provide an SNI value). If none
12304 of the hostnames in the certificate match the specified hostname, the
12305 handshake is aborted. The hostnames in the server-provided certificate may
12306 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070012307
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012308weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012309 The "weight" parameter is used to adjust the server's weight relative to
12310 other servers. All servers will receive a load proportional to their weight
12311 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020012312 load. The default weight is 1, and the maximal value is 256. A value of 0
12313 means the server will not participate in load-balancing but will still accept
12314 persistent connections. If this parameter is used to distribute the load
12315 according to server's capacity, it is recommended to start with values which
12316 can both grow and shrink, for instance between 10 and 100 to leave enough
12317 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012318
12319
Cyril Bonté46175dd2015-07-02 22:45:32 +0200123205.3. Server IP address resolution using DNS
12321-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012322
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012323HAProxy allows using a host name on the server line to retrieve its IP address
12324using name servers. By default, HAProxy resolves the name when parsing the
12325configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012326This is not sufficient in some cases, such as in Amazon where a server's IP
12327can change after a reboot or an ELB Virtual IP can change based on current
12328workload.
12329This chapter describes how HAProxy can be configured to process server's name
12330resolution at run time.
12331Whether run time server name resolution has been enable or not, HAProxy will
12332carry on doing the first resolution when parsing the configuration.
12333
12334
Cyril Bonté46175dd2015-07-02 22:45:32 +0200123355.3.1. Global overview
12336----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012337
12338As we've seen in introduction, name resolution in HAProxy occurs at two
12339different steps of the process life:
12340
12341 1. when starting up, HAProxy parses the server line definition and matches a
12342 host name. It uses libc functions to get the host name resolved. This
12343 resolution relies on /etc/resolv.conf file.
12344
Christopher Faulet67957bd2017-09-27 11:00:59 +020012345 2. at run time, HAProxy performs periodically name resolutions for servers
12346 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012347
12348A few other events can trigger a name resolution at run time:
12349 - when a server's health check ends up in a connection timeout: this may be
12350 because the server has a new IP address. So we need to trigger a name
12351 resolution to know this new IP.
12352
Christopher Faulet67957bd2017-09-27 11:00:59 +020012353When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012354HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020012355SRV label is specified, then the corresponding SRV records will be retrieved
12356from the DNS server, and the provided hostnames will be used. The SRV label
12357will be checked periodically, and if any server are added or removed, haproxy
12358will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020012359
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012360A few things important to notice:
John Roeslerfb2fce12019-07-10 15:45:51 -050012361 - all the name servers are queried in the meantime. HAProxy will process the
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012362 first valid response.
12363
12364 - a resolution is considered as invalid (NX, timeout, refused), when all the
12365 servers return an error.
12366
12367
Cyril Bonté46175dd2015-07-02 22:45:32 +0200123685.3.2. The resolvers section
12369----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012370
12371This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020012372HAProxy. There can be as many as resolvers section as needed. Each section can
12373contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012374
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012375When multiple name servers are configured in a resolvers section, then HAProxy
12376uses the first valid response. In case of invalid responses, only the last one
12377is treated. Purpose is to give the chance to a slow server to deliver a valid
12378answer after a fast faulty or outdated server.
12379
12380When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020012381used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012382
Christopher Faulet67957bd2017-09-27 11:00:59 +020012383 1. HAProxy retries the same DNS query with a new query type. The A queries are
12384 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
12385 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012386
Christopher Faulet67957bd2017-09-27 11:00:59 +020012387 2. When the fallback on the query type was done (or not applicable), HAProxy
12388 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012389
Christopher Faulet67957bd2017-09-27 11:00:59 +020012390 3. HAProxy retries previous steps <resolve_retires> times. If no valid
12391 response is received after that, it stops the DNS resolution and reports
12392 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012393
Christopher Faulet67957bd2017-09-27 11:00:59 +020012394For example, with 2 name servers configured in a resolvers section, the
12395following scenarios are possible:
12396
12397 - First response is valid and is applied directly, second response is
12398 ignored
12399
12400 - First response is invalid and second one is valid, then second response is
12401 applied
12402
12403 - First response is a NX domain and second one a truncated response, then
12404 HAProxy retries the query with a new type
12405
12406 - First response is a NX domain and second one is a timeout, then HAProxy
12407 retries the query with a new type
12408
12409 - Query timed out for both name servers, then HAProxy retries it with the
12410 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012411
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012412As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
12413a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020012414<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012415
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012416
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012417resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012418 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012419
12420A resolvers section accept the following parameters:
12421
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012422accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012423 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020012424 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012425 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
12426 by RFC 6891)
12427
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020012428 Note: the maximum allowed value is 8192.
12429
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012430nameserver <id> <ip>:<port>
12431 DNS server description:
12432 <id> : label of the server, should be unique
12433 <ip> : IP address of the server
12434 <port> : port where the DNS service actually runs
12435
Ben Draut44e609b2018-05-29 15:40:08 -060012436parse-resolv-conf
12437 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
12438 list. Ordered as if each nameserver in /etc/resolv.conf was individually
12439 placed in the resolvers section in place of this directive.
12440
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012441hold <status> <period>
12442 Defines <period> during which the last name resolution should be kept based
12443 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012444 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012445 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012446 <period> : interval between two successive name resolution when the last
12447 answer was in <status>. It follows the HAProxy time format.
12448 <period> is in milliseconds by default.
12449
Baptiste Assmann686408b2017-08-18 10:15:42 +020012450 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012451
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012452resolve_retries <nb>
12453 Defines the number <nb> of queries to send to resolve a server name before
12454 giving up.
12455 Default value: 3
12456
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012457 A retry occurs on name server timeout or when the full sequence of DNS query
12458 type failover is over and we need to start up from the default ANY query
12459 type.
12460
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012461timeout <event> <time>
12462 Defines timeouts related to name resolution
12463 <event> : the event on which the <time> timeout period applies to.
12464 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012465 - resolve : default time to trigger name resolutions when no
12466 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012467 Default value: 1s
12468 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012469 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012470 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012471 <time> : time related to the event. It follows the HAProxy time format.
12472 <time> is expressed in milliseconds.
12473
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012474 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012475
12476 resolvers mydns
12477 nameserver dns1 10.0.0.1:53
12478 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060012479 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012480 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020012481 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012482 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012483 hold other 30s
12484 hold refused 30s
12485 hold nx 30s
12486 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012487 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012488 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012489
12490
Christopher Faulet87f1f3d2019-07-18 14:51:20 +0200124916. Cache
12492---------
12493
12494HAProxy provides a cache, which was designed to perform cache on small objects
12495(favicon, css...). This is a minimalist low-maintenance cache which runs in
12496RAM.
12497
12498The cache is based on a memory which is shared between processes and threads,
12499this memory is split in blocks of 1k.
12500
12501If an object is not used anymore, it can be deleted to store a new object
12502independently of its expiration date. The oldest objects are deleted first
12503when we try to allocate a new one.
12504
12505The cache uses a hash of the host header and the URI as the key.
12506
12507It's possible to view the status of a cache using the Unix socket command
12508"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
12509for more details.
12510
12511When an object is delivered from the cache, the server name in the log is
12512replaced by "<CACHE>".
12513
12514
125156.1. Limitation
12516----------------
12517
12518The cache won't store and won't deliver objects in these cases:
12519
12520- If the response is not a 200
12521- If the response contains a Vary header
12522- If the Content-Length + the headers size is greater than "max-object-size"
12523- If the response is not cacheable
12524
12525- If the request is not a GET
12526- If the HTTP version of the request is smaller than 1.1
12527- If the request contains an Authorization header
12528
12529
125306.2. Setup
12531-----------
12532
12533To setup a cache, you must define a cache section and use it in a proxy with
12534the corresponding http-request and response actions.
12535
12536
125376.2.1. Cache section
12538---------------------
12539
12540cache <name>
12541 Declare a cache section, allocate a shared cache memory named <name>, the
12542 size of cache is mandatory.
12543
12544total-max-size <megabytes>
12545 Define the size in RAM of the cache in megabytes. This size is split in
12546 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
12547
12548max-object-size <bytes>
12549 Define the maximum size of the objects to be cached. Must not be greater than
12550 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
12551 All objects with sizes larger than "max-object-size" will not be cached.
12552
12553max-age <seconds>
12554 Define the maximum expiration duration. The expiration is set has the lowest
12555 value between the s-maxage or max-age (in this order) directive in the
12556 Cache-Control response header and this value. The default value is 60
12557 seconds, which means that you can't cache an object more than 60 seconds by
12558 default.
12559
12560
125616.2.2. Proxy section
12562---------------------
12563
12564http-request cache-use <name> [ { if | unless } <condition> ]
12565 Try to deliver a cached object from the cache <name>. This directive is also
12566 mandatory to store the cache as it calculates the cache hash. If you want to
12567 use a condition for both storage and delivering that's a good idea to put it
12568 after this one.
12569
12570http-response cache-store <name> [ { if | unless } <condition> ]
12571 Store an http-response within the cache. The storage of the response headers
12572 is done at this step, which means you can use others http-response actions
12573 to modify headers before or after the storage of the response. This action
12574 is responsible for the setup of the cache storage filter.
12575
12576
12577Example:
12578
12579 backend bck1
12580 mode http
12581
12582 http-request cache-use foobar
12583 http-response cache-store foobar
12584 server srv1 127.0.0.1:80
12585
12586 cache foobar
12587 total-max-size 4
12588 max-age 240
12589
12590
Willy Tarreau74ca5042013-06-11 23:12:07 +0200125917. Using ACLs and fetching samples
12592----------------------------------
12593
Davor Ocelice9ed2812017-12-25 17:49:28 +010012594HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020012595client or server information, from tables, environmental information etc...
12596The action of extracting such data is called fetching a sample. Once retrieved,
12597these samples may be used for various purposes such as a key to a stick-table,
12598but most common usages consist in matching them against predefined constant
12599data called patterns.
12600
12601
126027.1. ACL basics
12603---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012604
12605The use of Access Control Lists (ACL) provides a flexible solution to perform
12606content switching and generally to take decisions based on content extracted
12607from the request, the response or any environmental status. The principle is
12608simple :
12609
Willy Tarreau74ca5042013-06-11 23:12:07 +020012610 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012611 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020012612 - apply one or multiple pattern matching methods on this sample
12613 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012614
Willy Tarreau74ca5042013-06-11 23:12:07 +020012615The actions generally consist in blocking a request, selecting a backend, or
12616adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012617
12618In order to define a test, the "acl" keyword is used. The syntax is :
12619
Willy Tarreau74ca5042013-06-11 23:12:07 +020012620 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012621
12622This creates a new ACL <aclname> or completes an existing one with new tests.
12623Those tests apply to the portion of request/response specified in <criterion>
12624and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012625an operator which may be specified before the set of values. Optionally some
12626conversion operators may be applied to the sample, and they will be specified
12627as a comma-delimited list of keywords just after the first keyword. The values
12628are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012629
12630ACL names must be formed from upper and lower case letters, digits, '-' (dash),
12631'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
12632which means that "my_acl" and "My_Acl" are two different ACLs.
12633
12634There is no enforced limit to the number of ACLs. The unused ones do not affect
12635performance, they just consume a small amount of memory.
12636
Willy Tarreau74ca5042013-06-11 23:12:07 +020012637The criterion generally is the name of a sample fetch method, or one of its ACL
12638specific declinations. The default test method is implied by the output type of
12639this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012640methods of a same sample fetch method. The sample fetch methods are the only
12641ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012642
12643Sample fetch methods return data which can be of the following types :
12644 - boolean
12645 - integer (signed or unsigned)
12646 - IPv4 or IPv6 address
12647 - string
12648 - data block
12649
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012650Converters transform any of these data into any of these. For example, some
12651converters might convert a string to a lower-case string while other ones
12652would turn a string to an IPv4 address, or apply a netmask to an IP address.
12653The resulting sample is of the type of the last converter applied to the list,
12654which defaults to the type of the sample fetch method.
12655
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012656Each sample or converter returns data of a specific type, specified with its
12657keyword in this documentation. When an ACL is declared using a standard sample
12658fetch method, certain types automatically involved a default matching method
12659which are summarized in the table below :
12660
12661 +---------------------+-----------------+
12662 | Sample or converter | Default |
12663 | output type | matching method |
12664 +---------------------+-----------------+
12665 | boolean | bool |
12666 +---------------------+-----------------+
12667 | integer | int |
12668 +---------------------+-----------------+
12669 | ip | ip |
12670 +---------------------+-----------------+
12671 | string | str |
12672 +---------------------+-----------------+
12673 | binary | none, use "-m" |
12674 +---------------------+-----------------+
12675
12676Note that in order to match a binary samples, it is mandatory to specify a
12677matching method, see below.
12678
Willy Tarreau74ca5042013-06-11 23:12:07 +020012679The ACL engine can match these types against patterns of the following types :
12680 - boolean
12681 - integer or integer range
12682 - IP address / network
12683 - string (exact, substring, suffix, prefix, subdir, domain)
12684 - regular expression
12685 - hex block
12686
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012687The following ACL flags are currently supported :
12688
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012689 -i : ignore case during matching of all subsequent patterns.
12690 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012691 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010012692 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010012693 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010012694 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020012695 -- : force end of flags. Useful when a string looks like one of the flags.
12696
Willy Tarreau74ca5042013-06-11 23:12:07 +020012697The "-f" flag is followed by the name of a file from which all lines will be
12698read as individual values. It is even possible to pass multiple "-f" arguments
12699if the patterns are to be loaded from multiple files. Empty lines as well as
12700lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
12701will be stripped. If it is absolutely necessary to insert a valid pattern
12702beginning with a sharp, just prefix it with a space so that it is not taken for
12703a comment. Depending on the data type and match method, haproxy may load the
12704lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
12705exact string matching. In this case, duplicates will automatically be removed.
12706
Thierry FOURNIER9860c412014-01-29 14:23:29 +010012707The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
12708parsed as two column file. The first column contains the patterns used by the
12709ACL, and the second column contain the samples. The sample can be used later by
12710a map. This can be useful in some rare cases where an ACL would just be used to
12711check for the existence of a pattern in a map before a mapping is applied.
12712
Thierry FOURNIER3534d882014-01-20 17:01:44 +010012713The "-u" flag forces the unique id of the ACL. This unique id is used with the
12714socket interface to identify ACL and dynamically change its values. Note that a
12715file is always identified by its name even if an id is set.
12716
Willy Tarreau74ca5042013-06-11 23:12:07 +020012717Also, note that the "-i" flag applies to subsequent entries and not to entries
12718loaded from files preceding it. For instance :
12719
12720 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
12721
12722In this example, each line of "exact-ua.lst" will be exactly matched against
12723the "user-agent" header of the request. Then each line of "generic-ua" will be
12724case-insensitively matched. Then the word "test" will be insensitively matched
12725as well.
12726
12727The "-m" flag is used to select a specific pattern matching method on the input
12728sample. All ACL-specific criteria imply a pattern matching method and generally
12729do not need this flag. However, this flag is useful with generic sample fetch
12730methods to describe how they're going to be matched against the patterns. This
12731is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010012732obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020012733followed by a pattern matching method name, this method is used instead of the
12734default one for the criterion. This makes it possible to match contents in ways
12735that were not initially planned, or with sample fetch methods which return a
12736string. The matching method also affects the way the patterns are parsed.
12737
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010012738The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
12739By default, if the parser cannot parse ip address it considers that the parsed
12740string is maybe a domain name and try dns resolution. The flag "-n" disable this
12741resolution. It is useful for detecting malformed ip lists. Note that if the DNS
12742server is not reachable, the haproxy configuration parsing may last many minutes
John Roeslerfb2fce12019-07-10 15:45:51 -050012743waiting for the timeout. During this time no error messages are displayed. The
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010012744flag "-n" disable this behavior. Note also that during the runtime, this
12745function is disabled for the dynamic acl modifications.
12746
Willy Tarreau74ca5042013-06-11 23:12:07 +020012747There are some restrictions however. Not all methods can be used with all
12748sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
12749be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020012750
12751 - "found" : only check if the requested sample could be found in the stream,
12752 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020012753 to pass any pattern to avoid confusion. This matching method is
12754 particularly useful to detect presence of certain contents such
12755 as headers, cookies, etc... even if they are empty and without
12756 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012757
12758 - "bool" : check the value as a boolean. It can only be applied to fetches
12759 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012760 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012761
12762 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020012763 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012764
12765 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020012766 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012767
Davor Ocelice9ed2812017-12-25 17:49:28 +010012768 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020012769 binary sequence. This may be used with binary or string samples.
12770
12771 - "len" : match the sample's length as an integer. This may be used with
12772 binary or string samples.
12773
Willy Tarreau74ca5042013-06-11 23:12:07 +020012774 - "str" : exact match : match the contents against a string. This may be
12775 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012776
Willy Tarreau74ca5042013-06-11 23:12:07 +020012777 - "sub" : substring match : check that the contents contain at least one of
12778 the provided string patterns. This may be used with binary or
12779 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012780
Willy Tarreau74ca5042013-06-11 23:12:07 +020012781 - "reg" : regex match : match the contents against a list of regular
12782 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012783
Willy Tarreau74ca5042013-06-11 23:12:07 +020012784 - "beg" : prefix match : check that the contents begin like the provided
12785 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012786
Willy Tarreau74ca5042013-06-11 23:12:07 +020012787 - "end" : suffix match : check that the contents end like the provided
12788 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012789
Willy Tarreau74ca5042013-06-11 23:12:07 +020012790 - "dir" : subdir match : check that a slash-delimited portion of the
12791 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012792 This may be used with binary or string samples.
12793
Willy Tarreau74ca5042013-06-11 23:12:07 +020012794 - "dom" : domain match : check that a dot-delimited portion of the contents
12795 exactly match one of the provided string patterns. This may be
12796 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020012797
12798For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
12799request, it is possible to do :
12800
12801 acl jsess_present cook(JSESSIONID) -m found
12802
12803In order to apply a regular expression on the 500 first bytes of data in the
12804buffer, one would use the following acl :
12805
12806 acl script_tag payload(0,500) -m reg -i <script>
12807
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012808On systems where the regex library is much slower when using "-i", it is
12809possible to convert the sample to lowercase before matching, like this :
12810
12811 acl script_tag payload(0,500),lower -m reg <script>
12812
Willy Tarreau74ca5042013-06-11 23:12:07 +020012813All ACL-specific criteria imply a default matching method. Most often, these
12814criteria are composed by concatenating the name of the original sample fetch
12815method and the matching method. For example, "hdr_beg" applies the "beg" match
12816to samples retrieved using the "hdr" fetch method. Since all ACL-specific
12817criteria rely on a sample fetch method, it is always possible instead to use
12818the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012819
Willy Tarreau74ca5042013-06-11 23:12:07 +020012820If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012821the matching method is simply applied to the underlying sample fetch method.
12822For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012823
Willy Tarreau74ca5042013-06-11 23:12:07 +020012824 acl short_form hdr_beg(host) www.
12825 acl alternate1 hdr_beg(host) -m beg www.
12826 acl alternate2 hdr_dom(host) -m beg www.
12827 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012828
Willy Tarreau2b5285d2010-05-09 23:45:24 +020012829
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012830The table below summarizes the compatibility matrix between sample or converter
12831types and the pattern types to fetch against. It indicates for each compatible
12832combination the name of the matching method to be used, surrounded with angle
12833brackets ">" and "<" when the method is the default one and will work by
12834default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010012835
Willy Tarreau74ca5042013-06-11 23:12:07 +020012836 +-------------------------------------------------+
12837 | Input sample type |
12838 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012839 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012840 +----------------------+---------+---------+---------+---------+---------+
12841 | none (presence only) | found | found | found | found | found |
12842 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012843 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012844 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012845 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012846 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012847 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012848 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012849 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012850 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012851 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012852 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012853 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012854 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012855 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012856 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012857 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012858 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012859 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012860 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012861 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012862 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010012863 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020012864 +----------------------+---------+---------+---------+---------+---------+
12865 | hex block | | | | bin | bin |
12866 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020012867
12868
Willy Tarreau74ca5042013-06-11 23:12:07 +0200128697.1.1. Matching booleans
12870------------------------
12871
12872In order to match a boolean, no value is needed and all values are ignored.
12873Boolean matching is used by default for all fetch methods of type "boolean".
12874When boolean matching is used, the fetched value is returned as-is, which means
12875that a boolean "true" will always match and a boolean "false" will never match.
12876
12877Boolean matching may also be enforced using "-m bool" on fetch methods which
12878return an integer value. Then, integer value 0 is converted to the boolean
12879"false" and all other values are converted to "true".
12880
Willy Tarreau6a06a402007-07-15 20:15:28 +020012881
Willy Tarreau74ca5042013-06-11 23:12:07 +0200128827.1.2. Matching integers
12883------------------------
12884
12885Integer matching applies by default to integer fetch methods. It can also be
12886enforced on boolean fetches using "-m int". In this case, "false" is converted
12887to the integer 0, and "true" is converted to the integer 1.
12888
12889Integer matching also supports integer ranges and operators. Note that integer
12890matching only applies to positive values. A range is a value expressed with a
12891lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012892
12893For instance, "1024:65535" is a valid range to represent a range of
12894unprivileged ports, and "1024:" would also work. "0:1023" is a valid
12895representation of privileged ports, and ":1023" would also work.
12896
Willy Tarreau62644772008-07-16 18:36:06 +020012897As a special case, some ACL functions support decimal numbers which are in fact
12898two integers separated by a dot. This is used with some version checks for
12899instance. All integer properties apply to those decimal numbers, including
12900ranges and operators.
12901
Willy Tarreau6a06a402007-07-15 20:15:28 +020012902For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010012903operators with ranges does not make much sense and is strongly discouraged.
12904Similarly, it does not make much sense to perform order comparisons with a set
12905of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012906
Willy Tarreau0ba27502007-12-24 16:55:16 +010012907Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020012908
12909 eq : true if the tested value equals at least one value
12910 ge : true if the tested value is greater than or equal to at least one value
12911 gt : true if the tested value is greater than at least one value
12912 le : true if the tested value is less than or equal to at least one value
12913 lt : true if the tested value is less than at least one value
12914
Willy Tarreau0ba27502007-12-24 16:55:16 +010012915For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020012916
12917 acl negative-length hdr_val(content-length) lt 0
12918
Willy Tarreau62644772008-07-16 18:36:06 +020012919This one matches SSL versions between 3.0 and 3.1 (inclusive) :
12920
12921 acl sslv3 req_ssl_ver 3:3.1
12922
Willy Tarreau6a06a402007-07-15 20:15:28 +020012923
Willy Tarreau74ca5042013-06-11 23:12:07 +0200129247.1.3. Matching strings
12925-----------------------
12926
12927String matching applies to string or binary fetch methods, and exists in 6
12928different forms :
12929
12930 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012931 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020012932
12933 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012934 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020012935
12936 - prefix match (-m beg) : the patterns are compared with the beginning of
12937 the extracted string, and the ACL matches if any of them matches.
12938
12939 - suffix match (-m end) : the patterns are compared with the end of the
12940 extracted string, and the ACL matches if any of them matches.
12941
Baptiste Assmann33db6002016-03-06 23:32:10 +010012942 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020012943 string, delimited with slashes ("/"), and the ACL matches if any of them
12944 matches.
12945
12946 - domain match (-m dom) : the patterns are looked up inside the extracted
12947 string, delimited with dots ("."), and the ACL matches if any of them
12948 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012949
12950String matching applies to verbatim strings as they are passed, with the
12951exception of the backslash ("\") which makes it possible to escape some
12952characters such as the space. If the "-i" flag is passed before the first
12953string, then the matching will be performed ignoring the case. In order
12954to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010012955before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020012956
Mathias Weiersmuellercb250fc2019-12-02 09:43:40 +010012957Do not use string matches for binary fetches which might contain null bytes
12958(0x00), as the comparison stops at the occurrence of the first null byte.
12959Instead, convert the binary fetch to a hex string with the hex converter first.
12960
12961Example:
12962 # matches if the string <tag> is present in the binary sample
12963 acl tag_found req.payload(0,0),hex -m sub 3C7461673E
12964
Willy Tarreau6a06a402007-07-15 20:15:28 +020012965
Willy Tarreau74ca5042013-06-11 23:12:07 +0200129667.1.4. Matching regular expressions (regexes)
12967---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020012968
12969Just like with string matching, regex matching applies to verbatim strings as
12970they are passed, with the exception of the backslash ("\") which makes it
12971possible to escape some characters such as the space. If the "-i" flag is
12972passed before the first regex, then the matching will be performed ignoring
12973the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010012974the "--" flag before the first string. Same principle applies of course to
12975match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020012976
12977
Willy Tarreau74ca5042013-06-11 23:12:07 +0200129787.1.5. Matching arbitrary data blocks
12979-------------------------------------
12980
12981It is possible to match some extracted samples against a binary block which may
12982not safely be represented as a string. For this, the patterns must be passed as
12983a series of hexadecimal digits in an even number, when the match method is set
12984to binary. Each sequence of two digits will represent a byte. The hexadecimal
12985digits may be used upper or lower case.
12986
12987Example :
12988 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
12989 acl hello payload(0,6) -m bin 48656c6c6f0a
12990
12991
129927.1.6. Matching IPv4 and IPv6 addresses
12993---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020012994
12995IPv4 addresses values can be specified either as plain addresses or with a
12996netmask appended, in which case the IPv4 address matches whenever it is
12997within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010012998host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010012999difficult to read and debug configurations. If hostnames are used, you should
13000at least ensure that they are present in /etc/hosts so that the configuration
13001does not depend on any random DNS match at the moment the configuration is
13002parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013003
Daniel Schnellereba56342016-04-13 00:26:52 +020013004The dotted IPv4 address notation is supported in both regular as well as the
13005abbreviated form with all-0-octets omitted:
13006
13007 +------------------+------------------+------------------+
13008 | Example 1 | Example 2 | Example 3 |
13009 +------------------+------------------+------------------+
13010 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
13011 | 192.168.1 | 10.12 | 127.1 |
13012 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
13013 | 192.168.1/22 | 10.12/8 | 127.1/8 |
13014 +------------------+------------------+------------------+
13015
13016Notice that this is different from RFC 4632 CIDR address notation in which
13017192.168.42/24 would be equivalent to 192.168.42.0/24.
13018
Willy Tarreauceb4ac92012-04-28 00:41:46 +020013019IPv6 may be entered in their usual form, with or without a netmask appended.
13020Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
13021trouble with randomly resolved IP addresses, host names are never allowed in
13022IPv6 patterns.
13023
13024HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
13025following situations :
13026 - tested address is IPv4, pattern address is IPv4, the match applies
13027 in IPv4 using the supplied mask if any.
13028 - tested address is IPv6, pattern address is IPv6, the match applies
13029 in IPv6 using the supplied mask if any.
13030 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
13031 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
13032 ::IPV4 or ::ffff:IPV4, otherwise it fails.
13033 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
13034 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
13035 applied in IPv6 using the supplied IPv6 mask.
13036
Willy Tarreau74ca5042013-06-11 23:12:07 +020013037
130387.2. Using ACLs to form conditions
13039----------------------------------
13040
13041Some actions are only performed upon a valid condition. A condition is a
13042combination of ACLs with operators. 3 operators are supported :
13043
13044 - AND (implicit)
13045 - OR (explicit with the "or" keyword or the "||" operator)
13046 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020013047
Willy Tarreau74ca5042013-06-11 23:12:07 +020013048A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020013049
Willy Tarreau74ca5042013-06-11 23:12:07 +020013050 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020013051
Willy Tarreau74ca5042013-06-11 23:12:07 +020013052Such conditions are generally used after an "if" or "unless" statement,
13053indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020013054
Willy Tarreau74ca5042013-06-11 23:12:07 +020013055For instance, to block HTTP requests to the "*" URL with methods other than
13056"OPTIONS", as well as POST requests without content-length, and GET or HEAD
13057requests with a content-length greater than 0, and finally every request which
13058is not either GET/HEAD/POST/OPTIONS !
13059
13060 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013061 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
13062 http-request deny if METH_GET HTTP_CONTENT
13063 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020013064
13065To select a different backend for requests to static contents on the "www" site
13066and to every request on the "img", "video", "download" and "ftp" hosts :
13067
13068 acl url_static path_beg /static /images /img /css
13069 acl url_static path_end .gif .png .jpg .css .js
13070 acl host_www hdr_beg(host) -i www
13071 acl host_static hdr_beg(host) -i img. video. download. ftp.
13072
Davor Ocelice9ed2812017-12-25 17:49:28 +010013073 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020013074 # of host "www". Use backend "www" for the rest.
13075 use_backend static if host_static or host_www url_static
13076 use_backend www if host_www
13077
13078It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
13079expressions that are built on the fly without needing to be declared. They must
13080be enclosed between braces, with a space before and after each brace (because
13081the braces must be seen as independent words). Example :
13082
13083 The following rule :
13084
13085 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013086 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020013087
13088 Can also be written that way :
13089
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013090 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020013091
13092It is generally not recommended to use this construct because it's a lot easier
13093to leave errors in the configuration when written that way. However, for very
13094simple rules matching only one source IP address for instance, it can make more
13095sense to use them than to declare ACLs with random names. Another example of
13096good use is the following :
13097
13098 With named ACLs :
13099
13100 acl site_dead nbsrv(dynamic) lt 2
13101 acl site_dead nbsrv(static) lt 2
13102 monitor fail if site_dead
13103
13104 With anonymous ACLs :
13105
13106 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
13107
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013108See section 4.2 for detailed help on the "http-request deny" and "use_backend"
13109keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013110
13111
131127.3. Fetching samples
13113---------------------
13114
13115Historically, sample fetch methods were only used to retrieve data to match
13116against patterns using ACLs. With the arrival of stick-tables, a new class of
13117sample fetch methods was created, most often sharing the same syntax as their
13118ACL counterpart. These sample fetch methods are also known as "fetches". As
13119of now, ACLs and fetches have converged. All ACL fetch methods have been made
13120available as fetch methods, and ACLs may use any sample fetch method as well.
13121
13122This section details all available sample fetch methods and their output type.
13123Some sample fetch methods have deprecated aliases that are used to maintain
13124compatibility with existing configurations. They are then explicitly marked as
13125deprecated and should not be used in new setups.
13126
13127The ACL derivatives are also indicated when available, with their respective
13128matching methods. These ones all have a well defined default pattern matching
13129method, so it is never necessary (though allowed) to pass the "-m" option to
13130indicate how the sample will be matched using ACLs.
13131
13132As indicated in the sample type versus matching compatibility matrix above,
13133when using a generic sample fetch method in an ACL, the "-m" option is
13134mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
13135the same keyword exists as an ACL keyword and as a standard fetch method, the
13136ACL engine will automatically pick the ACL-only one by default.
13137
13138Some of these keywords support one or multiple mandatory arguments, and one or
13139multiple optional arguments. These arguments are strongly typed and are checked
13140when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010013141incorrect argument (e.g. an unresolved backend name). Fetch function arguments
13142are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020013143is optional, it will be indicated below between square brackets ('[ ]'). When
13144all arguments are optional, the parenthesis may be omitted.
13145
13146Thus, the syntax of a standard sample fetch method is one of the following :
13147 - name
13148 - name(arg1)
13149 - name(arg1,arg2)
13150
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013151
131527.3.1. Converters
13153-----------------
13154
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013155Sample fetch methods may be combined with transformations to be applied on top
13156of the fetched sample (also called "converters"). These combinations form what
13157is called "sample expressions" and the result is a "sample". Initially this
13158was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010013159has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013160unique-id-format, add-header, ...).
13161
13162These transformations are enumerated as a series of specific keywords after the
13163sample fetch method. These keywords may equally be appended immediately after
13164the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010013165support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013166
Willy Tarreau97707872015-01-27 15:12:13 +010013167A certain category of converters are bitwise and arithmetic operators which
13168support performing basic operations on integers. Some bitwise operations are
13169supported (and, or, xor, cpl) and some arithmetic operations are supported
13170(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
13171bool) which make it possible to report a match without having to write an ACL.
13172
Willy Tarreau74ca5042013-06-11 23:12:07 +020013173The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010013174
Ben Shillitof25e8e52016-12-02 14:25:37 +00001317551d.single(<prop>[,<prop>*])
13176 Returns values for the properties requested as a string, where values are
13177 separated by the delimiter specified with "51degrees-property-separator".
13178 The device is identified using the User-Agent header passed to the
13179 converter. The function can be passed up to five property names, and if a
13180 property name can't be found, the value "NoData" is returned.
13181
13182 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013183 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
13184 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000013185 # User-Agent passed to the converter.
13186 frontend http-in
13187 bind *:8081
13188 default_backend servers
13189 http-request set-header X-51D-DeviceTypeMobileTablet \
13190 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
13191
Willy Tarreau97707872015-01-27 15:12:13 +010013192add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013193 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013194 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013195 name. The name of the variable starts with an indication about its scope. The
13196 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013197 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013198 "sess" : the variable is shared with the whole session
13199 "txn" : the variable is shared with the transaction (request and response)
13200 "req" : the variable is shared only during request processing
13201 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013202 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013203 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013204
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010013205aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
13206 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
13207 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
13208 need to be base64 encoded and the returned result is in raw byte format.
13209 If the <aead_tag> validation fails, the converter doesn't return any data.
13210 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
13211 converter requires at least OpenSSL 1.0.1.
13212
13213 Example:
13214 http-response set-header X-Decrypted-Text %[var(txn.enc),\
13215 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
13216
Willy Tarreau97707872015-01-27 15:12:13 +010013217and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013218 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013219 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013220 numeric value or a variable name. The name of the variable starts with an
13221 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013222 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013223 "sess" : the variable is shared with the whole session
13224 "txn" : the variable is shared with the transaction (request and response)
13225 "req" : the variable is shared only during request processing
13226 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013227 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013228 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013229
Holger Just1bfc24b2017-05-06 00:56:53 +020013230b64dec
13231 Converts (decodes) a base64 encoded input string to its binary
13232 representation. It performs the inverse operation of base64().
13233
Emeric Brun53d1a982014-04-30 18:21:37 +020013234base64
13235 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010013236 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020013237 an SSL ID can be copied in a header).
13238
Willy Tarreau97707872015-01-27 15:12:13 +010013239bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013240 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013241 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013242 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013243 presence of a flag).
13244
Emeric Brun54c4ac82014-11-03 15:32:43 +010013245bytes(<offset>[,<length>])
13246 Extracts some bytes from an input binary sample. The result is a binary
13247 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010013248 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010013249
Willy Tarreau280f42b2018-02-19 15:34:12 +010013250concat([<start>],[<var>],[<end>])
13251 Concatenates up to 3 fields after the current sample which is then turned to
13252 a string. The first one, <start>, is a constant string, that will be appended
13253 immediately after the existing sample. It may be omitted if not used. The
13254 second one, <var>, is a variable name. The variable will be looked up, its
13255 contents converted to a string, and it will be appended immediately after the
13256 <first> part. If the variable is not found, nothing is appended. It may be
13257 omitted as well. The third field, <end> is a constant string that will be
13258 appended after the variable. It may also be omitted. Together, these elements
13259 allow to concatenate variables with delimiters to an existing set of
13260 variables. This can be used to build new variables made of a succession of
John Roeslerfb2fce12019-07-10 15:45:51 -050013261 other variables, such as colon-delimited values. Note that due to the config
Willy Tarreau280f42b2018-02-19 15:34:12 +010013262 parser, it is not possible to use a comma nor a closing parenthesis as
John Roeslerfb2fce12019-07-10 15:45:51 -050013263 delimiters.
Willy Tarreau280f42b2018-02-19 15:34:12 +010013264
13265 Example:
13266 tcp-request session set-var(sess.src) src
13267 tcp-request session set-var(sess.dn) ssl_c_s_dn
13268 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
13269 http-request set-header x-hap-sig %[var(txn.sig)]
13270
Willy Tarreau97707872015-01-27 15:12:13 +010013271cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013272 Takes the input value of type signed integer, applies a ones-complement
13273 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010013274
Willy Tarreau80599772015-01-20 19:35:24 +010013275crc32([<avalanche>])
13276 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
13277 hash function. Optionally, it is possible to apply a full avalanche hash
13278 function to the output if the optional <avalanche> argument equals 1. This
13279 converter uses the same functions as used by the various hash-based load
13280 balancing algorithms, so it will provide exactly the same results. It is
13281 provided for compatibility with other software which want a CRC32 to be
13282 computed on some input keys, so it follows the most common implementation as
13283 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
13284 but may provide a better or at least less predictable distribution. It must
13285 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013286 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
13287
13288crc32c([<avalanche>])
13289 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
13290 hash function. Optionally, it is possible to apply a full avalanche hash
13291 function to the output if the optional <avalanche> argument equals 1. This
13292 converter uses the same functions as described in RFC4960, Appendix B [8].
13293 It is provided for compatibility with other software which want a CRC32C to be
13294 computed on some input keys. It is slower than the other algorithms and it must
13295 not be used for security purposes as a 32-bit hash is trivial to break. See
13296 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010013297
David Carlier29b3ca32015-09-25 14:09:21 +010013298da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020013299 Asks the DeviceAtlas converter to identify the User Agent string passed on
13300 input, and to emit a string made of the concatenation of the properties
13301 enumerated in argument, delimited by the separator defined by the global
13302 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000013303 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020013304 configuration language.
13305
13306 Example:
13307 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013308 bind *:8881
13309 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000013310 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 +020013311
Willy Tarreau0851fd52019-12-17 10:07:25 +010013312debug([<prefix][,<destination>])
13313 This converter is used as debug tool. It takes a capture of the input sample
13314 and sends it to event sink <destination>, which may designate a ring buffer
13315 such as "buf0", as well as "stdout", or "stderr". Available sinks may be
13316 checked at run time by issuing "show events" on the CLI. When not specified,
13317 the output will be "buf0", which may be consulted via the CLI's "show events"
13318 command. An optional prefix <prefix> may be passed to help distinguish
13319 outputs from multiple expressions. It will then appear before the colon in
13320 the output message. The input sample is passed as-is on the output, so that
13321 it is safe to insert the debug converter anywhere in a chain, even with non-
13322 printable sample types.
13323
13324 Example:
13325 tcp-request connection track-sc0 src,debug(track-sc)
Thierry FOURNIER9687c772015-05-07 15:46:29 +020013326
Willy Tarreau97707872015-01-27 15:12:13 +010013327div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013328 Divides the input value of type signed integer by <value>, and returns the
13329 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013330 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010013331 variable name. The name of the variable starts with an indication about its
13332 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013333 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013334 "sess" : the variable is shared with the whole session
13335 "txn" : the variable is shared with the transaction (request and response)
13336 "req" : the variable is shared only during request processing
13337 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013338 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013339 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013340
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013341djb2([<avalanche>])
13342 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
13343 hash function. Optionally, it is possible to apply a full avalanche hash
13344 function to the output if the optional <avalanche> argument equals 1. This
13345 converter uses the same functions as used by the various hash-based load
13346 balancing algorithms, so it will provide exactly the same results. It is
13347 mostly intended for debugging, but can be used as a stick-table entry to
13348 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013349 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
13350 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013351
Willy Tarreau97707872015-01-27 15:12:13 +010013352even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013353 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010013354 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
13355
Marcin Deranek9631a282018-04-16 14:30:46 +020013356field(<index>,<delimiters>[,<count>])
13357 Extracts the substring at the given index counting from the beginning
13358 (positive index) or from the end (negative index) considering given delimiters
13359 from an input string. Indexes start at 1 or -1 and delimiters are a string
13360 formatted list of chars. Optionally you can specify <count> of fields to
13361 extract (default: 1). Value of 0 indicates extraction of all remaining
13362 fields.
13363
13364 Example :
13365 str(f1_f2_f3__f5),field(5,_) # f5
13366 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
13367 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
13368 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
13369 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010013370
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013371hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010013372 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013373 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010013374 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 +020013375 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010013376
Dragan Dosen3f957b22017-10-24 09:27:34 +020013377hex2i
13378 Converts a hex string containing two hex digits per input byte to an
John Roeslerfb2fce12019-07-10 15:45:51 -050013379 integer. If the input value cannot be converted, then zero is returned.
Dragan Dosen3f957b22017-10-24 09:27:34 +020013380
Cyril Bonté6bcd1822019-11-05 23:13:59 +010013381http_date([<offset],[<unit>])
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013382 Converts an integer supposed to contain a date since epoch to a string
13383 representing this date in a format suitable for use in HTTP header fields. If
Damien Claisseae6f1252019-10-30 15:57:28 +000013384 an offset value is specified, then it is added to the date before the
13385 conversion is operated. This is particularly useful to emit Date header fields,
13386 Expires values in responses when combined with a positive offset, or
13387 Last-Modified values when the offset is negative.
13388 If a unit value is specified, then consider the timestamp as either
13389 "s" for seconds (default behavior), "ms" for milliseconds, or "us" for
13390 microseconds since epoch. Offset is assumed to have the same unit as
13391 input timestamp.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013392
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013393in_table(<table>)
13394 Uses the string representation of the input sample to perform a look up in
13395 the specified table. If the key is not found in the table, a boolean false
13396 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010013397 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013398 or not a source IP address or an Authorization header was already seen).
13399
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013400ipmask(<mask4>, [<mask6>])
13401 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013402 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013403 table entries and as such use the same server. The mask4 can be passed in
13404 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
13405 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
13406 If no mask6 is given IPv6 addresses will fail to convert for backwards
13407 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013408
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013409json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010013410 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013411 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013412 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013413 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
13414 of errors:
13415 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
13416 bytes, ...)
13417 - invalid range (the decoded value is within a UTF-8 prohibited range),
13418 - code overlong (the value is encoded with more bytes than necessary).
13419
13420 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
13421 character is greater than 0xffff because the JSON string escape specification
13422 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
13423 in 4 variants designated by a combination of two suffix letters : "p" for
13424 "permissive" and "s" for "silently ignore". The behaviors of the decoders
13425 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013426 - "ascii" : never fails;
13427 - "utf8" : fails on any detected errors;
13428 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013429 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010013430 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013431 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
13432 characters corresponding to the other errors.
13433
13434 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010013435 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013436
13437 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013438 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013439 capture request header user-agent len 150
13440 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013441
13442 Input request from client 127.0.0.1:
13443 GET / HTTP/1.0
13444 User-Agent: Very "Ugly" UA 1/2
13445
13446 Output log:
13447 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
13448
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013449language(<value>[,<default>])
13450 Returns the value with the highest q-factor from a list as extracted from the
13451 "accept-language" header using "req.fhdr". Values with no q-factor have a
13452 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
13453 belong to the list of semi-colon delimited <values> will be considered. The
13454 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
13455 given list and a default value is provided, it is returned. Note that language
13456 names may have a variant after a dash ('-'). If this variant is present in the
13457 list, it will be matched, but if it is not, only the base language is checked.
13458 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010013459 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013460 ordering of the values in the request counts, as the first value among
13461 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020013462
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013463 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013464
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013465 # this configuration switches to the backend matching a
13466 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013467
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013468 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
13469 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
13470 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
13471 use_backend spanish if es
13472 use_backend french if fr
13473 use_backend english if en
13474 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020013475
Willy Tarreau60a2ee72017-12-15 07:13:48 +010013476length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010013477 Get the length of the string. This can only be placed after a string
13478 sample fetch function or after a transformation keyword returning a string
13479 type. The result is of type integer.
13480
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013481lower
13482 Convert a string sample to lower case. This can only be placed after a string
13483 sample fetch function or after a transformation keyword returning a string
13484 type. The result is of type string.
13485
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013486ltime(<format>[,<offset>])
13487 Converts an integer supposed to contain a date since epoch to a string
13488 representing this date in local time using a format defined by the <format>
13489 string using strftime(3). The purpose is to allow any date format to be used
13490 in logs. An optional <offset> in seconds may be applied to the input date
13491 (positive or negative). See the strftime() man page for the format supported
13492 by your operating system. See also the utime converter.
13493
13494 Example :
13495
13496 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010013497 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013498 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
13499
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013500map(<map_file>[,<default_value>])
13501map_<match_type>(<map_file>[,<default_value>])
13502map_<match_type>_<output_type>(<map_file>[,<default_value>])
13503 Search the input value from <map_file> using the <match_type> matching method,
13504 and return the associated value converted to the type <output_type>. If the
13505 input value cannot be found in the <map_file>, the converter returns the
13506 <default_value>. If the <default_value> is not set, the converter fails and
13507 acts as if no input value could be fetched. If the <match_type> is not set, it
13508 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
13509 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
13510 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013511
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013512 It is important to avoid overlapping between the keys : IP addresses and
13513 strings are stored in trees, so the first of the finest match will be used.
13514 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013515
Tim Düsterhus4896c442016-11-29 02:15:19 +010013516 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013517 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013518
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013519 input type | match method | output type str | output type int | output type ip
13520 -----------+--------------+-----------------+-----------------+---------------
13521 str | str | map_str | map_str_int | map_str_ip
13522 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020013523 str | beg | map_beg | map_beg_int | map_end_ip
13524 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013525 str | sub | map_sub | map_sub_int | map_sub_ip
13526 -----------+--------------+-----------------+-----------------+---------------
13527 str | dir | map_dir | map_dir_int | map_dir_ip
13528 -----------+--------------+-----------------+-----------------+---------------
13529 str | dom | map_dom | map_dom_int | map_dom_ip
13530 -----------+--------------+-----------------+-----------------+---------------
13531 str | end | map_end | map_end_int | map_end_ip
13532 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080013533 str | reg | map_reg | map_reg_int | map_reg_ip
13534 -----------+--------------+-----------------+-----------------+---------------
13535 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013536 -----------+--------------+-----------------+-----------------+---------------
13537 int | int | map_int | map_int_int | map_int_ip
13538 -----------+--------------+-----------------+-----------------+---------------
13539 ip | ip | map_ip | map_ip_int | map_ip_ip
13540 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013541
Thierry Fournier8feaa662016-02-10 22:55:20 +010013542 The special map called "map_regm" expect matching zone in the regular
13543 expression and modify the output replacing back reference (like "\1") by
13544 the corresponding match text.
13545
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013546 The file contains one key + value per line. Lines which start with '#' are
13547 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
13548 is then the first "word" (series of non-space/tabs characters), and the value
13549 is what follows this series of space/tab till the end of the line excluding
13550 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013551
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013552 Example :
13553
13554 # this is a comment and is ignored
13555 2.22.246.0/23 United Kingdom \n
13556 <-><-----------><--><------------><---->
13557 | | | | `- trailing spaces ignored
13558 | | | `---------- value
13559 | | `-------------------- middle spaces ignored
13560 | `---------------------------- key
13561 `------------------------------------ leading spaces ignored
13562
Willy Tarreau97707872015-01-27 15:12:13 +010013563mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013564 Divides the input value of type signed integer by <value>, and returns the
13565 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013566 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013567 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013568 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013569 "sess" : the variable is shared with the whole session
13570 "txn" : the variable is shared with the transaction (request and response)
13571 "req" : the variable is shared only during request processing
13572 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013573 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013574 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013575
13576mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013577 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020013578 the product as an signed integer. In case of overflow, the largest possible
13579 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013580 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013581 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013582 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013583 "sess" : the variable is shared with the whole session
13584 "txn" : the variable is shared with the transaction (request and response)
13585 "req" : the variable is shared only during request processing
13586 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013587 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013588 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013589
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010013590nbsrv
13591 Takes an input value of type string, interprets it as a backend name and
13592 returns the number of usable servers in that backend. Can be used in places
13593 where we want to look up a backend from a dynamic name, like a result of a
13594 map lookup.
13595
Willy Tarreau97707872015-01-27 15:12:13 +010013596neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013597 Takes the input value of type signed integer, computes the opposite value,
13598 and returns the remainder as an signed integer. 0 is identity. This operator
13599 is provided for reversed subtracts : in order to subtract the input from a
13600 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010013601
13602not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013603 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013604 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013605 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013606 absence of a flag).
13607
13608odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013609 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010013610 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
13611
13612or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013613 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013614 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013615 numeric value or a variable name. The name of the variable starts with an
13616 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013617 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013618 "sess" : the variable is shared with the whole session
13619 "txn" : the variable is shared with the transaction (request and response)
13620 "req" : the variable is shared only during request processing
13621 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013622 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013623 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013624
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010013625protobuf(<field_number>,[<field_type>])
13626 This extracts the protocol buffers message field in raw mode of an input binary
13627 sample representation of a protocol buffer message with <field_number> as field
13628 number (dotted notation) if <field_type> is not present, or as an integer sample
13629 if this field is present (see also "ungrpc" below).
13630 The list of the authorized types is the following one: "int32", "int64", "uint32",
13631 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
13632 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
13633 "float" for the wire type 5. Note that "string" is considered as a length-delimited
13634 type, so it does not require any <field_type> argument to be extracted.
13635 More information may be found here about the protocol buffers message field types:
13636 https://developers.google.com/protocol-buffers/docs/encoding
13637
Willy Tarreauc4dc3502015-01-23 20:39:28 +010013638regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010013639 Applies a regex-based substitution to the input string. It does the same
13640 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
13641 default it will replace in the input string the first occurrence of the
13642 largest part matching the regular expression <regex> with the substitution
13643 string <subst>. It is possible to replace all occurrences instead by adding
13644 the flag "g" in the third argument <flags>. It is also possible to make the
13645 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
13646 string, it is made up from the concatenation of all desired flags. Thus if
13647 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
13648 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010013649 configuration parser, some characters such as closing parenthesis, closing
13650 square brackets or comma are not possible to use in the arguments. The first
13651 use of this converter is to replace certain characters or sequence of
13652 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010013653
13654 Example :
13655
13656 # de-duplicate "/" in header "x-path".
13657 # input: x-path: /////a///b/c/xzxyz/
13658 # output: x-path: /a/b/c/xzxyz/
13659 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
13660
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013661capture-req(<id>)
13662 Capture the string entry in the request slot <id> and returns the entry as
13663 is. If the slot doesn't exist, the capture fails silently.
13664
13665 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013666 "http-response capture", "capture.req.hdr" and
13667 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013668
13669capture-res(<id>)
13670 Capture the string entry in the response slot <id> and returns the entry as
13671 is. If the slot doesn't exist, the capture fails silently.
13672
13673 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013674 "http-response capture", "capture.req.hdr" and
13675 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013676
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013677sdbm([<avalanche>])
13678 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
13679 hash function. Optionally, it is possible to apply a full avalanche hash
13680 function to the output if the optional <avalanche> argument equals 1. This
13681 converter uses the same functions as used by the various hash-based load
13682 balancing algorithms, so it will provide exactly the same results. It is
13683 mostly intended for debugging, but can be used as a stick-table entry to
13684 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013685 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
13686 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013687
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013688set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010013689 Sets a variable with the input content and returns the content on the output
13690 as-is. The variable keeps the value and the associated input type. The name of
13691 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013692 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013693 "sess" : the variable is shared with the whole session
13694 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013695 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013696 "req" : the variable is shared only during request processing,
13697 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013698 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013699 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013700
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020013701sha1
Tim Duesterhusd4376302019-06-17 12:41:44 +020013702 Converts a binary input sample to a SHA-1 digest. The result is a binary
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020013703 sample with length of 20 bytes.
13704
Tim Duesterhusd4376302019-06-17 12:41:44 +020013705sha2([<bits>])
13706 Converts a binary input sample to a digest in the SHA-2 family. The result
13707 is a binary sample with length of <bits>/8 bytes.
13708
13709 Valid values for <bits> are 224, 256, 384, 512, each corresponding to
13710 SHA-<bits>. The default value is 256.
13711
13712 Please note that this converter is only available when haproxy has been
13713 compiled with USE_OPENSSL.
13714
Nenad Merdanovic177adc92019-08-27 01:58:13 +020013715srv_queue
13716 Takes an input value of type string, either a server name or <backend>/<server>
13717 format and returns the number of queued sessions on that server. Can be used
13718 in places where we want to look up queued sessions from a dynamic name, like a
13719 cookie value (e.g. req.cook(SRVID),srv_queue) and then make a decision to break
13720 persistence or direct a request elsewhere.
13721
Tim Duesterhusca097c12018-04-27 21:18:45 +020013722strcmp(<var>)
13723 Compares the contents of <var> with the input value of type string. Returns
13724 the result as a signed integer compatible with strcmp(3): 0 if both strings
13725 are identical. A value less than 0 if the left string is lexicographically
13726 smaller than the right string or if the left string is shorter. A value greater
13727 than 0 otherwise (right string greater than left string or the right string is
13728 shorter).
13729
13730 Example :
13731
13732 http-request set-var(txn.host) hdr(host)
13733 # Check whether the client is attempting domain fronting.
13734 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
13735
13736
Willy Tarreau97707872015-01-27 15:12:13 +010013737sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013738 Subtracts <value> from the input value of type signed integer, and returns
13739 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013740 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010013741 or a variable name. The name of the variable starts with an indication about
13742 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013743 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013744 "sess" : the variable is shared with the whole session
13745 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013746 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010013747 "req" : the variable is shared only during request processing,
13748 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013749 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013750 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013751
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013752table_bytes_in_rate(<table>)
13753 Uses the string representation of the input sample to perform a look up in
13754 the specified table. If the key is not found in the table, integer value zero
13755 is returned. Otherwise the converter returns the average client-to-server
13756 bytes rate associated with the input sample in the designated table, measured
13757 in amount of bytes over the period configured in the table. See also the
13758 sc_bytes_in_rate sample fetch keyword.
13759
13760
13761table_bytes_out_rate(<table>)
13762 Uses the string representation of the input sample to perform a look up in
13763 the specified table. If the key is not found in the table, integer value zero
13764 is returned. Otherwise the converter returns the average server-to-client
13765 bytes rate associated with the input sample in the designated table, measured
13766 in amount of bytes over the period configured in the table. See also the
13767 sc_bytes_out_rate sample fetch keyword.
13768
13769table_conn_cnt(<table>)
13770 Uses the string representation of the input sample to perform a look up in
13771 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013772 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013773 connections associated with the input sample in the designated table. See
13774 also the sc_conn_cnt sample fetch keyword.
13775
13776table_conn_cur(<table>)
13777 Uses the string representation of the input sample to perform a look up in
13778 the specified table. If the key is not found in the table, integer value zero
13779 is returned. Otherwise the converter returns the current amount of concurrent
13780 tracked connections associated with the input sample in the designated table.
13781 See also the sc_conn_cur sample fetch keyword.
13782
13783table_conn_rate(<table>)
13784 Uses the string representation of the input sample to perform a look up in
13785 the specified table. If the key is not found in the table, integer value zero
13786 is returned. Otherwise the converter returns the average incoming connection
13787 rate associated with the input sample in the designated table. See also the
13788 sc_conn_rate sample fetch keyword.
13789
Thierry FOURNIER236657b2015-08-19 08:25:14 +020013790table_gpt0(<table>)
13791 Uses the string representation of the input sample to perform a look up in
13792 the specified table. If the key is not found in the table, boolean value zero
13793 is returned. Otherwise the converter returns the current value of the first
13794 general purpose tag associated with the input sample in the designated table.
13795 See also the sc_get_gpt0 sample fetch keyword.
13796
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013797table_gpc0(<table>)
13798 Uses the string representation of the input sample to perform a look up in
13799 the specified table. If the key is not found in the table, integer value zero
13800 is returned. Otherwise the converter returns the current value of the first
13801 general purpose counter associated with the input sample in the designated
13802 table. See also the sc_get_gpc0 sample fetch keyword.
13803
13804table_gpc0_rate(<table>)
13805 Uses the string representation of the input sample to perform a look up in
13806 the specified table. If the key is not found in the table, integer value zero
13807 is returned. Otherwise the converter returns the frequency which the gpc0
13808 counter was incremented over the configured period in the table, associated
13809 with the input sample in the designated table. See also the sc_get_gpc0_rate
13810 sample fetch keyword.
13811
Frédéric Lécaille6778b272018-01-29 15:22:53 +010013812table_gpc1(<table>)
13813 Uses the string representation of the input sample to perform a look up in
13814 the specified table. If the key is not found in the table, integer value zero
13815 is returned. Otherwise the converter returns the current value of the second
13816 general purpose counter associated with the input sample in the designated
13817 table. See also the sc_get_gpc1 sample fetch keyword.
13818
13819table_gpc1_rate(<table>)
13820 Uses the string representation of the input sample to perform a look up in
13821 the specified table. If the key is not found in the table, integer value zero
13822 is returned. Otherwise the converter returns the frequency which the gpc1
13823 counter was incremented over the configured period in the table, associated
13824 with the input sample in the designated table. See also the sc_get_gpc1_rate
13825 sample fetch keyword.
13826
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013827table_http_err_cnt(<table>)
13828 Uses the string representation of the input sample to perform a look up in
13829 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013830 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013831 errors associated with the input sample in the designated table. See also the
13832 sc_http_err_cnt sample fetch keyword.
13833
13834table_http_err_rate(<table>)
13835 Uses the string representation of the input sample to perform a look up in
13836 the specified table. If the key is not found in the table, integer value zero
13837 is returned. Otherwise the average rate of HTTP errors associated with the
13838 input sample in the designated table, measured in amount of errors over the
13839 period configured in the table. See also the sc_http_err_rate sample fetch
13840 keyword.
13841
13842table_http_req_cnt(<table>)
13843 Uses the string representation of the input sample to perform a look up in
13844 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013845 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013846 requests associated with the input sample in the designated table. See also
13847 the sc_http_req_cnt sample fetch keyword.
13848
13849table_http_req_rate(<table>)
13850 Uses the string representation of the input sample to perform a look up in
13851 the specified table. If the key is not found in the table, integer value zero
13852 is returned. Otherwise the average rate of HTTP requests associated with the
13853 input sample in the designated table, measured in amount of requests over the
13854 period configured in the table. See also the sc_http_req_rate sample fetch
13855 keyword.
13856
13857table_kbytes_in(<table>)
13858 Uses the string representation of the input sample to perform a look up in
13859 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013860 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013861 to-server data associated with the input sample in the designated table,
13862 measured in kilobytes. The test is currently performed on 32-bit integers,
13863 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
13864 keyword.
13865
13866table_kbytes_out(<table>)
13867 Uses the string representation of the input sample to perform a look up in
13868 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013869 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013870 to-client data associated with the input sample in the designated table,
13871 measured in kilobytes. The test is currently performed on 32-bit integers,
13872 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
13873 keyword.
13874
13875table_server_id(<table>)
13876 Uses the string representation of the input sample to perform a look up in
13877 the specified table. If the key is not found in the table, integer value zero
13878 is returned. Otherwise the converter returns the server ID associated with
13879 the input sample in the designated table. A server ID is associated to a
13880 sample by a "stick" rule when a connection to a server succeeds. A server ID
13881 zero means that no server is associated with this key.
13882
13883table_sess_cnt(<table>)
13884 Uses the string representation of the input sample to perform a look up in
13885 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010013886 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013887 sessions associated with the input sample in the designated table. Note that
13888 a session here refers to an incoming connection being accepted by the
13889 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
13890 keyword.
13891
13892table_sess_rate(<table>)
13893 Uses the string representation of the input sample to perform a look up in
13894 the specified table. If the key is not found in the table, integer value zero
13895 is returned. Otherwise the converter returns the average incoming session
13896 rate associated with the input sample in the designated table. Note that a
13897 session here refers to an incoming connection being accepted by the
13898 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
13899 keyword.
13900
13901table_trackers(<table>)
13902 Uses the string representation of the input sample to perform a look up in
13903 the specified table. If the key is not found in the table, integer value zero
13904 is returned. Otherwise the converter returns the current amount of concurrent
13905 connections tracking the same key as the input sample in the designated
13906 table. It differs from table_conn_cur in that it does not rely on any stored
13907 information but on the table's reference count (the "use" value which is
13908 returned by "show table" on the CLI). This may sometimes be more suited for
13909 layer7 tracking. It can be used to tell a server how many concurrent
13910 connections there are from a given address for example. See also the
13911 sc_trackers sample fetch keyword.
13912
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013913upper
13914 Convert a string sample to upper case. This can only be placed after a string
13915 sample fetch function or after a transformation keyword returning a string
13916 type. The result is of type string.
13917
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020013918url_dec
13919 Takes an url-encoded string provided as input and returns the decoded
13920 version as output. The input and the output are of type string.
13921
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013922ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010013923 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010013924 sample representation of a gRPC message with <field_number> as field number
13925 (dotted notation) if <field_type> is not present, or as an integer sample if this
13926 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013927 The list of the authorized types is the following one: "int32", "int64", "uint32",
13928 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
13929 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
13930 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010013931 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013932 More information may be found here about the protocol buffers message field types:
13933 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010013934
13935 Example:
13936 // with such a protocol buffer .proto file content adapted from
13937 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
13938
13939 message Point {
13940 int32 latitude = 1;
13941 int32 longitude = 2;
13942 }
13943
13944 message PPoint {
13945 Point point = 59;
13946 }
13947
13948 message Rectangle {
13949 // One corner of the rectangle.
13950 PPoint lo = 48;
13951 // The other corner of the rectangle.
13952 PPoint hi = 49;
13953 }
13954
13955 let's say a body request is made of a "Rectangle" object value (two PPoint
13956 protocol buffers messages), the four protocol buffers fields could be
13957 extracted with these "ungrpc" directives:
13958
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013959 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
13960 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
John Roeslerfb2fce12019-07-10 15:45:51 -050013961 req.body,ungrpc(49.59.1,int32) # "latitude" of "hi" second PPoint
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013962 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
13963
Frédéric Lécaille93d33162019-03-06 09:35:59 +010013964 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013965
Frédéric Lécaille93d33162019-03-06 09:35:59 +010013966 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010013967
John Roeslerfb2fce12019-07-10 15:45:51 -050013968 As a gRPC message is always made of a gRPC header followed by protocol buffers
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010013969 messages, in the previous example the "latitude" of "lo" first PPoint
13970 could be extracted with these equivalent directives:
13971
13972 req.body,ungrpc(48.59),protobuf(1,int32)
13973 req.body,ungrpc(48),protobuf(59.1,int32)
13974 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
13975
13976 Note that the first convert must be "ungrpc", the remaining ones must be
13977 "protobuf" and only the last one may have or not a second argument to
13978 interpret the previous binary sample.
13979
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010013980
Christopher Faulet85d79c92016-11-09 16:54:56 +010013981unset-var(<var name>)
13982 Unsets a variable if the input content is defined. The name of the variable
13983 starts with an indication about its scope. The scopes allowed are:
13984 "proc" : the variable is shared with the whole process
13985 "sess" : the variable is shared with the whole session
13986 "txn" : the variable is shared with the transaction (request and
13987 response),
13988 "req" : the variable is shared only during request processing,
13989 "res" : the variable is shared only during response processing.
13990 This prefix is followed by a name. The separator is a '.'. The name may only
13991 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
13992
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013993utime(<format>[,<offset>])
13994 Converts an integer supposed to contain a date since epoch to a string
13995 representing this date in UTC time using a format defined by the <format>
13996 string using strftime(3). The purpose is to allow any date format to be used
13997 in logs. An optional <offset> in seconds may be applied to the input date
13998 (positive or negative). See the strftime() man page for the format supported
13999 by your operating system. See also the ltime converter.
14000
14001 Example :
14002
14003 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014004 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014005 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
14006
Marcin Deranek9631a282018-04-16 14:30:46 +020014007word(<index>,<delimiters>[,<count>])
14008 Extracts the nth word counting from the beginning (positive index) or from
14009 the end (negative index) considering given delimiters from an input string.
14010 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
14011 Optionally you can specify <count> of words to extract (default: 1).
14012 Value of 0 indicates extraction of all remaining words.
14013
14014 Example :
14015 str(f1_f2_f3__f5),word(4,_) # f5
14016 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
14017 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
14018 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
14019 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010014020
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014021wt6([<avalanche>])
14022 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
14023 hash function. Optionally, it is possible to apply a full avalanche hash
14024 function to the output if the optional <avalanche> argument equals 1. This
14025 converter uses the same functions as used by the various hash-based load
14026 balancing algorithms, so it will provide exactly the same results. It is
14027 mostly intended for debugging, but can be used as a stick-table entry to
14028 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014029 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
14030 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014031
Willy Tarreau97707872015-01-27 15:12:13 +010014032xor(<value>)
14033 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014034 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014035 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014036 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014037 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014038 "sess" : the variable is shared with the whole session
14039 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014040 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014041 "req" : the variable is shared only during request processing,
14042 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014043 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014044 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014045
Thierry FOURNIER01e09742016-12-26 11:46:11 +010014046xxh32([<seed>])
14047 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
14048 variant of the XXHash hash function. This hash supports a seed which defaults
14049 to zero but a different value maybe passed as the <seed> argument. This hash
14050 is known to be very good and very fast so it can be used to hash URLs and/or
14051 URL parameters for use as stick-table keys to collect statistics with a low
14052 collision rate, though care must be taken as the algorithm is not considered
14053 as cryptographically secure.
14054
14055xxh64([<seed>])
14056 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
14057 variant of the XXHash hash function. This hash supports a seed which defaults
14058 to zero but a different value maybe passed as the <seed> argument. This hash
14059 is known to be very good and very fast so it can be used to hash URLs and/or
14060 URL parameters for use as stick-table keys to collect statistics with a low
14061 collision rate, though care must be taken as the algorithm is not considered
14062 as cryptographically secure.
14063
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014064
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200140657.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020014066--------------------------------------------
14067
14068A first set of sample fetch methods applies to internal information which does
14069not even relate to any client information. These ones are sometimes used with
14070"monitor-fail" directives to report an internal status to external watchers.
14071The sample fetch methods described in this section are usable anywhere.
14072
14073always_false : boolean
14074 Always returns the boolean "false" value. It may be used with ACLs as a
14075 temporary replacement for another one when adjusting configurations.
14076
14077always_true : boolean
14078 Always returns the boolean "true" value. It may be used with ACLs as a
14079 temporary replacement for another one when adjusting configurations.
14080
14081avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014082 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020014083 divided by the number of active servers. The current backend is used if no
14084 backend is specified. This is very similar to "queue" except that the size of
14085 the farm is considered, in order to give a more accurate measurement of the
14086 time it may take for a new connection to be processed. The main usage is with
14087 ACL to return a sorry page to new users when it becomes certain they will get
14088 a degraded service, or to pass to the backend servers in a header so that
14089 they decide to work in degraded mode or to disable some functions to speed up
14090 the processing a bit. Note that in the event there would not be any active
14091 server anymore, twice the number of queued connections would be considered as
14092 the measured value. This is a fair estimate, as we expect one server to get
14093 back soon anyway, but we still prefer to send new traffic to another backend
14094 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
14095 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010014096
Willy Tarreau74ca5042013-06-11 23:12:07 +020014097be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020014098 Applies to the number of currently established connections on the backend,
14099 possibly including the connection being evaluated. If no backend name is
14100 specified, the current one is used. But it is also possible to check another
14101 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014102 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
14103
14104be_conn_free([<backend>]) : integer
14105 Returns an integer value corresponding to the number of available connections
14106 across available servers in the backend. Queue slots are not included. Backup
14107 servers are also not included, unless all other servers are down. If no
14108 backend name is specified, the current one is used. But it is also possible
14109 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014110 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
14111 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014112
14113 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
14114 (meaning unlimited), then this fetch clearly does not make sense, in which
14115 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014116
Willy Tarreau74ca5042013-06-11 23:12:07 +020014117be_sess_rate([<backend>]) : integer
14118 Returns an integer value corresponding to the sessions creation rate on the
14119 backend, in number of new sessions per second. This is used with ACLs to
14120 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010014121 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014122 online dictionary). It can also be useful to add this element to logs using a
14123 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014124
14125 Example :
14126 # Redirect to an error page if the dictionary is requested too often
14127 backend dynamic
14128 mode http
14129 acl being_scanned be_sess_rate gt 100
14130 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010014131
Davor Ocelice9ed2812017-12-25 17:49:28 +010014132bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014133 Returns a binary chain. The input is the hexadecimal representation
14134 of the string.
14135
14136bool(<bool>) : bool
14137 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
14138 'false' and '0' are the same. 'true' and '1' are the same.
14139
Willy Tarreau74ca5042013-06-11 23:12:07 +020014140connslots([<backend>]) : integer
14141 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014142 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020014143 connections on all servers and the maximum queue size. This is probably only
14144 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050014145
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014146 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020014147 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014148 usage; see "use_backend" keyword) can be redirected to a different backend.
14149
Willy Tarreau55165fe2009-05-10 12:02:55 +020014150 'connslots' = number of available server connection slots, + number of
14151 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014152
Willy Tarreaua36af912009-10-10 12:02:45 +020014153 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020014154 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020014155 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020014156 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010014157 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020014158 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020014159 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014160
Willy Tarreau55165fe2009-05-10 12:02:55 +020014161 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
14162 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020014163 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020014164 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014165
Willy Tarreau70fe9442018-11-22 16:07:39 +010014166cpu_calls : integer
14167 Returns the number of calls to the task processing the stream or current
14168 request since it was allocated. This number is reset for each new request on
14169 the same connections in case of HTTP keep-alive. This value should usually be
14170 low and stable (around 2 calls for a typically simple request) but may become
14171 high if some processing (compression, caching or analysis) is performed. This
14172 is purely for performance monitoring purposes.
14173
14174cpu_ns_avg : integer
14175 Returns the average number of nanoseconds spent in each call to the task
14176 processing the stream or current request. This number is reset for each new
14177 request on the same connections in case of HTTP keep-alive. This value
14178 indicates the overall cost of processing the request or the connection for
14179 each call. There is no good nor bad value but the time spent in a call
14180 automatically causes latency for other processing (see lat_ns_avg below),
14181 and may affect other connection's apparent response time. Certain operations
14182 like compression, complex regex matching or heavy Lua operations may directly
14183 affect this value, and having it in the logs will make it easier to spot the
14184 faulty processing that needs to be fixed to recover decent performance.
14185 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
14186
14187cpu_ns_tot : integer
14188 Returns the total number of nanoseconds spent in each call to the task
14189 processing the stream or current request. This number is reset for each new
14190 request on the same connections in case of HTTP keep-alive. This value
14191 indicates the overall cost of processing the request or the connection for
14192 each call. There is no good nor bad value but the time spent in a call
14193 automatically causes latency for other processing (see lat_ns_avg below),
14194 induces CPU costs on the machine, and may affect other connection's apparent
14195 response time. Certain operations like compression, complex regex matching or
14196 heavy Lua operations may directly affect this value, and having it in the
14197 logs will make it easier to spot the faulty processing that needs to be fixed
14198 to recover decent performance. The value may be artificially high due to a
14199 high cpu_calls count, for example when processing many HTTP chunks, and for
14200 this reason it is often preferred to log cpu_ns_avg instead.
14201
Cyril Bonté6bcd1822019-11-05 23:13:59 +010014202date([<offset>],[<unit>]) : integer
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014203 Returns the current date as the epoch (number of seconds since 01/01/1970).
Damien Claisseae6f1252019-10-30 15:57:28 +000014204
14205 If an offset value is specified, then it is added to the current date before
14206 returning the value. This is particularly useful to compute relative dates,
14207 as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020014208 It is useful combined with the http_date converter.
14209
Damien Claisseae6f1252019-10-30 15:57:28 +000014210 <unit> is facultative, and can be set to "s" for seconds (default behavior),
14211 "ms" for milliseconds or "us" for microseconds.
14212 If unit is set, return value is an integer reflecting either seconds,
14213 milliseconds or microseconds since epoch, plus offset.
14214 It is useful when a time resolution of less than a second is needed.
14215
Willy Tarreau276fae92013-07-25 14:36:01 +020014216 Example :
14217
14218 # set an expires header to now+1 hour in every response
14219 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014220
Damien Claisseae6f1252019-10-30 15:57:28 +000014221 # set an expires header to now+1 hour in every response, with
14222 # millisecond granularity
14223 http-response set-header Expires %[date(3600000,ms),http_date(0,ms)]
14224
Etienne Carrierea792a0a2018-01-17 13:43:24 +010014225date_us : integer
14226 Return the microseconds part of the date (the "second" part is returned by
14227 date sample). This sample is coherent with the date sample as it is comes
14228 from the same timeval structure.
14229
Willy Tarreaud716f9b2017-10-13 11:03:15 +020014230distcc_body(<token>[,<occ>]) : binary
14231 Parses a distcc message and returns the body associated to occurrence #<occ>
14232 of the token <token>. Occurrences start at 1, and when unspecified, any may
14233 match though in practice only the first one is checked for now. This can be
14234 used to extract file names or arguments in files built using distcc through
14235 haproxy. Please refer to distcc's protocol documentation for the complete
14236 list of supported tokens.
14237
14238distcc_param(<token>[,<occ>]) : integer
14239 Parses a distcc message and returns the parameter associated to occurrence
14240 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
14241 any may match though in practice only the first one is checked for now. This
14242 can be used to extract certain information such as the protocol version, the
14243 file size or the argument in files built using distcc through haproxy.
14244 Another use case consists in waiting for the start of the preprocessed file
14245 contents before connecting to the server to avoid keeping idle connections.
14246 Please refer to distcc's protocol documentation for the complete list of
14247 supported tokens.
14248
14249 Example :
14250 # wait up to 20s for the pre-processed file to be uploaded
14251 tcp-request inspect-delay 20s
14252 tcp-request content accept if { distcc_param(DOTI) -m found }
14253 # send large files to the big farm
14254 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
14255
Willy Tarreau595ec542013-06-12 21:34:28 +020014256env(<name>) : string
14257 Returns a string containing the value of environment variable <name>. As a
14258 reminder, environment variables are per-process and are sampled when the
14259 process starts. This can be useful to pass some information to a next hop
14260 server, or with ACLs to take specific action when the process is started a
14261 certain way.
14262
14263 Examples :
14264 # Pass the Via header to next hop with the local hostname in it
14265 http-request add-header Via 1.1\ %[env(HOSTNAME)]
14266
14267 # reject cookie-less requests when the STOP environment variable is set
14268 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
14269
Willy Tarreau74ca5042013-06-11 23:12:07 +020014270fe_conn([<frontend>]) : integer
14271 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010014272 possibly including the connection being evaluated. If no frontend name is
14273 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014274 frontend. It can be used to return a sorry page before hard-blocking, or to
14275 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010014276 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020014277 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
14278 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020014279
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020014280fe_req_rate([<frontend>]) : integer
14281 Returns an integer value corresponding to the number of HTTP requests per
14282 second sent to a frontend. This number can differ from "fe_sess_rate" in
14283 situations where client-side keep-alive is enabled.
14284
Willy Tarreau74ca5042013-06-11 23:12:07 +020014285fe_sess_rate([<frontend>]) : integer
14286 Returns an integer value corresponding to the sessions creation rate on the
14287 frontend, in number of new sessions per second. This is used with ACLs to
14288 limit the incoming session rate to an acceptable range in order to prevent
14289 abuse of service at the earliest moment, for example when combined with other
14290 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
14291 down below the limit. It can also be useful to add this element to logs using
14292 a log-format directive. See also the "rate-limit sessions" directive for use
14293 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014294
14295 Example :
14296 # This frontend limits incoming mails to 10/s with a max of 100
14297 # concurrent connections. We accept any connection below 10/s, and
14298 # force excess clients to wait for 100 ms. Since clients are limited to
14299 # 100 max, there cannot be more than 10 incoming mails per second.
14300 frontend mail
14301 bind :25
14302 mode tcp
14303 maxconn 100
14304 acl too_fast fe_sess_rate ge 10
14305 tcp-request inspect-delay 100ms
14306 tcp-request content accept if ! too_fast
14307 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010014308
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010014309hostname : string
14310 Returns the system hostname.
14311
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014312int(<integer>) : signed integer
14313 Returns a signed integer.
14314
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014315ipv4(<ipv4>) : ipv4
14316 Returns an ipv4.
14317
14318ipv6(<ipv6>) : ipv6
14319 Returns an ipv6.
14320
Willy Tarreau70fe9442018-11-22 16:07:39 +010014321lat_ns_avg : integer
14322 Returns the average number of nanoseconds spent between the moment the task
14323 handling the stream is woken up and the moment it is effectively called. This
14324 number is reset for each new request on the same connections in case of HTTP
14325 keep-alive. This value indicates the overall latency inflicted to the current
14326 request by all other requests being processed in parallel, and is a direct
14327 indicator of perceived performance due to noisy neighbours. In order to keep
14328 the value low, it is possible to reduce the scheduler's run queue depth using
14329 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14330 once using "tune.maxpollevents", to decrease the stream's nice value using
14331 the "nice" option on the "bind" lines or in the frontend, or to look for
14332 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14333 whose processing needs to be adjusted or fixed. Compression of large buffers
14334 could be a culprit, like heavy regex or long lists of regex.
14335 Note: this value is exactly lat_ns_tot divided by cpu_calls.
14336
14337lat_ns_tot : integer
14338 Returns the total number of nanoseconds spent between the moment the task
14339 handling the stream is woken up and the moment it is effectively called. This
14340 number is reset for each new request on the same connections in case of HTTP
14341 keep-alive. This value indicates the overall latency inflicted to the current
14342 request by all other requests being processed in parallel, and is a direct
14343 indicator of perceived performance due to noisy neighbours. In order to keep
14344 the value low, it is possible to reduce the scheduler's run queue depth using
14345 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14346 once using "tune.maxpollevents", to decrease the stream's nice value using
14347 the "nice" option on the "bind" lines or in the frontend, or to look for
14348 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14349 whose processing needs to be adjusted or fixed. Compression of large buffers
14350 could be a culprit, like heavy regex or long lists of regex. Note: while it
14351 may intuitively seem that the total latency adds to a transfer time, it is
14352 almost never true because while a task waits for the CPU, network buffers
14353 continue to fill up and the next call will process more at once. The value
14354 may be artificially high due to a high cpu_calls count, for example when
14355 processing many HTTP chunks, and for this reason it is often preferred to log
14356 lat_ns_avg instead, which is a more relevant performance indicator.
14357
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014358meth(<method>) : method
14359 Returns a method.
14360
Willy Tarreau0f30d262014-11-24 16:02:05 +010014361nbproc : integer
14362 Returns an integer value corresponding to the number of processes that were
14363 started (it equals the global "nbproc" setting). This is useful for logging
14364 and debugging purposes.
14365
Willy Tarreau74ca5042013-06-11 23:12:07 +020014366nbsrv([<backend>]) : integer
14367 Returns an integer value corresponding to the number of usable servers of
14368 either the current backend or the named backend. This is mostly used with
14369 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010014370 switch to an alternate backend when the number of servers is too low to
14371 to handle some load. It is useful to report a failure when combined with
14372 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014373
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040014374prio_class : integer
14375 Returns the priority class of the current session for http mode or connection
14376 for tcp mode. The value will be that set by the last call to "http-request
14377 set-priority-class" or "tcp-request content set-priority-class".
14378
14379prio_offset : integer
14380 Returns the priority offset of the current session for http mode or
14381 connection for tcp mode. The value will be that set by the last call to
14382 "http-request set-priority-offset" or "tcp-request content
14383 set-priority-offset".
14384
Willy Tarreau0f30d262014-11-24 16:02:05 +010014385proc : integer
14386 Returns an integer value corresponding to the position of the process calling
14387 the function, between 1 and global.nbproc. This is useful for logging and
14388 debugging purposes.
14389
Willy Tarreau74ca5042013-06-11 23:12:07 +020014390queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014391 Returns the total number of queued connections of the designated backend,
14392 including all the connections in server queues. If no backend name is
14393 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014394 one. This is useful with ACLs or to pass statistics to backend servers. This
14395 can be used to take actions when queuing goes above a known level, generally
14396 indicating a surge of traffic or a massive slowdown on the servers. One
14397 possible action could be to reject new users but still accept old ones. See
14398 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
14399
Willy Tarreau84310e22014-02-14 11:59:04 +010014400rand([<range>]) : integer
14401 Returns a random integer value within a range of <range> possible values,
14402 starting at zero. If the range is not specified, it defaults to 2^32, which
14403 gives numbers between 0 and 4294967295. It can be useful to pass some values
14404 needed to take some routing decisions for example, or just for debugging
14405 purposes. This random must not be used for security purposes.
14406
Luca Schimweg8a694b82019-09-10 15:42:52 +020014407uuid([<version>]) : string
14408 Returns a UUID following the RFC4122 standard. If the version is not
14409 specified, a UUID version 4 (fully random) is returned.
14410 Currently, only version 4 is supported.
14411
Willy Tarreau74ca5042013-06-11 23:12:07 +020014412srv_conn([<backend>/]<server>) : integer
14413 Returns an integer value corresponding to the number of currently established
14414 connections on the designated server, possibly including the connection being
14415 evaluated. If <backend> is omitted, then the server is looked up in the
14416 current backend. It can be used to use a specific farm when one server is
14417 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014418 connections with it. See also the "fe_conn", "be_conn", "queue", and
14419 "srv_conn_free" fetch methods.
14420
14421srv_conn_free([<backend>/]<server>) : integer
14422 Returns an integer value corresponding to the number of available connections
14423 on the designated server, possibly including the connection being evaluated.
14424 The value does not include queue slots. If <backend> is omitted, then the
14425 server is looked up in the current backend. It can be used to use a specific
14426 farm when one server is full, or to inform the server about our view of the
14427 number of active connections with it. See also the "be_conn_free" and
14428 "srv_conn" fetch methods.
14429
14430 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
14431 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014432
14433srv_is_up([<backend>/]<server>) : boolean
14434 Returns true when the designated server is UP, and false when it is either
14435 DOWN or in maintenance mode. If <backend> is omitted, then the server is
14436 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010014437 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014438 availability). Another possible use which is more of a hack consists in
14439 using dummy servers as boolean variables that can be enabled or disabled from
14440 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
14441
Willy Tarreauff2b7af2017-10-13 11:46:26 +020014442srv_queue([<backend>/]<server>) : integer
14443 Returns an integer value corresponding to the number of connections currently
14444 pending in the designated server's queue. If <backend> is omitted, then the
14445 server is looked up in the current backend. It can sometimes be used together
14446 with the "use-server" directive to force to use a known faster server when it
14447 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
14448 fetch methods.
14449
Willy Tarreau74ca5042013-06-11 23:12:07 +020014450srv_sess_rate([<backend>/]<server>) : integer
14451 Returns an integer corresponding to the sessions creation rate on the
14452 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014453 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020014454 used with ACLs but can make sense with logs too. This is used to switch to an
14455 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010014456 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020014457 overloading servers).
14458
14459 Example :
14460 # Redirect to a separate back
14461 acl srv1_full srv_sess_rate(be1/srv1) gt 50
14462 acl srv2_full srv_sess_rate(be1/srv2) gt 50
14463 use_backend be2 if srv1_full or srv2_full
14464
Willy Tarreau0f30d262014-11-24 16:02:05 +010014465stopping : boolean
14466 Returns TRUE if the process calling the function is currently stopping. This
14467 can be useful for logging, or for relaxing certain checks or helping close
14468 certain connections upon graceful shutdown.
14469
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014470str(<string>) : string
14471 Returns a string.
14472
Willy Tarreau74ca5042013-06-11 23:12:07 +020014473table_avl([<table>]) : integer
14474 Returns the total number of available entries in the current proxy's
14475 stick-table or in the designated stick-table. See also table_cnt.
14476
14477table_cnt([<table>]) : integer
14478 Returns the total number of entries currently in use in the current proxy's
14479 stick-table or in the designated stick-table. See also src_conn_cnt and
14480 table_avl for other entry counting methods.
14481
Christopher Faulet34adb2a2017-11-21 21:45:38 +010014482thread : integer
14483 Returns an integer value corresponding to the position of the thread calling
14484 the function, between 0 and (global.nbthread-1). This is useful for logging
14485 and debugging purposes.
14486
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014487var(<var-name>) : undefined
14488 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010014489 sample fetch fails. The name of the variable starts with an indication
14490 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014491 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014492 "sess" : the variable is shared with the whole session
14493 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014494 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014495 "req" : the variable is shared only during request processing,
14496 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014497 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014498 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014499
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200145007.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020014501----------------------------------
14502
14503The layer 4 usually describes just the transport layer which in haproxy is
14504closest to the connection, where no content is yet made available. The fetch
14505methods described here are usable as low as the "tcp-request connection" rule
14506sets unless they require some future information. Those generally include
14507TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014508the incoming connection. For retrieving a value from a sticky counters, the
14509counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020014510"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
14511used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
14512can be specified as the first integer argument when using the "sc_" prefix.
14513Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
14514table may be specified with the "sc*" form, in which case the currently
14515tracked key will be looked up into this alternate table instead of the table
14516currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014517
Jérôme Magnin35e53a62019-01-16 14:38:37 +010014518bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010014519 Returns the backend connection's HTTP major version encoding, which may be 1
14520 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14521 encoding and not the version present in the request header.
14522
Willy Tarreau74ca5042013-06-11 23:12:07 +020014523be_id : integer
14524 Returns an integer containing the current backend's id. It can be used in
14525 frontends with responses to check which backend processed the request.
14526
Marcin Deranekd2471c22016-12-12 14:08:05 +010014527be_name : string
14528 Returns a string containing the current backend's name. It can be used in
14529 frontends with responses to check which backend processed the request.
14530
Willy Tarreau74ca5042013-06-11 23:12:07 +020014531dst : ip
14532 This is the destination IPv4 address of the connection on the client side,
14533 which is the address the client connected to. It can be useful when running
14534 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
14535 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010014536 RFC 4291. When the incoming connection passed through address translation or
14537 redirection involving connection tracking, the original destination address
14538 before the redirection will be reported. On Linux systems, the source and
14539 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
14540 is set, because a late response may reopen a timed out connection and switch
14541 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014542
14543dst_conn : integer
14544 Returns an integer value corresponding to the number of currently established
14545 connections on the same socket including the one being evaluated. It is
14546 normally used with ACLs but can as well be used to pass the information to
14547 servers in an HTTP header or in logs. It can be used to either return a sorry
14548 page before hard-blocking, or to use a specific backend to drain new requests
14549 when the socket is considered saturated. This offers the ability to assign
14550 different limits to different listening ports or addresses. See also the
14551 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014552
Willy Tarreau16e01562016-08-09 16:46:18 +020014553dst_is_local : boolean
14554 Returns true if the destination address of the incoming connection is local
14555 to the system, or false if the address doesn't exist on the system, meaning
14556 that it was intercepted in transparent mode. It can be useful to apply
14557 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010014558 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020014559 be delivered only on this address, or SSH access could be locally redirected.
14560 Please note that the check involves a few system calls, so it's better to do
14561 it only once per connection.
14562
Willy Tarreau74ca5042013-06-11 23:12:07 +020014563dst_port : integer
14564 Returns an integer value corresponding to the destination TCP port of the
14565 connection on the client side, which is the port the client connected to.
14566 This might be used when running in transparent mode, when assigning dynamic
14567 ports to some clients for a whole application session, to stick all users to
14568 a same server, or to pass the destination port information to a server using
14569 an HTTP header.
14570
Willy Tarreau60ca10a2017-08-18 15:26:54 +020014571fc_http_major : integer
14572 Reports the front connection's HTTP major version encoding, which may be 1
14573 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14574 encoding and not on the version present in the request header.
14575
Geoff Simmons7185b782019-08-27 18:31:16 +020014576fc_pp_authority : string
14577 Returns the authority TLV sent by the client in the PROXY protocol header,
14578 if any.
14579
Emeric Brun4f603012017-01-05 15:11:44 +010014580fc_rcvd_proxy : boolean
14581 Returns true if the client initiated the connection with a PROXY protocol
14582 header.
14583
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020014584fc_rtt(<unit>) : integer
14585 Returns the Round Trip Time (RTT) measured by the kernel for the client
14586 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
14587 can be set to "ms" for milliseconds or "us" for microseconds. If the server
14588 connection is not established, if the connection is not TCP or if the
14589 operating system does not support TCP_INFO, for example Linux kernels before
14590 2.4, the sample fetch fails.
14591
14592fc_rttvar(<unit>) : integer
14593 Returns the Round Trip Time (RTT) variance measured by the kernel for the
14594 client connection. <unit> is facultative, by default the unit is milliseconds.
14595 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
14596 server connection is not established, if the connection is not TCP or if the
14597 operating system does not support TCP_INFO, for example Linux kernels before
14598 2.4, the sample fetch fails.
14599
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014600fc_unacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014601 Returns the unacked counter measured by the kernel for the client connection.
14602 If the server connection is not established, if the connection is not TCP or
14603 if the operating system does not support TCP_INFO, for example Linux kernels
14604 before 2.4, the sample fetch fails.
14605
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014606fc_sacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014607 Returns the sacked counter measured by the kernel for the client connection.
14608 If the server connection is not established, if the connection is not TCP or
14609 if the operating system does not support TCP_INFO, for example Linux kernels
14610 before 2.4, the sample fetch fails.
14611
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014612fc_retrans : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014613 Returns the retransmits counter measured by the kernel for the client
14614 connection. If the server connection is not established, if the connection is
14615 not TCP or if the operating system does not support TCP_INFO, for example
14616 Linux kernels before 2.4, the sample fetch fails.
14617
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014618fc_fackets : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014619 Returns the fack counter measured by the kernel for the client
14620 connection. If the server connection is not established, if the connection is
14621 not TCP or if the operating system does not support TCP_INFO, for example
14622 Linux kernels before 2.4, the sample fetch fails.
14623
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014624fc_lost : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014625 Returns the lost counter measured by the kernel for the client
14626 connection. If the server connection is not established, if the connection is
14627 not TCP or if the operating system does not support TCP_INFO, for example
14628 Linux kernels before 2.4, the sample fetch fails.
14629
Christopher Fauletba0c53e2019-10-17 14:40:48 +020014630fc_reordering : integer
Joe Williams30fcd392016-08-10 07:06:44 -070014631 Returns the reordering counter measured by the kernel for the client
14632 connection. If the server connection is not established, if the connection is
14633 not TCP or if the operating system does not support TCP_INFO, for example
14634 Linux kernels before 2.4, the sample fetch fails.
14635
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020014636fe_defbe : string
14637 Returns a string containing the frontend's default backend name. It can be
14638 used in frontends to check which backend will handle requests by default.
14639
Willy Tarreau74ca5042013-06-11 23:12:07 +020014640fe_id : integer
14641 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010014642 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020014643 coming via a same frontend to the same server.
14644
Marcin Deranekd2471c22016-12-12 14:08:05 +010014645fe_name : string
14646 Returns a string containing the current frontend's name. It can be used in
14647 backends to check from which frontend it was called, or to stick all users
14648 coming via a same frontend to the same server.
14649
Cyril Bonté62ba8702014-04-22 23:52:25 +020014650sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014651sc0_bytes_in_rate([<table>]) : integer
14652sc1_bytes_in_rate([<table>]) : integer
14653sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014654 Returns the average client-to-server bytes rate from the currently tracked
14655 counters, measured in amount of bytes over the period configured in the
14656 table. See also src_bytes_in_rate.
14657
Cyril Bonté62ba8702014-04-22 23:52:25 +020014658sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014659sc0_bytes_out_rate([<table>]) : integer
14660sc1_bytes_out_rate([<table>]) : integer
14661sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014662 Returns the average server-to-client bytes rate from the currently tracked
14663 counters, measured in amount of bytes over the period configured in the
14664 table. See also src_bytes_out_rate.
14665
Cyril Bonté62ba8702014-04-22 23:52:25 +020014666sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014667sc0_clr_gpc0([<table>]) : integer
14668sc1_clr_gpc0([<table>]) : integer
14669sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020014670 Clears the first General Purpose Counter associated to the currently tracked
14671 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010014672 stored value is zero, so first invocation will always return zero. This is
14673 typically used as a second ACL in an expression in order to mark a connection
14674 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020014675
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014676 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020014677 # block if 5 consecutive requests continue to come faster than 10 sess
14678 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014679 acl abuse sc0_http_req_rate gt 10
14680 acl kill sc0_inc_gpc0 gt 5
14681 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020014682 tcp-request connection accept if !abuse save
14683 tcp-request connection reject if abuse kill
14684
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014685sc_clr_gpc1(<ctr>[,<table>]) : integer
14686sc0_clr_gpc1([<table>]) : integer
14687sc1_clr_gpc1([<table>]) : integer
14688sc2_clr_gpc1([<table>]) : integer
14689 Clears the second General Purpose Counter associated to the currently tracked
14690 counters, and returns its previous value. Before the first invocation, the
14691 stored value is zero, so first invocation will always return zero. This is
14692 typically used as a second ACL in an expression in order to mark a connection
14693 when a first ACL was verified.
14694
Cyril Bonté62ba8702014-04-22 23:52:25 +020014695sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014696sc0_conn_cnt([<table>]) : integer
14697sc1_conn_cnt([<table>]) : integer
14698sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014699 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014700 counters. See also src_conn_cnt.
14701
Cyril Bonté62ba8702014-04-22 23:52:25 +020014702sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014703sc0_conn_cur([<table>]) : integer
14704sc1_conn_cur([<table>]) : integer
14705sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014706 Returns the current amount of concurrent connections tracking the same
14707 tracked counters. This number is automatically incremented when tracking
14708 begins and decremented when tracking stops. See also src_conn_cur.
14709
Cyril Bonté62ba8702014-04-22 23:52:25 +020014710sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014711sc0_conn_rate([<table>]) : integer
14712sc1_conn_rate([<table>]) : integer
14713sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014714 Returns the average connection rate from the currently tracked counters,
14715 measured in amount of connections over the period configured in the table.
14716 See also src_conn_rate.
14717
Cyril Bonté62ba8702014-04-22 23:52:25 +020014718sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014719sc0_get_gpc0([<table>]) : integer
14720sc1_get_gpc0([<table>]) : integer
14721sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014722 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014723 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014724
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014725sc_get_gpc1(<ctr>[,<table>]) : integer
14726sc0_get_gpc1([<table>]) : integer
14727sc1_get_gpc1([<table>]) : integer
14728sc2_get_gpc1([<table>]) : integer
14729 Returns the value of the second General Purpose Counter associated to the
14730 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
14731
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014732sc_get_gpt0(<ctr>[,<table>]) : integer
14733sc0_get_gpt0([<table>]) : integer
14734sc1_get_gpt0([<table>]) : integer
14735sc2_get_gpt0([<table>]) : integer
14736 Returns the value of the first General Purpose Tag associated to the
14737 currently tracked counters. See also src_get_gpt0.
14738
Cyril Bonté62ba8702014-04-22 23:52:25 +020014739sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014740sc0_gpc0_rate([<table>]) : integer
14741sc1_gpc0_rate([<table>]) : integer
14742sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014743 Returns the average increment rate of the first General Purpose Counter
14744 associated to the currently tracked counters. It reports the frequency
14745 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014746 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
14747 that the "gpc0_rate" counter must be stored in the stick-table for a value to
14748 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020014749
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014750sc_gpc1_rate(<ctr>[,<table>]) : integer
14751sc0_gpc1_rate([<table>]) : integer
14752sc1_gpc1_rate([<table>]) : integer
14753sc2_gpc1_rate([<table>]) : integer
14754 Returns the average increment rate of the second General Purpose Counter
14755 associated to the currently tracked counters. It reports the frequency
14756 which the gpc1 counter was incremented over the configured period. See also
14757 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
14758 that the "gpc1_rate" counter must be stored in the stick-table for a value to
14759 be returned, as "gpc1" only holds the event count.
14760
Cyril Bonté62ba8702014-04-22 23:52:25 +020014761sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014762sc0_http_err_cnt([<table>]) : integer
14763sc1_http_err_cnt([<table>]) : integer
14764sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014765 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014766 counters. This includes the both request errors and 4xx error responses.
14767 See also src_http_err_cnt.
14768
Cyril Bonté62ba8702014-04-22 23:52:25 +020014769sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014770sc0_http_err_rate([<table>]) : integer
14771sc1_http_err_rate([<table>]) : integer
14772sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014773 Returns the average rate of HTTP errors from the currently tracked counters,
14774 measured in amount of errors over the period configured in the table. This
14775 includes the both request errors and 4xx error responses. See also
14776 src_http_err_rate.
14777
Cyril Bonté62ba8702014-04-22 23:52:25 +020014778sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014779sc0_http_req_cnt([<table>]) : integer
14780sc1_http_req_cnt([<table>]) : integer
14781sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014782 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014783 counters. This includes every started request, valid or not. See also
14784 src_http_req_cnt.
14785
Cyril Bonté62ba8702014-04-22 23:52:25 +020014786sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014787sc0_http_req_rate([<table>]) : integer
14788sc1_http_req_rate([<table>]) : integer
14789sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014790 Returns the average rate of HTTP requests from the currently tracked
14791 counters, measured in amount of requests over the period configured in
14792 the table. This includes every started request, valid or not. See also
14793 src_http_req_rate.
14794
Cyril Bonté62ba8702014-04-22 23:52:25 +020014795sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014796sc0_inc_gpc0([<table>]) : integer
14797sc1_inc_gpc0([<table>]) : integer
14798sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014799 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010014800 tracked counters, and returns its new value. Before the first invocation,
14801 the stored value is zero, so first invocation will increase it to 1 and will
14802 return 1. This is typically used as a second ACL in an expression in order
14803 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020014804
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014805 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014806 acl abuse sc0_http_req_rate gt 10
14807 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020014808 tcp-request connection reject if abuse kill
14809
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014810sc_inc_gpc1(<ctr>[,<table>]) : integer
14811sc0_inc_gpc1([<table>]) : integer
14812sc1_inc_gpc1([<table>]) : integer
14813sc2_inc_gpc1([<table>]) : integer
14814 Increments the second General Purpose Counter associated to the currently
14815 tracked counters, and returns its new value. Before the first invocation,
14816 the stored value is zero, so first invocation will increase it to 1 and will
14817 return 1. This is typically used as a second ACL in an expression in order
14818 to mark a connection when a first ACL was verified.
14819
Cyril Bonté62ba8702014-04-22 23:52:25 +020014820sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014821sc0_kbytes_in([<table>]) : integer
14822sc1_kbytes_in([<table>]) : integer
14823sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014824 Returns the total amount of client-to-server data from the currently tracked
14825 counters, measured in kilobytes. The test is currently performed on 32-bit
14826 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020014827
Cyril Bonté62ba8702014-04-22 23:52:25 +020014828sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014829sc0_kbytes_out([<table>]) : integer
14830sc1_kbytes_out([<table>]) : integer
14831sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020014832 Returns the total amount of server-to-client data from the currently tracked
14833 counters, measured in kilobytes. The test is currently performed on 32-bit
14834 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020014835
Cyril Bonté62ba8702014-04-22 23:52:25 +020014836sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014837sc0_sess_cnt([<table>]) : integer
14838sc1_sess_cnt([<table>]) : integer
14839sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014840 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020014841 into sessions, which means that they were accepted by a "tcp-request
14842 connection" rule, from the currently tracked counters. A backend may count
14843 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014844 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020014845 with the client. See also src_sess_cnt.
14846
Cyril Bonté62ba8702014-04-22 23:52:25 +020014847sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014848sc0_sess_rate([<table>]) : integer
14849sc1_sess_rate([<table>]) : integer
14850sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014851 Returns the average session rate from the currently tracked counters,
14852 measured in amount of sessions over the period configured in the table. A
14853 session is a connection that got past the early "tcp-request connection"
14854 rules. A backend may count more sessions than connections because each
14855 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040014856 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020014857
Cyril Bonté62ba8702014-04-22 23:52:25 +020014858sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020014859sc0_tracked([<table>]) : boolean
14860sc1_tracked([<table>]) : boolean
14861sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020014862 Returns true if the designated session counter is currently being tracked by
14863 the current session. This can be useful when deciding whether or not we want
14864 to set some values in a header passed to the server.
14865
Cyril Bonté62ba8702014-04-22 23:52:25 +020014866sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014867sc0_trackers([<table>]) : integer
14868sc1_trackers([<table>]) : integer
14869sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010014870 Returns the current amount of concurrent connections tracking the same
14871 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014872 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010014873 that it does not rely on any stored information but on the table's reference
14874 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020014875 may sometimes be more suited for layer7 tracking. It can be used to tell a
14876 server how many concurrent connections there are from a given address for
14877 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010014878
Willy Tarreau74ca5042013-06-11 23:12:07 +020014879so_id : integer
14880 Returns an integer containing the current listening socket's id. It is useful
14881 in frontends involving many "bind" lines, or to stick all users coming via a
14882 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014883
Willy Tarreau74ca5042013-06-11 23:12:07 +020014884src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010014885 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020014886 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
14887 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
14888 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010014889 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
14890 directive is used, it can be the address of a client behind another
14891 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010014892 "tcp-request connection" which sees the real address. When the incoming
14893 connection passed through address translation or redirection involving
14894 connection tracking, the original destination address before the redirection
14895 will be reported. On Linux systems, the source and destination may seldom
14896 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
14897 response may reopen a timed out connection and switch what is believed to be
14898 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014899
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014900 Example:
14901 # add an HTTP header in requests with the originating address' country
14902 http-request set-header X-Country %[src,map_ip(geoip.lst)]
14903
Willy Tarreau74ca5042013-06-11 23:12:07 +020014904src_bytes_in_rate([<table>]) : integer
14905 Returns the average bytes rate from the incoming connection's source address
14906 in the current proxy's stick-table or in the designated stick-table, measured
14907 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014908 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014909
Willy Tarreau74ca5042013-06-11 23:12:07 +020014910src_bytes_out_rate([<table>]) : integer
14911 Returns the average bytes rate to the incoming connection's source address in
14912 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014913 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014914 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014915
Willy Tarreau74ca5042013-06-11 23:12:07 +020014916src_clr_gpc0([<table>]) : integer
14917 Clears the first General Purpose Counter associated to the incoming
14918 connection's source address in the current proxy's stick-table or in the
14919 designated stick-table, and returns its previous value. If the address is not
14920 found, an entry is created and 0 is returned. This is typically used as a
14921 second ACL in an expression in order to mark a connection when a first ACL
14922 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020014923
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014924 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020014925 # block if 5 consecutive requests continue to come faster than 10 sess
14926 # per second, and reset the counter as soon as the traffic slows down.
14927 acl abuse src_http_req_rate gt 10
14928 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010014929 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020014930 tcp-request connection accept if !abuse save
14931 tcp-request connection reject if abuse kill
14932
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014933src_clr_gpc1([<table>]) : integer
14934 Clears the second General Purpose Counter associated to the incoming
14935 connection's source address in the current proxy's stick-table or in the
14936 designated stick-table, and returns its previous value. If the address is not
14937 found, an entry is created and 0 is returned. This is typically used as a
14938 second ACL in an expression in order to mark a connection when a first ACL
14939 was verified.
14940
Willy Tarreau74ca5042013-06-11 23:12:07 +020014941src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014942 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020014943 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014944 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014945 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014946
Willy Tarreau74ca5042013-06-11 23:12:07 +020014947src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020014948 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020014949 current incoming connection's source address in the current proxy's
14950 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014951 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014952
Willy Tarreau74ca5042013-06-11 23:12:07 +020014953src_conn_rate([<table>]) : integer
14954 Returns the average connection rate from the incoming connection's source
14955 address in the current proxy's stick-table or in the designated stick-table,
14956 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014957 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014958
Willy Tarreau74ca5042013-06-11 23:12:07 +020014959src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020014960 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020014961 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020014962 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014963 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020014964
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014965src_get_gpc1([<table>]) : integer
14966 Returns the value of the second General Purpose Counter associated to the
14967 incoming connection's source address in the current proxy's stick-table or in
14968 the designated stick-table. If the address is not found, zero is returned.
14969 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
14970
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014971src_get_gpt0([<table>]) : integer
14972 Returns the value of the first General Purpose Tag associated to the
14973 incoming connection's source address in the current proxy's stick-table or in
14974 the designated stick-table. If the address is not found, zero is returned.
14975 See also sc/sc0/sc1/sc2_get_gpt0.
14976
Willy Tarreau74ca5042013-06-11 23:12:07 +020014977src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014978 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020014979 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014980 stick-table or in the designated stick-table. It reports the frequency
14981 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014982 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
14983 that the "gpc0_rate" counter must be stored in the stick-table for a value to
14984 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014985
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014986src_gpc1_rate([<table>]) : integer
14987 Returns the average increment rate of the second General Purpose Counter
14988 associated to the incoming connection's source address in the current proxy's
14989 stick-table or in the designated stick-table. It reports the frequency
14990 which the gpc1 counter was incremented over the configured period. See also
14991 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
14992 that the "gpc1_rate" counter must be stored in the stick-table for a value to
14993 be returned, as "gpc1" only holds the event count.
14994
Willy Tarreau74ca5042013-06-11 23:12:07 +020014995src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014996 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014997 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020014998 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014999 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020015000 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015001
Willy Tarreau74ca5042013-06-11 23:12:07 +020015002src_http_err_rate([<table>]) : integer
15003 Returns the average rate of HTTP errors from the incoming connection's source
15004 address in the current proxy's stick-table or in the designated stick-table,
15005 measured in amount of errors over the period configured in the table. This
15006 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015007 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015008
Willy Tarreau74ca5042013-06-11 23:12:07 +020015009src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015010 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015011 source address in the current proxy's stick-table or in the designated stick-
15012 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015013 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015014
Willy Tarreau74ca5042013-06-11 23:12:07 +020015015src_http_req_rate([<table>]) : integer
15016 Returns the average rate of HTTP requests from the incoming connection's
15017 source address in the current proxy's stick-table or in the designated stick-
15018 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020015019 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015020 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015021
Willy Tarreau74ca5042013-06-11 23:12:07 +020015022src_inc_gpc0([<table>]) : integer
15023 Increments the first General Purpose Counter associated to the incoming
15024 connection's source address in the current proxy's stick-table or in the
15025 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015026 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015027 This is typically used as a second ACL in an expression in order to mark a
15028 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020015029
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015030 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020015031 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010015032 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015033 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020015034
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015035src_inc_gpc1([<table>]) : integer
15036 Increments the second General Purpose Counter associated to the incoming
15037 connection's source address in the current proxy's stick-table or in the
15038 designated stick-table, and returns its new value. If the address is not
15039 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
15040 This is typically used as a second ACL in an expression in order to mark a
15041 connection when a first ACL was verified.
15042
Willy Tarreau16e01562016-08-09 16:46:18 +020015043src_is_local : boolean
15044 Returns true if the source address of the incoming connection is local to the
15045 system, or false if the address doesn't exist on the system, meaning that it
15046 comes from a remote machine. Note that UNIX addresses are considered local.
15047 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015048 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020015049 note that the check involves a few system calls, so it's better to do it only
15050 once per connection.
15051
Willy Tarreau74ca5042013-06-11 23:12:07 +020015052src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015053 Returns the total amount of data received from the incoming connection's
15054 source address in the current proxy's stick-table or in the designated
15055 stick-table, measured in kilobytes. If the address is not found, zero is
15056 returned. The test is currently performed on 32-bit integers, which limits
15057 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015058
Willy Tarreau74ca5042013-06-11 23:12:07 +020015059src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015060 Returns the total amount of data sent to the incoming connection's source
15061 address in the current proxy's stick-table or in the designated stick-table,
15062 measured in kilobytes. If the address is not found, zero is returned. The
15063 test is currently performed on 32-bit integers, which limits values to 4
15064 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015065
Willy Tarreau74ca5042013-06-11 23:12:07 +020015066src_port : integer
15067 Returns an integer value corresponding to the TCP source port of the
15068 connection on the client side, which is the port the client connected from.
15069 Usage of this function is very limited as modern protocols do not care much
15070 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010015071
Willy Tarreau74ca5042013-06-11 23:12:07 +020015072src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015073 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020015074 connection's source IPv4 address in the current proxy's stick-table or in the
15075 designated stick-table, that were transformed into sessions, which means that
15076 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015077 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015078
Willy Tarreau74ca5042013-06-11 23:12:07 +020015079src_sess_rate([<table>]) : integer
15080 Returns the average session rate from the incoming connection's source
15081 address in the current proxy's stick-table or in the designated stick-table,
15082 measured in amount of sessions over the period configured in the table. A
15083 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015084 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015085
Willy Tarreau74ca5042013-06-11 23:12:07 +020015086src_updt_conn_cnt([<table>]) : integer
15087 Creates or updates the entry associated to the incoming connection's source
15088 address in the current proxy's stick-table or in the designated stick-table.
15089 This table must be configured to store the "conn_cnt" data type, otherwise
15090 the match will be ignored. The current count is incremented by one, and the
15091 expiration timer refreshed. The updated count is returned, so this match
15092 can't return zero. This was used to reject service abusers based on their
15093 source address. Note: it is recommended to use the more complete "track-sc*"
15094 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015095
15096 Example :
15097 # This frontend limits incoming SSH connections to 3 per 10 second for
15098 # each source address, and rejects excess connections until a 10 second
15099 # silence is observed. At most 20 addresses are tracked.
15100 listen ssh
15101 bind :22
15102 mode tcp
15103 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020015104 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020015105 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015106 server local 127.0.0.1:22
15107
Willy Tarreau74ca5042013-06-11 23:12:07 +020015108srv_id : integer
15109 Returns an integer containing the server's id when processing the response.
15110 While it's almost only used with ACLs, it may be used for logging or
15111 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020015112
vkill1dfd1652019-10-30 16:58:14 +080015113srv_name : string
15114 Returns a string containing the server's name when processing the response.
15115 While it's almost only used with ACLs, it may be used for logging or
15116 debugging.
15117
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200151187.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020015119----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020015120
Willy Tarreau74ca5042013-06-11 23:12:07 +020015121The layer 5 usually describes just the session layer which in haproxy is
15122closest to the session once all the connection handshakes are finished, but
15123when no content is yet made available. The fetch methods described here are
15124usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015125future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020015126
Ben Shillitof25e8e52016-12-02 14:25:37 +00001512751d.all(<prop>[,<prop>*]) : string
15128 Returns values for the properties requested as a string, where values are
15129 separated by the delimiter specified with "51degrees-property-separator".
15130 The device is identified using all the important HTTP headers from the
15131 request. The function can be passed up to five property names, and if a
15132 property name can't be found, the value "NoData" is returned.
15133
15134 Example :
15135 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
15136 # containing the three properties requested using all relevant headers from
15137 # the request.
15138 frontend http-in
15139 bind *:8081
15140 default_backend servers
15141 http-request set-header X-51D-DeviceTypeMobileTablet \
15142 %[51d.all(DeviceType,IsMobile,IsTablet)]
15143
Emeric Brun645ae792014-04-30 14:21:06 +020015144ssl_bc : boolean
15145 Returns true when the back connection was made via an SSL/TLS transport
15146 layer and is locally deciphered. This means the outgoing connection was made
15147 other a server with the "ssl" option.
15148
15149ssl_bc_alg_keysize : integer
15150 Returns the symmetric cipher key size supported in bits when the outgoing
15151 connection was made over an SSL/TLS transport layer.
15152
Olivier Houchard6b77f492018-11-22 18:18:29 +010015153ssl_bc_alpn : string
15154 This extracts the Application Layer Protocol Negotiation field from an
15155 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020015156 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010015157 server. The SSL library must have been built with support for TLS
15158 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15159 not advertised unless the "alpn" keyword on the "server" line specifies a
15160 protocol list. Also, nothing forces the server to pick a protocol from this
15161 list, any other one may be requested. The TLS ALPN extension is meant to
15162 replace the TLS NPN extension. See also "ssl_bc_npn".
15163
Emeric Brun645ae792014-04-30 14:21:06 +020015164ssl_bc_cipher : string
15165 Returns the name of the used cipher when the outgoing connection was made
15166 over an SSL/TLS transport layer.
15167
Patrick Hemmer65674662019-06-04 08:13:03 -040015168ssl_bc_client_random : binary
15169 Returns the client random of the back connection when the incoming connection
15170 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15171 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15172
Emeric Brun74f7ffa2018-02-19 16:14:12 +010015173ssl_bc_is_resumed : boolean
15174 Returns true when the back connection was made over an SSL/TLS transport
15175 layer and the newly created SSL session was resumed using a cached
15176 session or a TLS ticket.
15177
Olivier Houchard6b77f492018-11-22 18:18:29 +010015178ssl_bc_npn : string
15179 This extracts the Next Protocol Negotiation field from an outgoing connection
15180 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020015181 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010015182 built with support for TLS extensions enabled (check haproxy -vv). Note that
15183 the TLS NPN extension is not advertised unless the "npn" keyword on the
15184 "server" line specifies a protocol list. Also, nothing forces the server to
15185 pick a protocol from this list, any other one may be used. Please note that
15186 the TLS NPN extension was replaced with ALPN.
15187
Emeric Brun645ae792014-04-30 14:21:06 +020015188ssl_bc_protocol : string
15189 Returns the name of the used protocol when the outgoing connection was made
15190 over an SSL/TLS transport layer.
15191
Emeric Brunb73a9b02014-04-30 18:49:19 +020015192ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020015193 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015194 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15195 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020015196
Patrick Hemmer65674662019-06-04 08:13:03 -040015197ssl_bc_server_random : binary
15198 Returns the server random of the back connection when the incoming connection
15199 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15200 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15201
Emeric Brun645ae792014-04-30 14:21:06 +020015202ssl_bc_session_id : binary
15203 Returns the SSL ID of the back connection when the outgoing connection was
15204 made over an SSL/TLS transport layer. It is useful to log if we want to know
15205 if session was reused or not.
15206
Patrick Hemmere0275472018-04-28 19:15:51 -040015207ssl_bc_session_key : binary
15208 Returns the SSL session master key of the back connection when the outgoing
15209 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15210 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15211 BoringSSL.
15212
Emeric Brun645ae792014-04-30 14:21:06 +020015213ssl_bc_use_keysize : integer
15214 Returns the symmetric cipher key size used in bits when the outgoing
15215 connection was made over an SSL/TLS transport layer.
15216
Willy Tarreau74ca5042013-06-11 23:12:07 +020015217ssl_c_ca_err : integer
15218 When the incoming connection was made over an SSL/TLS transport layer,
15219 returns the ID of the first error detected during verification of the client
15220 certificate at depth > 0, or 0 if no error was encountered during this
15221 verification process. Please refer to your SSL library's documentation to
15222 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020015223
Willy Tarreau74ca5042013-06-11 23:12:07 +020015224ssl_c_ca_err_depth : integer
15225 When the incoming connection was made over an SSL/TLS transport layer,
15226 returns the depth in the CA chain of the first error detected during the
15227 verification of the client certificate. If no error is encountered, 0 is
15228 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015229
Emeric Brun43e79582014-10-29 19:03:26 +010015230ssl_c_der : binary
15231 Returns the DER formatted certificate presented by the client when the
15232 incoming connection was made over an SSL/TLS transport layer. When used for
15233 an ACL, the value(s) to match against can be passed in hexadecimal form.
15234
Willy Tarreau74ca5042013-06-11 23:12:07 +020015235ssl_c_err : integer
15236 When the incoming connection was made over an SSL/TLS transport layer,
15237 returns the ID of the first error detected during verification at depth 0, or
15238 0 if no error was encountered during this verification process. Please refer
15239 to your SSL library's documentation to find the exhaustive list of error
15240 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020015241
Willy Tarreau74ca5042013-06-11 23:12:07 +020015242ssl_c_i_dn([<entry>[,<occ>]]) : string
15243 When the incoming connection was made over an SSL/TLS transport layer,
15244 returns the full distinguished name of the issuer of the certificate
15245 presented by the client when no <entry> is specified, or the value of the
15246 first given entry found from the beginning of the DN. If a positive/negative
15247 occurrence number is specified as the optional second argument, it returns
15248 the value of the nth given entry value from the beginning/end of the DN.
15249 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
15250 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020015251
Willy Tarreau74ca5042013-06-11 23:12:07 +020015252ssl_c_key_alg : string
15253 Returns the name of the algorithm used to generate the key of the certificate
15254 presented by the client when the incoming connection was made over an SSL/TLS
15255 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020015256
Willy Tarreau74ca5042013-06-11 23:12:07 +020015257ssl_c_notafter : string
15258 Returns the end date presented by the client as a formatted string
15259 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15260 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020015261
Willy Tarreau74ca5042013-06-11 23:12:07 +020015262ssl_c_notbefore : string
15263 Returns the start date presented by the client as a formatted string
15264 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15265 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015266
Willy Tarreau74ca5042013-06-11 23:12:07 +020015267ssl_c_s_dn([<entry>[,<occ>]]) : string
15268 When the incoming connection was made over an SSL/TLS transport layer,
15269 returns the full distinguished name of the subject of the certificate
15270 presented by the client when no <entry> is specified, or the value of the
15271 first given entry found from the beginning of the DN. If a positive/negative
15272 occurrence number is specified as the optional second argument, it returns
15273 the value of the nth given entry value from the beginning/end of the DN.
15274 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
15275 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015276
Willy Tarreau74ca5042013-06-11 23:12:07 +020015277ssl_c_serial : binary
15278 Returns the serial of the certificate presented by the client when the
15279 incoming connection was made over an SSL/TLS transport layer. When used for
15280 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015281
Willy Tarreau74ca5042013-06-11 23:12:07 +020015282ssl_c_sha1 : binary
15283 Returns the SHA-1 fingerprint of the certificate presented by the client when
15284 the incoming connection was made over an SSL/TLS transport layer. This can be
15285 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015286 Note that the output is binary, so if you want to pass that signature to the
15287 server, you need to encode it in hex or base64, such as in the example below:
15288
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015289 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015290 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020015291
Willy Tarreau74ca5042013-06-11 23:12:07 +020015292ssl_c_sig_alg : string
15293 Returns the name of the algorithm used to sign the certificate presented by
15294 the client when the incoming connection was made over an SSL/TLS transport
15295 layer.
Emeric Brun87855892012-10-17 17:39:35 +020015296
Willy Tarreau74ca5042013-06-11 23:12:07 +020015297ssl_c_used : boolean
15298 Returns true if current SSL session uses a client certificate even if current
15299 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020015300
Willy Tarreau74ca5042013-06-11 23:12:07 +020015301ssl_c_verify : integer
15302 Returns the verify result error ID when the incoming connection was made over
15303 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
15304 refer to your SSL library's documentation for an exhaustive list of error
15305 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020015306
Willy Tarreau74ca5042013-06-11 23:12:07 +020015307ssl_c_version : integer
15308 Returns the version of the certificate presented by the client when the
15309 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020015310
Emeric Brun43e79582014-10-29 19:03:26 +010015311ssl_f_der : binary
15312 Returns the DER formatted certificate presented by the frontend when the
15313 incoming connection was made over an SSL/TLS transport layer. When used for
15314 an ACL, the value(s) to match against can be passed in hexadecimal form.
15315
Willy Tarreau74ca5042013-06-11 23:12:07 +020015316ssl_f_i_dn([<entry>[,<occ>]]) : string
15317 When the incoming connection was made over an SSL/TLS transport layer,
15318 returns the full distinguished name of the issuer of the certificate
15319 presented by the frontend when no <entry> is specified, or the value of the
15320 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020015321 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020015322 the value of the nth given entry value from the beginning/end of the DN.
15323 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
15324 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020015325
Willy Tarreau74ca5042013-06-11 23:12:07 +020015326ssl_f_key_alg : string
15327 Returns the name of the algorithm used to generate the key of the certificate
15328 presented by the frontend when the incoming connection was made over an
15329 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015330
Willy Tarreau74ca5042013-06-11 23:12:07 +020015331ssl_f_notafter : string
15332 Returns the end date presented by the frontend as a formatted string
15333 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15334 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015335
Willy Tarreau74ca5042013-06-11 23:12:07 +020015336ssl_f_notbefore : string
15337 Returns the start date presented by the frontend as a formatted string
15338 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15339 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020015340
Willy Tarreau74ca5042013-06-11 23:12:07 +020015341ssl_f_s_dn([<entry>[,<occ>]]) : string
15342 When the incoming connection was made over an SSL/TLS transport layer,
15343 returns the full distinguished name of the subject of the certificate
15344 presented by the frontend when no <entry> is specified, or the value of the
15345 first given entry found from the beginning of the DN. If a positive/negative
15346 occurrence number is specified as the optional second argument, it returns
15347 the value of the nth given entry value from the beginning/end of the DN.
15348 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
15349 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020015350
Willy Tarreau74ca5042013-06-11 23:12:07 +020015351ssl_f_serial : binary
15352 Returns the serial of the certificate presented by the frontend when the
15353 incoming connection was made over an SSL/TLS transport layer. When used for
15354 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020015355
Emeric Brun55f4fa82014-04-30 17:11:25 +020015356ssl_f_sha1 : binary
15357 Returns the SHA-1 fingerprint of the certificate presented by the frontend
15358 when the incoming connection was made over an SSL/TLS transport layer. This
15359 can be used to know which certificate was chosen using SNI.
15360
Willy Tarreau74ca5042013-06-11 23:12:07 +020015361ssl_f_sig_alg : string
15362 Returns the name of the algorithm used to sign the certificate presented by
15363 the frontend when the incoming connection was made over an SSL/TLS transport
15364 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015365
Willy Tarreau74ca5042013-06-11 23:12:07 +020015366ssl_f_version : integer
15367 Returns the version of the certificate presented by the frontend when the
15368 incoming connection was made over an SSL/TLS transport layer.
15369
15370ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015371 Returns true when the front connection was made via an SSL/TLS transport
15372 layer and is locally deciphered. This means it has matched a socket declared
15373 with a "bind" line having the "ssl" option.
15374
Willy Tarreau74ca5042013-06-11 23:12:07 +020015375 Example :
15376 # This passes "X-Proto: https" to servers when client connects over SSL
15377 listen http-https
15378 bind :80
15379 bind :443 ssl crt /etc/haproxy.pem
15380 http-request add-header X-Proto https if { ssl_fc }
15381
15382ssl_fc_alg_keysize : integer
15383 Returns the symmetric cipher key size supported in bits when the incoming
15384 connection was made over an SSL/TLS transport layer.
15385
15386ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015387 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020015388 incoming connection made via a TLS transport layer and locally deciphered by
15389 haproxy. The result is a string containing the protocol name advertised by
15390 the client. The SSL library must have been built with support for TLS
15391 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15392 not advertised unless the "alpn" keyword on the "bind" line specifies a
15393 protocol list. Also, nothing forces the client to pick a protocol from this
15394 list, any other one may be requested. The TLS ALPN extension is meant to
15395 replace the TLS NPN extension. See also "ssl_fc_npn".
15396
Willy Tarreau74ca5042013-06-11 23:12:07 +020015397ssl_fc_cipher : string
15398 Returns the name of the used cipher when the incoming connection was made
15399 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020015400
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015401ssl_fc_cipherlist_bin : binary
15402 Returns the binary form of the client hello cipher list. The maximum returned
15403 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015404 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015405
15406ssl_fc_cipherlist_hex : string
15407 Returns the binary form of the client hello cipher list encoded as
15408 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015409 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015410
15411ssl_fc_cipherlist_str : string
15412 Returns the decoded text form of the client hello cipher list. The maximum
15413 number of ciphers returned is according with the value of
15414 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010015415 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020015416 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015417
15418ssl_fc_cipherlist_xxh : integer
15419 Returns a xxh64 of the cipher list. This hash can be return only is the value
15420 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015421 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015422
Patrick Hemmer65674662019-06-04 08:13:03 -040015423ssl_fc_client_random : binary
15424 Returns the client random of the front connection when the incoming connection
15425 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15426 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15427
Willy Tarreau74ca5042013-06-11 23:12:07 +020015428ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015429 Returns true if a client certificate is present in an incoming connection over
15430 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010015431 Note: on SSL session resumption with Session ID or TLS ticket, client
15432 certificate is not present in the current connection but may be retrieved
15433 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
15434 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015435
Olivier Houchardccaa7de2017-10-02 11:51:03 +020015436ssl_fc_has_early : boolean
15437 Returns true if early data were sent, and the handshake didn't happen yet. As
15438 it has security implications, it is useful to be able to refuse those, or
15439 wait until the handshake happened.
15440
Willy Tarreau74ca5042013-06-11 23:12:07 +020015441ssl_fc_has_sni : boolean
15442 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020015443 in an incoming connection was made over an SSL/TLS transport layer. Returns
15444 true when the incoming connection presents a TLS SNI field. This requires
John Roeslerfb2fce12019-07-10 15:45:51 -050015445 that the SSL library is built with support for TLS extensions enabled (check
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020015446 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020015447
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020015448ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015449 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010015450 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
15451 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015452
Willy Tarreau74ca5042013-06-11 23:12:07 +020015453ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015454 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020015455 made via a TLS transport layer and locally deciphered by haproxy. The result
15456 is a string containing the protocol name advertised by the client. The SSL
15457 library must have been built with support for TLS extensions enabled (check
15458 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
15459 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
15460 forces the client to pick a protocol from this list, any other one may be
15461 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020015462
Willy Tarreau74ca5042013-06-11 23:12:07 +020015463ssl_fc_protocol : string
15464 Returns the name of the used protocol when the incoming connection was made
15465 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020015466
Emeric Brunb73a9b02014-04-30 18:49:19 +020015467ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040015468 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015469 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15470 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040015471
Patrick Hemmer65674662019-06-04 08:13:03 -040015472ssl_fc_server_random : binary
15473 Returns the server random of the front connection when the incoming connection
15474 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15475 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15476
Willy Tarreau74ca5042013-06-11 23:12:07 +020015477ssl_fc_session_id : binary
15478 Returns the SSL ID of the front connection when the incoming connection was
15479 made over an SSL/TLS transport layer. It is useful to stick a given client to
15480 a server. It is important to note that some browsers refresh their session ID
15481 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020015482
Patrick Hemmere0275472018-04-28 19:15:51 -040015483ssl_fc_session_key : binary
15484 Returns the SSL session master key of the front connection when the incoming
15485 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15486 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15487 BoringSSL.
15488
15489
Willy Tarreau74ca5042013-06-11 23:12:07 +020015490ssl_fc_sni : string
15491 This extracts the Server Name Indication TLS extension (SNI) field from an
15492 incoming connection made via an SSL/TLS transport layer and locally
15493 deciphered by haproxy. The result (when present) typically is a string
15494 matching the HTTPS host name (253 chars or less). The SSL library must have
15495 been built with support for TLS extensions enabled (check haproxy -vv).
15496
15497 This fetch is different from "req_ssl_sni" above in that it applies to the
15498 connection being deciphered by haproxy and not to SSL contents being blindly
15499 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
John Roeslerfb2fce12019-07-10 15:45:51 -050015500 requires that the SSL library is built with support for TLS extensions
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020015501 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020015502
Willy Tarreau74ca5042013-06-11 23:12:07 +020015503 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020015504 ssl_fc_sni_end : suffix match
15505 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020015506
Willy Tarreau74ca5042013-06-11 23:12:07 +020015507ssl_fc_use_keysize : integer
15508 Returns the symmetric cipher key size used in bits when the incoming
15509 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015510
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015511
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200155127.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020015513------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015514
Willy Tarreau74ca5042013-06-11 23:12:07 +020015515Fetching samples from buffer contents is a bit different from the previous
15516sample fetches above because the sampled data are ephemeral. These data can
15517only be used when they're available and will be lost when they're forwarded.
15518For this reason, samples fetched from buffer contents during a request cannot
15519be used in a response for example. Even while the data are being fetched, they
15520can change. Sometimes it is necessary to set some delays or combine multiple
15521sample fetch methods to ensure that the expected data are complete and usable,
15522for example through TCP request content inspection. Please see the "tcp-request
15523content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020015524
Willy Tarreau74ca5042013-06-11 23:12:07 +020015525payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010015526 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015527 "stick on", "stick match"), and for "res.payload" when used in the context of
15528 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015529
Willy Tarreau74ca5042013-06-11 23:12:07 +020015530payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
15531 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010015532 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015533 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015534
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020015535req.hdrs : string
15536 Returns the current request headers as string including the last empty line
15537 separating headers from the request body. The last empty line can be used to
15538 detect a truncated header block. This sample fetch is useful for some SPOE
15539 headers analyzers and for advanced logging.
15540
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020015541req.hdrs_bin : binary
15542 Returns the current request headers contained in preparsed binary form. This
15543 is useful for offloading some processing with SPOE. Each string is described
15544 by a length followed by the number of bytes indicated in the length. The
15545 length is represented using the variable integer encoding detailed in the
15546 SPOE documentation. The end of the list is marked by a couple of empty header
15547 names and values (length of 0 for both).
15548
15549 *(<str:header-name><str:header-value>)<empty string><empty string>
15550
15551 int: refer to the SPOE documentation for the encoding
15552 str: <int:length><bytes>
15553
Willy Tarreau74ca5042013-06-11 23:12:07 +020015554req.len : integer
15555req_len : integer (deprecated)
15556 Returns an integer value corresponding to the number of bytes present in the
15557 request buffer. This is mostly used in ACL. It is important to understand
15558 that this test does not return false as long as the buffer is changing. This
15559 means that a check with equality to zero will almost always immediately match
15560 at the beginning of the session, while a test for more data will wait for
15561 that data to come in and return false only when haproxy is certain that no
15562 more data will come in. This test was designed to be used with TCP request
15563 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015564
Willy Tarreau74ca5042013-06-11 23:12:07 +020015565req.payload(<offset>,<length>) : binary
15566 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015567 in the request buffer. As a special case, if the <length> argument is zero,
15568 the the whole buffer from <offset> to the end is extracted. This can be used
15569 with ACLs in order to check for the presence of some content in a buffer at
15570 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015571
Willy Tarreau74ca5042013-06-11 23:12:07 +020015572 ACL alternatives :
15573 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015574
Willy Tarreau74ca5042013-06-11 23:12:07 +020015575req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
15576 This extracts a binary block whose size is specified at <offset1> for <length>
15577 bytes, and which starts at <offset2> if specified or just after the length in
15578 the request buffer. The <offset2> parameter also supports relative offsets if
15579 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015580
Willy Tarreau74ca5042013-06-11 23:12:07 +020015581 ACL alternatives :
15582 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015583
Willy Tarreau74ca5042013-06-11 23:12:07 +020015584 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015585
Willy Tarreau74ca5042013-06-11 23:12:07 +020015586req.proto_http : boolean
15587req_proto_http : boolean (deprecated)
15588 Returns true when data in the request buffer look like HTTP and correctly
15589 parses as such. It is the same parser as the common HTTP request parser which
15590 is used so there should be no surprises. The test does not match until the
15591 request is complete, failed or timed out. This test may be used to report the
15592 protocol in TCP logs, but the biggest use is to block TCP request analysis
15593 until a complete HTTP request is present in the buffer, for example to track
15594 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015595
Willy Tarreau74ca5042013-06-11 23:12:07 +020015596 Example:
15597 # track request counts per "base" (concatenation of Host+URL)
15598 tcp-request inspect-delay 10s
15599 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015600 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015601
Willy Tarreau74ca5042013-06-11 23:12:07 +020015602req.rdp_cookie([<name>]) : string
15603rdp_cookie([<name>]) : string (deprecated)
15604 When the request buffer looks like the RDP protocol, extracts the RDP cookie
15605 <name>, or any cookie if unspecified. The parser only checks for the first
15606 cookie, as illustrated in the RDP protocol specification. The cookie name is
15607 case insensitive. Generally the "MSTS" cookie name will be used, as it can
15608 contain the user name of the client connecting to the server if properly
15609 configured on the client. The "MSTSHASH" cookie is often used as well for
15610 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015611
Willy Tarreau74ca5042013-06-11 23:12:07 +020015612 This differs from "balance rdp-cookie" in that any balancing algorithm may be
15613 used and thus the distribution of clients to backend servers is not linked to
15614 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
15615 such as "balance roundrobin" or "balance leastconn" will lead to a more even
15616 distribution of clients to backend servers than the hash used by "balance
15617 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015618
Willy Tarreau74ca5042013-06-11 23:12:07 +020015619 ACL derivatives :
15620 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015621
Willy Tarreau74ca5042013-06-11 23:12:07 +020015622 Example :
15623 listen tse-farm
15624 bind 0.0.0.0:3389
15625 # wait up to 5s for an RDP cookie in the request
15626 tcp-request inspect-delay 5s
15627 tcp-request content accept if RDP_COOKIE
15628 # apply RDP cookie persistence
15629 persist rdp-cookie
15630 # Persist based on the mstshash cookie
15631 # This is only useful makes sense if
15632 # balance rdp-cookie is not used
15633 stick-table type string size 204800
15634 stick on req.rdp_cookie(mstshash)
15635 server srv1 1.1.1.1:3389
15636 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015637
Willy Tarreau74ca5042013-06-11 23:12:07 +020015638 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
15639 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015640
Willy Tarreau74ca5042013-06-11 23:12:07 +020015641req.rdp_cookie_cnt([name]) : integer
15642rdp_cookie_cnt([name]) : integer (deprecated)
15643 Tries to parse the request buffer as RDP protocol, then returns an integer
15644 corresponding to the number of RDP cookies found. If an optional cookie name
15645 is passed, only cookies matching this name are considered. This is mostly
15646 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015647
Willy Tarreau74ca5042013-06-11 23:12:07 +020015648 ACL derivatives :
15649 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015650
Alex Zorin4afdd132018-12-30 13:56:28 +110015651req.ssl_alpn : string
15652 Returns a string containing the values of the Application-Layer Protocol
15653 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
15654 ClientHello message. Note that this only applies to raw contents found in the
15655 request buffer and not to the contents deciphered via an SSL data layer, so
15656 this will not work with "bind" lines having the "ssl" option. This is useful
15657 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020015658 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110015659
15660 Examples :
15661 # Wait for a client hello for at most 5 seconds
15662 tcp-request inspect-delay 5s
15663 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020015664 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110015665 default_backend bk_default
15666
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020015667req.ssl_ec_ext : boolean
15668 Returns a boolean identifying if client sent the Supported Elliptic Curves
15669 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020015670 message. This can be used to present ECC compatible clients with EC
15671 certificate and to use RSA for all others, on the same IP address. Note that
15672 this only applies to raw contents found in the request buffer and not to
15673 contents deciphered via an SSL data layer, so this will not work with "bind"
15674 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020015675
Willy Tarreau74ca5042013-06-11 23:12:07 +020015676req.ssl_hello_type : integer
15677req_ssl_hello_type : integer (deprecated)
15678 Returns an integer value containing the type of the SSL hello message found
15679 in the request buffer if the buffer contains data that parse as a complete
15680 SSL (v3 or superior) client hello message. Note that this only applies to raw
15681 contents found in the request buffer and not to contents deciphered via an
15682 SSL data layer, so this will not work with "bind" lines having the "ssl"
15683 option. This is mostly used in ACL to detect presence of an SSL hello message
15684 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015685
Willy Tarreau74ca5042013-06-11 23:12:07 +020015686req.ssl_sni : string
15687req_ssl_sni : string (deprecated)
15688 Returns a string containing the value of the Server Name TLS extension sent
15689 by a client in a TLS stream passing through the request buffer if the buffer
15690 contains data that parse as a complete SSL (v3 or superior) client hello
15691 message. Note that this only applies to raw contents found in the request
15692 buffer and not to contents deciphered via an SSL data layer, so this will not
15693 work with "bind" lines having the "ssl" option. SNI normally contains the
15694 name of the host the client tries to connect to (for recent browsers). SNI is
15695 useful for allowing or denying access to certain hosts when SSL/TLS is used
15696 by the client. This test was designed to be used with TCP request content
15697 inspection. If content switching is needed, it is recommended to first wait
15698 for a complete client hello (type 1), like in the example below. See also
15699 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015700
Willy Tarreau74ca5042013-06-11 23:12:07 +020015701 ACL derivatives :
15702 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015703
Willy Tarreau74ca5042013-06-11 23:12:07 +020015704 Examples :
15705 # Wait for a client hello for at most 5 seconds
15706 tcp-request inspect-delay 5s
15707 tcp-request content accept if { req_ssl_hello_type 1 }
15708 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
15709 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015710
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053015711req.ssl_st_ext : integer
15712 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
15713 Returns 1 if the client sent SessionTicket TLS Extension
15714 Returns 2 if the client also sent non-zero length TLS SessionTicket
15715 Note that this only applies to raw contents found in the request buffer and
15716 not to contents deciphered via an SSL data layer, so this will not work with
15717 "bind" lines having the "ssl" option. This can for example be used to detect
15718 whether the client sent a SessionTicket or not and stick it accordingly, if
15719 no SessionTicket then stick on SessionID or don't stick as there's no server
15720 side state is there when SessionTickets are in use.
15721
Willy Tarreau74ca5042013-06-11 23:12:07 +020015722req.ssl_ver : integer
15723req_ssl_ver : integer (deprecated)
15724 Returns an integer value containing the version of the SSL/TLS protocol of a
15725 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
15726 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
15727 composed of the major version multiplied by 65536, added to the minor
15728 version. Note that this only applies to raw contents found in the request
15729 buffer and not to contents deciphered via an SSL data layer, so this will not
15730 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010015731 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015732 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015733
Willy Tarreau74ca5042013-06-11 23:12:07 +020015734 ACL derivatives :
15735 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010015736
Willy Tarreau47e8eba2013-09-11 23:28:46 +020015737res.len : integer
15738 Returns an integer value corresponding to the number of bytes present in the
15739 response buffer. This is mostly used in ACL. It is important to understand
15740 that this test does not return false as long as the buffer is changing. This
15741 means that a check with equality to zero will almost always immediately match
15742 at the beginning of the session, while a test for more data will wait for
15743 that data to come in and return false only when haproxy is certain that no
15744 more data will come in. This test was designed to be used with TCP response
15745 content inspection.
15746
Willy Tarreau74ca5042013-06-11 23:12:07 +020015747res.payload(<offset>,<length>) : binary
15748 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015749 in the response buffer. As a special case, if the <length> argument is zero,
15750 the the whole buffer from <offset> to the end is extracted. This can be used
15751 with ACLs in order to check for the presence of some content in a buffer at
15752 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015753
Willy Tarreau74ca5042013-06-11 23:12:07 +020015754res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
15755 This extracts a binary block whose size is specified at <offset1> for <length>
15756 bytes, and which starts at <offset2> if specified or just after the length in
15757 the response buffer. The <offset2> parameter also supports relative offsets
15758 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015759
Willy Tarreau74ca5042013-06-11 23:12:07 +020015760 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015761
Willy Tarreau971f7b62015-09-29 14:06:59 +020015762res.ssl_hello_type : integer
15763rep_ssl_hello_type : integer (deprecated)
15764 Returns an integer value containing the type of the SSL hello message found
15765 in the response buffer if the buffer contains data that parses as a complete
15766 SSL (v3 or superior) hello message. Note that this only applies to raw
15767 contents found in the response buffer and not to contents deciphered via an
15768 SSL data layer, so this will not work with "server" lines having the "ssl"
15769 option. This is mostly used in ACL to detect presence of an SSL hello message
15770 that is supposed to contain an SSL session ID usable for stickiness.
15771
Willy Tarreau74ca5042013-06-11 23:12:07 +020015772wait_end : boolean
15773 This fetch either returns true when the inspection period is over, or does
15774 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010015775 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020015776 actions, such as a delayed reject for some special addresses. Since it either
15777 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010015778 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020015779 "WAIT_END" is always usable without prior declaration. This test was designed
15780 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015781
Willy Tarreau74ca5042013-06-11 23:12:07 +020015782 Examples :
15783 # delay every incoming request by 2 seconds
15784 tcp-request inspect-delay 2s
15785 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010015786
Willy Tarreau74ca5042013-06-11 23:12:07 +020015787 # don't immediately tell bad guys they are rejected
15788 tcp-request inspect-delay 10s
15789 acl goodguys src 10.0.0.0/24
15790 acl badguys src 10.0.1.0/24
15791 tcp-request content accept if goodguys
15792 tcp-request content reject if badguys WAIT_END
15793 tcp-request content reject
15794
15795
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200157967.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020015797--------------------------------------
15798
15799It is possible to fetch samples from HTTP contents, requests and responses.
15800This application layer is also called layer 7. It is only possible to fetch the
15801data in this section when a full HTTP request or response has been parsed from
15802its respective request or response buffer. This is always the case with all
15803HTTP specific rules and for sections running with "mode http". When using TCP
15804content inspection, it may be necessary to support an inspection delay in order
15805to let the request or response come in first. These fetches may require a bit
15806more CPU resources than the layer 4 ones, but not much since the request and
15807response are indexed.
15808
15809base : string
15810 This returns the concatenation of the first Host header and the path part of
15811 the request, which starts at the first slash and ends before the question
15812 mark. It can be useful in virtual hosted environments to detect URL abuses as
15813 well as to improve shared caches efficiency. Using this with a limited size
15814 stick table also allows one to collect statistics about most commonly
15815 requested objects by host/path. With ACLs it can allow simple content
15816 switching rules involving the host and the path at the same time, such as
15817 "www.example.com/favicon.ico". See also "path" and "uri".
15818
15819 ACL derivatives :
15820 base : exact string match
15821 base_beg : prefix match
15822 base_dir : subdir match
15823 base_dom : domain match
15824 base_end : suffix match
15825 base_len : length match
15826 base_reg : regex match
15827 base_sub : substring match
15828
15829base32 : integer
15830 This returns a 32-bit hash of the value returned by the "base" fetch method
15831 above. This is useful to track per-URL activity on high traffic sites without
15832 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020015833 memory. The output type is an unsigned integer. The hash function used is
15834 SDBM with full avalanche on the output. Technically, base32 is exactly equal
15835 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020015836
15837base32+src : binary
15838 This returns the concatenation of the base32 fetch above and the src fetch
15839 below. The resulting type is of type binary, with a size of 8 or 20 bytes
15840 depending on the source address family. This can be used to track per-IP,
15841 per-URL counters.
15842
William Lallemand65ad6e12014-01-31 15:08:02 +010015843capture.req.hdr(<idx>) : string
15844 This extracts the content of the header captured by the "capture request
15845 header", idx is the position of the capture keyword in the configuration.
15846 The first entry is an index of 0. See also: "capture request header".
15847
15848capture.req.method : string
15849 This extracts the METHOD of an HTTP request. It can be used in both request
15850 and response. Unlike "method", it can be used in both request and response
15851 because it's allocated.
15852
15853capture.req.uri : string
15854 This extracts the request's URI, which starts at the first slash and ends
15855 before the first space in the request (without the host part). Unlike "path"
15856 and "url", it can be used in both request and response because it's
15857 allocated.
15858
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020015859capture.req.ver : string
15860 This extracts the request's HTTP version and returns either "HTTP/1.0" or
15861 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
15862 logs because it relies on a persistent flag.
15863
William Lallemand65ad6e12014-01-31 15:08:02 +010015864capture.res.hdr(<idx>) : string
15865 This extracts the content of the header captured by the "capture response
15866 header", idx is the position of the capture keyword in the configuration.
15867 The first entry is an index of 0.
15868 See also: "capture response header"
15869
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020015870capture.res.ver : string
15871 This extracts the response's HTTP version and returns either "HTTP/1.0" or
15872 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
15873 persistent flag.
15874
Willy Tarreaua5910cc2015-05-02 00:46:08 +020015875req.body : binary
15876 This returns the HTTP request's available body as a block of data. It
15877 requires that the request body has been buffered made available using
15878 "option http-buffer-request". In case of chunked-encoded body, currently only
15879 the first chunk is analyzed.
15880
Thierry FOURNIER9826c772015-05-20 15:50:54 +020015881req.body_param([<name>) : string
15882 This fetch assumes that the body of the POST request is url-encoded. The user
15883 can check if the "content-type" contains the value
15884 "application/x-www-form-urlencoded". This extracts the first occurrence of the
15885 parameter <name> in the body, which ends before '&'. The parameter name is
15886 case-sensitive. If no name is given, any parameter will match, and the first
15887 one will be returned. The result is a string corresponding to the value of the
15888 parameter <name> as presented in the request body (no URL decoding is
15889 performed). Note that the ACL version of this fetch iterates over multiple
15890 parameters and will iteratively report all parameters values if no name is
15891 given.
15892
Willy Tarreaua5910cc2015-05-02 00:46:08 +020015893req.body_len : integer
15894 This returns the length of the HTTP request's available body in bytes. It may
15895 be lower than the advertised length if the body is larger than the buffer. It
15896 requires that the request body has been buffered made available using
15897 "option http-buffer-request".
15898
15899req.body_size : integer
15900 This returns the advertised length of the HTTP request's body in bytes. It
15901 will represent the advertised Content-Length header, or the size of the first
15902 chunk in case of chunked encoding. In order to parse the chunks, it requires
15903 that the request body has been buffered made available using
15904 "option http-buffer-request".
15905
Willy Tarreau74ca5042013-06-11 23:12:07 +020015906req.cook([<name>]) : string
15907cook([<name>]) : string (deprecated)
15908 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15909 header line from the request, and returns its value as string. If no name is
15910 specified, the first cookie value is returned. When used with ACLs, all
15911 matching cookies are evaluated. Spaces around the name and the value are
15912 ignored as requested by the Cookie header specification (RFC6265). The cookie
15913 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
15914 well return an empty value if it is present. Use the "found" match to detect
15915 presence. Use the res.cook() variant for response cookies sent by the server.
15916
15917 ACL derivatives :
15918 cook([<name>]) : exact string match
15919 cook_beg([<name>]) : prefix match
15920 cook_dir([<name>]) : subdir match
15921 cook_dom([<name>]) : domain match
15922 cook_end([<name>]) : suffix match
15923 cook_len([<name>]) : length match
15924 cook_reg([<name>]) : regex match
15925 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010015926
Willy Tarreau74ca5042013-06-11 23:12:07 +020015927req.cook_cnt([<name>]) : integer
15928cook_cnt([<name>]) : integer (deprecated)
15929 Returns an integer value representing the number of occurrences of the cookie
15930 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015931
Willy Tarreau74ca5042013-06-11 23:12:07 +020015932req.cook_val([<name>]) : integer
15933cook_val([<name>]) : integer (deprecated)
15934 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15935 header line from the request, and converts its value to an integer which is
15936 returned. If no name is specified, the first cookie value is returned. When
15937 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020015938
Willy Tarreau74ca5042013-06-11 23:12:07 +020015939cookie([<name>]) : string (deprecated)
15940 This extracts the last occurrence of the cookie name <name> on a "Cookie"
15941 header line from the request, or a "Set-Cookie" header from the response, and
15942 returns its value as a string. A typical use is to get multiple clients
15943 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020015944 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020015945 multi-peer synchronization and state keeping across restarts. If no name is
15946 specified, the first cookie value is returned. This fetch should not be used
15947 anymore and should be replaced by req.cook() or res.cook() instead as it
15948 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015949
Willy Tarreau74ca5042013-06-11 23:12:07 +020015950hdr([<name>[,<occ>]]) : string
15951 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
15952 used on responses. Please refer to these respective fetches for more details.
15953 In case of doubt about the fetch direction, please use the explicit ones.
15954 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015955 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015956
Willy Tarreau74ca5042013-06-11 23:12:07 +020015957req.fhdr(<name>[,<occ>]) : string
15958 This extracts the last occurrence of header <name> in an HTTP request. When
15959 used from an ACL, all occurrences are iterated over until a match is found.
15960 Optionally, a specific occurrence might be specified as a position number.
15961 Positive values indicate a position from the first occurrence, with 1 being
15962 the first one. Negative values indicate positions relative to the last one,
15963 with -1 being the last one. It differs from req.hdr() in that any commas
15964 present in the value are returned and are not used as delimiters. This is
15965 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015966
Willy Tarreau74ca5042013-06-11 23:12:07 +020015967req.fhdr_cnt([<name>]) : integer
15968 Returns an integer value representing the number of occurrences of request
15969 header field name <name>, or the total number of header fields if <name> is
15970 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
15971 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015972
Willy Tarreau74ca5042013-06-11 23:12:07 +020015973req.hdr([<name>[,<occ>]]) : string
15974 This extracts the last occurrence of header <name> in an HTTP request. When
15975 used from an ACL, all occurrences are iterated over until a match is found.
15976 Optionally, a specific occurrence might be specified as a position number.
15977 Positive values indicate a position from the first occurrence, with 1 being
15978 the first one. Negative values indicate positions relative to the last one,
15979 with -1 being the last one. A typical use is with the X-Forwarded-For header
15980 once converted to IP, associated with an IP stick-table. The function
15981 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000015982 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020015983 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010015984 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010015985
Willy Tarreau74ca5042013-06-11 23:12:07 +020015986 ACL derivatives :
15987 hdr([<name>[,<occ>]]) : exact string match
15988 hdr_beg([<name>[,<occ>]]) : prefix match
15989 hdr_dir([<name>[,<occ>]]) : subdir match
15990 hdr_dom([<name>[,<occ>]]) : domain match
15991 hdr_end([<name>[,<occ>]]) : suffix match
15992 hdr_len([<name>[,<occ>]]) : length match
15993 hdr_reg([<name>[,<occ>]]) : regex match
15994 hdr_sub([<name>[,<occ>]]) : substring match
15995
15996req.hdr_cnt([<name>]) : integer
15997hdr_cnt([<header>]) : integer (deprecated)
15998 Returns an integer value representing the number of occurrences of request
15999 header field name <name>, or the total number of header field values if
16000 <name> is not specified. It is important to remember that one header line may
16001 count as several headers if it has several values. The function considers any
16002 comma as a delimiter for distinct values. If full-line headers are desired
16003 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
16004 detect presence, absence or abuse of a specific header, as well as to block
16005 request smuggling attacks by rejecting requests which contain more than one
16006 of certain headers. See "req.hdr" for more information on header matching.
16007
16008req.hdr_ip([<name>[,<occ>]]) : ip
16009hdr_ip([<name>[,<occ>]]) : ip (deprecated)
16010 This extracts the last occurrence of header <name> in an HTTP request,
16011 converts it to an IPv4 or IPv6 address and returns this address. When used
16012 with ACLs, all occurrences are checked, and if <name> is omitted, every value
16013 of every header is checked. Optionally, a specific occurrence might be
16014 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010016015 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020016016 positions relative to the last one, with -1 being the last one. A typical use
16017 is with the X-Forwarded-For and X-Client-IP headers.
16018
16019req.hdr_val([<name>[,<occ>]]) : integer
16020hdr_val([<name>[,<occ>]]) : integer (deprecated)
16021 This extracts the last occurrence of header <name> in an HTTP request, and
16022 converts it to an integer value. When used with ACLs, all occurrences are
16023 checked, and if <name> is omitted, every value of every header is checked.
16024 Optionally, a specific occurrence might be specified as a position number.
16025 Positive values indicate a position from the first occurrence, with 1 being
16026 the first one. Negative values indicate positions relative to the last one,
16027 with -1 being the last one. A typical use is with the X-Forwarded-For header.
16028
Frédéric Lécailleec891192019-02-26 15:02:35 +010016029
16030
Willy Tarreau74ca5042013-06-11 23:12:07 +020016031http_auth(<userlist>) : boolean
16032 Returns a boolean indicating whether the authentication data received from
16033 the client match a username & password stored in the specified userlist. This
16034 fetch function is not really useful outside of ACLs. Currently only http
16035 basic auth is supported.
16036
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016037http_auth_group(<userlist>) : string
16038 Returns a string corresponding to the user name found in the authentication
16039 data received from the client if both the user name and password are valid
16040 according to the specified userlist. The main purpose is to use it in ACLs
16041 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016042 This fetch function is not really useful outside of ACLs. Currently only http
16043 basic auth is supported.
16044
16045 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016046 http_auth_group(<userlist>) : group ...
16047 Returns true when the user extracted from the request and whose password is
16048 valid according to the specified userlist belongs to at least one of the
16049 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016050
Christopher Fauleta4063562019-08-02 11:51:37 +020016051http_auth_pass : string
16052 Returns the user's password found in the authentication data received from
16053 the client, as supplied in the Authorization header. Not checks are
16054 performed by this sample fetch. Only Basic authentication is supported.
16055
16056http_auth_type : string
16057 Returns the authentication method found in the authentication data received from
16058 the client, as supplied in the Authorization header. Not checks are
16059 performed by this sample fetch. Only Basic authentication is supported.
16060
16061http_auth_user : string
16062 Returns the user name found in the authentication data received from the
16063 client, as supplied in the Authorization header. Not checks are performed by
16064 this sample fetch. Only Basic authentication is supported.
16065
Willy Tarreau74ca5042013-06-11 23:12:07 +020016066http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020016067 Returns true when the request being processed is the first one of the
16068 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020016069 from some requests when a request is not the first one, or to help grouping
16070 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020016071
Willy Tarreau74ca5042013-06-11 23:12:07 +020016072method : integer + string
16073 Returns an integer value corresponding to the method in the HTTP request. For
16074 example, "GET" equals 1 (check sources to establish the matching). Value 9
16075 means "other method" and may be converted to a string extracted from the
16076 stream. This should not be used directly as a sample, this is only meant to
16077 be used from ACLs, which transparently convert methods from patterns to these
16078 integer + string values. Some predefined ACL already check for most common
16079 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016080
Willy Tarreau74ca5042013-06-11 23:12:07 +020016081 ACL derivatives :
16082 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016083
Willy Tarreau74ca5042013-06-11 23:12:07 +020016084 Example :
16085 # only accept GET and HEAD requests
16086 acl valid_method method GET HEAD
16087 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020016088
Willy Tarreau74ca5042013-06-11 23:12:07 +020016089path : string
16090 This extracts the request's URL path, which starts at the first slash and
16091 ends before the question mark (without the host part). A typical use is with
16092 prefetch-capable caches, and with portals which need to aggregate multiple
16093 information from databases and keep them in caches. Note that with outgoing
16094 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010016095 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020016096 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016097
Willy Tarreau74ca5042013-06-11 23:12:07 +020016098 ACL derivatives :
16099 path : exact string match
16100 path_beg : prefix match
16101 path_dir : subdir match
16102 path_dom : domain match
16103 path_end : suffix match
16104 path_len : length match
16105 path_reg : regex match
16106 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016107
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016108query : string
16109 This extracts the request's query string, which starts after the first
16110 question mark. If no question mark is present, this fetch returns nothing. If
16111 a question mark is present but nothing follows, it returns an empty string.
16112 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010016113 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016114 which stops before the question mark.
16115
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016116req.hdr_names([<delim>]) : string
16117 This builds a string made from the concatenation of all header names as they
16118 appear in the request when the rule is evaluated. The default delimiter is
16119 the comma (',') but it may be overridden as an optional argument <delim>. In
16120 this case, only the first character of <delim> is considered.
16121
Willy Tarreau74ca5042013-06-11 23:12:07 +020016122req.ver : string
16123req_ver : string (deprecated)
16124 Returns the version string from the HTTP request, for example "1.1". This can
16125 be useful for logs, but is mostly there for ACL. Some predefined ACL already
16126 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016127
Willy Tarreau74ca5042013-06-11 23:12:07 +020016128 ACL derivatives :
16129 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020016130
Willy Tarreau74ca5042013-06-11 23:12:07 +020016131res.comp : boolean
16132 Returns the boolean "true" value if the response has been compressed by
16133 HAProxy, otherwise returns boolean "false". This may be used to add
16134 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016135
Willy Tarreau74ca5042013-06-11 23:12:07 +020016136res.comp_algo : string
16137 Returns a string containing the name of the algorithm used if the response
16138 was compressed by HAProxy, for example : "deflate". This may be used to add
16139 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016140
Willy Tarreau74ca5042013-06-11 23:12:07 +020016141res.cook([<name>]) : string
16142scook([<name>]) : string (deprecated)
16143 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16144 header line from the response, and returns its value as string. If no name is
16145 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016146
Willy Tarreau74ca5042013-06-11 23:12:07 +020016147 ACL derivatives :
16148 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016149
Willy Tarreau74ca5042013-06-11 23:12:07 +020016150res.cook_cnt([<name>]) : integer
16151scook_cnt([<name>]) : integer (deprecated)
16152 Returns an integer value representing the number of occurrences of the cookie
16153 <name> in the response, or all cookies if <name> is not specified. This is
16154 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016155
Willy Tarreau74ca5042013-06-11 23:12:07 +020016156res.cook_val([<name>]) : integer
16157scook_val([<name>]) : integer (deprecated)
16158 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16159 header line from the response, and converts its value to an integer which is
16160 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016161
Willy Tarreau74ca5042013-06-11 23:12:07 +020016162res.fhdr([<name>[,<occ>]]) : string
16163 This extracts the last occurrence of header <name> in an HTTP response, or of
16164 the last header if no <name> is specified. Optionally, a specific occurrence
16165 might be specified as a position number. Positive values indicate a position
16166 from the first occurrence, with 1 being the first one. Negative values
16167 indicate positions relative to the last one, with -1 being the last one. It
16168 differs from res.hdr() in that any commas present in the value are returned
16169 and are not used as delimiters. If this is not desired, the res.hdr() fetch
16170 should be used instead. This is sometimes useful with headers such as Date or
16171 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016172
Willy Tarreau74ca5042013-06-11 23:12:07 +020016173res.fhdr_cnt([<name>]) : integer
16174 Returns an integer value representing the number of occurrences of response
16175 header field name <name>, or the total number of header fields if <name> is
16176 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
16177 the number of full line headers and does not stop on commas. If this is not
16178 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016179
Willy Tarreau74ca5042013-06-11 23:12:07 +020016180res.hdr([<name>[,<occ>]]) : string
16181shdr([<name>[,<occ>]]) : string (deprecated)
16182 This extracts the last occurrence of header <name> in an HTTP response, or of
16183 the last header if no <name> is specified. Optionally, a specific occurrence
16184 might be specified as a position number. Positive values indicate a position
16185 from the first occurrence, with 1 being the first one. Negative values
16186 indicate positions relative to the last one, with -1 being the last one. This
16187 can be useful to learn some data into a stick-table. The function considers
16188 any comma as a delimiter for distinct values. If this is not desired, the
16189 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016190
Willy Tarreau74ca5042013-06-11 23:12:07 +020016191 ACL derivatives :
16192 shdr([<name>[,<occ>]]) : exact string match
16193 shdr_beg([<name>[,<occ>]]) : prefix match
16194 shdr_dir([<name>[,<occ>]]) : subdir match
16195 shdr_dom([<name>[,<occ>]]) : domain match
16196 shdr_end([<name>[,<occ>]]) : suffix match
16197 shdr_len([<name>[,<occ>]]) : length match
16198 shdr_reg([<name>[,<occ>]]) : regex match
16199 shdr_sub([<name>[,<occ>]]) : substring match
16200
16201res.hdr_cnt([<name>]) : integer
16202shdr_cnt([<name>]) : integer (deprecated)
16203 Returns an integer value representing the number of occurrences of response
16204 header field name <name>, or the total number of header fields if <name> is
16205 not specified. The function considers any comma as a delimiter for distinct
16206 values. If this is not desired, the res.fhdr_cnt() fetch should be used
16207 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016208
Willy Tarreau74ca5042013-06-11 23:12:07 +020016209res.hdr_ip([<name>[,<occ>]]) : ip
16210shdr_ip([<name>[,<occ>]]) : ip (deprecated)
16211 This extracts the last occurrence of header <name> in an HTTP response,
16212 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
16213 specific occurrence might be specified as a position number. Positive values
16214 indicate a position from the first occurrence, with 1 being the first one.
16215 Negative values indicate positions relative to the last one, with -1 being
16216 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016217
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016218res.hdr_names([<delim>]) : string
16219 This builds a string made from the concatenation of all header names as they
16220 appear in the response when the rule is evaluated. The default delimiter is
16221 the comma (',') but it may be overridden as an optional argument <delim>. In
16222 this case, only the first character of <delim> is considered.
16223
Willy Tarreau74ca5042013-06-11 23:12:07 +020016224res.hdr_val([<name>[,<occ>]]) : integer
16225shdr_val([<name>[,<occ>]]) : integer (deprecated)
16226 This extracts the last occurrence of header <name> in an HTTP response, and
16227 converts it to an integer value. Optionally, a specific occurrence might be
16228 specified as a position number. Positive values indicate a position from the
16229 first occurrence, with 1 being the first one. Negative values indicate
16230 positions relative to the last one, with -1 being the last one. This can be
16231 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016232
Willy Tarreau74ca5042013-06-11 23:12:07 +020016233res.ver : string
16234resp_ver : string (deprecated)
16235 Returns the version string from the HTTP response, for example "1.1". This
16236 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020016237
Willy Tarreau74ca5042013-06-11 23:12:07 +020016238 ACL derivatives :
16239 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016240
Willy Tarreau74ca5042013-06-11 23:12:07 +020016241set-cookie([<name>]) : string (deprecated)
16242 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16243 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020016244 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020016245 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016246
Willy Tarreau74ca5042013-06-11 23:12:07 +020016247 This fetch function is deprecated and has been superseded by the "res.cook"
16248 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016249
Willy Tarreau74ca5042013-06-11 23:12:07 +020016250status : integer
16251 Returns an integer containing the HTTP status code in the HTTP response, for
16252 example, 302. It is mostly used within ACLs and integer ranges, for example,
16253 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016254
Thierry Fournier0e00dca2016-04-07 15:47:40 +020016255unique-id : string
16256 Returns the unique-id attached to the request. The directive
16257 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
16258 fails. Note that the unique-id is usually used with HTTP requests, however this
16259 sample fetch can be used with other protocols. Obviously, if it is used with
16260 other protocols than HTTP, the unique-id-format directive must not contain
16261 HTTP parts. See: unique-id-format and unique-id-header
16262
Willy Tarreau74ca5042013-06-11 23:12:07 +020016263url : string
16264 This extracts the request's URL as presented in the request. A typical use is
16265 with prefetch-capable caches, and with portals which need to aggregate
16266 multiple information from databases and keep them in caches. With ACLs, using
16267 "path" is preferred over using "url", because clients may send a full URL as
16268 is normally done with proxies. The only real use is to match "*" which does
16269 not match in "path", and for which there is already a predefined ACL. See
16270 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016271
Willy Tarreau74ca5042013-06-11 23:12:07 +020016272 ACL derivatives :
16273 url : exact string match
16274 url_beg : prefix match
16275 url_dir : subdir match
16276 url_dom : domain match
16277 url_end : suffix match
16278 url_len : length match
16279 url_reg : regex match
16280 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016281
Willy Tarreau74ca5042013-06-11 23:12:07 +020016282url_ip : ip
16283 This extracts the IP address from the request's URL when the host part is
16284 presented as an IP address. Its use is very limited. For instance, a
16285 monitoring system might use this field as an alternative for the source IP in
16286 order to test what path a given source address would follow, or to force an
16287 entry in a table for a given source address. With ACLs it can be used to
16288 restrict access to certain systems through a proxy, for example when combined
16289 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016290
Willy Tarreau74ca5042013-06-11 23:12:07 +020016291url_port : integer
16292 This extracts the port part from the request's URL. Note that if the port is
16293 not specified in the request, port 80 is assumed. With ACLs it can be used to
16294 restrict access to certain systems through a proxy, for example when combined
16295 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016296
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016297urlp([<name>[,<delim>]]) : string
16298url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020016299 This extracts the first occurrence of the parameter <name> in the query
16300 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016301 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
16302 any parameter will match, and the first one will be returned. The result is
16303 a string corresponding to the value of the parameter <name> as presented in
16304 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020016305 stickiness based on a client ID, to extract an application cookie passed as a
16306 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016307 this fetch iterates over multiple parameters and will iteratively report all
16308 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016309
Willy Tarreau74ca5042013-06-11 23:12:07 +020016310 ACL derivatives :
16311 urlp(<name>[,<delim>]) : exact string match
16312 urlp_beg(<name>[,<delim>]) : prefix match
16313 urlp_dir(<name>[,<delim>]) : subdir match
16314 urlp_dom(<name>[,<delim>]) : domain match
16315 urlp_end(<name>[,<delim>]) : suffix match
16316 urlp_len(<name>[,<delim>]) : length match
16317 urlp_reg(<name>[,<delim>]) : regex match
16318 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016319
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016320
Willy Tarreau74ca5042013-06-11 23:12:07 +020016321 Example :
16322 # match http://example.com/foo?PHPSESSIONID=some_id
16323 stick on urlp(PHPSESSIONID)
16324 # match http://example.com/foo;JSESSIONID=some_id
16325 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016326
Jarno Huuskonen676f6222017-03-30 09:19:45 +030016327urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020016328 See "urlp" above. This one extracts the URL parameter <name> in the request
16329 and converts it to an integer value. This can be used for session stickiness
16330 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020016331
Dragan Dosen0070cd52016-06-16 12:19:49 +020016332url32 : integer
16333 This returns a 32-bit hash of the value obtained by concatenating the first
16334 Host header and the whole URL including parameters (not only the path part of
16335 the request, as in the "base32" fetch above). This is useful to track per-URL
16336 activity. A shorter hash is stored, saving a lot of memory. The output type
16337 is an unsigned integer.
16338
16339url32+src : binary
16340 This returns the concatenation of the "url32" fetch and the "src" fetch. The
16341 resulting type is of type binary, with a size of 8 or 20 bytes depending on
16342 the source address family. This can be used to track per-IP, per-URL counters.
16343
Willy Tarreau198a7442008-01-17 12:05:32 +010016344
Willy Tarreau74ca5042013-06-11 23:12:07 +0200163457.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016346---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016347
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016348Some predefined ACLs are hard-coded so that they do not have to be declared in
16349every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020016350order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010016351
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016352ACL name Equivalent to Usage
16353---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016354FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020016355HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016356HTTP_1.0 req_ver 1.0 match HTTP version 1.0
16357HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010016358HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
16359HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
16360HTTP_URL_SLASH url_beg / match URL beginning with "/"
16361HTTP_URL_STAR url * match URL equal to "*"
16362LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016363METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016364METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016365METH_GET method GET HEAD match HTTP GET or HEAD method
16366METH_HEAD method HEAD match HTTP HEAD method
16367METH_OPTIONS method OPTIONS match HTTP OPTIONS method
16368METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016369METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016370METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020016371RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016372REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010016373TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016374WAIT_END wait_end wait for end of content analysis
16375---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016376
Willy Tarreaub937b7e2010-01-12 15:27:54 +010016377
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200163788. Logging
16379----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010016380
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016381One of HAProxy's strong points certainly lies is its precise logs. It probably
16382provides the finest level of information available for such a product, which is
16383very important for troubleshooting complex environments. Standard information
16384provided in logs include client ports, TCP/HTTP state timers, precise session
16385state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016386to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016387headers.
16388
16389In order to improve administrators reactivity, it offers a great transparency
16390about encountered problems, both internal and external, and it is possible to
16391send logs to different sources at the same time with different level filters :
16392
16393 - global process-level logs (system errors, start/stop, etc..)
16394 - per-instance system and internal errors (lack of resource, bugs, ...)
16395 - per-instance external troubles (servers up/down, max connections)
16396 - per-instance activity (client connections), either at the establishment or
16397 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010016398 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060016399 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016400
16401The ability to distribute different levels of logs to different log servers
16402allow several production teams to interact and to fix their problems as soon
16403as possible. For example, the system team might monitor system-wide errors,
16404while the application team might be monitoring the up/down for their servers in
16405real time, and the security team might analyze the activity logs with one hour
16406delay.
16407
16408
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164098.1. Log levels
16410---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016411
Simon Hormandf791f52011-05-29 15:01:10 +090016412TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016413source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090016414HTTP request, HTTP return code, number of bytes transmitted, conditions
16415in which the session ended, and even exchanged cookies values. For example
16416track a particular user's problems. All messages may be sent to up to two
16417syslog servers. Check the "log" keyword in section 4.2 for more information
16418about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016419
16420
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164218.2. Log formats
16422----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016423
William Lallemand48940402012-01-30 16:47:22 +010016424HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090016425and will be detailed in the following sections. A few of them may vary
16426slightly with the configuration, due to indicators specific to certain
16427options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016428
16429 - the default format, which is very basic and very rarely used. It only
16430 provides very basic information about the incoming connection at the moment
16431 it is accepted : source IP:port, destination IP:port, and frontend-name.
16432 This mode will eventually disappear so it will not be described to great
16433 extents.
16434
16435 - the TCP format, which is more advanced. This format is enabled when "option
16436 tcplog" is set on the frontend. HAProxy will then usually wait for the
16437 connection to terminate before logging. This format provides much richer
16438 information, such as timers, connection counts, queue size, etc... This
16439 format is recommended for pure TCP proxies.
16440
16441 - the HTTP format, which is the most advanced for HTTP proxying. This format
16442 is enabled when "option httplog" is set on the frontend. It provides the
16443 same information as the TCP format with some HTTP-specific fields such as
16444 the request, the status code, and captures of headers and cookies. This
16445 format is recommended for HTTP proxies.
16446
Emeric Brun3a058f32009-06-30 18:26:00 +020016447 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
16448 fields arranged in the same order as the CLF format. In this mode, all
16449 timers, captures, flags, etc... appear one per field after the end of the
16450 common fields, in the same order they appear in the standard HTTP format.
16451
William Lallemand48940402012-01-30 16:47:22 +010016452 - the custom log format, allows you to make your own log line.
16453
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016454Next sections will go deeper into details for each of these formats. Format
16455specification will be performed on a "field" basis. Unless stated otherwise, a
16456field is a portion of text delimited by any number of spaces. Since syslog
16457servers are susceptible of inserting fields at the beginning of a line, it is
16458always assumed that the first field is the one containing the process name and
16459identifier.
16460
16461Note : Since log lines may be quite long, the log examples in sections below
16462 might be broken into multiple lines. The example log lines will be
16463 prefixed with 3 closing angle brackets ('>>>') and each time a log is
16464 broken into multiple lines, each non-final line will end with a
16465 backslash ('\') and the next line will start indented by two characters.
16466
16467
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200164688.2.1. Default log format
16469-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016470
16471This format is used when no specific option is set. The log is emitted as soon
16472as the connection is accepted. One should note that this currently is the only
16473format which logs the request's destination IP and ports.
16474
16475 Example :
16476 listen www
16477 mode http
16478 log global
16479 server srv1 127.0.0.1:8000
16480
16481 >>> Feb 6 12:12:09 localhost \
16482 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
16483 (www/HTTP)
16484
16485 Field Format Extract from the example above
16486 1 process_name '[' pid ']:' haproxy[14385]:
16487 2 'Connect from' Connect from
16488 3 source_ip ':' source_port 10.0.1.2:33312
16489 4 'to' to
16490 5 destination_ip ':' destination_port 10.0.3.31:8012
16491 6 '(' frontend_name '/' mode ')' (www/HTTP)
16492
16493Detailed fields description :
16494 - "source_ip" is the IP address of the client which initiated the connection.
16495 - "source_port" is the TCP port of the client which initiated the connection.
16496 - "destination_ip" is the IP address the client connected to.
16497 - "destination_port" is the TCP port the client connected to.
16498 - "frontend_name" is the name of the frontend (or listener) which received
16499 and processed the connection.
16500 - "mode is the mode the frontend is operating (TCP or HTTP).
16501
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016502In case of a UNIX socket, the source and destination addresses are marked as
16503"unix:" and the ports reflect the internal ID of the socket which accepted the
16504connection (the same ID as reported in the stats).
16505
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016506It is advised not to use this deprecated format for newer installations as it
16507will eventually disappear.
16508
16509
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200165108.2.2. TCP log format
16511---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016512
16513The TCP format is used when "option tcplog" is specified in the frontend, and
16514is the recommended format for pure TCP proxies. It provides a lot of precious
16515information for troubleshooting. Since this format includes timers and byte
16516counts, the log is normally emitted at the end of the session. It can be
16517emitted earlier if "option logasap" is specified, which makes sense in most
16518environments with long sessions such as remote terminals. Sessions which match
16519the "monitor" rules are never logged. It is also possible not to emit logs for
16520sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016521specifying "option dontlognull" in the frontend. Successful connections will
16522not be logged if "option dontlog-normal" is specified in the frontend. A few
16523fields may slightly vary depending on some configuration options, those are
16524marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016525
16526 Example :
16527 frontend fnt
16528 mode tcp
16529 option tcplog
16530 log global
16531 default_backend bck
16532
16533 backend bck
16534 server srv1 127.0.0.1:8000
16535
16536 >>> Feb 6 12:12:56 localhost \
16537 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
16538 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
16539
16540 Field Format Extract from the example above
16541 1 process_name '[' pid ']:' haproxy[14387]:
16542 2 client_ip ':' client_port 10.0.1.2:33313
16543 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
16544 4 frontend_name fnt
16545 5 backend_name '/' server_name bck/srv1
16546 6 Tw '/' Tc '/' Tt* 0/0/5007
16547 7 bytes_read* 212
16548 8 termination_state --
16549 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
16550 10 srv_queue '/' backend_queue 0/0
16551
16552Detailed fields description :
16553 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016554 connection to haproxy. If the connection was accepted on a UNIX socket
16555 instead, the IP address would be replaced with the word "unix". Note that
16556 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016557 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016558 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016559 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016560
16561 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016562 If the connection was accepted on a UNIX socket instead, the port would be
16563 replaced with the ID of the accepting socket, which is also reported in the
16564 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016565
16566 - "accept_date" is the exact date when the connection was received by haproxy
16567 (which might be very slightly different from the date observed on the
16568 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020016569 the same date which may appear in any upstream firewall's log. When used in
16570 HTTP mode, the accept_date field will be reset to the first moment the
16571 connection is ready to receive a new request (end of previous response for
16572 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016573
16574 - "frontend_name" is the name of the frontend (or listener) which received
16575 and processed the connection.
16576
16577 - "backend_name" is the name of the backend (or listener) which was selected
16578 to manage the connection to the server. This will be the same as the
16579 frontend if no switching rule has been applied, which is common for TCP
16580 applications.
16581
16582 - "server_name" is the name of the last server to which the connection was
16583 sent, which might differ from the first one if there were connection errors
16584 and a redispatch occurred. Note that this server belongs to the backend
16585 which processed the request. If the connection was aborted before reaching
16586 a server, "<NOSRV>" is indicated instead of a server name.
16587
16588 - "Tw" is the total time in milliseconds spent waiting in the various queues.
16589 It can be "-1" if the connection was aborted before reaching the queue.
16590 See "Timers" below for more details.
16591
16592 - "Tc" is the total time in milliseconds spent waiting for the connection to
16593 establish to the final server, including retries. It can be "-1" if the
16594 connection was aborted before a connection could be established. See
16595 "Timers" below for more details.
16596
16597 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016598 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016599 "option logasap" was specified, then the time counting stops at the moment
16600 the log is emitted. In this case, a '+' sign is prepended before the value,
16601 indicating that the final one will be larger. See "Timers" below for more
16602 details.
16603
16604 - "bytes_read" is the total number of bytes transmitted from the server to
16605 the client when the log is emitted. If "option logasap" is specified, the
16606 this value will be prefixed with a '+' sign indicating that the final one
16607 may be larger. Please note that this value is a 64-bit counter, so log
16608 analysis tools must be able to handle it without overflowing.
16609
16610 - "termination_state" is the condition the session was in when the session
16611 ended. This indicates the session state, which side caused the end of
16612 session to happen, and for what reason (timeout, error, ...). The normal
16613 flags should be "--", indicating the session was closed by either end with
16614 no data remaining in buffers. See below "Session state at disconnection"
16615 for more details.
16616
16617 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016618 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016619 limits have been reached. For instance, if actconn is close to 512 when
16620 multiple connection errors occur, chances are high that the system limits
16621 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016622 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016623
16624 - "feconn" is the total number of concurrent connections on the frontend when
16625 the session was logged. It is useful to estimate the amount of resource
16626 required to sustain high loads, and to detect when the frontend's "maxconn"
16627 has been reached. Most often when this value increases by huge jumps, it is
16628 because there is congestion on the backend servers, but sometimes it can be
16629 caused by a denial of service attack.
16630
16631 - "beconn" is the total number of concurrent connections handled by the
16632 backend when the session was logged. It includes the total number of
16633 concurrent connections active on servers as well as the number of
16634 connections pending in queues. It is useful to estimate the amount of
16635 additional servers needed to support high loads for a given application.
16636 Most often when this value increases by huge jumps, it is because there is
16637 congestion on the backend servers, but sometimes it can be caused by a
16638 denial of service attack.
16639
16640 - "srv_conn" is the total number of concurrent connections still active on
16641 the server when the session was logged. It can never exceed the server's
16642 configured "maxconn" parameter. If this value is very often close or equal
16643 to the server's "maxconn", it means that traffic regulation is involved a
16644 lot, meaning that either the server's maxconn value is too low, or that
16645 there aren't enough servers to process the load with an optimal response
16646 time. When only one of the server's "srv_conn" is high, it usually means
16647 that this server has some trouble causing the connections to take longer to
16648 be processed than on other servers.
16649
16650 - "retries" is the number of connection retries experienced by this session
16651 when trying to connect to the server. It must normally be zero, unless a
16652 server is being stopped at the same moment the connection was attempted.
16653 Frequent retries generally indicate either a network problem between
16654 haproxy and the server, or a misconfigured system backlog on the server
16655 preventing new connections from being queued. This field may optionally be
16656 prefixed with a '+' sign, indicating that the session has experienced a
16657 redispatch after the maximal retry count has been reached on the initial
16658 server. In this case, the server name appearing in the log is the one the
16659 connection was redispatched to, and not the first one, though both may
16660 sometimes be the same in case of hashing for instance. So as a general rule
16661 of thumb, when a '+' is present in front of the retry count, this count
16662 should not be attributed to the logged server.
16663
16664 - "srv_queue" is the total number of requests which were processed before
16665 this one in the server queue. It is zero when the request has not gone
16666 through the server queue. It makes it possible to estimate the approximate
16667 server's response time by dividing the time spent in queue by the number of
16668 requests in the queue. It is worth noting that if a session experiences a
16669 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016670 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016671 backend queue unless a redispatch occurs.
16672
16673 - "backend_queue" is the total number of requests which were processed before
16674 this one in the backend's global queue. It is zero when the request has not
16675 gone through the global queue. It makes it possible to estimate the average
16676 queue length, which easily translates into a number of missing servers when
16677 divided by a server's "maxconn" parameter. It is worth noting that if a
16678 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010016679 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016680 through both the server queue and the backend queue unless a redispatch
16681 occurs.
16682
16683
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166848.2.3. HTTP log format
16685----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016686
16687The HTTP format is the most complete and the best suited for HTTP proxies. It
16688is enabled by when "option httplog" is specified in the frontend. It provides
16689the same level of information as the TCP format with additional features which
16690are specific to the HTTP protocol. Just like the TCP format, the log is usually
16691emitted at the end of the session, unless "option logasap" is specified, which
16692generally only makes sense for download sites. A session which matches the
16693"monitor" rules will never logged. It is also possible not to log sessions for
16694which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016695frontend. Successful connections will not be logged if "option dontlog-normal"
16696is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016697
16698Most fields are shared with the TCP log, some being different. A few fields may
16699slightly vary depending on some configuration options. Those ones are marked
16700with a star ('*') after the field name below.
16701
16702 Example :
16703 frontend http-in
16704 mode http
16705 option httplog
16706 log global
16707 default_backend bck
16708
16709 backend static
16710 server srv1 127.0.0.1:8000
16711
16712 >>> Feb 6 12:14:14 localhost \
16713 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
16714 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 +010016715 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016716
16717 Field Format Extract from the example above
16718 1 process_name '[' pid ']:' haproxy[14389]:
16719 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016720 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016721 4 frontend_name http-in
16722 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016723 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016724 7 status_code 200
16725 8 bytes_read* 2750
16726 9 captured_request_cookie -
16727 10 captured_response_cookie -
16728 11 termination_state ----
16729 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
16730 13 srv_queue '/' backend_queue 0/0
16731 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
16732 15 '{' captured_response_headers* '}' {}
16733 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010016734
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016735Detailed fields description :
16736 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016737 connection to haproxy. If the connection was accepted on a UNIX socket
16738 instead, the IP address would be replaced with the word "unix". Note that
16739 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016740 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016741 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016742 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016743
16744 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016745 If the connection was accepted on a UNIX socket instead, the port would be
16746 replaced with the ID of the accepting socket, which is also reported in the
16747 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016748
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016749 - "request_date" is the exact date when the first byte of the HTTP request
16750 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016751
16752 - "frontend_name" is the name of the frontend (or listener) which received
16753 and processed the connection.
16754
16755 - "backend_name" is the name of the backend (or listener) which was selected
16756 to manage the connection to the server. This will be the same as the
16757 frontend if no switching rule has been applied.
16758
16759 - "server_name" is the name of the last server to which the connection was
16760 sent, which might differ from the first one if there were connection errors
16761 and a redispatch occurred. Note that this server belongs to the backend
16762 which processed the request. If the request was aborted before reaching a
16763 server, "<NOSRV>" is indicated instead of a server name. If the request was
16764 intercepted by the stats subsystem, "<STATS>" is indicated instead.
16765
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016766 - "TR" is the total time in milliseconds spent waiting for a full HTTP
16767 request from the client (not counting body) after the first byte was
16768 received. It can be "-1" if the connection was aborted before a complete
John Roeslerfb2fce12019-07-10 15:45:51 -050016769 request could be received or a bad request was received. It should
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016770 always be very small because a request generally fits in one single packet.
16771 Large times here generally indicate network issues between the client and
Willy Tarreau590a0512018-09-05 11:56:48 +020016772 haproxy or requests being typed by hand. See section 8.4 "Timing Events"
16773 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016774
16775 - "Tw" is the total time in milliseconds spent waiting in the various queues.
16776 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020016777 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016778
16779 - "Tc" is the total time in milliseconds spent waiting for the connection to
16780 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020016781 request was aborted before a connection could be established. See section
16782 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016783
16784 - "Tr" is the total time in milliseconds spent waiting for the server to send
16785 a full HTTP response, not counting data. It can be "-1" if the request was
16786 aborted before a complete response could be received. It generally matches
16787 the server's processing time for the request, though it may be altered by
16788 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020016789 "GET" requests generally indicate an overloaded server. See section 8.4
16790 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016791
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016792 - "Ta" is the time the request remained active in haproxy, which is the total
16793 time in milliseconds elapsed between the first byte of the request was
16794 received and the last byte of response was sent. It covers all possible
16795 processing except the handshake (see Th) and idle time (see Ti). There is
16796 one exception, if "option logasap" was specified, then the time counting
16797 stops at the moment the log is emitted. In this case, a '+' sign is
16798 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020016799 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016800
16801 - "status_code" is the HTTP status code returned to the client. This status
16802 is generally set by the server, but it might also be set by haproxy when
16803 the server cannot be reached or when its response is blocked by haproxy.
16804
16805 - "bytes_read" is the total number of bytes transmitted to the client when
16806 the log is emitted. This does include HTTP headers. If "option logasap" is
John Roeslerfb2fce12019-07-10 15:45:51 -050016807 specified, this value will be prefixed with a '+' sign indicating that
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016808 the final one may be larger. Please note that this value is a 64-bit
16809 counter, so log analysis tools must be able to handle it without
16810 overflowing.
16811
16812 - "captured_request_cookie" is an optional "name=value" entry indicating that
16813 the client had this cookie in the request. The cookie name and its maximum
16814 length are defined by the "capture cookie" statement in the frontend
16815 configuration. The field is a single dash ('-') when the option is not
16816 set. Only one cookie may be captured, it is generally used to track session
16817 ID exchanges between a client and a server to detect session crossing
16818 between clients due to application bugs. For more details, please consult
16819 the section "Capturing HTTP headers and cookies" below.
16820
16821 - "captured_response_cookie" is an optional "name=value" entry indicating
16822 that the server has returned a cookie with its response. The cookie name
16823 and its maximum length are defined by the "capture cookie" statement in the
16824 frontend configuration. The field is a single dash ('-') when the option is
16825 not set. Only one cookie may be captured, it is generally used to track
16826 session ID exchanges between a client and a server to detect session
16827 crossing between clients due to application bugs. For more details, please
16828 consult the section "Capturing HTTP headers and cookies" below.
16829
16830 - "termination_state" is the condition the session was in when the session
16831 ended. This indicates the session state, which side caused the end of
16832 session to happen, for what reason (timeout, error, ...), just like in TCP
16833 logs, and information about persistence operations on cookies in the last
16834 two characters. The normal flags should begin with "--", indicating the
16835 session was closed by either end with no data remaining in buffers. See
16836 below "Session state at disconnection" for more details.
16837
16838 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016839 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016840 limits have been reached. For instance, if actconn is close to 512 or 1024
16841 when multiple connection errors occur, chances are high that the system
16842 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016843 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016844 system.
16845
16846 - "feconn" is the total number of concurrent connections on the frontend when
16847 the session was logged. It is useful to estimate the amount of resource
16848 required to sustain high loads, and to detect when the frontend's "maxconn"
16849 has been reached. Most often when this value increases by huge jumps, it is
16850 because there is congestion on the backend servers, but sometimes it can be
16851 caused by a denial of service attack.
16852
16853 - "beconn" is the total number of concurrent connections handled by the
16854 backend when the session was logged. It includes the total number of
16855 concurrent connections active on servers as well as the number of
16856 connections pending in queues. It is useful to estimate the amount of
16857 additional servers needed to support high loads for a given application.
16858 Most often when this value increases by huge jumps, it is because there is
16859 congestion on the backend servers, but sometimes it can be caused by a
16860 denial of service attack.
16861
16862 - "srv_conn" is the total number of concurrent connections still active on
16863 the server when the session was logged. It can never exceed the server's
16864 configured "maxconn" parameter. If this value is very often close or equal
16865 to the server's "maxconn", it means that traffic regulation is involved a
16866 lot, meaning that either the server's maxconn value is too low, or that
16867 there aren't enough servers to process the load with an optimal response
16868 time. When only one of the server's "srv_conn" is high, it usually means
16869 that this server has some trouble causing the requests to take longer to be
16870 processed than on other servers.
16871
16872 - "retries" is the number of connection retries experienced by this session
16873 when trying to connect to the server. It must normally be zero, unless a
16874 server is being stopped at the same moment the connection was attempted.
16875 Frequent retries generally indicate either a network problem between
16876 haproxy and the server, or a misconfigured system backlog on the server
16877 preventing new connections from being queued. This field may optionally be
16878 prefixed with a '+' sign, indicating that the session has experienced a
16879 redispatch after the maximal retry count has been reached on the initial
16880 server. In this case, the server name appearing in the log is the one the
16881 connection was redispatched to, and not the first one, though both may
16882 sometimes be the same in case of hashing for instance. So as a general rule
16883 of thumb, when a '+' is present in front of the retry count, this count
16884 should not be attributed to the logged server.
16885
16886 - "srv_queue" is the total number of requests which were processed before
16887 this one in the server queue. It is zero when the request has not gone
16888 through the server queue. It makes it possible to estimate the approximate
16889 server's response time by dividing the time spent in queue by the number of
16890 requests in the queue. It is worth noting that if a session experiences a
16891 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016892 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016893 backend queue unless a redispatch occurs.
16894
16895 - "backend_queue" is the total number of requests which were processed before
16896 this one in the backend's global queue. It is zero when the request has not
16897 gone through the global queue. It makes it possible to estimate the average
16898 queue length, which easily translates into a number of missing servers when
16899 divided by a server's "maxconn" parameter. It is worth noting that if a
16900 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010016901 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016902 through both the server queue and the backend queue unless a redispatch
16903 occurs.
16904
16905 - "captured_request_headers" is a list of headers captured in the request due
16906 to the presence of the "capture request header" statement in the frontend.
16907 Multiple headers can be captured, they will be delimited by a vertical bar
16908 ('|'). When no capture is enabled, the braces do not appear, causing a
16909 shift of remaining fields. It is important to note that this field may
16910 contain spaces, and that using it requires a smarter log parser than when
16911 it's not used. Please consult the section "Capturing HTTP headers and
16912 cookies" below for more details.
16913
16914 - "captured_response_headers" is a list of headers captured in the response
16915 due to the presence of the "capture response header" statement in the
16916 frontend. Multiple headers can be captured, they will be delimited by a
16917 vertical bar ('|'). When no capture is enabled, the braces do not appear,
16918 causing a shift of remaining fields. It is important to note that this
16919 field may contain spaces, and that using it requires a smarter log parser
16920 than when it's not used. Please consult the section "Capturing HTTP headers
16921 and cookies" below for more details.
16922
16923 - "http_request" is the complete HTTP request line, including the method,
16924 request and HTTP version string. Non-printable characters are encoded (see
16925 below the section "Non-printable characters"). This is always the last
16926 field, and it is always delimited by quotes and is the only one which can
16927 contain quotes. If new fields are added to the log format, they will be
16928 added before this field. This field might be truncated if the request is
16929 huge and does not fit in the standard syslog buffer (1024 characters). This
16930 is the reason why this field must always remain the last one.
16931
16932
Cyril Bontédc4d9032012-04-08 21:57:39 +0200169338.2.4. Custom log format
16934------------------------
William Lallemand48940402012-01-30 16:47:22 +010016935
Willy Tarreau2beef582012-12-20 17:22:52 +010016936The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010016937mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010016938
Davor Ocelice9ed2812017-12-25 17:49:28 +010016939HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010016940Variables can take arguments using braces ('{}'), and multiple arguments are
16941separated by commas within the braces. Flags may be added or removed by
16942prefixing them with a '+' or '-' sign.
16943
16944Special variable "%o" may be used to propagate its flags to all other
16945variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010016946("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010016947
Willy Tarreauc8368452012-12-21 00:09:23 +010016948If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020016949as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010016950less common information such as the client's SSL certificate's DN, or to log
16951the key that would be used to store an entry into a stick table.
16952
William Lallemand48940402012-01-30 16:47:22 +010016953Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016954In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010016955in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010016956
Dragan Dosen835b9212016-02-12 13:23:03 +010016957Note: when using the RFC5424 syslog message format, the characters '"',
16958'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
16959https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
16960such cases, the use of the flag "E" should be considered.
16961
William Lallemand48940402012-01-30 16:47:22 +010016962Flags are :
16963 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016964 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010016965 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
16966 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010016967
16968 Example:
16969
16970 log-format %T\ %t\ Some\ Text
16971 log-format %{+Q}o\ %t\ %s\ %{-Q}r
16972
Dragan Dosen835b9212016-02-12 13:23:03 +010016973 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
16974
William Lallemand48940402012-01-30 16:47:22 +010016975At the moment, the default HTTP format is defined this way :
16976
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016977 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
16978 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010016979
William Lallemandbddd4fd2012-02-27 11:23:10 +010016980the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010016981
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016982 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
16983 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
16984 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010016985
William Lallemandbddd4fd2012-02-27 11:23:10 +010016986and the default TCP format is defined this way :
16987
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016988 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
16989 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010016990
William Lallemand48940402012-01-30 16:47:22 +010016991Please refer to the table below for currently defined variables :
16992
William Lallemandbddd4fd2012-02-27 11:23:10 +010016993 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020016994 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010016995 +---+------+-----------------------------------------------+-------------+
16996 | | %o | special variable, apply flags on all next var | |
16997 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010016998 | | %B | bytes_read (from server to client) | numeric |
16999 | H | %CC | captured_request_cookie | string |
17000 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020017001 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017002 | H | %HM | HTTP method (ex: POST) | string |
17003 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000017004 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017005 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
17006 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010017007 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020017008 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017009 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017010 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017011 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020017012 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080017013 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017014 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
17015 | H | %Ti | idle time before the HTTP request | numeric |
17016 | H | %Tq | Th + Ti + TR | numeric |
17017 | H | %TR | time to receive the full request from 1st byte| numeric |
17018 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017019 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017020 | | %Tt | Tt | numeric |
17021 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017022 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017023 | | %ac | actconn | numeric |
17024 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017025 | | %bc | beconn (backend concurrent connections) | numeric |
17026 | | %bi | backend_source_ip (connecting address) | IP |
17027 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017028 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017029 | | %ci | client_ip (accepted address) | IP |
17030 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017031 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017032 | | %fc | feconn (frontend concurrent connections) | numeric |
17033 | | %fi | frontend_ip (accepting address) | IP |
17034 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020017035 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020017036 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017037 | | %hr | captured_request_headers default style | string |
17038 | | %hrl | captured_request_headers CLF style | string list |
17039 | | %hs | captured_response_headers default style | string |
17040 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020017041 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017042 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017043 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017044 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010017045 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017046 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017047 | | %sc | srv_conn (server concurrent connections) | numeric |
17048 | | %si | server_IP (target address) | IP |
17049 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017050 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017051 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
17052 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017053 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017054 | H | %tr | date_time of HTTP request | date |
17055 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020017056 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017057 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017058 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017059 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010017060
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017061 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010017062
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017063
170648.2.5. Error log format
17065-----------------------
17066
17067When an incoming connection fails due to an SSL handshake or an invalid PROXY
17068protocol header, haproxy will log the event using a shorter, fixed line format.
17069By default, logs are emitted at the LOG_INFO level, unless the option
17070"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017071will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017072logged if the "dontlognull" option is set.
17073
17074The format looks like this :
17075
17076 >>> Dec 3 18:27:14 localhost \
17077 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
17078 Connection error during SSL handshake
17079
17080 Field Format Extract from the example above
17081 1 process_name '[' pid ']:' haproxy[6103]:
17082 2 client_ip ':' client_port 127.0.0.1:56059
17083 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
17084 4 frontend_name "/" bind_name ":" frt/f1:
17085 5 message Connection error during SSL handshake
17086
17087These fields just provide minimal information to help debugging connection
17088failures.
17089
17090
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170918.3. Advanced logging options
17092-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017093
17094Some advanced logging options are often looked for but are not easy to find out
17095just by looking at the various options. Here is an entry point for the few
17096options which can enable better logging. Please refer to the keywords reference
17097for more information about their usage.
17098
17099
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171008.3.1. Disabling logging of external tests
17101------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017102
17103It is quite common to have some monitoring tools perform health checks on
17104haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
17105commercial load-balancer, and sometimes it will simply be a more complete
17106monitoring system such as Nagios. When the tests are very frequent, users often
17107ask how to disable logging for those checks. There are three possibilities :
17108
17109 - if connections come from everywhere and are just TCP probes, it is often
17110 desired to simply disable logging of connections without data exchange, by
17111 setting "option dontlognull" in the frontend. It also disables logging of
17112 port scans, which may or may not be desired.
17113
17114 - if the connection come from a known source network, use "monitor-net" to
17115 declare this network as monitoring only. Any host in this network will then
17116 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017117 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017118 such as other load-balancers.
17119
17120 - if the tests are performed on a known URI, use "monitor-uri" to declare
17121 this URI as dedicated to monitoring. Any host sending this request will
17122 only get the result of a health-check, and the request will not be logged.
17123
17124
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171258.3.2. Logging before waiting for the session to terminate
17126----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017127
17128The problem with logging at end of connection is that you have no clue about
17129what is happening during very long sessions, such as remote terminal sessions
17130or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010017131"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017132just before data transfer begins. This means that in case of TCP, it will still
17133log the connection status to the server, and in case of HTTP, it will log just
17134after processing the server headers. In this case, the number of bytes reported
17135is the number of header bytes sent to the client. In order to avoid confusion
17136with normal logs, the total time field and the number of bytes are prefixed
17137with a '+' sign which means that real numbers are certainly larger.
17138
17139
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171408.3.3. Raising log level upon errors
17141------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017142
17143Sometimes it is more convenient to separate normal traffic from errors logs,
17144for instance in order to ease error monitoring from log files. When the option
17145"log-separate-errors" is used, connections which experience errors, timeouts,
17146retries, redispatches or HTTP status codes 5xx will see their syslog level
17147raised from "info" to "err". This will help a syslog daemon store the log in
17148a separate file. It is very important to keep the errors in the normal traffic
17149file too, so that log ordering is not altered. You should also be careful if
17150you already have configured your syslog daemon to store all logs higher than
17151"notice" in an "admin" file, because the "err" level is higher than "notice".
17152
17153
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171548.3.4. Disabling logging of successful connections
17155--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017156
17157Although this may sound strange at first, some large sites have to deal with
17158multiple thousands of logs per second and are experiencing difficulties keeping
17159them intact for a long time or detecting errors within them. If the option
17160"dontlog-normal" is set on the frontend, all normal connections will not be
17161logged. In this regard, a normal connection is defined as one without any
17162error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
17163and a response with a status 5xx is not considered normal and will be logged
17164too. Of course, doing is is really discouraged as it will remove most of the
17165useful information from the logs. Do this only if you have no other
17166alternative.
17167
17168
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171698.4. Timing events
17170------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017171
17172Timers provide a great help in troubleshooting network problems. All values are
17173reported in milliseconds (ms). These timers should be used in conjunction with
17174the session termination flags. In TCP mode with "option tcplog" set on the
17175frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017176mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
17177addition, three other measures are provided, "Th", "Ti", and "Tq".
17178
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010017179Timings events in HTTP mode:
17180
17181 first request 2nd request
17182 |<-------------------------------->|<-------------- ...
17183 t tr t tr ...
17184 ---|----|----|----|----|----|----|----|----|--
17185 : Th Ti TR Tw Tc Tr Td : Ti ...
17186 :<---- Tq ---->: :
17187 :<-------------- Tt -------------->:
17188 :<--------- Ta --------->:
17189
17190Timings events in TCP mode:
17191
17192 TCP session
17193 |<----------------->|
17194 t t
17195 ---|----|----|----|----|---
17196 | Th Tw Tc Td |
17197 |<------ Tt ------->|
17198
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017199 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017200 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017201 only happen once during the whole connection's lifetime. A large time here
17202 may indicate that the client only pre-established the connection without
17203 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010017204 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020017205 SSL handshake was very expensive to compute. Please note that this time is
17206 reported only before the first request, so it is safe to average it over
17207 all request to calculate the amortized value. The second and subsequent
17208 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017209
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017210 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
17211 counts between the end of the handshakes and the first byte of the HTTP
17212 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020017213 to count after the end of the transmission the previous response. When a
17214 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
17215 after the previous request. Some browsers pre-establish connections to a
17216 server in order to reduce the latency of a future request, and keep them
17217 pending until they need it. This delay will be reported as the idle time. A
17218 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017219
17220 - TR: total time to get the client request (HTTP mode only). It's the time
17221 elapsed between the first bytes received and the moment the proxy received
17222 the empty line marking the end of the HTTP headers. The value "-1"
17223 indicates that the end of headers has never been seen. This happens when
17224 the client closes prematurely or times out. This time is usually very short
17225 since most requests fit in a single packet. A large time may indicate a
17226 request typed by hand during a test.
17227
17228 - Tq: total time to get the client request from the accept date or since the
17229 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010017230 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 +020017231 returns -1 as well. This timer used to be very useful before the arrival of
17232 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
17233 it in favor of TR nowadays, as the idle time adds a lot of noise to the
17234 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017235
17236 - Tw: total time spent in the queues waiting for a connection slot. It
17237 accounts for backend queue as well as the server queues, and depends on the
17238 queue size, and the time needed for the server to complete previous
17239 requests. The value "-1" means that the request was killed before reaching
17240 the queue, which is generally what happens with invalid or denied requests.
17241
17242 - Tc: total time to establish the TCP connection to the server. It's the time
17243 elapsed between the moment the proxy sent the connection request, and the
17244 moment it was acknowledged by the server, or between the TCP SYN packet and
17245 the matching SYN/ACK packet in return. The value "-1" means that the
17246 connection never established.
17247
17248 - Tr: server response time (HTTP mode only). It's the time elapsed between
17249 the moment the TCP connection was established to the server and the moment
17250 the server sent its complete response headers. It purely shows its request
17251 processing time, without the network overhead due to the data transmission.
17252 It is worth noting that when the client has data to send to the server, for
17253 instance during a POST request, the time already runs, and this can distort
17254 apparent response time. For this reason, it's generally wise not to trust
17255 too much this field for POST requests initiated from clients behind an
17256 untrusted network. A value of "-1" here means that the last the response
17257 header (empty line) was never seen, most likely because the server timeout
17258 stroke before the server managed to process the request.
17259
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017260 - Ta: total active time for the HTTP request, between the moment the proxy
17261 received the first byte of the request header and the emission of the last
17262 byte of the response body. The exception is when the "logasap" option is
17263 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
17264 a '+' sign. From this field, we can deduce "Td", the data transmission time,
17265 by subtracting other timers when valid :
17266
17267 Td = Ta - (TR + Tw + Tc + Tr)
17268
17269 Timers with "-1" values have to be excluded from this equation. Note that
17270 "Ta" can never be negative.
17271
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017272 - Tt: total session duration time, between the moment the proxy accepted it
17273 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017274 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
17275 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017276 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017277
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017278 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017279
17280 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017281 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
17282 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017283
17284These timers provide precious indications on trouble causes. Since the TCP
17285protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
17286that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017287due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
17288"Tt" is close to a timeout value specified in the configuration, it often means
17289that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017290
17291Most common cases :
17292
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017293 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
17294 the client and the proxy. This is very rare on local networks but might
17295 happen when clients are on far remote networks and send large requests. It
17296 may happen that values larger than usual appear here without any network
17297 cause. Sometimes, during an attack or just after a resource starvation has
17298 ended, haproxy may accept thousands of connections in a few milliseconds.
17299 The time spent accepting these connections will inevitably slightly delay
17300 processing of other connections, and it can happen that request times in the
17301 order of a few tens of milliseconds are measured after a few thousands of
17302 new connections have been accepted at once. Using one of the keep-alive
17303 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020017304 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017305
17306 - If "Tc" is close to 3000, a packet has probably been lost between the
17307 server and the proxy during the server connection phase. This value should
17308 always be very low, such as 1 ms on local networks and less than a few tens
17309 of ms on remote networks.
17310
Willy Tarreau55165fe2009-05-10 12:02:55 +020017311 - If "Tr" is nearly always lower than 3000 except some rare values which seem
17312 to be the average majored by 3000, there are probably some packets lost
17313 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017314
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017315 - If "Ta" is large even for small byte counts, it generally is because
17316 neither the client nor the server decides to close the connection while
17317 haproxy is running in tunnel mode and both have agreed on a keep-alive
17318 connection mode. In order to solve this issue, it will be needed to specify
17319 one of the HTTP options to manipulate keep-alive or close options on either
17320 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
17321 important when connection regulation is used with the "maxconn" option on
17322 the servers, since no new connection will be sent to the server until
17323 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017324
17325Other noticeable HTTP log cases ('xx' means any value to be ignored) :
17326
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017327 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017328 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017329 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017330
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017331 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017332 or it aborted too early. Check the session termination flags
17333 then "timeout http-request" and "timeout client" settings.
17334
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017335 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017336 servers were out of order, because the request was invalid
17337 or forbidden by ACL rules. Check the session termination
17338 flags.
17339
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017340 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
17341 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017342 Check the session termination flags, then check the
17343 "timeout connect" setting. Note that the tarpit action might
17344 return similar-looking patterns, with "Tw" equal to the time
17345 the client connection was maintained open.
17346
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017347 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017348 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017349 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017350 termination flags, then check the "timeout server" setting.
17351
17352
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173538.5. Session state at disconnection
17354-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017355
17356TCP and HTTP logs provide a session termination indicator in the
17357"termination_state" field, just before the number of active connections. It is
173582-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
17359each of which has a special meaning :
17360
17361 - On the first character, a code reporting the first event which caused the
17362 session to terminate :
17363
17364 C : the TCP session was unexpectedly aborted by the client.
17365
17366 S : the TCP session was unexpectedly aborted by the server, or the
17367 server explicitly refused it.
17368
17369 P : the session was prematurely aborted by the proxy, because of a
17370 connection limit enforcement, because a DENY filter was matched,
17371 because of a security check which detected and blocked a dangerous
17372 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010017373 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020017374
17375 L : the session was locally processed by haproxy and was not passed to
17376 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017377
17378 R : a resource on the proxy has been exhausted (memory, sockets, source
17379 ports, ...). Usually, this appears during the connection phase, and
17380 system logs should contain a copy of the precise error. If this
17381 happens, it must be considered as a very serious anomaly which
17382 should be fixed as soon as possible by any means.
17383
17384 I : an internal error was identified by the proxy during a self-check.
17385 This should NEVER happen, and you are encouraged to report any log
17386 containing this, because this would almost certainly be a bug. It
17387 would be wise to preventively restart the process after such an
17388 event too, in case it would be caused by memory corruption.
17389
Simon Horman752dc4a2011-06-21 14:34:59 +090017390 D : the session was killed by haproxy because the server was detected
17391 as down and was configured to kill all connections when going down.
17392
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070017393 U : the session was killed by haproxy on this backup server because an
17394 active server was detected as up and was configured to kill all
17395 backup connections when going up.
17396
Willy Tarreaua2a64e92011-09-07 23:01:56 +020017397 K : the session was actively killed by an admin operating on haproxy.
17398
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017399 c : the client-side timeout expired while waiting for the client to
17400 send or receive data.
17401
17402 s : the server-side timeout expired while waiting for the server to
17403 send or receive data.
17404
17405 - : normal session completion, both the client and the server closed
17406 with nothing left in the buffers.
17407
17408 - on the second character, the TCP or HTTP session state when it was closed :
17409
Willy Tarreauf7b30a92010-12-06 22:59:17 +010017410 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017411 (HTTP mode only). Nothing was sent to any server.
17412
17413 Q : the proxy was waiting in the QUEUE for a connection slot. This can
17414 only happen when servers have a 'maxconn' parameter set. It can
17415 also happen in the global queue after a redispatch consecutive to
17416 a failed attempt to connect to a dying server. If no redispatch is
17417 reported, then no connection attempt was made to any server.
17418
17419 C : the proxy was waiting for the CONNECTION to establish on the
17420 server. The server might at most have noticed a connection attempt.
17421
17422 H : the proxy was waiting for complete, valid response HEADERS from the
17423 server (HTTP only).
17424
17425 D : the session was in the DATA phase.
17426
17427 L : the proxy was still transmitting LAST data to the client while the
17428 server had already finished. This one is very rare as it can only
17429 happen when the client dies while receiving the last packets.
17430
17431 T : the request was tarpitted. It has been held open with the client
17432 during the whole "timeout tarpit" duration or until the client
17433 closed, both of which will be reported in the "Tw" timer.
17434
17435 - : normal session completion after end of data transfer.
17436
17437 - the third character tells whether the persistence cookie was provided by
17438 the client (only in HTTP mode) :
17439
17440 N : the client provided NO cookie. This is usually the case for new
17441 visitors, so counting the number of occurrences of this flag in the
17442 logs generally indicate a valid trend for the site frequentation.
17443
17444 I : the client provided an INVALID cookie matching no known server.
17445 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020017446 cookies between HTTP/HTTPS sites, persistence conditionally
17447 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017448
17449 D : the client provided a cookie designating a server which was DOWN,
17450 so either "option persist" was used and the client was sent to
17451 this server, or it was not set and the client was redispatched to
17452 another server.
17453
Willy Tarreau996a92c2010-10-13 19:30:47 +020017454 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017455 server.
17456
Willy Tarreau996a92c2010-10-13 19:30:47 +020017457 E : the client provided a valid cookie, but with a last date which was
17458 older than what is allowed by the "maxidle" cookie parameter, so
17459 the cookie is consider EXPIRED and is ignored. The request will be
17460 redispatched just as if there was no cookie.
17461
17462 O : the client provided a valid cookie, but with a first date which was
17463 older than what is allowed by the "maxlife" cookie parameter, so
17464 the cookie is consider too OLD and is ignored. The request will be
17465 redispatched just as if there was no cookie.
17466
Willy Tarreauc89ccb62012-04-05 21:18:22 +020017467 U : a cookie was present but was not used to select the server because
17468 some other server selection mechanism was used instead (typically a
17469 "use-server" rule).
17470
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017471 - : does not apply (no cookie set in configuration).
17472
17473 - the last character reports what operations were performed on the persistence
17474 cookie returned by the server (only in HTTP mode) :
17475
17476 N : NO cookie was provided by the server, and none was inserted either.
17477
17478 I : no cookie was provided by the server, and the proxy INSERTED one.
17479 Note that in "cookie insert" mode, if the server provides a cookie,
17480 it will still be overwritten and reported as "I" here.
17481
Willy Tarreau996a92c2010-10-13 19:30:47 +020017482 U : the proxy UPDATED the last date in the cookie that was presented by
17483 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017484 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020017485 date indicated in the cookie. If any other change happens, such as
17486 a redispatch, then the cookie will be marked as inserted instead.
17487
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017488 P : a cookie was PROVIDED by the server and transmitted as-is.
17489
17490 R : the cookie provided by the server was REWRITTEN by the proxy, which
17491 happens in "cookie rewrite" or "cookie prefix" modes.
17492
17493 D : the cookie provided by the server was DELETED by the proxy.
17494
17495 - : does not apply (no cookie set in configuration).
17496
Willy Tarreau996a92c2010-10-13 19:30:47 +020017497The combination of the two first flags gives a lot of information about what
17498was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017499helpful to detect server saturation, network troubles, local system resource
17500starvation, attacks, etc...
17501
17502The most common termination flags combinations are indicated below. They are
17503alphabetically sorted, with the lowercase set just after the upper case for
17504easier finding and understanding.
17505
17506 Flags Reason
17507
17508 -- Normal termination.
17509
17510 CC The client aborted before the connection could be established to the
17511 server. This can happen when haproxy tries to connect to a recently
17512 dead (or unchecked) server, and the client aborts while haproxy is
17513 waiting for the server to respond or for "timeout connect" to expire.
17514
17515 CD The client unexpectedly aborted during data transfer. This can be
17516 caused by a browser crash, by an intermediate equipment between the
17517 client and haproxy which decided to actively break the connection,
17518 by network routing issues between the client and haproxy, or by a
17519 keep-alive session between the server and the client terminated first
17520 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010017521
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017522 cD The client did not send nor acknowledge any data for as long as the
17523 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020017524 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017525
17526 CH The client aborted while waiting for the server to start responding.
17527 It might be the server taking too long to respond or the client
17528 clicking the 'Stop' button too fast.
17529
17530 cH The "timeout client" stroke while waiting for client data during a
17531 POST request. This is sometimes caused by too large TCP MSS values
17532 for PPPoE networks which cannot transport full-sized packets. It can
17533 also happen when client timeout is smaller than server timeout and
17534 the server takes too long to respond.
17535
17536 CQ The client aborted while its session was queued, waiting for a server
17537 with enough empty slots to accept it. It might be that either all the
17538 servers were saturated or that the assigned server was taking too
17539 long a time to respond.
17540
17541 CR The client aborted before sending a full HTTP request. Most likely
17542 the request was typed by hand using a telnet client, and aborted
17543 too early. The HTTP status code is likely a 400 here. Sometimes this
17544 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020017545 and the client. "option http-ignore-probes" can be used to ignore
17546 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017547
17548 cR The "timeout http-request" stroke before the client sent a full HTTP
17549 request. This is sometimes caused by too large TCP MSS values on the
17550 client side for PPPoE networks which cannot transport full-sized
17551 packets, or by clients sending requests by hand and not typing fast
17552 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020017553 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020017554 some browsers started to implement a "pre-connect" feature consisting
17555 in speculatively connecting to some recently visited web sites just
17556 in case the user would like to visit them. This results in many
17557 connections being established to web sites, which end up in 408
17558 Request Timeout if the timeout strikes first, or 400 Bad Request when
17559 the browser decides to close them first. These ones pollute the log
17560 and feed the error counters. Some versions of some browsers have even
17561 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010017562 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020017563 http-ignore-probes" in the frontend, resulting in connections with
17564 zero data transfer to be totally ignored. This will definitely hide
17565 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017566
17567 CT The client aborted while its session was tarpitted. It is important to
17568 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020017569 wrong tarpit rules have been written. If a lot of them happen, it
17570 might make sense to lower the "timeout tarpit" value to something
17571 closer to the average reported "Tw" timer, in order not to consume
17572 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017573
Willy Tarreau570f2212013-06-10 16:42:09 +020017574 LR The request was intercepted and locally handled by haproxy. Generally
17575 it means that this was a redirect or a stats request.
17576
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017577 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017578 the TCP connection (the proxy received a TCP RST or an ICMP message
17579 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010017580 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017581 or no ARP response on local network). When this happens in HTTP mode,
17582 the status code is likely a 502 or 503 here.
17583
17584 sC The "timeout connect" stroke before a connection to the server could
17585 complete. When this happens in HTTP mode, the status code is likely a
17586 503 or 504 here.
17587
17588 SD The connection to the server died with an error during the data
17589 transfer. This usually means that haproxy has received an RST from
17590 the server or an ICMP message from an intermediate equipment while
17591 exchanging data with the server. This can be caused by a server crash
17592 or by a network issue on an intermediate equipment.
17593
17594 sD The server did not send nor acknowledge any data for as long as the
17595 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010017596 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017597 load-balancers, ...), as well as keep-alive sessions maintained
17598 between the client and the server expiring first on haproxy.
17599
17600 SH The server aborted before sending its full HTTP response headers, or
17601 it crashed while processing the request. Since a server aborting at
17602 this moment is very rare, it would be wise to inspect its logs to
17603 control whether it crashed and why. The logged request may indicate a
17604 small set of faulty requests, demonstrating bugs in the application.
17605 Sometimes this might also be caused by an IDS killing the connection
17606 between haproxy and the server.
17607
17608 sH The "timeout server" stroke before the server could return its
17609 response headers. This is the most common anomaly, indicating too
17610 long transactions, probably caused by server or database saturation.
17611 The immediate workaround consists in increasing the "timeout server"
17612 setting, but it is important to keep in mind that the user experience
17613 will suffer from these long response times. The only long term
17614 solution is to fix the application.
17615
17616 sQ The session spent too much time in queue and has been expired. See
17617 the "timeout queue" and "timeout connect" settings to find out how to
17618 fix this if it happens too often. If it often happens massively in
17619 short periods, it may indicate general problems on the affected
17620 servers due to I/O or database congestion, or saturation caused by
17621 external attacks.
17622
17623 PC The proxy refused to establish a connection to the server because the
17624 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020017625 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017626 so that it does not happen anymore. This status is very rare and
17627 might happen when the global "ulimit-n" parameter is forced by hand.
17628
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017629 PD The proxy blocked an incorrectly formatted chunked encoded message in
17630 a request or a response, after the server has emitted its headers. In
17631 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010017632 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020017633 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017634
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017635 PH The proxy blocked the server's response, because it was invalid,
17636 incomplete, dangerous (cache control), or matched a security filter.
17637 In any case, an HTTP 502 error is sent to the client. One possible
17638 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017639 containing unauthorized characters. It is also possible but quite
17640 rare, that the proxy blocked a chunked-encoding request from the
17641 client due to an invalid syntax, before the server responded. In this
17642 case, an HTTP 400 error is sent to the client and reported in the
17643 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017644
17645 PR The proxy blocked the client's HTTP request, either because of an
17646 invalid HTTP syntax, in which case it returned an HTTP 400 error to
17647 the client, or because a deny filter matched, in which case it
17648 returned an HTTP 403 error.
17649
17650 PT The proxy blocked the client's request and has tarpitted its
17651 connection before returning it a 500 server error. Nothing was sent
17652 to the server. The connection was maintained open for as long as
17653 reported by the "Tw" timer field.
17654
17655 RC A local resource has been exhausted (memory, sockets, source ports)
17656 preventing the connection to the server from establishing. The error
17657 logs will tell precisely what was missing. This is very rare and can
17658 only be solved by proper system tuning.
17659
Willy Tarreau996a92c2010-10-13 19:30:47 +020017660The combination of the two last flags gives a lot of information about how
17661persistence was handled by the client, the server and by haproxy. This is very
17662important to troubleshoot disconnections, when users complain they have to
17663re-authenticate. The commonly encountered flags are :
17664
17665 -- Persistence cookie is not enabled.
17666
17667 NN No cookie was provided by the client, none was inserted in the
17668 response. For instance, this can be in insert mode with "postonly"
17669 set on a GET request.
17670
17671 II A cookie designating an invalid server was provided by the client,
17672 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017673 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020017674 value can be presented by a client when no other server knows it.
17675
17676 NI No cookie was provided by the client, one was inserted in the
17677 response. This typically happens for first requests from every user
17678 in "insert" mode, which makes it an easy way to count real users.
17679
17680 VN A cookie was provided by the client, none was inserted in the
17681 response. This happens for most responses for which the client has
17682 already got a cookie.
17683
17684 VU A cookie was provided by the client, with a last visit date which is
17685 not completely up-to-date, so an updated cookie was provided in
17686 response. This can also happen if there was no date at all, or if
17687 there was a date but the "maxidle" parameter was not set, so that the
17688 cookie can be switched to unlimited time.
17689
17690 EI A cookie was provided by the client, with a last visit date which is
17691 too old for the "maxidle" parameter, so the cookie was ignored and a
17692 new cookie was inserted in the response.
17693
17694 OI A cookie was provided by the client, with a first visit date which is
17695 too old for the "maxlife" parameter, so the cookie was ignored and a
17696 new cookie was inserted in the response.
17697
17698 DI The server designated by the cookie was down, a new server was
17699 selected and a new cookie was emitted in the response.
17700
17701 VI The server designated by the cookie was not marked dead but could not
17702 be reached. A redispatch happened and selected another one, which was
17703 then advertised in the response.
17704
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017705
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177068.6. Non-printable characters
17707-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017708
17709In order not to cause trouble to log analysis tools or terminals during log
17710consulting, non-printable characters are not sent as-is into log files, but are
17711converted to the two-digits hexadecimal representation of their ASCII code,
17712prefixed by the character '#'. The only characters that can be logged without
17713being escaped are comprised between 32 and 126 (inclusive). Obviously, the
17714escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
17715is the same for the character '"' which becomes "#22", as well as '{', '|' and
17716'}' when logging headers.
17717
17718Note that the space character (' ') is not encoded in headers, which can cause
17719issues for tools relying on space count to locate fields. A typical header
17720containing spaces is "User-Agent".
17721
17722Last, it has been observed that some syslog daemons such as syslog-ng escape
17723the quote ('"') with a backslash ('\'). The reverse operation can safely be
17724performed since no quote may appear anywhere else in the logs.
17725
17726
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177278.7. Capturing HTTP cookies
17728---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017729
17730Cookie capture simplifies the tracking a complete user session. This can be
17731achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017732section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017733cookie will simultaneously be checked in the request ("Cookie:" header) and in
17734the response ("Set-Cookie:" header). The respective values will be reported in
17735the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017736locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017737not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
17738user switches to a new session for example, because the server will reassign it
17739a new cookie. It is also possible to detect if a server unexpectedly sets a
17740wrong cookie to a client, leading to session crossing.
17741
17742 Examples :
17743 # capture the first cookie whose name starts with "ASPSESSION"
17744 capture cookie ASPSESSION len 32
17745
17746 # capture the first cookie whose name is exactly "vgnvisitor"
17747 capture cookie vgnvisitor= len 32
17748
17749
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177508.8. Capturing HTTP headers
17751---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017752
17753Header captures are useful to track unique request identifiers set by an upper
17754proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
17755the response, one can search for information about the response length, how the
17756server asked the cache to behave, or an object location during a redirection.
17757
17758Header captures are performed using the "capture request header" and "capture
17759response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017760section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017761
17762It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017763time. Non-existent headers are logged as empty strings, and if one header
17764appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017765are grouped within braces '{' and '}' in the same order as they were declared,
17766and delimited with a vertical bar '|' without any space. Response headers
17767follow the same representation, but are displayed after a space following the
17768request headers block. These blocks are displayed just before the HTTP request
17769in the logs.
17770
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017771As a special case, it is possible to specify an HTTP header capture in a TCP
17772frontend. The purpose is to enable logging of headers which will be parsed in
17773an HTTP backend if the request is then switched to this HTTP backend.
17774
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017775 Example :
17776 # This instance chains to the outgoing proxy
17777 listen proxy-out
17778 mode http
17779 option httplog
17780 option logasap
17781 log global
17782 server cache1 192.168.1.1:3128
17783
17784 # log the name of the virtual server
17785 capture request header Host len 20
17786
17787 # log the amount of data uploaded during a POST
17788 capture request header Content-Length len 10
17789
17790 # log the beginning of the referrer
17791 capture request header Referer len 20
17792
17793 # server name (useful for outgoing proxies only)
17794 capture response header Server len 20
17795
17796 # logging the content-length is useful with "option logasap"
17797 capture response header Content-Length len 10
17798
Davor Ocelice9ed2812017-12-25 17:49:28 +010017799 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017800 capture response header Cache-Control len 8
17801
17802 # the Via header will report the next proxy's name
17803 capture response header Via len 20
17804
17805 # log the URL location during a redirection
17806 capture response header Location len 20
17807
17808 >>> Aug 9 20:26:09 localhost \
17809 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
17810 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
17811 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
17812 "GET http://fr.adserver.yahoo.com/"
17813
17814 >>> Aug 9 20:30:46 localhost \
17815 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
17816 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
17817 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017818 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017819
17820 >>> Aug 9 20:30:46 localhost \
17821 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
17822 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
17823 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
17824 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017825 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017826
17827
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200178288.9. Examples of logs
17829---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017830
17831These are real-world examples of logs accompanied with an explanation. Some of
17832them have been made up by hand. The syslog part has been removed for better
17833reading. Their sole purpose is to explain how to decipher them.
17834
17835 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
17836 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
17837 "HEAD / HTTP/1.0"
17838
17839 => long request (6.5s) entered by hand through 'telnet'. The server replied
17840 in 147 ms, and the session ended normally ('----')
17841
17842 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
17843 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
17844 0/9 "HEAD / HTTP/1.0"
17845
17846 => Idem, but the request was queued in the global queue behind 9 other
17847 requests, and waited there for 1230 ms.
17848
17849 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
17850 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
17851 "GET /image.iso HTTP/1.0"
17852
17853 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017854 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017855 14 ms, 243 bytes of headers were sent to the client, and total time from
17856 accept to first data byte is 30 ms.
17857
17858 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
17859 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
17860 "GET /cgi-bin/bug.cgi? HTTP/1.0"
17861
Christopher Faulet87f1f3d2019-07-18 14:51:20 +020017862 => the proxy blocked a server response either because of an "http-response
17863 deny" rule, or because the response was improperly formatted and not
17864 HTTP-compliant, or because it blocked sensitive information which risked
17865 being cached. In this case, the response is replaced with a "502 bad
17866 gateway". The flags ("PH--") tell us that it was haproxy who decided to
17867 return the 502 and not the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017868
17869 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017870 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 +010017871
17872 => the client never completed its request and aborted itself ("C---") after
17873 8.5s, while the proxy was waiting for the request headers ("-R--").
17874 Nothing was sent to any server.
17875
17876 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
17877 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
17878
17879 => The client never completed its request, which was aborted by the
17880 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010017881 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017882 send a 408 return code to the client.
17883
17884 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
17885 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
17886
17887 => This log was produced with "option tcplog". The client timed out after
17888 5 seconds ("c----").
17889
17890 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
17891 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010017892 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017893
17894 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017895 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017896 (config says 'retries 3'), and no redispatch (otherwise we would have
17897 seen "/+3"). Status code 503 was returned to the client. There were 115
17898 connections on this server, 202 connections on this proxy, and 205 on
17899 the global process. It is possible that the server refused the
17900 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010017901
Willy Tarreau52b2d222011-09-07 23:48:48 +020017902
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200179039. Supported filters
17904--------------------
17905
17906Here are listed officially supported filters with the list of parameters they
17907accept. Depending on compile options, some of these filters might be
17908unavailable. The list of available filters is reported in haproxy -vv.
17909
17910See also : "filter"
17911
179129.1. Trace
17913----------
17914
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010017915filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017916
17917 Arguments:
17918 <name> is an arbitrary name that will be reported in
17919 messages. If no name is provided, "TRACE" is used.
17920
17921 <random-parsing> enables the random parsing of data exchanged between
17922 the client and the server. By default, this filter
17923 parses all available data. With this parameter, it
17924 only parses a random amount of the available data.
17925
Davor Ocelice9ed2812017-12-25 17:49:28 +010017926 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017927 default, this filter forwards all previously parsed
17928 data. With this parameter, it only forwards a random
17929 amount of the parsed data.
17930
Davor Ocelice9ed2812017-12-25 17:49:28 +010017931 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010017932
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017933This filter can be used as a base to develop new filters. It defines all
17934callbacks and print a message on the standard error stream (stderr) with useful
17935information for all of them. It may be useful to debug the activity of other
17936filters or, quite simply, HAProxy's activity.
17937
17938Using <random-parsing> and/or <random-forwarding> parameters is a good way to
17939tests the behavior of a filter that parses data exchanged between a client and
17940a server by adding some latencies in the processing.
17941
17942
179439.2. HTTP compression
17944---------------------
17945
17946filter compression
17947
17948The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
17949keyword must still be used to enable and configure the HTTP compression. And
Christopher Fauletb30b3102019-09-12 23:03:09 +020017950when no other filter is used, it is enough. When used with the cache or the
17951fcgi-app enabled, it is also enough. In this case, the compression is always
17952done after the response is stored in the cache. But it is mandatory to
17953explicitly use a filter line to enable the HTTP compression when at least one
17954filter other than the cache or the fcgi-app is used for the same
17955listener/frontend/backend. This is important to know the filters evaluation
17956order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017957
Christopher Fauletb30b3102019-09-12 23:03:09 +020017958See also : "compression", section 9.4 about the cache filter and section 9.5
17959 about the fcgi-app filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020017960
17961
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200179629.3. Stream Processing Offload Engine (SPOE)
17963--------------------------------------------
17964
17965filter spoe [engine <name>] config <file>
17966
17967 Arguments :
17968
17969 <name> is the engine name that will be used to find the right scope in
17970 the configuration file. If not provided, all the file will be
17971 parsed.
17972
17973 <file> is the path of the engine configuration file. This file can
17974 contain configuration of several engines. In this case, each
17975 part must be placed in its own scope.
17976
17977The Stream Processing Offload Engine (SPOE) is a filter communicating with
17978external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010017979streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020017980exchanged with them are configured in dedicated files, for the main part. It
17981also requires dedicated backends, defined in HAProxy configuration.
17982
17983SPOE communicates with external components using an in-house binary protocol,
17984the Stream Processing Offload Protocol (SPOP).
17985
Tim Düsterhus4896c442016-11-29 02:15:19 +010017986For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020017987"doc/SPOE.txt".
17988
17989Important note:
17990 The SPOE filter is highly experimental for now and was not heavily
17991 tested. It is really not production ready. So use it carefully.
17992
Christopher Faulet99a17a22018-12-11 09:18:27 +0100179939.4. Cache
17994----------
17995
17996filter cache <name>
17997
17998 Arguments :
17999
18000 <name> is name of the cache section this filter will use.
18001
18002The cache uses a filter to store cacheable responses. The HTTP rules
18003"cache-store" and "cache-use" must be used to define how and when to use a
John Roeslerfb2fce12019-07-10 15:45:51 -050018004cache. By default the corresponding filter is implicitly defined. And when no
Christopher Fauletb30b3102019-09-12 23:03:09 +020018005other filters than fcgi-app or compression are used, it is enough. In such
18006case, the compression filter is always evaluated after the cache filter. But it
18007is mandatory to explicitly use a filter line to use a cache when at least one
18008filter other than the compression or the fcgi-app is used for the same
Christopher Faulet27d93c32018-12-15 22:32:02 +010018009listener/frontend/backend. This is important to know the filters evaluation
18010order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018011
Christopher Fauletb30b3102019-09-12 23:03:09 +020018012See also : section 9.2 about the compression filter, section 9.5 about the
18013 fcgi-app filter and section 6 about cache.
18014
18015
180169.5. Fcgi-app
18017-------------
18018
18019filter fcg-app <name>
18020
18021 Arguments :
18022
18023 <name> is name of the fcgi-app section this filter will use.
18024
18025The FastCGI application uses a filter to evaluate all custom parameters on the
18026request path, and to process the headers on the response path. the <name> must
18027reference an existing fcgi-app section. The directive "use-fcgi-app" should be
18028used to define the application to use. By default the corresponding filter is
18029implicitly defined. And when no other filters than cache or compression are
18030used, it is enough. But it is mandatory to explicitly use a filter line to a
18031fcgi-app when at least one filter other than the compression or the cache is
18032used for the same backend. This is important to know the filters evaluation
18033order.
18034
18035See also: "use-fcgi-app", section 9.2 about the compression filter, section 9.4
18036 about the cache filter and section 10 about FastCGI application.
18037
18038
1803910. FastCGI applications
18040-------------------------
18041
18042HAProxy is able to send HTTP requests to Responder FastCGI applications. This
18043feature was added in HAProxy 2.1. To do so, servers must be configured to use
18044the FastCGI protocol (using the keyword "proto fcgi" on the server line) and a
18045FastCGI application must be configured and used by the backend managing these
18046servers (using the keyword "use-fcgi-app" into the proxy section). Several
18047FastCGI applications may be defined, but only one can be used at a time by a
18048backend.
18049
18050HAProxy implements all features of the FastCGI specification for Responder
18051application. Especially it is able to multiplex several requests on a simple
18052connection.
18053
1805410.1. Setup
18055-----------
18056
1805710.1.1. Fcgi-app section
18058--------------------------
18059
18060fcgi-app <name>
18061 Declare a FastCGI application named <name>. To be valid, at least the
18062 document root must be defined.
18063
18064acl <aclname> <criterion> [flags] [operator] <value> ...
18065 Declare or complete an access list.
18066
18067 See "acl" keyword in section 4.2 and section 7 about ACL usage for
18068 details. ACLs defined for a FastCGI application are private. They cannot be
18069 used by any other application or by any proxy. In the same way, ACLs defined
18070 in any other section are not usable by a FastCGI application. However,
18071 Pre-defined ACLs are available.
18072
18073docroot <path>
18074 Define the document root on the remote host. <path> will be used to build
18075 the default value of FastCGI parameters SCRIPT_FILENAME and
18076 PATH_TRANSLATED. It is a mandatory setting.
18077
18078index <script-name>
18079 Define the script name that will be appended after an URI that ends with a
18080 slash ("/") to set the default value of the FastCGI parameter SCRIPT_NAME. It
18081 is an optional setting.
18082
18083 Example :
18084 index index.php
18085
18086log-stderr global
18087log-stderr <address> [len <length>] [format <format>]
18088 [sample <ranges>:<smp_size>] <facility> [<level> [<minlevel>]]
18089 Enable logging of STDERR messages reported by the FastCGI application.
18090
18091 See "log" keyword in section 4.2 for details. It is an optional setting. By
18092 default STDERR messages are ignored.
18093
18094pass-header <name> [ { if | unless } <condition> ]
18095 Specify the name of a request header which will be passed to the FastCGI
18096 application. It may optionally be followed by an ACL-based condition, in
18097 which case it will only be evaluated if the condition is true.
18098
18099 Most request headers are already available to the FastCGI application,
18100 prefixed with "HTTP_". Thus, this directive is only required to pass headers
18101 that are purposefully omitted. Currently, the headers "Authorization",
18102 "Proxy-Authorization" and hop-by-hop headers are omitted.
18103
18104 Note that the headers "Content-type" and "Content-length" are never passed to
18105 the FastCGI application because they are already converted into parameters.
18106
18107path-info <regex>
18108 Define a regular expression to extract the script-name and the path-info
18109 from the URI. Thus, <regex> should have two captures: the first one to
18110 capture the script name and the second one to capture the path-info. It is an
18111 optional setting. If it is not defined, no matching is performed on the
18112 URI. and the FastCGI parameters PATH_INFO and PATH_TRANSLATED are not filled.
18113
18114 Example :
18115 path-info ^(/.+\.php)(/.*)?$
18116
18117option get-values
18118no option get-values
18119 Enable or disable the retrieve of variables about connection management.
18120
18121 HAproxy is able to send the record FCGI_GET_VALUES on connection
18122 establishment to retrieve the value for following variables:
18123
18124 * FCGI_MAX_REQS The maximum number of concurrent requests this
18125 application will accept.
18126
William Lallemand93e548e2019-09-30 13:54:02 +020018127 * FCGI_MPXS_CONNS "0" if this application does not multiplex connections,
18128 "1" otherwise.
Christopher Fauletb30b3102019-09-12 23:03:09 +020018129
18130 Some FastCGI applications does not support this feature. Some others close
18131 the connexion immediately after sending their response. So, by default, this
18132 option is disabled.
18133
18134 Note that the maximum number of concurrent requests accepted by a FastCGI
18135 application is a connection variable. It only limits the number of streams
18136 per connection. If the global load must be limited on the application, the
18137 server parameters "maxconn" and "pool-max-conn" must be set. In addition, if
18138 an application does not support connection multiplexing, the maximum number
18139 of concurrent requests is automatically set to 1.
18140
18141option keep-conn
18142no option keep-conn
18143 Instruct the FastCGI application to keep the connection open or not after
18144 sending a response.
18145
18146 If disabled, the FastCGI application closes the connection after responding
18147 to this request. By default, this option is enabled.
18148
18149option max-reqs <reqs>
18150 Define the maximum number of concurrent requests this application will
18151 accept.
18152
18153 This option may be overwritten if the variable FCGI_MAX_REQS is retrieved
18154 during connection establishment. Furthermore, if the application does not
18155 support connection multiplexing, this option will be ignored. By default set
18156 to 1.
18157
18158option mpxs-conns
18159no option mpxs-conns
18160 Enable or disable the support of connection multiplexing.
18161
18162 This option may be overwritten if the variable FCGI_MPXS_CONNS is retrieved
18163 during connection establishment. It is disabled by default.
18164
18165set-param <name> <fmt> [ { if | unless } <condition> ]
18166 Set a FastCGI parameter that should be passed to this application. Its
18167 value, defined by <fmt> must follows the log-format rules (see section 8.2.4
18168 "Custom Log format"). It may optionally be followed by an ACL-based
18169 condition, in which case it will only be evaluated if the condition is true.
18170
18171 With this directive, it is possible to overwrite the value of default FastCGI
18172 parameters. If the value is evaluated to an empty string, the rule is
18173 ignored. These directives are evaluated in their declaration order.
18174
18175 Example :
18176 # PHP only, required if PHP was built with --enable-force-cgi-redirect
18177 set-param REDIRECT_STATUS 200
18178
18179 set-param PHP_AUTH_DIGEST %[req.hdr(Authorization)]
18180
18181
1818210.1.2. Proxy section
18183---------------------
18184
18185use-fcgi-app <name>
18186 Define the FastCGI application to use for the backend.
18187
18188 Arguments :
18189 <name> is the name of the FastCGI application to use.
18190
18191 This keyword is only available for HTTP proxies with the backend capability
18192 and with at least one FastCGI server. However, FastCGI servers can be mixed
18193 with HTTP servers. But except there is a good reason to do so, it is not
18194 recommended (see section 10.3 about the limitations for details). Only one
18195 application may be defined at a time per backend.
18196
18197 Note that, once a FastCGI application is referenced for a backend, depending
18198 on the configuration some processing may be done even if the request is not
18199 sent to a FastCGI server. Rules to set parameters or pass headers to an
18200 application are evaluated.
18201
18202
1820310.1.3. Example
18204---------------
18205
18206 frontend front-http
18207 mode http
18208 bind *:80
18209 bind *:
18210
18211 use_backend back-dynamic if { path_reg ^/.+\.php(/.*)?$ }
18212 default_backend back-static
18213
18214 backend back-static
18215 mode http
18216 server www A.B.C.D:80
18217
18218 backend back-dynamic
18219 mode http
18220 use-fcgi-app php-fpm
18221 server php-fpm A.B.C.D:9000 proto fcgi
18222
18223 fcgi-app php-fpm
18224 log-stderr global
18225 option keep-conn
18226
18227 docroot /var/www/my-app
18228 index index.php
18229 path-info ^(/.+\.php)(/.*)?$
18230
18231
1823210.2. Default parameters
18233------------------------
18234
18235A Responder FastCGI application has the same purpose as a CGI/1.1 program. In
18236the CGI/1.1 specification (RFC3875), several variables must be passed to the
18237scipt. So HAProxy set them and some others commonly used by FastCGI
18238applications. All these variables may be overwritten, with caution though.
18239
18240 +-------------------+-----------------------------------------------------+
18241 | AUTH_TYPE | Identifies the mechanism, if any, used by HAProxy |
18242 | | to authenticate the user. Concretely, only the |
18243 | | BASIC authentication mechanism is supported. |
18244 | | |
18245 +-------------------+-----------------------------------------------------+
18246 | CONTENT_LENGTH | Contains the size of the message-body attached to |
18247 | | the request. It means only requests with a known |
18248 | | size are considered as valid and sent to the |
18249 | | application. |
18250 | | |
18251 +-------------------+-----------------------------------------------------+
18252 | CONTENT_TYPE | Contains the type of the message-body attached to |
18253 | | the request. It may not be set. |
18254 | | |
18255 +-------------------+-----------------------------------------------------+
18256 | DOCUMENT_ROOT | Contains the document root on the remote host under |
18257 | | which the script should be executed, as defined in |
18258 | | the application's configuration. |
18259 | | |
18260 +-------------------+-----------------------------------------------------+
18261 | GATEWAY_INTERFACE | Contains the dialect of CGI being used by HAProxy |
18262 | | to communicate with the FastCGI application. |
18263 | | Concretely, it is set to "CGI/1.1". |
18264 | | |
18265 +-------------------+-----------------------------------------------------+
18266 | PATH_INFO | Contains the portion of the URI path hierarchy |
18267 | | following the part that identifies the script |
18268 | | itself. To be set, the directive "path-info" must |
18269 | | be defined. |
18270 | | |
18271 +-------------------+-----------------------------------------------------+
18272 | PATH_TRANSLATED | If PATH_INFO is set, it is its translated version. |
18273 | | It is the concatenation of DOCUMENT_ROOT and |
18274 | | PATH_INFO. If PATH_INFO is not set, this parameters |
18275 | | is not set too. |
18276 | | |
18277 +-------------------+-----------------------------------------------------+
18278 | QUERY_STRING | Contains the request's query string. It may not be |
18279 | | set. |
18280 | | |
18281 +-------------------+-----------------------------------------------------+
18282 | REMOTE_ADDR | Contains the network address of the client sending |
18283 | | the request. |
18284 | | |
18285 +-------------------+-----------------------------------------------------+
18286 | REMOTE_USER | Contains the user identification string supplied by |
18287 | | client as part of user authentication. |
18288 | | |
18289 +-------------------+-----------------------------------------------------+
18290 | REQUEST_METHOD | Contains the method which should be used by the |
18291 | | script to process the request. |
18292 | | |
18293 +-------------------+-----------------------------------------------------+
18294 | REQUEST_URI | Contains the request's URI. |
18295 | | |
18296 +-------------------+-----------------------------------------------------+
18297 | SCRIPT_FILENAME | Contains the absolute pathname of the script. it is |
18298 | | the concatenation of DOCUMENT_ROOT and SCRIPT_NAME. |
18299 | | |
18300 +-------------------+-----------------------------------------------------+
18301 | SCRIPT_NAME | Contains the name of the script. If the directive |
18302 | | "path-info" is defined, it is the first part of the |
18303 | | URI path hierarchy, ending with the script name. |
18304 | | Otherwise, it is the entire URI path. |
18305 | | |
18306 +-------------------+-----------------------------------------------------+
18307 | SERVER_NAME | Contains the name of the server host to which the |
18308 | | client request is directed. It is the value of the |
18309 | | header "Host", if defined. Otherwise, the |
18310 | | destination address of the connection on the client |
18311 | | side. |
18312 | | |
18313 +-------------------+-----------------------------------------------------+
18314 | SERVER_PORT | Contains the destination TCP port of the connection |
18315 | | on the client side, which is the port the client |
18316 | | connected to. |
18317 | | |
18318 +-------------------+-----------------------------------------------------+
18319 | SERVER_PROTOCOL | Contains the request's protocol. |
18320 | | |
18321 +-------------------+-----------------------------------------------------+
18322 | HTTPS | Set to a non-empty value ("on") if the script was |
18323 | | queried through the HTTPS protocol. |
18324 | | |
18325 +-------------------+-----------------------------------------------------+
18326
18327
1832810.3. Limitations
18329------------------
18330
18331The current implementation have some limitations. The first one is about the
18332way some request headers are hidden to the FastCGI applications. This happens
18333during the headers analysis, on the backend side, before the connection
18334establishment. At this stage, HAProxy know the backend is using a FastCGI
18335application but it don't know if the request will be routed to a FastCGI server
18336or not. But to hide request headers, it simply removes them from the HTX
18337message. So, if the request is finally routed to an HTTP server, it never see
18338these headers. For this reason, it is not recommended to mix FastCGI servers
18339and HTTP servers under the same backend.
18340
18341Similarly, the rules "set-param" and "pass-header" are evaluated during the
18342request headers analysis. So the evaluation is always performed, even if the
18343requests is finally forwarded to an HTTP server.
18344
18345About the rules "set-param", when a rule is applied, a pseudo header is added
18346into the HTX message. So, the same way than for HTTP header rewrites, it may
18347fail if the buffer is full. The rules "set-param" will compete with
18348"http-request" ones.
18349
18350Finally, all FastCGI params and HTTP headers are sent into a unique record
18351FCGI_PARAM. Encoding of this record must be done in one pass, otherwise a
18352processing error is returned. It means the record FCGI_PARAM, once encoded,
18353must not exceeds the size of a buffer. However, there is no reserve to respect
18354here.
William Lallemand86d0df02017-11-24 21:36:45 +010018355
Willy Tarreau0ba27502007-12-24 16:55:16 +010018356/*
18357 * Local variables:
18358 * fill-column: 79
18359 * End:
18360 */