blob: b2ee83545b1b231c73fb21e4b1aaeed469f10a6a [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 Tarreaufba74ea2018-12-22 11:19:45 +01005 version 2.0
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreaud8498662020-02-13 07:58:50 +01007 2020/02/13
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
676. HTTP header manipulation
68
Willy Tarreau74ca5042013-06-11 23:12:07 +0200697. Using ACLs and fetching samples
707.1. ACL basics
717.1.1. Matching booleans
727.1.2. Matching integers
737.1.3. Matching strings
747.1.4. Matching regular expressions (regexes)
757.1.5. Matching arbitrary data blocks
767.1.6. Matching IPv4 and IPv6 addresses
777.2. Using ACLs to form conditions
787.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200797.3.1. Converters
807.3.2. Fetching samples from internal states
817.3.3. Fetching samples at Layer 4
827.3.4. Fetching samples at Layer 5
837.3.5. Fetching samples from buffer contents (Layer 6)
847.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200857.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020086
878. Logging
888.1. Log levels
898.2. Log formats
908.2.1. Default log format
918.2.2. TCP log format
928.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100938.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100948.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200958.3. Advanced logging options
968.3.1. Disabling logging of external tests
978.3.2. Logging before waiting for the session to terminate
988.3.3. Raising log level upon errors
998.3.4. Disabling logging of successful connections
1008.4. Timing events
1018.5. Session state at disconnection
1028.6. Non-printable characters
1038.7. Capturing HTTP cookies
1048.8. Capturing HTTP headers
1058.9. Examples of logs
106
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001079. Supported filters
1089.1. Trace
1099.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001109.3. Stream Processing Offload Engine (SPOE)
Christopher Faulet99a17a22018-12-11 09:18:27 +01001119.4. Cache
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200112
William Lallemand86d0df02017-11-24 21:36:45 +010011310. Cache
Cyril Bonté7b888f12017-11-26 22:24:31 +010011410.1. Limitation
11510.2. Setup
11610.2.1. Cache section
11710.2.2. Proxy section
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200118
1191. Quick reminder about HTTP
120----------------------------
121
Davor Ocelice9ed2812017-12-25 17:49:28 +0100122When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200123fully analyzed and indexed, thus it becomes possible to build matching criteria
124on almost anything found in the contents.
125
126However, it is important to understand how HTTP requests and responses are
127formed, and how HAProxy decomposes them. It will then become easier to write
128correct rules and to debug existing configurations.
129
130
1311.1. The HTTP transaction model
132-------------------------------
133
134The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100135to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100136from the client to the server, a request is sent by the client through the
137connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200138will involve a new connection :
139
140 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
141
142In this mode, called the "HTTP close" mode, there are as many connection
143establishments as there are HTTP transactions. Since the connection is closed
144by the server after the response, the client does not need to know the content
145length.
146
147Due to the transactional nature of the protocol, it was possible to improve it
148to avoid closing a connection between two subsequent transactions. In this mode
149however, it is mandatory that the server indicates the content length for each
150response so that the client does not wait indefinitely. For this, a special
151header is used: "Content-length". This mode is called the "keep-alive" mode :
152
153 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
154
155Its advantages are a reduced latency between transactions, and less processing
156power required on the server side. It is generally better than the close mode,
157but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200158a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200159
Willy Tarreau95c4e142017-11-26 12:18:55 +0100160Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200161keep-alive, but the client does not wait for the first response to send the
162second request. This is useful for fetching large number of images composing a
163page :
164
165 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
166
167This can obviously have a tremendous benefit on performance because the network
168latency is eliminated between subsequent requests. Many HTTP agents do not
169correctly support pipelining since there is no way to associate a response with
170the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100171server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200172
Willy Tarreau95c4e142017-11-26 12:18:55 +0100173The next improvement is the multiplexed mode, as implemented in HTTP/2. This
174time, each transaction is assigned a single stream identifier, and all streams
175are multiplexed over an existing connection. Many requests can be sent in
176parallel by the client, and responses can arrive in any order since they also
177carry the stream identifier.
178
Willy Tarreau70dffda2014-01-30 03:07:23 +0100179By default HAProxy operates in keep-alive mode with regards to persistent
180connections: for each connection it processes each request and response, and
181leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100182start of a new request. When it receives HTTP/2 connections from a client, it
183processes all the requests in parallel and leaves the connection idling,
184waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200185
Christopher Faulet315b39c2018-09-21 16:26:19 +0200186HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +0100187 - keep alive : all requests and responses are processed (default)
188 - tunnel : only the first request and response are processed,
Christopher Faulet6c9bbb22019-03-26 21:37:23 +0100189 everything else is forwarded with no analysis (deprecated).
Willy Tarreau70dffda2014-01-30 03:07:23 +0100190 - server close : the server-facing connection is closed after the response.
Christopher Faulet315b39c2018-09-21 16:26:19 +0200191 - close : the connection is actively closed after end of response.
Willy Tarreau70dffda2014-01-30 03:07:23 +0100192
Davor Ocelice9ed2812017-12-25 17:49:28 +0100193For HTTP/2, the connection mode resembles more the "server close" mode : given
194the independence of all streams, there is currently no place to hook the idle
Willy Tarreau95c4e142017-11-26 12:18:55 +0100195server connection after a response, so it is closed after the response. HTTP/2
196is only supported for incoming connections, not on connections going to
197servers.
198
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200199
2001.2. HTTP request
201-----------------
202
203First, let's consider this HTTP request :
204
205 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100206 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200207 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
208 2 Host: www.mydomain.com
209 3 User-agent: my small browser
210 4 Accept: image/jpeg, image/gif
211 5 Accept: image/png
212
213
2141.2.1. The Request line
215-----------------------
216
217Line 1 is the "request line". It is always composed of 3 fields :
218
219 - a METHOD : GET
220 - a URI : /serv/login.php?lang=en&profile=2
221 - a version tag : HTTP/1.1
222
223All of them are delimited by what the standard calls LWS (linear white spaces),
224which are commonly spaces, but can also be tabs or line feeds/carriage returns
225followed by spaces/tabs. The method itself cannot contain any colon (':') and
226is limited to alphabetic letters. All those various combinations make it
227desirable that HAProxy performs the splitting itself rather than leaving it to
228the user to write a complex or inaccurate regular expression.
229
230The URI itself can have several forms :
231
232 - A "relative URI" :
233
234 /serv/login.php?lang=en&profile=2
235
236 It is a complete URL without the host part. This is generally what is
237 received by servers, reverse proxies and transparent proxies.
238
239 - An "absolute URI", also called a "URL" :
240
241 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
242
243 It is composed of a "scheme" (the protocol name followed by '://'), a host
244 name or address, optionally a colon (':') followed by a port number, then
245 a relative URI beginning at the first slash ('/') after the address part.
246 This is generally what proxies receive, but a server supporting HTTP/1.1
247 must accept this form too.
248
249 - a star ('*') : this form is only accepted in association with the OPTIONS
250 method and is not relayable. It is used to inquiry a next hop's
251 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100252
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200253 - an address:port combination : 192.168.0.12:80
254 This is used with the CONNECT method, which is used to establish TCP
255 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
256 other protocols too.
257
258In a relative URI, two sub-parts are identified. The part before the question
259mark is called the "path". It is typically the relative path to static objects
260on the server. The part after the question mark is called the "query string".
261It is mostly used with GET requests sent to dynamic scripts and is very
262specific to the language, framework or application in use.
263
Willy Tarreau95c4e142017-11-26 12:18:55 +0100264HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100265assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100266However, haproxy natively processes HTTP/1.x requests and headers, so requests
267received over an HTTP/2 connection are transcoded to HTTP/1.1 before being
268processed. This explains why they still appear as "HTTP/1.1" in haproxy's logs
269as well as in server logs.
270
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200271
2721.2.2. The request headers
273--------------------------
274
275The headers start at the second line. They are composed of a name at the
276beginning of the line, immediately followed by a colon (':'). Traditionally,
277an LWS is added after the colon but that's not required. Then come the values.
278Multiple identical headers may be folded into one single line, delimiting the
279values with commas, provided that their order is respected. This is commonly
280encountered in the "Cookie:" field. A header may span over multiple lines if
281the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
282define a total of 3 values for the "Accept:" header.
283
Davor Ocelice9ed2812017-12-25 17:49:28 +0100284Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200285their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100286"Connection:" header). In HTTP/2, header names are always sent in lower case,
287as can be seen when running in debug mode.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200288
289The end of the headers is indicated by the first empty line. People often say
290that it's a double line feed, which is not exact, even if a double line feed
291is one valid form of empty line.
292
293Fortunately, HAProxy takes care of all these complex combinations when indexing
294headers, checking values and counting them, so there is no reason to worry
295about the way they could be written, but it is important not to accuse an
296application of being buggy if it does unusual, valid things.
297
298Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000299 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200300 in the middle of headers by LWS in order to join multi-line headers. This
301 is necessary for proper analysis and helps less capable HTTP parsers to work
302 correctly and not to be fooled by such complex constructs.
303
304
3051.3. HTTP response
306------------------
307
308An HTTP response looks very much like an HTTP request. Both are called HTTP
309messages. Let's consider this HTTP response :
310
311 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100312 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200313 1 HTTP/1.1 200 OK
314 2 Content-length: 350
315 3 Content-Type: text/html
316
Willy Tarreau816b9792009-09-15 21:25:21 +0200317As a special case, HTTP supports so called "Informational responses" as status
318codes 1xx. These messages are special in that they don't convey any part of the
319response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100320continue to post its request for instance. In the case of a status 100 response
321the requested information will be carried by the next non-100 response message
322following the informational one. This implies that multiple responses may be
323sent to a single request, and that this only works when keep-alive is enabled
324(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
325correctly forward and skip them, and only process the next non-100 response. As
326such, these messages are neither logged nor transformed, unless explicitly
327state otherwise. Status 101 messages indicate that the protocol is changing
328over the same connection and that haproxy must switch to tunnel mode, just as
329if a CONNECT had occurred. Then the Upgrade header would contain additional
330information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200331
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200332
Davor Ocelice9ed2812017-12-25 17:49:28 +01003331.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200334------------------------
335
336Line 1 is the "response line". It is always composed of 3 fields :
337
338 - a version tag : HTTP/1.1
339 - a status code : 200
340 - a reason : OK
341
342The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100343 - 1xx = informational message to be skipped (e.g. 100, 101)
344 - 2xx = OK, content is following (e.g. 200, 206)
345 - 3xx = OK, no content following (e.g. 302, 304)
346 - 4xx = error caused by the client (e.g. 401, 403, 404)
347 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200348
Lukas Tribus23953682017-04-28 13:24:30 +0000349Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100350"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200351found there, but it's a common practice to respect the well-established
352messages. It can be composed of one or multiple words, such as "OK", "Found",
353or "Authentication Required".
354
Davor Ocelice9ed2812017-12-25 17:49:28 +0100355HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200356
357 Code When / reason
358 200 access to stats page, and when replying to monitoring requests
359 301 when performing a redirection, depending on the configured code
360 302 when performing a redirection, depending on the configured code
361 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100362 307 when performing a redirection, depending on the configured code
363 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200364 400 for an invalid or too large request
365 401 when an authentication is required to perform the action (when
366 accessing the stats page)
367 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
368 408 when the request timeout strikes before the request is complete
369 500 when haproxy encounters an unrecoverable internal error, such as a
370 memory allocation failure, which should never happen
371 502 when the server returns an empty, invalid or incomplete response, or
372 when an "rspdeny" filter blocks the response.
373 503 when no server was available to handle the request, or in response to
374 monitoring requests which match the "monitor fail" condition
375 504 when the response timeout strikes before the server responds
376
377The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3784.2).
379
380
3811.3.2. The response headers
382---------------------------
383
384Response headers work exactly like request headers, and as such, HAProxy uses
385the same parsing function for both. Please refer to paragraph 1.2.2 for more
386details.
387
388
3892. Configuring HAProxy
390----------------------
391
3922.1. Configuration file format
393------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200394
395HAProxy's configuration process involves 3 major sources of parameters :
396
397 - the arguments from the command-line, which always take precedence
398 - the "global" section, which sets process-wide parameters
399 - the proxies sections which can take form of "defaults", "listen",
400 "frontend" and "backend".
401
Willy Tarreau0ba27502007-12-24 16:55:16 +0100402The configuration file syntax consists in lines beginning with a keyword
403referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200404delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100405
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200406
William Lallemandf9873ba2015-05-05 17:37:14 +02004072.2. Quoting and escaping
408-------------------------
409
410HAProxy's configuration introduces a quoting and escaping system similar to
411many programming languages. The configuration file supports 3 types: escaping
412with a backslash, weak quoting with double quotes, and strong quoting with
413single quotes.
414
415If spaces have to be entered in strings, then they must be escaped by preceding
416them by a backslash ('\') or by quoting them. Backslashes also have to be
417escaped by doubling or strong quoting them.
418
419Escaping is achieved by preceding a special character by a backslash ('\'):
420
421 \ to mark a space and differentiate it from a delimiter
422 \# to mark a hash and differentiate it from a comment
423 \\ to use a backslash
424 \' to use a single quote and differentiate it from strong quoting
425 \" to use a double quote and differentiate it from weak quoting
426
427Weak quoting is achieved by using double quotes (""). Weak quoting prevents
428the interpretation of:
429
430 space as a parameter separator
431 ' single quote as a strong quoting delimiter
432 # hash as a comment start
433
William Lallemandb2f07452015-05-12 14:27:13 +0200434Weak quoting permits the interpretation of variables, if you want to use a non
435-interpreted dollar within a double quoted string, you should escape it with a
436backslash ("\$"), it does not work outside weak quoting.
437
438Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200439quoting.
440
441Strong quoting is achieved by using single quotes (''). Inside single quotes,
442nothing is interpreted, it's the efficient way to quote regexes.
443
444Quoted and escaped strings are replaced in memory by their interpreted
445equivalent, it allows you to perform concatenation.
446
447 Example:
448 # those are equivalents:
449 log-format %{+Q}o\ %t\ %s\ %{-Q}r
450 log-format "%{+Q}o %t %s %{-Q}r"
451 log-format '%{+Q}o %t %s %{-Q}r'
452 log-format "%{+Q}o %t"' %s %{-Q}r'
453 log-format "%{+Q}o %t"' %s'\ %{-Q}r
454
455 # those are equivalents:
456 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
457 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
458 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
459 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
460
461
William Lallemandb2f07452015-05-12 14:27:13 +02004622.3. Environment variables
463--------------------------
464
465HAProxy's configuration supports environment variables. Those variables are
466interpreted only within double quotes. Variables are expanded during the
467configuration parsing. Variable names must be preceded by a dollar ("$") and
468optionally enclosed with braces ("{}") similarly to what is done in Bourne
469shell. Variable names can contain alphanumerical characters or the character
470underscore ("_") but should not start with a digit.
471
472 Example:
473
474 bind "fd@${FD_APP1}"
475
476 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
477
478 user "$HAPROXY_USER"
479
William Lallemand4d03e432019-06-14 15:35:37 +0200480Some variables are defined by HAProxy, they can be used in the configuration
481file, or could be inherited by a program (See 3.7. Programs):
William Lallemanddaf4cd22018-04-17 16:46:13 +0200482
William Lallemand4d03e432019-06-14 15:35:37 +0200483* HAPROXY_LOCALPEER: defined at the startup of the process which contains the
484 name of the local peer. (See "-L" in the management guide.)
485
486* HAPROXY_CFGFILES: list of the configuration files loaded by HAProxy,
487 separated by semicolons. Can be useful in the case you specified a
488 directory.
489
490* HAPROXY_MWORKER: In master-worker mode, this variable is set to 1.
491
John Roesler27c754c2019-07-10 15:45:51 -0500492* HAPROXY_CLI: configured listeners addresses of the stats socket for every
William Lallemand4d03e432019-06-14 15:35:37 +0200493 processes, separated by semicolons.
494
John Roesler27c754c2019-07-10 15:45:51 -0500495* HAPROXY_MASTER_CLI: In master-worker mode, listeners addresses of the master
William Lallemand4d03e432019-06-14 15:35:37 +0200496 CLI, separated by semicolons.
497
498See also "external-check command" for other variables.
William Lallemandb2f07452015-05-12 14:27:13 +0200499
5002.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200501----------------
502
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100503Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100504values are generally expressed in milliseconds (unless explicitly stated
505otherwise) but may be expressed in any other unit by suffixing the unit to the
506numeric value. It is important to consider this because it will not be repeated
507for every keyword. Supported units are :
508
509 - us : microseconds. 1 microsecond = 1/1000000 second
510 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
511 - s : seconds. 1s = 1000ms
512 - m : minutes. 1m = 60s = 60000ms
513 - h : hours. 1h = 60m = 3600s = 3600000ms
514 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
515
516
Lukas Tribusaa83a312017-03-21 09:25:09 +00005172.5. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200518-------------
519
520 # Simple configuration for an HTTP proxy listening on port 80 on all
521 # interfaces and forwarding requests to a single backend "servers" with a
522 # single server "server1" listening on 127.0.0.1:8000
523 global
524 daemon
525 maxconn 256
526
527 defaults
528 mode http
529 timeout connect 5000ms
530 timeout client 50000ms
531 timeout server 50000ms
532
533 frontend http-in
534 bind *:80
535 default_backend servers
536
537 backend servers
538 server server1 127.0.0.1:8000 maxconn 32
539
540
541 # The same configuration defined with a single listen block. Shorter but
542 # less expressive, especially in HTTP mode.
543 global
544 daemon
545 maxconn 256
546
547 defaults
548 mode http
549 timeout connect 5000ms
550 timeout client 50000ms
551 timeout server 50000ms
552
553 listen http-in
554 bind *:80
555 server server1 127.0.0.1:8000 maxconn 32
556
557
558Assuming haproxy is in $PATH, test these configurations in a shell with:
559
Willy Tarreauccb289d2010-12-11 20:19:38 +0100560 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200561
562
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005633. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200564--------------------
565
566Parameters in the "global" section are process-wide and often OS-specific. They
567are generally set once for all and do not need being changed once correct. Some
568of them have command-line equivalents.
569
570The following keywords are supported in the "global" section :
571
572 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200573 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200574 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200575 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200576 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200577 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200578 - description
579 - deviceatlas-json-file
580 - deviceatlas-log-level
581 - deviceatlas-separator
582 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900583 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200584 - gid
585 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100586 - hard-stop-after
Christopher Fauleta99ff4d2019-07-22 16:18:24 +0200587 - h1-case-adjust
588 - h1-case-adjust-file
Willy Tarreau6a06a402007-07-15 20:15:28 +0200589 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200590 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100591 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200592 - lua-load
William Lallemand27edc4b2019-05-07 17:49:33 +0200593 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200594 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200595 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200596 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200597 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100598 - presetenv
599 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200600 - uid
601 - ulimit-n
602 - user
Willy Tarreau636848a2019-04-15 19:38:50 +0200603 - set-dumpable
Willy Tarreau1d549722016-02-16 12:41:57 +0100604 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200605 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200606 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200607 - ssl-default-bind-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200608 - ssl-default-bind-options
609 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200610 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200611 - ssl-default-server-options
612 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100613 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100614 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100615 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100616 - 51degrees-data-file
617 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200618 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200619 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200620 - wurfl-data-file
621 - wurfl-information-list
622 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200623 - wurfl-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100624
Willy Tarreau6a06a402007-07-15 20:15:28 +0200625 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200626 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200627 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200628 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100629 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100630 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100631 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200632 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200633 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200634 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200635 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200636 - noepoll
637 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +0000638 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +0200639 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100640 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300641 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000642 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +0100643 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +0200644 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200645 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200646 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000647 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000648 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200649 - tune.buffers.limit
650 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200651 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200652 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100653 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200654 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200655 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200656 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100657 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200658 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200659 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100660 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100661 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100662 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100663 - tune.lua.session-timeout
664 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200665 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100666 - tune.maxaccept
667 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200668 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200669 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200670 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100671 - tune.rcvbuf.client
672 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100673 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200674 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100675 - tune.sndbuf.client
676 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100677 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100678 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200679 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100680 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200681 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200682 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100683 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200684 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100685 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200686 - tune.vars.reqres-max-size
687 - tune.vars.sess-max-size
688 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100689 - tune.zlib.memlevel
690 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100691
Willy Tarreau6a06a402007-07-15 20:15:28 +0200692 * Debugging
693 - debug
694 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200695
696
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006973.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200698------------------------------------
699
Emeric Brunc8e8d122012-10-02 18:42:10 +0200700ca-base <dir>
701 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200702 relative path is used with "ca-file" or "crl-file" directives. Absolute
703 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200704
Willy Tarreau6a06a402007-07-15 20:15:28 +0200705chroot <jail dir>
706 Changes current directory to <jail dir> and performs a chroot() there before
707 dropping privileges. This increases the security level in case an unknown
708 vulnerability would be exploited, since it would make it very hard for the
709 attacker to exploit the system. This only works when the process is started
710 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100711 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100712
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100713cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
714 On Linux 2.6 and above, it is possible to bind a process or a thread to a
715 specific CPU set. This means that the process or the thread will never run on
716 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
717 sets. The first argument is a process set, eventually followed by a thread
718 set. These sets have the format
719
720 all | odd | even | number[-[number]]
721
722 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100723 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100724 ignored. It is possible to specify a range with two such number delimited by
725 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100726 "all", only odd numbers using "odd" or even numbers using "even", just like
727 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100728 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 +0100729 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100730 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100731 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100732 specified. Each "cpu-map" directive will replace the previous ones when they
733 overlap. A thread will be bound on the intersection of its mapping and the
734 one of the process on which it is attached. If the intersection is null, no
735 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100736
Christopher Fauletff4121f2017-11-22 16:38:49 +0100737 Ranges can be partially defined. The higher bound can be omitted. In such
738 case, it is replaced by the corresponding maximum value, 32 or 64 depending
739 on the machine's word size.
740
Christopher Faulet26028f62017-11-22 15:01:51 +0100741 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100742 automatically bind a process or a thread to a CPU by incrementing
743 process/thread and CPU sets. To be valid, both sets must have the same
744 size. No matter the declaration order of the CPU sets, it will be bound from
745 the lowest to the highest bound. Having a process and a thread range with the
746 "auto:" prefix is not supported. Only one range is supported, the other one
747 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100748
749 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100750 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
751
752 cpu-map 1/all 0-3 # bind all threads of the first process on the
753 # first 4 CPUs
754
755 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
756 # or "cpu-map 1-32 0-31" depending on the machine's
757 # word size.
758
Christopher Faulet26028f62017-11-22 15:01:51 +0100759 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100760 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100761 cpu-map auto:1-4 0-3
762 cpu-map auto:1-4 0-1 2-3
763 cpu-map auto:1-4 3 2 1 0
764
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100765 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
766 # and so on.
767 cpu-map auto:1/1-4 0-3
768 cpu-map auto:1/1-4 0-1 2-3
769 cpu-map auto:1/1-4 3 2 1 0
770
Davor Ocelice9ed2812017-12-25 17:49:28 +0100771 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100772 cpu-map auto:all 0-63
773 cpu-map auto:even 0-31
774 cpu-map auto:odd 32-63
775
776 # invalid cpu-map because process and CPU sets have different sizes.
777 cpu-map auto:1-4 0 # invalid
778 cpu-map auto:1 0-3 # invalid
779
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100780 # invalid cpu-map because automatic binding is used with a process range
781 # and a thread range.
782 cpu-map auto:all/all 0 # invalid
783 cpu-map auto:all/1-4 0 # invalid
784 cpu-map auto:1-4/all 0 # invalid
785
Emeric Brunc8e8d122012-10-02 18:42:10 +0200786crt-base <dir>
787 Assigns a default directory to fetch SSL certificates from when a relative
788 path is used with "crtfile" directives. Absolute locations specified after
789 "crtfile" prevail and ignore "crt-base".
790
Willy Tarreau6a06a402007-07-15 20:15:28 +0200791daemon
792 Makes the process fork into background. This is the recommended mode of
793 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100794 disabled by the command line "-db" argument. This option is ignored in
795 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200796
David Carlier8167f302015-06-01 13:50:06 +0200797deviceatlas-json-file <path>
798 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100799 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200800
801deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100802 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200803 optional and set to 0 by default if not set.
804
805deviceatlas-separator <char>
806 Sets the character separator for the API properties results. This directive
807 is optional and set to | by default if not set.
808
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100809deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200810 Sets the client cookie's name used for the detection if the DeviceAtlas
811 Client-side component was used during the request. This directive is optional
812 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100813
Simon Horman98637e52014-06-20 12:30:16 +0900814external-check
815 Allows the use of an external agent to perform health checks.
816 This is disabled by default as a security precaution.
817 See "option external-check".
818
Willy Tarreau6a06a402007-07-15 20:15:28 +0200819gid <number>
820 Changes the process' group ID to <number>. It is recommended that the group
821 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
822 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100823 Note that if haproxy is started from a user having supplementary groups, it
824 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200825 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100826
Willy Tarreau8b852462019-12-03 08:29:22 +0100827group <group name>
828 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
829 See also "gid" and "user".
830
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100831hard-stop-after <time>
832 Defines the maximum time allowed to perform a clean soft-stop.
833
834 Arguments :
835 <time> is the maximum time (by default in milliseconds) for which the
836 instance will remain alive when a soft-stop is received via the
837 SIGUSR1 signal.
838
839 This may be used to ensure that the instance will quit even if connections
840 remain opened during a soft-stop (for example with long timeouts for a proxy
841 in tcp mode). It applies both in TCP and HTTP mode.
842
843 Example:
844 global
845 hard-stop-after 30s
846
Christopher Fauleta99ff4d2019-07-22 16:18:24 +0200847h1-case-adjust <from> <to>
848 Defines the case adjustment to apply, when enabled, to the header name
849 <from>, to change it to <to> before sending it to HTTP/1 clients or
850 servers. <from> must be in lower case, and <from> and <to> must not differ
851 except for their case. It may be repeated if several header names need to be
852 ajusted. Duplicate entries are not allowed. If a lot of header names have to
853 be adjusted, it might be more convenient to use "h1-case-adjust-file".
854 Please note that no transformation will be applied unless "option
855 h1-case-adjust-bogus-client" or "option h1-case-adjust-bogus-server" is
856 specified in a proxy.
857
858 There is no standard case for header names because, as stated in RFC7230,
859 they are case-insensitive. So applications must handle them in a case-
860 insensitive manner. But some bogus applications violate the standards and
861 erroneously rely on the cases most commonly used by browsers. This problem
862 becomes critical with HTTP/2 because all header names must be exchanged in
863 lower case, and HAProxy follows the same convention. All header names are
864 sent in lower case to clients and servers, regardless of the HTTP version.
865
866 Applications which fail to properly process requests or responses may require
867 to temporarily use such workarounds to adjust header names sent to them for
868 the time it takes the application to be fixed. Please note that an
869 application which requires such workarounds might be vulnerable to content
870 smuggling attacks and must absolutely be fixed.
871
872 Example:
873 global
874 h1-case-adjust content-length Content-Length
875
876 See "h1-case-adjust-file", "option h1-case-adjust-bogus-client" and
877 "option h1-case-adjust-bogus-server".
878
879h1-case-adjust-file <hdrs-file>
880 Defines a file containing a list of key/value pairs used to adjust the case
881 of some header names before sending them to HTTP/1 clients or servers. The
882 file <hdrs-file> must contain 2 header names per line. The first one must be
883 in lower case and both must not differ except for their case. Lines which
884 start with '#' are ignored, just like empty lines. Leading and trailing tabs
885 and spaces are stripped. Duplicate entries are not allowed. Please note that
886 no transformation will be applied unless "option h1-case-adjust-bogus-client"
887 or "option h1-case-adjust-bogus-server" is specified in a proxy.
888
889 If this directive is repeated, only the last one will be processed. It is an
890 alternative to the directive "h1-case-adjust" if a lot of header names need
891 to be adjusted. Please read the risks associated with using this.
892
893 See "h1-case-adjust", "option h1-case-adjust-bogus-client" and
894 "option h1-case-adjust-bogus-server".
895
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200896log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
897 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100898 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100899 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100900 configured with "log global".
901
902 <address> can be one of:
903
Willy Tarreau2769aa02007-12-27 18:26:09 +0100904 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100905 no port is specified, 514 is used by default (the standard syslog
906 port).
907
David du Colombier24bb5f52011-03-17 10:40:23 +0100908 - An IPv6 address followed by a colon and optionally a UDP port. If
909 no port is specified, 514 is used by default (the standard syslog
910 port).
911
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100912 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +0100913 considerations for chroot (be sure the path is accessible inside
914 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100915 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100916
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100917 - A file descriptor number in the form "fd@<number>", which may point
918 to a pipe, terminal, or socket. In this case unbuffered logs are used
919 and one writev() call per log is performed. This is a bit expensive
920 but acceptable for most workloads. Messages sent this way will not be
921 truncated but may be dropped, in which case the DroppedLogs counter
922 will be incremented. The writev() call is atomic even on pipes for
923 messages up to PIPE_BUF size, which POSIX recommends to be at least
924 512 and which is 4096 bytes on most modern operating systems. Any
925 larger message may be interleaved with messages from other processes.
926 Exceptionally for debugging purposes the file descriptor may also be
927 directed to a file, but doing so will significantly slow haproxy down
928 as non-blocking calls will be ignored. Also there will be no way to
929 purge nor rotate this file without restarting the process. Note that
930 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +0100931 for use with a TCP syslog server. See also the "short" and "raw"
932 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100933
934 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
935 "fd@2", see above.
936
William Lallemandb2f07452015-05-12 14:27:13 +0200937 You may want to reference some environment variables in the address
938 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100939
Willy Tarreau18324f52014-06-27 18:10:07 +0200940 <length> is an optional maximum line length. Log lines larger than this value
941 will be truncated before being sent. The reason is that syslog
942 servers act differently on log line length. All servers support the
943 default value of 1024, but some servers simply drop larger lines
944 while others do log them. If a server supports long lines, it may
945 make sense to set this value here in order to avoid truncating long
946 lines. Similarly, if a server drops long lines, it is preferable to
947 truncate them before sending them. Accepted values are 80 to 65535
948 inclusive. The default value of 1024 is generally fine for all
949 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +0100950 JSON-formatted logs may require larger values. You may also need to
951 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +0200952
Dragan Dosen7ad31542015-09-28 17:16:47 +0200953 <format> is the log format used when generating syslog messages. It may be
954 one of the following :
955
956 rfc3164 The RFC3164 syslog message format. This is the default.
957 (https://tools.ietf.org/html/rfc3164)
958
959 rfc5424 The RFC5424 syslog message format.
960 (https://tools.ietf.org/html/rfc5424)
961
Willy Tarreaue8746a02018-11-12 08:45:00 +0100962 short A message containing only a level between angle brackets such as
963 '<3>', followed by the text. The PID, date, time, process name
964 and system name are omitted. This is designed to be used with a
965 local log server. This format is compatible with what the systemd
966 logger consumes.
967
Willy Tarreauc1b06452018-11-12 11:57:56 +0100968 raw A message containing only the text. The level, PID, date, time,
969 process name and system name are omitted. This is designed to be
970 used in containers or during development, where the severity only
971 depends on the file descriptor used (stdout/stderr).
972
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200973 <ranges> A list of comma-separated ranges to identify the logs to sample.
974 This is used to balance the load of the logs to send to the log
975 server. The limits of the ranges cannot be null. They are numbered
976 from 1. The size or period (in number of logs) of the sample must be
977 set with <sample_size> parameter.
978
979 <sample_size>
980 The size of the sample in number of logs to consider when balancing
981 their logging loads. It is used to balance the load of the logs to
982 send to the syslog server. This size must be greater or equal to the
983 maximum of the high limits of the ranges.
984 (see also <ranges> parameter).
985
Robert Tsai81ae1952007-12-05 10:47:29 +0100986 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200987
Willy Tarreaue8746a02018-11-12 08:45:00 +0100988 kern user mail daemon auth syslog lpr news
989 uucp cron auth2 ftp ntp audit alert cron2
990 local0 local1 local2 local3 local4 local5 local6 local7
991
Willy Tarreauc1b06452018-11-12 11:57:56 +0100992 Note that the facility is ignored for the "short" and "raw"
993 formats, but still required as a positional field. It is
994 recommended to use "daemon" in this case to make it clear that
995 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200996
997 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200998 all messages are sent. If a maximum level is specified, only messages with a
999 severity at least as important as this level will be sent. An optional minimum
1000 level can be specified. If it is set, logs emitted with a more severe level
1001 than this one will be capped to this level. This is used to avoid sending
1002 "emerg" messages on all terminals on some default syslog configurations.
1003 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +02001004
Cyril Bontédc4d9032012-04-08 21:57:39 +02001005 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +02001006
Joe Williamsdf5b38f2010-12-29 17:05:48 +01001007log-send-hostname [<string>]
1008 Sets the hostname field in the syslog header. If optional "string" parameter
1009 is set the header is set to the string contents, otherwise uses the hostname
1010 of the system. Generally used if one is not relaying logs through an
1011 intermediate syslog server or for simply customizing the hostname printed in
1012 the logs.
1013
Kevinm48936af2010-12-22 16:08:21 +00001014log-tag <string>
1015 Sets the tag field in the syslog header to this string. It defaults to the
1016 program name as launched from the command line, which usually is "haproxy".
1017 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +01001018 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +00001019
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001020lua-load <file>
1021 This global directive loads and executes a Lua file. This directive can be
1022 used multiple times.
1023
William Lallemand4cfede82017-11-24 22:02:34 +01001024master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +02001025 Master-worker mode. It is equivalent to the command line "-W" argument.
1026 This mode will launch a "master" which will monitor the "workers". Using
1027 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001028 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +02001029 or daemon mode. It is recommended to use this mode with multiprocess and
1030 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +01001031 By default, if a worker exits with a bad return code, in the case of a
1032 segfault for example, all workers will be killed, and the master will leave.
1033 It is convenient to combine this behavior with Restart=on-failure in a
1034 systemd unit file in order to relaunch the whole process. If you don't want
1035 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +02001036
William Lallemand4cfede82017-11-24 22:02:34 +01001037 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +02001038
William Lallemand27edc4b2019-05-07 17:49:33 +02001039mworker-max-reloads <number>
1040 In master-worker mode, this option limits the number of time a worker can
John Roesler27c754c2019-07-10 15:45:51 -05001041 survive to a reload. If the worker did not leave after a reload, once its
William Lallemand27edc4b2019-05-07 17:49:33 +02001042 number of reloads is greater than this number, the worker will receive a
1043 SIGTERM. This option helps to keep under control the number of workers.
1044 See also "show proc" in the Management Guide.
1045
Willy Tarreau6a06a402007-07-15 20:15:28 +02001046nbproc <number>
1047 Creates <number> processes when going daemon. This requires the "daemon"
1048 mode. By default, only one process is created, which is the recommended mode
1049 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +01001050 process, it may be needed to fork multiple daemons. When set to a value
1051 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreau1f672a82019-01-26 14:20:55 +01001052 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon" and
1053 "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001054
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001055nbthread <number>
1056 This setting is only available when support for threads was built in. It
Willy Tarreau26f6ae12019-02-02 12:56:15 +01001057 makes haproxy run on <number> threads. This is exclusive with "nbproc". While
1058 "nbproc" historically used to be the only way to use multiple processors, it
1059 also involved a number of shortcomings related to the lack of synchronization
1060 between processes (health-checks, peers, stick-tables, stats, ...) which do
1061 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +01001062 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
1063 HAProxy is started in foreground. On some platforms supporting CPU affinity,
1064 when nbproc is not used, the default "nbthread" value is automatically set to
1065 the number of CPUs the process is bound to upon startup. This means that the
1066 thread count can easily be adjusted from the calling process using commands
1067 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1068 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001069
Willy Tarreau6a06a402007-07-15 20:15:28 +02001070pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001071 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +02001072 the "-p" command line argument. The file must be accessible to the user
1073 starting the process. See also "daemon".
1074
Willy Tarreau1d549722016-02-16 12:41:57 +01001075presetenv <name> <value>
1076 Sets environment variable <name> to value <value>. If the variable exists, it
1077 is NOT overwritten. The changes immediately take effect so that the next line
1078 in the configuration file sees the new value. See also "setenv", "resetenv",
1079 and "unsetenv".
1080
1081resetenv [<name> ...]
1082 Removes all environment variables except the ones specified in argument. It
1083 allows to use a clean controlled environment before setting new values with
1084 setenv or unsetenv. Please note that some internal functions may make use of
1085 some environment variables, such as time manipulation functions, but also
1086 OpenSSL or even external checks. This must be used with extreme care and only
1087 after complete validation. The changes immediately take effect so that the
1088 next line in the configuration file sees the new environment. See also
1089 "setenv", "presetenv", and "unsetenv".
1090
Christopher Fauletff4121f2017-11-22 16:38:49 +01001091stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001092 Limits the stats socket to a certain set of processes numbers. By default the
1093 stats socket is bound to all processes, causing a warning to be emitted when
1094 nbproc is greater than 1 because there is no way to select the target process
1095 when connecting. However, by using this setting, it becomes possible to pin
1096 the stats socket to a specific set of processes, typically the first one. The
1097 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001098 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001099 word size (32 or 64). Ranges can be partially defined. The higher bound can
1100 be omitted. In such case, it is replaced by the corresponding maximum
1101 value. A better option consists in using the "process" setting of the "stats
1102 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001103
Baptiste Assmann5626f482015-08-23 10:00:10 +02001104server-state-base <directory>
1105 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001106 file names which do not start with a '/'. See also "server-state-file",
1107 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001108
1109server-state-file <file>
1110 Specifies the path to the file containing state of servers. If the path starts
1111 with a slash ('/'), it is considered absolute, otherwise it is considered
1112 relative to the directory specified using "server-state-base" (if set) or to
1113 the current directory. Before reloading HAProxy, it is possible to save the
1114 servers' current state using the stats command "show servers state". The
1115 output of this command must be written in the file pointed by <file>. When
1116 starting up, before handling traffic, HAProxy will read, load and apply state
1117 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001118 configuration. See also "server-state-base" and "show servers state",
1119 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001120
Willy Tarreau1d549722016-02-16 12:41:57 +01001121setenv <name> <value>
1122 Sets environment variable <name> to value <value>. If the variable exists, it
1123 is overwritten. The changes immediately take effect so that the next line in
1124 the configuration file sees the new value. See also "presetenv", "resetenv",
1125 and "unsetenv".
1126
Willy Tarreau636848a2019-04-15 19:38:50 +02001127set-dumpable
1128 This option is better left disabled by default and enabled only upon a
1129 developer's request. It has no impact on performance nor stability but will
1130 try hard to re-enable core dumps that were possibly disabled by file size
1131 limitations (ulimit -f), core size limitations (ulimit -c), or "dumpability"
1132 of a process after changing its UID/GID (such as /proc/sys/fs/suid_dumpable
1133 on Linux). Core dumps might still be limited by the current directory's
1134 permissions (check what directory the file is started from), the chroot
1135 directory's permission (it may be needed to temporarily disable the chroot
1136 directive or to move it to a dedicated writable location), or any other
1137 system-specific constraint. For example, some Linux flavours are notorious
1138 for replacing the default core file with a path to an executable not even
1139 installed on the system (check /proc/sys/kernel/core_pattern). Often, simply
1140 writing "core", "core.%p" or "/var/log/core/core.%p" addresses the issue.
1141 When trying to enable this option waiting for a rare issue to re-appear, it's
1142 often a good idea to first try to obtain such a dump by issuing, for example,
1143 "kill -11" to the haproxy process and verify that it leaves a core where
1144 expected when dying.
1145
Willy Tarreau610f04b2014-02-13 11:36:41 +01001146ssl-default-bind-ciphers <ciphers>
1147 This setting is only available when support for OpenSSL was built in. It sets
1148 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001149 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001150 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001151 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1152 information and recommendations see e.g.
1153 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1154 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1155 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1156 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001157
1158ssl-default-bind-ciphersuites <ciphersuites>
1159 This setting is only available when support for OpenSSL was built in and
1160 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1161 describing the list of cipher algorithms ("cipher suite") that are negotiated
1162 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1163 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001164 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1165 cipher configuration for TLSv1.2 and earlier, please check the
1166 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001167 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001168
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001169ssl-default-bind-options [<option>]...
1170 This setting is only available when support for OpenSSL was built in. It sets
1171 default ssl-options to force on all "bind" lines. Please check the "bind"
1172 keyword to see available options.
1173
1174 Example:
1175 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001176 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001177
Willy Tarreau610f04b2014-02-13 11:36:41 +01001178ssl-default-server-ciphers <ciphers>
1179 This setting is only available when support for OpenSSL was built in. It
1180 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001181 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001182 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001183 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1184 information and recommendations see e.g.
1185 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1186 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1187 For TLSv1.3 cipher configuration, please check the
1188 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1189 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001190
1191ssl-default-server-ciphersuites <ciphersuites>
1192 This setting is only available when support for OpenSSL was built in and
1193 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1194 string describing the list of cipher algorithms that are negotiated during
1195 the TLSv1.3 handshake with the server, for all "server" lines which do not
1196 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001197 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1198 cipher configuration for TLSv1.2 and earlier, please check the
1199 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1200 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001201
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001202ssl-default-server-options [<option>]...
1203 This setting is only available when support for OpenSSL was built in. It sets
1204 default ssl-options to force on all "server" lines. Please check the "server"
1205 keyword to see available options.
1206
Remi Gacogne47783ef2015-05-29 15:53:22 +02001207ssl-dh-param-file <file>
1208 This setting is only available when support for OpenSSL was built in. It sets
1209 the default DH parameters that are used during the SSL/TLS handshake when
1210 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001211 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001212 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001213 are not specified either by using ssl-dh-param-file or by setting them
1214 directly in the certificate file, pre-generated DH parameters of the size
1215 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1216 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001217 Custom DH parameters may be generated by using the OpenSSL command
1218 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1219 parameters should not be considered secure anymore.
1220
Emeric Brun850efd52014-01-29 12:24:34 +01001221ssl-server-verify [none|required]
1222 The default behavior for SSL verify on servers side. If specified to 'none',
1223 servers certificates are not verified. The default is 'required' except if
1224 forced using cmdline option '-dV'.
1225
Willy Tarreauabb175f2012-09-24 12:43:26 +02001226stats socket [<address:port>|<path>] [param*]
1227 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1228 Connections to this socket will return various statistics outputs and even
1229 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001230 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001231 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001232
Willy Tarreauabb175f2012-09-24 12:43:26 +02001233 All parameters supported by "bind" lines are supported, for instance to
1234 restrict access to some users or their access rights. Please consult
1235 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001236
1237stats timeout <timeout, in milliseconds>
1238 The default timeout on the stats socket is set to 10 seconds. It is possible
1239 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001240 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001241
1242stats maxconn <connections>
1243 By default, the stats socket is limited to 10 concurrent connections. It is
1244 possible to change this value with "stats maxconn".
1245
Willy Tarreau6a06a402007-07-15 20:15:28 +02001246uid <number>
1247 Changes the process' user ID to <number>. It is recommended that the user ID
1248 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1249 be started with superuser privileges in order to be able to switch to another
1250 one. See also "gid" and "user".
1251
1252ulimit-n <number>
1253 Sets the maximum number of per-process file-descriptors to <number>. By
1254 default, it is automatically computed, so it is recommended not to use this
1255 option.
1256
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001257unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1258 [ group <group> ] [ gid <gid> ]
1259
1260 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1261 This is mainly used to simplify declaration of those UNIX sockets and reduce
1262 the risk of errors, since those settings are most commonly required but are
1263 also process-specific. The <prefix> setting can be used to force all socket
1264 path to be relative to that directory. This might be needed to access another
1265 component's chroot. Note that those paths are resolved before haproxy chroots
1266 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1267 all have the same meaning as their homonyms used by the "bind" statement. If
1268 both are specified, the "bind" statement has priority, meaning that the
1269 "unix-bind" settings may be seen as process-wide default settings.
1270
Willy Tarreau1d549722016-02-16 12:41:57 +01001271unsetenv [<name> ...]
1272 Removes environment variables specified in arguments. This can be useful to
1273 hide some sensitive information that are occasionally inherited from the
1274 user's environment during some operations. Variables which did not exist are
1275 silently ignored so that after the operation, it is certain that none of
1276 these variables remain. The changes immediately take effect so that the next
1277 line in the configuration file will not see these variables. See also
1278 "setenv", "presetenv", and "resetenv".
1279
Willy Tarreau6a06a402007-07-15 20:15:28 +02001280user <user name>
1281 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1282 See also "uid" and "group".
1283
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001284node <name>
1285 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1286
1287 This statement is useful in HA configurations where two or more processes or
1288 servers share the same IP address. By setting a different node-name on all
1289 nodes, it becomes easy to immediately spot what server is handling the
1290 traffic.
1291
1292description <text>
1293 Add a text that describes the instance.
1294
1295 Please note that it is required to escape certain characters (# for example)
1296 and this text is inserted into a html page so you should avoid using
1297 "<" and ">" characters.
1298
Thomas Holmesdb04f192015-05-18 13:21:39 +0100129951degrees-data-file <file path>
1300 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001301 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001302
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001303 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001304 compiled with USE_51DEGREES.
1305
Ben Shillitof25e8e52016-12-02 14:25:37 +0000130651degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001307 A list of 51Degrees property names to be load from the dataset. A full list
1308 of names is available on the 51Degrees website:
1309 https://51degrees.com/resources/property-dictionary
1310
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001311 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001312 compiled with USE_51DEGREES.
1313
Dragan Dosen93b38d92015-06-29 16:43:25 +0200131451degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001315 A char that will be appended to every property value in a response header
1316 containing 51Degrees results. If not set that will be set as ','.
1317
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001318 Please note that this option is only available when haproxy has been
1319 compiled with USE_51DEGREES.
1320
132151degrees-cache-size <number>
1322 Sets the size of the 51Degrees converter cache to <number> entries. This
1323 is an LRU cache which reminds previous device detections and their results.
1324 By default, this cache is disabled.
1325
1326 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001327 compiled with USE_51DEGREES.
1328
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001329wurfl-data-file <file path>
1330 The path of the WURFL data file to provide device detection services. The
1331 file should be accessible by HAProxy with relevant permissions.
1332
1333 Please note that this option is only available when haproxy has been compiled
1334 with USE_WURFL=1.
1335
1336wurfl-information-list [<capability>]*
1337 A space-delimited list of WURFL capabilities, virtual capabilities, property
1338 names we plan to use in injected headers. A full list of capability and
1339 virtual capability names is available on the Scientiamobile website :
1340
1341 https://www.scientiamobile.com/wurflCapability
1342
1343 Valid WURFL properties are:
1344 - wurfl_id Contains the device ID of the matched device.
1345
1346 - wurfl_root_id Contains the device root ID of the matched
1347 device.
1348
1349 - wurfl_isdevroot Tells if the matched device is a root device.
1350 Possible values are "TRUE" or "FALSE".
1351
1352 - wurfl_useragent The original useragent coming with this
1353 particular web request.
1354
1355 - wurfl_api_version Contains a string representing the currently
1356 used Libwurfl API version.
1357
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001358 - wurfl_info A string containing information on the parsed
1359 wurfl.xml and its full path.
1360
1361 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1362 WURFL has been loaded successfully.
1363
1364 - wurfl_normalized_useragent The normalized useragent.
1365
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001366 Please note that this option is only available when haproxy has been compiled
1367 with USE_WURFL=1.
1368
1369wurfl-information-list-separator <char>
1370 A char that will be used to separate values in a response header containing
1371 WURFL results. If not set that a comma (',') will be used by default.
1372
1373 Please note that this option is only available when haproxy has been compiled
1374 with USE_WURFL=1.
1375
1376wurfl-patch-file [<file path>]
1377 A list of WURFL patch file paths. Note that patches are loaded during startup
1378 thus before the chroot.
1379
1380 Please note that this option is only available when haproxy has been compiled
1381 with USE_WURFL=1.
1382
paulborilebad132c2019-04-18 11:57:04 +02001383wurfl-cache-size <size>
1384 Sets the WURFL Useragent cache size. For faster lookups, already processed user
1385 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001386 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02001387 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001388
1389 Please note that this option is only available when haproxy has been compiled
1390 with USE_WURFL=1.
1391
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013923.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001393-----------------------
1394
Willy Tarreaubeb859a2018-11-22 18:07:59 +01001395busy-polling
1396 In some situations, especially when dealing with low latency on processors
1397 supporting a variable frequency or when running inside virtual machines, each
1398 time the process waits for an I/O using the poller, the processor goes back
1399 to sleep or is offered to another VM for a long time, and it causes
1400 excessively high latencies. This option provides a solution preventing the
1401 processor from sleeping by always using a null timeout on the pollers. This
1402 results in a significant latency reduction (30 to 100 microseconds observed)
1403 at the expense of a risk to overheat the processor. It may even be used with
1404 threads, in which case improperly bound threads may heavily conflict,
1405 resulting in a worse performance and high values for the CPU stolen fields
1406 in "show info" output, indicating which threads are misconfigured. It is
1407 important not to let the process run on the same processor as the network
1408 interrupts when this option is used. It is also better to avoid using it on
1409 multiple CPU threads sharing the same core. This option is disabled by
1410 default. If it has been enabled, it may still be forcibly disabled by
1411 prefixing it with the "no" keyword. It is ignored by the "select" and
1412 "poll" pollers.
1413
William Dauchy857b9432019-12-28 15:36:02 +01001414 This option is automatically disabled on old processes in the context of
1415 seamless reload; it avoids too much cpu conflicts when multiple processes
1416 stay around for some time waiting for the end of their current connections.
1417
Willy Tarreau1746eec2014-04-25 10:46:47 +02001418max-spread-checks <delay in milliseconds>
1419 By default, haproxy tries to spread the start of health checks across the
1420 smallest health check interval of all the servers in a farm. The principle is
1421 to avoid hammering services running on the same server. But when using large
1422 check intervals (10 seconds or more), the last servers in the farm take some
1423 time before starting to be tested, which can be a problem. This parameter is
1424 used to enforce an upper bound on delay between the first and the last check,
1425 even if the servers' check intervals are larger. When servers run with
1426 shorter intervals, their intervals will be respected though.
1427
Willy Tarreau6a06a402007-07-15 20:15:28 +02001428maxconn <number>
1429 Sets the maximum per-process number of concurrent connections to <number>. It
1430 is equivalent to the command-line argument "-n". Proxies will stop accepting
1431 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001432 automatically adjusted according to this value. See also "ulimit-n". Note:
1433 the "select" poller cannot reliably use more than 1024 file descriptors on
1434 some platforms. If your platform only supports select and reports "select
1435 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01001436 below 500 in general). If this value is not set, it will automatically be
1437 calculated based on the current file descriptors limit reported by the
1438 "ulimit -n" command, possibly reduced to a lower value if a memory limit
1439 is enforced, based on the buffer size, memory allocated to compression, SSL
1440 cache size, and use or not of SSL and the associated maxsslconn (which can
1441 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001442
Willy Tarreau81c25d02011-09-07 15:17:21 +02001443maxconnrate <number>
1444 Sets the maximum per-process number of connections per second to <number>.
1445 Proxies will stop accepting connections when this limit is reached. It can be
1446 used to limit the global capacity regardless of each frontend capacity. It is
1447 important to note that this can only be used as a service protection measure,
1448 as there will not necessarily be a fair share between frontends when the
1449 limit is reached, so it's a good idea to also limit each frontend to some
1450 value close to its expected share. Also, lowering tune.maxaccept can improve
1451 fairness.
1452
William Lallemandd85f9172012-11-09 17:05:39 +01001453maxcomprate <number>
1454 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001455 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001456 level will be decreased during the session. If the maximum is reached at the
1457 beginning of a session, the session will not compress at all. If the maximum
1458 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001459 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001460 default value.
1461
William Lallemand072a2bf2012-11-20 17:01:01 +01001462maxcompcpuusage <number>
1463 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1464 for new requests or decreasing the compression level of current requests.
1465 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1466 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1467 case of multiple processes (nbproc > 1), each process manages its individual
1468 usage. A value of 100 disable the limit. The default value is 100. Setting
1469 a lower value will prevent the compression work from slowing the whole
1470 process down and from introducing high latencies.
1471
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001472maxpipes <number>
1473 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1474 are only used by kernel-based tcp splicing. Since a pipe contains two file
1475 descriptors, the "ulimit-n" value will be increased accordingly. The default
1476 value is maxconn/4, which seems to be more than enough for most heavy usages.
1477 The splice code dynamically allocates and releases pipes, and can fall back
1478 to standard copy, so setting this value too low may only impact performance.
1479
Willy Tarreau93e7c002013-10-07 18:51:07 +02001480maxsessrate <number>
1481 Sets the maximum per-process number of sessions per second to <number>.
1482 Proxies will stop accepting connections when this limit is reached. It can be
1483 used to limit the global capacity regardless of each frontend capacity. It is
1484 important to note that this can only be used as a service protection measure,
1485 as there will not necessarily be a fair share between frontends when the
1486 limit is reached, so it's a good idea to also limit each frontend to some
1487 value close to its expected share. Also, lowering tune.maxaccept can improve
1488 fairness.
1489
Willy Tarreau403edff2012-09-06 11:58:37 +02001490maxsslconn <number>
1491 Sets the maximum per-process number of concurrent SSL connections to
1492 <number>. By default there is no SSL-specific limit, which means that the
1493 global maxconn setting will apply to all connections. Setting this limit
1494 avoids having openssl use too much memory and crash when malloc returns NULL
1495 (since it unfortunately does not reliably check for such conditions). Note
1496 that the limit applies both to incoming and outgoing connections, so one
1497 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001498 If this value is not set, but a memory limit is enforced, this value will be
1499 automatically computed based on the memory limit, maxconn, the buffer size,
1500 memory allocated to compression, SSL cache size, and use of SSL in either
1501 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1502 when there is a memory limit, haproxy will automatically adjust these values
1503 so that 100% of the connections can be made over SSL with no risk, and will
1504 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001505
Willy Tarreaue43d5322013-10-07 20:01:52 +02001506maxsslrate <number>
1507 Sets the maximum per-process number of SSL sessions per second to <number>.
1508 SSL listeners will stop accepting connections when this limit is reached. It
1509 can be used to limit the global SSL CPU usage regardless of each frontend
1510 capacity. It is important to note that this can only be used as a service
1511 protection measure, as there will not necessarily be a fair share between
1512 frontends when the limit is reached, so it's a good idea to also limit each
1513 frontend to some value close to its expected share. It is also important to
1514 note that the sessions are accounted before they enter the SSL stack and not
1515 after, which also protects the stack against bad handshakes. Also, lowering
1516 tune.maxaccept can improve fairness.
1517
William Lallemand9d5f5482012-11-07 16:12:57 +01001518maxzlibmem <number>
1519 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1520 When the maximum amount is reached, future sessions will not compress as long
1521 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001522 The default value is 0. The value is available in bytes on the UNIX socket
1523 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1524 "ZlibMemUsage" in bytes.
1525
Willy Tarreau6a06a402007-07-15 20:15:28 +02001526noepoll
1527 Disables the use of the "epoll" event polling system on Linux. It is
1528 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001529 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001530
1531nokqueue
1532 Disables the use of the "kqueue" event polling system on BSD. It is
1533 equivalent to the command-line argument "-dk". The next polling system
1534 used will generally be "poll". See also "nopoll".
1535
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001536noevports
1537 Disables the use of the event ports event polling system on SunOS systems
1538 derived from Solaris 10 and later. It is equivalent to the command-line
1539 argument "-dv". The next polling system used will generally be "poll". See
1540 also "nopoll".
1541
Willy Tarreau6a06a402007-07-15 20:15:28 +02001542nopoll
1543 Disables the use of the "poll" event polling system. It is equivalent to the
1544 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001545 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001546 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
1547 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001548
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001549nosplice
1550 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001551 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001552 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001553 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001554 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1555 be used. This option makes it easier to globally disable kernel splicing in
1556 case of doubt. See also "option splice-auto", "option splice-request" and
1557 "option splice-response".
1558
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001559nogetaddrinfo
1560 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1561 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1562
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001563noreuseport
1564 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1565 command line argument "-dR".
1566
Willy Tarreaud2d33482019-04-25 17:09:07 +02001567profiling.tasks { auto | on | off }
1568 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
1569 the profiling automatically turns on a thread when it starts to suffer from
1570 an average latency of 1000 microseconds or higher as reported in the
1571 "avg_loop_us" activity field, and automatically turns off when the latency
John Roesler27c754c2019-07-10 15:45:51 -05001572 returns below 990 microseconds (this value is an average over the last 1024
Willy Tarreaud2d33482019-04-25 17:09:07 +02001573 loops so it does not vary quickly and tends to significantly smooth short
1574 spikes). It may also spontaneously trigger from time to time on overloaded
1575 systems, containers, or virtual machines, or when the system swaps (which
1576 must absolutely never happen on a load balancer).
1577
1578 CPU profiling per task can be very convenient to report where the time is
1579 spent and which requests have what effect on which other request. Enabling
1580 it will typically affect the overall's performance by less than 1%, thus it
1581 is recommended to leave it to the default 'auto' value so that it only
1582 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01001583 supporting the clock_gettime(2) syscall with clock identifiers
1584 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
1585 be zero. This option may be changed at run time using "set profiling" on the
1586 CLI.
1587
Willy Tarreaufe255b72007-10-14 23:09:26 +02001588spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001589 Sometimes it is desirable to avoid sending agent and health checks to
1590 servers at exact intervals, for instance when many logical servers are
1591 located on the same physical server. With the help of this parameter, it
1592 becomes possible to add some randomness in the check interval between 0
1593 and +/- 50%. A value between 2 and 5 seems to show good results. The
1594 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001595
Davor Ocelice9ed2812017-12-25 17:49:28 +01001596ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001597 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001598 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001599 multiple times, it will simply enable multiple crypto engines. Referencing an
1600 unsupported engine will prevent haproxy from starting. Note that many engines
1601 will lead to lower HTTPS performance than pure software with recent
1602 processors. The optional command "algo" sets the default algorithms an ENGINE
1603 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001604 of "ALL" uses the engine for all cryptographic operations. If no list of
1605 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001606 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1607 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1608 openssl configuration file uses:
1609 https://www.openssl.org/docs/man1.0.2/apps/config.html
1610
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001611ssl-mode-async
1612 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001613 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001614 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1615 doesn't support moving read/write buffers and is not compliant with
1616 haproxy's buffer management. So the asynchronous mode is disabled on
John Roesler27c754c2019-07-10 15:45:51 -05001617 read/write operations (it is only enabled during initial and renegotiation
Emeric Brunb5e42a82017-06-06 12:35:14 +00001618 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001619
Willy Tarreau33cb0652014-12-23 22:52:37 +01001620tune.buffers.limit <number>
1621 Sets a hard limit on the number of buffers which may be allocated per process.
1622 The default value is zero which means unlimited. The minimum non-zero value
1623 will always be greater than "tune.buffers.reserve" and should ideally always
1624 be about twice as large. Forcing this value can be particularly useful to
1625 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001626 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001627 another one to be released by another session. Since buffers are dynamically
1628 allocated and released, the waiting time is very short and not perceptible
1629 provided that limits remain reasonable. In fact sometimes reducing the limit
1630 may even increase performance by increasing the CPU cache's efficiency. Tests
1631 have shown good results on average HTTP traffic with a limit to 1/10 of the
1632 expected global maxconn setting, which also significantly reduces memory
1633 usage. The memory savings come from the fact that a number of connections
1634 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1635 advised to do so by an haproxy core developer.
1636
Willy Tarreau1058ae72014-12-23 22:40:40 +01001637tune.buffers.reserve <number>
1638 Sets the number of buffers which are pre-allocated and reserved for use only
1639 during memory shortage conditions resulting in failed memory allocations. The
1640 minimum value is 2 and is also the default. There is no reason a user would
1641 want to change this value, it's mostly aimed at haproxy core developers.
1642
Willy Tarreau27a674e2009-08-17 07:23:33 +02001643tune.bufsize <number>
1644 Sets the buffer size to this size (in bytes). Lower values allow more
1645 sessions to coexist in the same amount of RAM, and higher values allow some
1646 applications with very large cookies to work. The default value is 16384 and
1647 can be changed at build time. It is strongly recommended not to change this
1648 from the default value, as very low values will break some services such as
1649 statistics, and values larger than default size will increase memory usage,
1650 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001651 parameter should be decreased by the same factor as this one is increased. In
1652 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1653 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001654 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Willy Tarreauc77d3642018-12-12 06:19:42 +01001655 than this size, haproxy will return HTTP 502 (Bad Gateway). Note that the
1656 value set using this parameter will automatically be rounded up to the next
1657 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02001658
Willy Tarreau43961d52010-10-04 20:39:20 +02001659tune.chksize <number>
1660 Sets the check buffer size to this size (in bytes). Higher values may help
1661 find string or regex patterns in very large pages, though doing so may imply
1662 more memory and CPU usage. The default value is 16384 and can be changed at
1663 build time. It is not recommended to change this value, but to use better
1664 checks whenever possible.
1665
William Lallemandf3747832012-11-09 12:33:10 +01001666tune.comp.maxlevel <number>
1667 Sets the maximum compression level. The compression level affects CPU
1668 usage during compression. This value affects CPU usage during compression.
1669 Each session using compression initializes the compression algorithm with
1670 this value. The default value is 1.
1671
Willy Tarreauc299e1e2019-02-27 11:35:12 +01001672tune.fail-alloc
1673 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
1674 allocation attempt fails. Must be between 0 (no failure) and 100 (no
1675 success). This is useful to debug and make sure memory failures are handled
1676 gracefully.
1677
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001678tune.h2.header-table-size <number>
1679 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1680 cannot be larger than 65536 bytes. A larger value may help certain clients
1681 send more compact requests, depending on their capabilities. This amount of
1682 memory is consumed for each HTTP/2 connection. It is recommended not to
1683 change it.
1684
Willy Tarreaue6baec02017-07-27 11:45:11 +02001685tune.h2.initial-window-size <number>
1686 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001687 can upload before waiting for an acknowledgment from haproxy. This setting
1688 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001689 The default value is 65535, which roughly allows up to 5 Mbps of upload
1690 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1691 over a 1-ms local network. It can make sense to increase this value to allow
1692 faster uploads, or to reduce it to increase fairness when dealing with many
1693 clients. It doesn't affect resource usage.
1694
Willy Tarreau5242ef82017-07-27 11:47:28 +02001695tune.h2.max-concurrent-streams <number>
1696 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1697 number of outstanding requests on a single connection). The default value is
1698 100. A larger one may slightly improve page load time for complex sites when
1699 visited over high latency networks, but increases the amount of resources a
1700 single client may allocate. A value of zero disables the limit so a single
1701 client may create as many streams as allocatable by haproxy. It is highly
1702 recommended not to change this value.
1703
Willy Tarreaua24b35c2019-02-21 13:24:36 +01001704tune.h2.max-frame-size <number>
1705 Sets the HTTP/2 maximum frame size that haproxy announces it is willing to
1706 receive to its peers. The default value is the largest between 16384 and the
1707 buffer size (tune.bufsize). In any case, haproxy will not announce support
1708 for frame sizes larger than buffers. The main purpose of this setting is to
1709 allow to limit the maximum frame size setting when using large buffers. Too
1710 large frame sizes might have performance impact or cause some peers to
1711 misbehave. It is highly recommended not to change this value.
1712
Willy Tarreau193b8c62012-11-22 00:17:38 +01001713tune.http.cookielen <number>
1714 Sets the maximum length of captured cookies. This is the maximum value that
1715 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1716 will automatically be truncated to this one. It is important not to set too
1717 high a value because all cookie captures still allocate this size whatever
1718 their configured value (they share a same pool). This value is per request
1719 per response, so the memory allocated is twice this value per connection.
1720 When not specified, the limit is set to 63 characters. It is recommended not
1721 to change this value.
1722
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001723tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001724 Sets the maximum length of request URI in logs. This prevents truncating long
1725 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001726 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001727 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001728 configuration directives too.
1729 The default value is 1024.
1730
Willy Tarreauac1932d2011-10-24 19:14:41 +02001731tune.http.maxhdr <number>
1732 Sets the maximum number of headers in a request. When a request comes with a
1733 number of headers greater than this value (including the first line), it is
1734 rejected with a "400 Bad Request" status code. Similarly, too large responses
1735 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1736 for all usages, considering that the widely deployed Apache server uses the
1737 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001738 a buggy application to work by the time it gets fixed. The accepted range is
1739 1..32767. Keep in mind that each new header consumes 32bits of memory for
1740 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001741
Willy Tarreau7e312732014-02-12 16:35:14 +01001742tune.idletimer <timeout>
1743 Sets the duration after which haproxy will consider that an empty buffer is
1744 probably associated with an idle stream. This is used to optimally adjust
1745 some packet sizes while forwarding large and small data alternatively. The
1746 decision to use splice() or to send large buffers in SSL is modulated by this
1747 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1748 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001749 which seems to correctly detect end user pauses (e.g. read a page before
John Roesler27c754c2019-07-10 15:45:51 -05001750 clicking). There should be no reason for changing this value. Please check
Willy Tarreau7e312732014-02-12 16:35:14 +01001751 tune.ssl.maxrecord below.
1752
Willy Tarreau7ac908b2019-02-27 12:02:18 +01001753tune.listener.multi-queue { on | off }
1754 Enables ('on') or disables ('off') the listener's multi-queue accept which
1755 spreads the incoming traffic to all threads a "bind" line is allowed to run
1756 on instead of taking them for itself. This provides a smoother traffic
1757 distribution and scales much better, especially in environments where threads
1758 may be unevenly loaded due to external activity (network interrupts colliding
1759 with one thread for example). This option is enabled by default, but it may
1760 be forcefully disabled for troubleshooting or for situations where it is
1761 estimated that the operating system already provides a good enough
1762 distribution and connections are extremely short-lived.
1763
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001764tune.lua.forced-yield <number>
1765 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001766 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001767 HAProxy scheduler to process other tasks like accepting connections or
1768 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001769 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001770 lowered. If the Lua code is quite long and its result is absolutely required
1771 to process the data, the <number> can be increased.
1772
Willy Tarreau32f61e22015-03-18 17:54:59 +01001773tune.lua.maxmem
1774 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1775 default it is zero which means unlimited. It is important to set a limit to
1776 ensure that a bug in a script will not result in the system running out of
1777 memory.
1778
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001779tune.lua.session-timeout <timeout>
1780 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001781 preventing infinite loops or spending too much time in Lua. This timeout
1782 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001783 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001784
1785tune.lua.task-timeout <timeout>
1786 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1787 dedicated to the tasks. By default, this timeout isn't set because a task may
1788 remain alive during of the lifetime of HAProxy. For example, a task used to
1789 check servers.
1790
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001791tune.lua.service-timeout <timeout>
1792 This is the execution timeout for the Lua services. This is useful for
1793 preventing infinite loops or spending too much time in Lua. This timeout
1794 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001795 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001796
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001797tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001798 Sets the maximum number of consecutive connections a process may accept in a
1799 row before switching to other work. In single process mode, higher numbers
1800 give better performance at high connection rates. However in multi-process
1801 modes, keeping a bit of fairness between processes generally is better to
1802 increase performance. This value applies individually to each listener, so
1803 that the number of processes a listener is bound to is taken into account.
1804 This value defaults to 64. In multi-process mode, it is divided by twice
1805 the number of processes the listener is bound to. Setting this value to -1
1806 completely disables the limitation. It should normally not be needed to tweak
1807 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001808
1809tune.maxpollevents <number>
1810 Sets the maximum amount of events that can be processed at once in a call to
1811 the polling system. The default value is adapted to the operating system. It
1812 has been noticed that reducing it below 200 tends to slightly decrease
1813 latency at the expense of network bandwidth, and increasing it above 200
1814 tends to trade latency for slightly increased bandwidth.
1815
Willy Tarreau27a674e2009-08-17 07:23:33 +02001816tune.maxrewrite <number>
1817 Sets the reserved buffer space to this size in bytes. The reserved space is
1818 used for header rewriting or appending. The first reads on sockets will never
1819 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1820 bufsize, though that does not make much sense since there are rarely large
1821 numbers of headers to add. Setting it too high prevents processing of large
1822 requests or responses. Setting it too low prevents addition of new headers
1823 to already large requests or to POST requests. It is generally wise to set it
1824 to about 1024. It is automatically readjusted to half of bufsize if it is
1825 larger than that. This means you don't have to worry about it when changing
1826 bufsize.
1827
Willy Tarreauf3045d22015-04-29 16:24:50 +02001828tune.pattern.cache-size <number>
1829 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1830 cache which reminds previous lookups and their results. It is used by ACLs
1831 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1832 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1833 strings. It applies to pattern expressions which means that it will be able
1834 to memorize the result of a lookup among all the patterns specified on a
1835 configuration line (including all those loaded from files). It automatically
1836 invalidates entries which are updated using HTTP actions or on the CLI. The
1837 default cache size is set to 10000 entries, which limits its footprint to
Willy Tarreau7fdd81c2019-10-23 06:59:31 +02001838 about 5 MB per process/thread on 32-bit systems and 8 MB per process/thread
1839 on 64-bit systems, as caches are thread/process local. There is a very low
Willy Tarreauf3045d22015-04-29 16:24:50 +02001840 risk of collision in this cache, which is in the order of the size of the
1841 cache divided by 2^64. Typically, at 10000 requests per second with the
1842 default cache size of 10000 entries, there's 1% chance that a brute force
1843 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1844 This is considered much lower than the risk of a memory corruption caused by
1845 aging components. If this is not acceptable, the cache can be disabled by
1846 setting this parameter to 0.
1847
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001848tune.pipesize <number>
1849 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1850 are the default size for the system. But sometimes when using TCP splicing,
1851 it can improve performance to increase pipe sizes, especially if it is
1852 suspected that pipes are not filled and that many calls to splice() are
1853 performed. This has an impact on the kernel's memory footprint, so this must
1854 not be changed if impacts are not understood.
1855
Olivier Houchard88698d92019-04-16 19:07:22 +02001856tune.pool-low-fd-ratio <number>
1857 This setting sets the max number of file descriptors (in percentage) used by
1858 haproxy globally against the maximum number of file descriptors haproxy can
1859 use before we stop putting connection into the idle pool for reuse. The
1860 default is 20.
1861
1862tune.pool-high-fd-ratio <number>
1863 This setting sets the max number of file descriptors (in percentage) used by
1864 haproxy globally against the maximum number of file descriptors haproxy can
1865 use before we start killing idle connections when we can't reuse a connection
1866 and we have to create a new one. The default is 25 (one quarter of the file
1867 descriptor will mean that roughly half of the maximum front connections can
1868 keep an idle connection behind, anything beyond this probably doesn't make
John Roesler27c754c2019-07-10 15:45:51 -05001869 much sense in the general case when targeting connection reuse).
Olivier Houchard88698d92019-04-16 19:07:22 +02001870
Willy Tarreaue803de22010-01-21 17:43:04 +01001871tune.rcvbuf.client <number>
1872tune.rcvbuf.server <number>
1873 Forces the kernel socket receive buffer size on the client or the server side
1874 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1875 and backends. It should normally never be set, and the default size (0) lets
John Roesler27c754c2019-07-10 15:45:51 -05001876 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001877 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001878 order to save kernel memory by preventing it from buffering too large amounts
1879 of received data. Lower values will significantly increase CPU usage though.
1880
Willy Tarreaub22fc302015-12-14 12:04:35 +01001881tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001882 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01001883 socket buffers. One of them is that a read returns more than <recv_enough>
1884 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1885 may be changed by this setting to better deal with workloads involving lots
1886 of short messages such as telnet or SSH sessions.
1887
Olivier Houchard1599b802018-05-24 18:59:04 +02001888tune.runqueue-depth <number>
John Roesler27c754c2019-07-10 15:45:51 -05001889 Sets the maximum amount of task that can be processed at once when running
Olivier Houchard1599b802018-05-24 18:59:04 +02001890 tasks. The default value is 200. Increasing it may incur latency when
1891 dealing with I/Os, making it too small can incur extra overhead.
1892
Willy Tarreaue803de22010-01-21 17:43:04 +01001893tune.sndbuf.client <number>
1894tune.sndbuf.server <number>
1895 Forces the kernel socket send buffer size on the client or the server side to
1896 the specified value in bytes. This value applies to all TCP/HTTP frontends
1897 and backends. It should normally never be set, and the default size (0) lets
John Roesler27c754c2019-07-10 15:45:51 -05001898 the kernel auto-tune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001899 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001900 order to save kernel memory by preventing it from buffering too large amounts
1901 of received data. Lower values will significantly increase CPU usage though.
1902 Another use case is to prevent write timeouts with extremely slow clients due
1903 to the kernel waiting for a large part of the buffer to be read before
1904 notifying haproxy again.
1905
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001906tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001907 Sets the size of the global SSL session cache, in a number of blocks. A block
1908 is large enough to contain an encoded session without peer certificate.
1909 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001910 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001911 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01001912 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01001913 and reassigned. Higher values reduce the occurrence of such a purge, hence
1914 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1915 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001916 and are shared between all processes if "nbproc" is greater than 1. Setting
1917 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001918
Emeric Brun8dc60392014-05-09 13:52:00 +02001919tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02001920 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02001921 should normally not be used since it will force many renegotiations due to
1922 clients hitting a random process. But it may be required on some operating
1923 systems where none of the SSL cache synchronization method may be used. In
1924 this case, adding a first layer of hash-based load balancing before the SSL
1925 layer might limit the impact of the lack of session sharing.
1926
Emeric Brun4f65bff2012-11-16 15:11:00 +01001927tune.ssl.lifetime <timeout>
1928 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001929 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001930 does not guarantee that sessions will last that long, because if the cache is
1931 full, the longest idle sessions will be purged despite their configured
1932 lifetime. The real usefulness of this setting is to prevent sessions from
1933 being used for too long.
1934
Willy Tarreaubfd59462013-02-21 07:46:09 +01001935tune.ssl.maxrecord <number>
1936 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1937 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1938 data only once it has received a full record. With large records, it means
1939 that clients might have to download up to 16kB of data before starting to
1940 process them. Limiting the value can improve page load times on browsers
1941 located over high latency or low bandwidth networks. It is suggested to find
1942 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1943 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1944 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1945 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001946 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01001947 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001948
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001949tune.ssl.default-dh-param <number>
1950 Sets the maximum size of the Diffie-Hellman parameters used for generating
1951 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1952 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1953 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1954 this maximum value. Default value if 1024. Only 1024 or higher values are
1955 allowed. Higher values will increase the CPU load, and values greater than
1956 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001957 used if static Diffie-Hellman parameters are supplied either directly
1958 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001959
Christopher Faulet31af49d2015-06-09 17:29:50 +02001960tune.ssl.ssl-ctx-cache-size <number>
1961 Sets the size of the cache used to store generated certificates to <number>
1962 entries. This is a LRU cache. Because generating a SSL certificate
1963 dynamically is expensive, they are cached. The default cache size is set to
1964 1000 entries.
1965
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01001966tune.ssl.capture-cipherlist-size <number>
1967 Sets the maximum size of the buffer used for capturing client-hello cipher
1968 list. If the value is 0 (default value) the capture is disabled, otherwise
1969 a buffer is allocated for each SSL/TLS connection.
1970
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001971tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001972tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001973tune.vars.reqres-max-size <size>
1974tune.vars.sess-max-size <size>
1975tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001976 These five tunes help to manage the maximum amount of memory used by the
1977 variables system. "global" limits the overall amount of memory available for
1978 all scopes. "proc" limits the memory for the process scope, "sess" limits the
1979 memory for the session scope, "txn" for the transaction scope, and "reqres"
1980 limits the memory for each request or response processing.
1981 Memory accounting is hierarchical, meaning more coarse grained limits include
1982 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
1983 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001984
Daniel Schneller0b547052016-03-21 20:46:57 +01001985 For example, when "tune.vars.sess-max-size" is limited to 100,
1986 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1987 100 either. If we create a variable "txn.var" that contains 100 bytes,
1988 all available space is consumed.
1989 Notice that exceeding the limits at runtime will not result in an error
1990 message, but values might be cut off or corrupted. So make sure to accurately
1991 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001992
William Lallemanda509e4c2012-11-07 16:54:34 +01001993tune.zlib.memlevel <number>
1994 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001995 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001996 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01001997 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01001998 between 1 and 9. The default value is 8.
1999
2000tune.zlib.windowsize <number>
2001 Sets the window size (the size of the history buffer) as a parameter of the
2002 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01002003 in better compression at the expense of memory usage. Can be a value between
2004 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002005
Willy Tarreauc57f0e22009-05-10 13:12:33 +020020063.3. Debugging
2007--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02002008
2009debug
2010 Enables debug mode which dumps to stdout all exchanges, and disables forking
2011 into background. It is the equivalent of the command-line argument "-d". It
2012 should never be used in a production configuration since it may prevent full
2013 system startup.
2014
2015quiet
2016 Do not display any message during startup. It is equivalent to the command-
2017 line argument "-q".
2018
Emeric Brunf099e792010-09-27 12:05:28 +02002019
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010020203.4. Userlists
2021--------------
2022It is possible to control access to frontend/backend/listen sections or to
2023http stats by allowing only authenticated and authorized users. To do this,
2024it is required to create at least one userlist and to define users.
2025
2026userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01002027 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002028 used to store authentication & authorization data for independent customers.
2029
2030group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01002031 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002032 attach users to this group by using a comma separated list of names
2033 proceeded by "users" keyword.
2034
Cyril Bontéf0c60612010-02-06 14:44:47 +01002035user <username> [password|insecure-password <password>]
2036 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002037 Adds user <username> to the current userlist. Both secure (encrypted) and
2038 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002039 evaluated using the crypt(3) function, so depending on the system's
2040 capabilities, different algorithms are supported. For example, modern Glibc
2041 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
2042 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002043
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01002044 Attention: Be aware that using encrypted passwords might cause significantly
2045 increased CPU usage, depending on the number of requests, and the algorithm
2046 used. For any of the hashed variants, the password for each request must
2047 be processed through the chosen algorithm, before it can be compared to the
2048 value specified in the config file. Most current algorithms are deliberately
2049 designed to be expensive to compute to achieve resistance against brute
2050 force attacks. They do not simply salt/hash the clear text password once,
2051 but thousands of times. This can quickly become a major factor in haproxy's
2052 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002053
2054 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01002055 userlist L1
2056 group G1 users tiger,scott
2057 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002058
Cyril Bontéf0c60612010-02-06 14:44:47 +01002059 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
2060 user scott insecure-password elgato
2061 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002062
Cyril Bontéf0c60612010-02-06 14:44:47 +01002063 userlist L2
2064 group G1
2065 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002066
Cyril Bontéf0c60612010-02-06 14:44:47 +01002067 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
2068 user scott insecure-password elgato groups G1,G2
2069 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002070
2071 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002072
Emeric Brunf099e792010-09-27 12:05:28 +02002073
20743.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002075----------
Emeric Brun94900952015-06-11 18:25:54 +02002076It is possible to propagate entries of any data-types in stick-tables between
2077several haproxy instances over TCP connections in a multi-master fashion. Each
2078instance pushes its local updates and insertions to remote peers. The pushed
2079values overwrite remote ones without aggregation. Interrupted exchanges are
2080automatically detected and recovered from the last known point.
2081In addition, during a soft restart, the old process connects to the new one
2082using such a TCP connection to push all its entries before the new process
2083tries to connect to other peers. That ensures very fast replication during a
2084reload, it typically takes a fraction of a second even for large tables.
2085Note that Server IDs are used to identify servers remotely, so it is important
2086that configurations look similar or at least that the same IDs are forced on
2087each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002088
2089peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002090 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002091 which is referenced by one or more stick-tables.
2092
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002093bind [<address>]:<port_range> [, ...] [param*]
2094 Defines the binding parameters of the local peer of this "peers" section.
2095 Such lines are not supported with "peer" line in the same "peers" section.
2096
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002097disabled
2098 Disables a peers section. It disables both listening and any synchronization
2099 related to this section. This is provided to disable synchronization of stick
2100 tables without having to comment out all "peers" references.
2101
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002102default-bind [param*]
2103 Defines the binding parameters for the local peer, excepted its address.
2104
2105default-server [param*]
2106 Change default options for a server in a "peers" section.
2107
2108 Arguments:
2109 <param*> is a list of parameters for this server. The "default-server"
2110 keyword accepts an important number of options and has a complete
2111 section dedicated to it. Please refer to section 5 for more
2112 details.
2113
2114
2115 See also: "server" and section 5 about server options
2116
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002117enable
2118 This re-enables a disabled peers section which was previously disabled.
2119
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002120peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002121 Defines a peer inside a peers section.
2122 If <peername> is set to the local peer name (by default hostname, or forced
2123 using "-L" command line option), haproxy will listen for incoming remote peer
2124 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
2125 to join the remote peer, and <peername> is used at the protocol level to
2126 identify and validate the remote peer on the server side.
2127
2128 During a soft restart, local peer <ip>:<port> is used by the old instance to
2129 connect the new one and initiate a complete replication (teaching process).
2130
2131 It is strongly recommended to have the exact same peers declaration on all
2132 peers and to only rely on the "-L" command line argument to change the local
2133 peer name. This makes it easier to maintain coherent configuration files
2134 across all peers.
2135
William Lallemandb2f07452015-05-12 14:27:13 +02002136 You may want to reference some environment variables in the address
2137 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01002138
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002139 Note: "peer" keyword may transparently be replaced by "server" keyword (see
2140 "server" keyword explanation below).
2141
2142server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02002143 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002144 with a support for all "server" parameters found in 5.2 paragraph.
2145 If the underlying peer is local, <ip>:<port> parameters must not be present.
2146 These parameters must be provided on a "bind" line (see "bind" keyword
2147 of this "peers" section).
2148 Some of these parameters are irrelevant for "peers" sections.
2149
2150
Cyril Bontédc4d9032012-04-08 21:57:39 +02002151 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002152 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02002153 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002154 peer haproxy1 192.168.0.1:1024
2155 peer haproxy2 192.168.0.2:1024
2156 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02002157
2158 backend mybackend
2159 mode tcp
2160 balance roundrobin
2161 stick-table type ip size 20k peers mypeers
2162 stick on src
2163
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002164 server srv1 192.168.0.30:80
2165 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02002166
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002167 Example:
2168 peers mypeers
2169 bind 127.0.0.11:10001 ssl crt mycerts/pem
2170 default-server ssl verify none
2171 server hostA 127.0.0.10:10000
2172 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02002173
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002174
2175table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
2176 size <size> [expire <expire>] [nopurge] [store <data_type>]*
2177
2178 Configure a stickiness table for the current section. This line is parsed
2179 exactly the same way as the "stick-table" keyword in others section, except
John Roesler27c754c2019-07-10 15:45:51 -05002180 for the "peers" argument which is not required here and with an additional
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002181 mandatory first parameter to designate the stick-table. Contrary to others
2182 sections, there may be several "table" lines in "peers" sections (see also
2183 "stick-table" keyword).
2184
2185 Also be aware of the fact that "peers" sections have their own stick-table
2186 namespaces to avoid collisions between stick-table names identical in
2187 different "peers" section. This is internally handled prepending the "peers"
2188 sections names to the name of the stick-tables followed by a '/' character.
2189 If somewhere else in the configuration file you have to refer to such
2190 stick-tables declared in "peers" sections you must use the prefixed version
2191 of the stick-table name as follows:
2192
2193 peers mypeers
2194 peer A ...
2195 peer B ...
2196 table t1 ...
2197
2198 frontend fe1
2199 tcp-request content track-sc0 src table mypeers/t1
2200
2201 This is also this prefixed version of the stick-table names which must be
2202 used to refer to stick-tables through the CLI.
2203
2204 About "peers" protocol, as only "peers" belonging to the same section may
2205 communicate with each others, there is no need to do such a distinction.
2206 Several "peers" sections may declare stick-tables with the same name.
2207 This is shorter version of the stick-table name which is sent over the network.
2208 There is only a '/' character as prefix to avoid stick-table name collisions between
2209 stick-tables declared as backends and stick-table declared in "peers" sections
2210 as follows in this weird but supported configuration:
2211
2212 peers mypeers
2213 peer A ...
2214 peer B ...
2215 table t1 type string size 10m store gpc0
2216
2217 backend t1
2218 stick-table type string size 10m store gpc0 peers mypeers
2219
2220 Here "t1" table declared in "mypeeers" section has "mypeers/t1" as global name.
2221 "t1" table declared as a backend as "t1" as global name. But at peer protocol
2222 level the former table is named "/t1", the latter is again named "t1".
2223
Simon Horman51a1cf62015-02-03 13:00:44 +090022243.6. Mailers
2225------------
2226It is possible to send email alerts when the state of servers changes.
2227If configured email alerts are sent to each mailer that is configured
2228in a mailers section. Email is sent to mailers using SMTP.
2229
Pieter Baauw386a1272015-08-16 15:26:24 +02002230mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09002231 Creates a new mailer list with the name <mailersect>. It is an
2232 independent section which is referenced by one or more proxies.
2233
2234mailer <mailername> <ip>:<port>
2235 Defines a mailer inside a mailers section.
2236
2237 Example:
2238 mailers mymailers
2239 mailer smtp1 192.168.0.1:587
2240 mailer smtp2 192.168.0.2:587
2241
2242 backend mybackend
2243 mode tcp
2244 balance roundrobin
2245
2246 email-alert mailers mymailers
2247 email-alert from test1@horms.org
2248 email-alert to test2@horms.org
2249
2250 server srv1 192.168.0.30:80
2251 server srv2 192.168.0.31:80
2252
Pieter Baauw235fcfc2016-02-13 15:33:40 +01002253timeout mail <time>
2254 Defines the time available for a mail/connection to be made and send to
2255 the mail-server. If not defined the default value is 10 seconds. To allow
2256 for at least two SYN-ACK packets to be send during initial TCP handshake it
2257 is advised to keep this value above 4 seconds.
2258
2259 Example:
2260 mailers mymailers
2261 timeout mail 20s
2262 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09002263
William Lallemandc9515522019-06-12 16:32:11 +020022643.7. Programs
2265-------------
2266In master-worker mode, it is possible to launch external binaries with the
2267master, these processes are called programs. These programs are launched and
2268managed the same way as the workers.
2269
2270During a reload of HAProxy, those processes are dealing with the same
2271sequence as a worker:
2272
2273 - the master is re-executed
2274 - the master sends a SIGUSR1 signal to the program
2275 - if "option start-on-reload" is not disabled, the master launches a new
2276 instance of the program
2277
2278During a stop, or restart, a SIGTERM is sent to the programs.
2279
2280program <name>
2281 This is a new program section, this section will create an instance <name>
2282 which is visible in "show proc" on the master CLI. (See "9.4. Master CLI" in
2283 the management guide).
2284
2285command <command> [arguments*]
2286 Define the command to start with optional arguments. The command is looked
2287 up in the current PATH if it does not include an absolute path. This is a
2288 mandatory option of the program section. Arguments containing spaces must
2289 be enclosed in quotes or double quotes or be prefixed by a backslash.
2290
2291option start-on-reload
2292no option start-on-reload
2293 Start (or not) a new instance of the program upon a reload of the master.
2294 The default is to start a new instance. This option may only be used in a
2295 program section.
2296
2297
Willy Tarreauc57f0e22009-05-10 13:12:33 +020022984. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02002299----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002300
Willy Tarreau6a06a402007-07-15 20:15:28 +02002301Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02002302 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02002303 - frontend <name>
2304 - backend <name>
2305 - listen <name>
2306
2307A "defaults" section sets default parameters for all other sections following
2308its declaration. Those default parameters are reset by the next "defaults"
2309section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01002310section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002311
2312A "frontend" section describes a set of listening sockets accepting client
2313connections.
2314
2315A "backend" section describes a set of servers to which the proxy will connect
2316to forward incoming connections.
2317
2318A "listen" section defines a complete proxy with its frontend and backend
2319parts combined in one section. It is generally useful for TCP-only traffic.
2320
Willy Tarreau0ba27502007-12-24 16:55:16 +01002321All proxy names must be formed from upper and lower case letters, digits,
2322'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
2323case-sensitive, which means that "www" and "WWW" are two different proxies.
2324
2325Historically, all proxy names could overlap, it just caused troubles in the
2326logs. Since the introduction of content switching, it is mandatory that two
2327proxies with overlapping capabilities (frontend/backend) have different names.
2328However, it is still permitted that a frontend and a backend share the same
2329name, as this configuration seems to be commonly encountered.
2330
2331Right now, two major proxy modes are supported : "tcp", also known as layer 4,
2332and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002333bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002334protocol, and can interact with it by allowing, blocking, switching, adding,
2335modifying, or removing arbitrary contents in requests or responses, based on
2336arbitrary criteria.
2337
Willy Tarreau70dffda2014-01-30 03:07:23 +01002338In HTTP mode, the processing applied to requests and responses flowing over
2339a connection depends in the combination of the frontend's HTTP options and
Julien Pivotto599788e2019-12-10 13:11:17 +01002340the backend's. HAProxy supports 3 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01002341
2342 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
2343 requests and responses are processed, and connections remain open but idle
2344 between responses and new requests.
2345
2346 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
2347 1.0 to 1.5-dev21 : only the first request and response are processed, and
2348 everything else is forwarded with no analysis at all. This mode should not
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002349 be used as it creates lots of trouble with logging and HTTP processing.
2350 And because it cannot work in HTTP/2, this option is deprecated and it is
2351 only supported on legacy HTTP frontends. In HTX, it is ignored and a
2352 warning is emitted during HAProxy startup.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002353
Willy Tarreau70dffda2014-01-30 03:07:23 +01002354 - SCL: server close ("option http-server-close") : the server-facing
2355 connection is closed after the end of the response is received, but the
2356 client-facing connection remains open.
2357
Christopher Faulet315b39c2018-09-21 16:26:19 +02002358 - CLO: close ("option httpclose"): the connection is closed after the end of
2359 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002360
2361The effective mode that will be applied to a connection passing through a
2362frontend and a backend can be determined by both proxy modes according to the
2363following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02002364weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002365
Christopher Faulet315b39c2018-09-21 16:26:19 +02002366 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01002367
Christopher Faulet315b39c2018-09-21 16:26:19 +02002368 | KAL | SCL | CLO
2369 ----+-----+-----+----
2370 KAL | KAL | SCL | CLO
2371 ----+-----+-----+----
2372 TUN | TUN | SCL | CLO
2373 Frontend ----+-----+-----+----
2374 mode SCL | SCL | SCL | CLO
2375 ----+-----+-----+----
2376 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01002377
Willy Tarreau0ba27502007-12-24 16:55:16 +01002378
Willy Tarreau70dffda2014-01-30 03:07:23 +01002379
Willy Tarreauc57f0e22009-05-10 13:12:33 +020023804.1. Proxy keywords matrix
2381--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002382
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002383The following list of keywords is supported. Most of them may only be used in a
2384limited set of section types. Some of them are marked as "deprecated" because
2385they are inherited from an old syntax which may be confusing or functionally
2386limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002387marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002388option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002389and must be disabled for a specific instance. Such options may also be prefixed
2390with "default" in order to restore default settings regardless of what has been
2391specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002392
Willy Tarreau6a06a402007-07-15 20:15:28 +02002393
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002394 keyword defaults frontend listen backend
2395------------------------------------+----------+----------+---------+---------
2396acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002397backlog X X X -
2398balance X - X X
2399bind - X X -
2400bind-process X X X X
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002401block (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002402capture cookie - X X -
2403capture request header - X X -
2404capture response header - X X -
2405clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002406compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002407contimeout (deprecated) X - X X
2408cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002409declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002410default-server X - X X
2411default_backend X X X -
2412description - X X X
2413disabled X X X X
2414dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002415email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002416email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002417email-alert mailers X X X X
2418email-alert myhostname X X X X
2419email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002420enabled X X X X
2421errorfile X X X X
2422errorloc X X X X
2423errorloc302 X X X X
2424-- keyword -------------------------- defaults - frontend - listen -- backend -
2425errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002426force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002427filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002428fullconn X - X X
2429grace X X X X
2430hash-type X - X X
2431http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002432http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002433http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002434http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002435http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002436http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002437http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002438id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002439ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002440load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002441log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002442log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002443log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002444log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002445max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002446maxconn X X X -
2447mode X X X X
2448monitor fail - X X -
2449monitor-net X X X -
2450monitor-uri X X X -
2451option abortonclose (*) X - X X
2452option accept-invalid-http-request (*) X X X -
2453option accept-invalid-http-response (*) X - X X
2454option allbackups (*) X - X X
2455option checkcache (*) X - X X
2456option clitcpka (*) X X X -
2457option contstats (*) X X X -
2458option dontlog-normal (*) X X X -
2459option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002460-- keyword -------------------------- defaults - frontend - listen -- backend -
2461option forwardfor X X X X
Christopher Fauleta99ff4d2019-07-22 16:18:24 +02002462option h1-case-adjust-bogus-client (*) X X X -
2463option h1-case-adjust-bogus-server (*) X - X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002464option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002465option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002466option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002467option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02002468option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002469option http-server-close (*) X X X X
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002470option http-tunnel (deprecated) (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002471option http-use-proxy-header (*) X X X -
Willy Tarreau68ad3a42018-10-22 11:49:15 +02002472option http-use-htx (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002473option httpchk X - X X
2474option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01002475option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002476option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002477option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002478option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002479option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002480option log-health-checks (*) X - X X
2481option log-separate-errors (*) X X X -
2482option logasap (*) X X X -
2483option mysql-check X - X X
2484option nolinger (*) X X X X
2485option originalto X X X X
2486option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002487option pgsql-check X - X X
2488option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002489option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002490option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002491option smtpchk X - X X
2492option socket-stats (*) X X X -
2493option splice-auto (*) X X X X
2494option splice-request (*) X X X X
2495option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002496option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002497option srvtcpka (*) X - X X
2498option ssl-hello-chk X - X X
2499-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002500option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002501option tcp-smart-accept (*) X X X -
2502option tcp-smart-connect (*) X - X X
2503option tcpka X X X X
2504option tcplog X X X X
2505option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002506external-check command X - X X
2507external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002508persist rdp-cookie X - X X
2509rate-limit sessions X X X -
2510redirect - X X X
2511redisp (deprecated) X - X X
2512redispatch (deprecated) X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002513reqadd (deprecated) - X X X
2514reqallow (deprecated) - X X X
2515reqdel (deprecated) - X X X
2516reqdeny (deprecated) - X X X
2517reqiallow (deprecated) - X X X
2518reqidel (deprecated) - X X X
2519reqideny (deprecated) - X X X
2520reqipass (deprecated) - X X X
2521reqirep (deprecated) - X X X
2522reqitarpit (deprecated) - X X X
2523reqpass (deprecated) - X X X
2524reqrep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002525-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau96d51952019-05-22 20:34:35 +02002526reqtarpit (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002527retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02002528retry-on X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002529rspadd (deprecated) - X X X
2530rspdel (deprecated) - X X X
2531rspdeny (deprecated) - X X X
2532rspidel (deprecated) - X X X
2533rspideny (deprecated) - X X X
2534rspirep (deprecated) - X X X
2535rsprep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002536server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002537server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002538server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002539source X - X X
2540srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002541stats admin - X X X
2542stats auth X X X X
2543stats enable X X X X
2544stats hide-version X X X X
2545stats http-request - X X X
2546stats realm X X X X
2547stats refresh X X X X
2548stats scope X X X X
2549stats show-desc X X X X
2550stats show-legends X X X X
2551stats show-node X X X X
2552stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002553-- keyword -------------------------- defaults - frontend - listen -- backend -
2554stick match - - X X
2555stick on - - X X
2556stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002557stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002558stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002559tcp-check connect - - X X
2560tcp-check expect - - X X
2561tcp-check send - - X X
2562tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002563tcp-request connection - X X -
2564tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002565tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002566tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002567tcp-response content - - X X
2568tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002569timeout check X - X X
2570timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002571timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002572timeout clitimeout (deprecated) X X X -
2573timeout connect X - X X
2574timeout contimeout (deprecated) X - X X
2575timeout http-keep-alive X X X X
2576timeout http-request X X X X
2577timeout queue X - X X
2578timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002579timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002580timeout srvtimeout (deprecated) X - X X
2581timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002582timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002583transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002584unique-id-format X X X -
2585unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002586use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002587use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002588------------------------------------+----------+----------+---------+---------
2589 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002590
Willy Tarreau0ba27502007-12-24 16:55:16 +01002591
Willy Tarreauc57f0e22009-05-10 13:12:33 +020025924.2. Alphabetically sorted keywords reference
2593---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002594
2595This section provides a description of each keyword and its usage.
2596
2597
2598acl <aclname> <criterion> [flags] [operator] <value> ...
2599 Declare or complete an access list.
2600 May be used in sections : defaults | frontend | listen | backend
2601 no | yes | yes | yes
2602 Example:
2603 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2604 acl invalid_src src_port 0:1023
2605 acl local_dst hdr(host) -i localhost
2606
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002607 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002608
2609
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002610backlog <conns>
2611 Give hints to the system about the approximate listen backlog desired size
2612 May be used in sections : defaults | frontend | listen | backend
2613 yes | yes | yes | no
2614 Arguments :
2615 <conns> is the number of pending connections. Depending on the operating
2616 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002617 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002618
2619 In order to protect against SYN flood attacks, one solution is to increase
2620 the system's SYN backlog size. Depending on the system, sometimes it is just
2621 tunable via a system parameter, sometimes it is not adjustable at all, and
2622 sometimes the system relies on hints given by the application at the time of
2623 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2624 to the listen() syscall. On systems which can make use of this value, it can
2625 sometimes be useful to be able to specify a different value, hence this
2626 backlog parameter.
2627
2628 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2629 used as a hint and the system accepts up to the smallest greater power of
2630 two, and never more than some limits (usually 32768).
2631
2632 See also : "maxconn" and the target operating system's tuning guide.
2633
2634
Willy Tarreau0ba27502007-12-24 16:55:16 +01002635balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002636balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002637 Define the load balancing algorithm to be used in a backend.
2638 May be used in sections : defaults | frontend | listen | backend
2639 yes | no | yes | yes
2640 Arguments :
2641 <algorithm> is the algorithm used to select a server when doing load
2642 balancing. This only applies when no persistence information
2643 is available, or when a connection is redispatched to another
2644 server. <algorithm> may be one of the following :
2645
2646 roundrobin Each server is used in turns, according to their weights.
2647 This is the smoothest and fairest algorithm when the server's
2648 processing time remains equally distributed. This algorithm
2649 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002650 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002651 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002652 large farms, when a server becomes up after having been down
2653 for a very short time, it may sometimes take a few hundreds
2654 requests for it to be re-integrated into the farm and start
2655 receiving traffic. This is normal, though very rare. It is
2656 indicated here in case you would have the chance to observe
2657 it, so that you don't worry.
2658
2659 static-rr Each server is used in turns, according to their weights.
2660 This algorithm is as similar to roundrobin except that it is
2661 static, which means that changing a server's weight on the
2662 fly will have no effect. On the other hand, it has no design
2663 limitation on the number of servers, and when a server goes
2664 up, it is always immediately reintroduced into the farm, once
2665 the full map is recomputed. It also uses slightly less CPU to
2666 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002667
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002668 leastconn The server with the lowest number of connections receives the
2669 connection. Round-robin is performed within groups of servers
2670 of the same load to ensure that all servers will be used. Use
2671 of this algorithm is recommended where very long sessions are
2672 expected, such as LDAP, SQL, TSE, etc... but is not very well
2673 suited for protocols using short sessions such as HTTP. This
2674 algorithm is dynamic, which means that server weights may be
2675 adjusted on the fly for slow starts for instance.
2676
Willy Tarreauf09c6602012-02-13 17:12:08 +01002677 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002678 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002679 identifier to the highest (see server parameter "id"), which
2680 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002681 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002682 not make sense to use this algorithm without setting maxconn.
2683 The purpose of this algorithm is to always use the smallest
2684 number of servers so that extra servers can be powered off
2685 during non-intensive hours. This algorithm ignores the server
2686 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002687 or IMAP than HTTP, though it can be useful there too. In
2688 order to use this algorithm efficiently, it is recommended
2689 that a cloud controller regularly checks server usage to turn
2690 them off when unused, and regularly checks backend queue to
2691 turn new servers on when the queue inflates. Alternatively,
2692 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002693
Willy Tarreau0ba27502007-12-24 16:55:16 +01002694 source The source IP address is hashed and divided by the total
2695 weight of the running servers to designate which server will
2696 receive the request. This ensures that the same client IP
2697 address will always reach the same server as long as no
2698 server goes down or up. If the hash result changes due to the
2699 number of running servers changing, many clients will be
2700 directed to a different server. This algorithm is generally
2701 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002702 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002703 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002704 static by default, which means that changing a server's
2705 weight on the fly will have no effect, but this can be
2706 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002707
Oskar Stolc8dc41842012-05-19 10:19:54 +01002708 uri This algorithm hashes either the left part of the URI (before
2709 the question mark) or the whole URI (if the "whole" parameter
2710 is present) and divides the hash value by the total weight of
2711 the running servers. The result designates which server will
2712 receive the request. This ensures that the same URI will
2713 always be directed to the same server as long as no server
2714 goes up or down. This is used with proxy caches and
2715 anti-virus proxies in order to maximize the cache hit rate.
2716 Note that this algorithm may only be used in an HTTP backend.
2717 This algorithm is static by default, which means that
2718 changing a server's weight on the fly will have no effect,
2719 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002720
Oskar Stolc8dc41842012-05-19 10:19:54 +01002721 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002722 "depth", both followed by a positive integer number. These
2723 options may be helpful when it is needed to balance servers
2724 based on the beginning of the URI only. The "len" parameter
2725 indicates that the algorithm should only consider that many
2726 characters at the beginning of the URI to compute the hash.
2727 Note that having "len" set to 1 rarely makes sense since most
2728 URIs start with a leading "/".
2729
2730 The "depth" parameter indicates the maximum directory depth
2731 to be used to compute the hash. One level is counted for each
2732 slash in the request. If both parameters are specified, the
2733 evaluation stops when either is reached.
2734
Willy Tarreau0ba27502007-12-24 16:55:16 +01002735 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002736 the query string of each HTTP GET request.
2737
2738 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002739 request entity will be searched for the parameter argument,
2740 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002741 ('?') in the URL. The message body will only start to be
2742 analyzed once either the advertised amount of data has been
2743 received or the request buffer is full. In the unlikely event
2744 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002745 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002746 be randomly balanced if at all. This keyword used to support
2747 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002748
2749 If the parameter is found followed by an equal sign ('=') and
2750 a value, then the value is hashed and divided by the total
2751 weight of the running servers. The result designates which
2752 server will receive the request.
2753
2754 This is used to track user identifiers in requests and ensure
2755 that a same user ID will always be sent to the same server as
2756 long as no server goes up or down. If no value is found or if
2757 the parameter is not found, then a round robin algorithm is
2758 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002759 backend. This algorithm is static by default, which means
2760 that changing a server's weight on the fly will have no
2761 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002762
Cyril Bontédc4d9032012-04-08 21:57:39 +02002763 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2764 request. Just as with the equivalent ACL 'hdr()' function,
2765 the header name in parenthesis is not case sensitive. If the
2766 header is absent or if it does not contain any value, the
2767 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002768
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002769 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002770 reducing the hash algorithm to the main domain part with some
2771 specific headers such as 'Host'. For instance, in the Host
2772 value "haproxy.1wt.eu", only "1wt" will be considered.
2773
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002774 This algorithm is static by default, which means that
2775 changing a server's weight on the fly will have no effect,
2776 but this can be changed using "hash-type".
2777
Willy Tarreau21c741a2019-01-14 18:14:27 +01002778 random
2779 random(<draws>)
2780 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02002781 hashing function. This means that the servers' weights are
2782 respected, dynamic weight changes immediately take effect, as
2783 well as new server additions. Random load balancing can be
2784 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01002785 or removed as it may avoid the hammering effect that could
2786 result from roundrobin or leastconn in this situation. The
2787 hash-balance-factor directive can be used to further improve
2788 fairness of the load balancing, especially in situations
2789 where servers show highly variable response times. When an
2790 argument <draws> is present, it must be an integer value one
2791 or greater, indicating the number of draws before selecting
2792 the least loaded of these servers. It was indeed demonstrated
2793 that picking the least loaded of two servers is enough to
2794 significantly improve the fairness of the algorithm, by
2795 always avoiding to pick the most loaded server within a farm
2796 and getting rid of any bias that could be induced by the
2797 unfair distribution of the consistent list. Higher values N
2798 will take away N-1 of the highest loaded servers at the
2799 expense of performance. With very high values, the algorithm
2800 will converge towards the leastconn's result but much slower.
2801 The default value is 2, which generally shows very good
2802 distribution and performance. This algorithm is also known as
2803 the Power of Two Random Choices and is described here :
2804 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02002805
Emeric Brun736aa232009-06-30 17:56:00 +02002806 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002807 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002808 The RDP cookie <name> (or "mstshash" if omitted) will be
2809 looked up and hashed for each incoming TCP request. Just as
2810 with the equivalent ACL 'req_rdp_cookie()' function, the name
2811 is not case-sensitive. This mechanism is useful as a degraded
2812 persistence mode, as it makes it possible to always send the
2813 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002814 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002815 used instead.
2816
2817 Note that for this to work, the frontend must ensure that an
2818 RDP cookie is already present in the request buffer. For this
2819 you must use 'tcp-request content accept' rule combined with
2820 a 'req_rdp_cookie_cnt' ACL.
2821
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002822 This algorithm is static by default, which means that
2823 changing a server's weight on the fly will have no effect,
2824 but this can be changed using "hash-type".
2825
Cyril Bontédc4d9032012-04-08 21:57:39 +02002826 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002827
Willy Tarreau0ba27502007-12-24 16:55:16 +01002828 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002829 algorithms. Right now, only "url_param" and "uri" support an
2830 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002831
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002832 The load balancing algorithm of a backend is set to roundrobin when no other
2833 algorithm, mode nor option have been set. The algorithm may only be set once
2834 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002835
Lukas Tribus80512b12018-10-27 20:07:40 +02002836 With authentication schemes that require the same connection like NTLM, URI
John Roesler27c754c2019-07-10 15:45:51 -05002837 based algorithms must not be used, as they would cause subsequent requests
Lukas Tribus80512b12018-10-27 20:07:40 +02002838 to be routed to different backend servers, breaking the invalid assumptions
2839 NTLM relies on.
2840
Willy Tarreau0ba27502007-12-24 16:55:16 +01002841 Examples :
2842 balance roundrobin
2843 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002844 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002845 balance hdr(User-Agent)
2846 balance hdr(host)
2847 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002848
2849 Note: the following caveats and limitations on using the "check_post"
2850 extension with "url_param" must be considered :
2851
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002852 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002853 to determine if the parameters will be found in the body or entity which
2854 may contain binary data. Therefore another method may be required to
2855 restrict consideration of POST requests that have no URL parameters in
2856 the body. (see acl reqideny http_end)
2857
2858 - using a <max_wait> value larger than the request buffer size does not
2859 make sense and is useless. The buffer size is set at build time, and
2860 defaults to 16 kB.
2861
2862 - Content-Encoding is not supported, the parameter search will probably
2863 fail; and load balancing will fall back to Round Robin.
2864
2865 - Expect: 100-continue is not supported, load balancing will fall back to
2866 Round Robin.
2867
Lukas Tribus23953682017-04-28 13:24:30 +00002868 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002869 If the entire parameter value is not present in the first chunk, the
2870 selection of server is undefined (actually, defined by how little
2871 actually appeared in the first chunk).
2872
2873 - This feature does not support generation of a 100, 411 or 501 response.
2874
2875 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002876 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002877 white space or control characters are found, indicating the end of what
2878 might be a URL parameter list. This is probably not a concern with SGML
2879 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002880
Willy Tarreau294d0f02015-08-10 19:40:12 +02002881 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002882
2883
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002884bind [<address>]:<port_range> [, ...] [param*]
2885bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002886 Define one or several listening addresses and/or ports in a frontend.
2887 May be used in sections : defaults | frontend | listen | backend
2888 no | yes | yes | no
2889 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002890 <address> is optional and can be a host name, an IPv4 address, an IPv6
2891 address, or '*'. It designates the address the frontend will
2892 listen on. If unset, all IPv4 addresses of the system will be
2893 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002894 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002895 Optionally, an address family prefix may be used before the
2896 address to force the family regardless of the address format,
2897 which can be useful to specify a path to a unix socket with
2898 no slash ('/'). Currently supported prefixes are :
2899 - 'ipv4@' -> address is always IPv4
2900 - 'ipv6@' -> address is always IPv6
2901 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002902 - 'abns@' -> address is in abstract namespace (Linux only).
2903 Note: since abstract sockets are not "rebindable", they
2904 do not cope well with multi-process mode during
2905 soft-restart, so it is better to avoid them if
2906 nbproc is greater than 1. The effect is that if the
2907 new process fails to start, only one of the old ones
2908 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002909 - 'fd@<n>' -> use file descriptor <n> inherited from the
2910 parent. The fd must be bound and may or may not already
2911 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02002912 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
2913 connected unix socket or of a socketpair. The bind waits
2914 to receive a FD over the unix socket and uses it as if it
2915 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02002916 You may want to reference some environment variables in the
2917 address parameter, see section 2.3 about environment
2918 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002919
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002920 <port_range> is either a unique TCP port, or a port range for which the
2921 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002922 above. The port is mandatory for TCP listeners. Note that in
2923 the case of an IPv6 address, the port is always the number
2924 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002925 - a numerical port (ex: '80')
2926 - a dash-delimited ports range explicitly stating the lower
2927 and upper bounds (ex: '2000-2100') which are included in
2928 the range.
2929
2930 Particular care must be taken against port ranges, because
2931 every <address:port> couple consumes one socket (= a file
2932 descriptor), so it's easy to consume lots of descriptors
2933 with a simple range, and to run out of sockets. Also, each
2934 <address:port> couple must be used only once among all
2935 instances running on a same system. Please note that binding
2936 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002937 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002938 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002939
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002940 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002941 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002942 receive UNIX connections on the socket located at this place.
2943 The path must begin with a slash and by default is absolute.
2944 It can be relative to the prefix defined by "unix-bind" in
2945 the global section. Note that the total length of the prefix
2946 followed by the socket path cannot exceed some system limits
2947 for UNIX sockets, which commonly are set to 107 characters.
2948
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002949 <param*> is a list of parameters common to all sockets declared on the
2950 same line. These numerous parameters depend on OS and build
2951 options and have a complete section dedicated to them. Please
2952 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002953
Willy Tarreau0ba27502007-12-24 16:55:16 +01002954 It is possible to specify a list of address:port combinations delimited by
2955 commas. The frontend will then listen on all of these addresses. There is no
2956 fixed limit to the number of addresses and ports which can be listened on in
2957 a frontend, as well as there is no limit to the number of "bind" statements
2958 in a frontend.
2959
2960 Example :
2961 listen http_proxy
2962 bind :80,:443
2963 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002964 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002965
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002966 listen http_https_proxy
2967 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002968 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002969
Willy Tarreau24709282013-03-10 21:32:12 +01002970 listen http_https_proxy_explicit
2971 bind ipv6@:80
2972 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2973 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2974
Willy Tarreaudad36a32013-03-11 01:20:04 +01002975 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002976 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002977
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002978 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2979 sun_path length is used for the address length. Some other programs
2980 such as socat use the string length only by default. Pass the option
2981 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2982 make it compatible with HAProxy's.
2983
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002984 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002985 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002986
2987
Christopher Fauletff4121f2017-11-22 16:38:49 +01002988bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002989 Limit visibility of an instance to a certain set of processes numbers.
2990 May be used in sections : defaults | frontend | listen | backend
2991 yes | yes | yes | yes
2992 Arguments :
2993 all All process will see this instance. This is the default. It
2994 may be used to override a default value.
2995
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002996 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002997 option may be combined with other numbers.
2998
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002999 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003000 option may be combined with other numbers. Do not use it
3001 with less than 2 processes otherwise some instances might be
3002 missing from all processes.
3003
Christopher Fauletff4121f2017-11-22 16:38:49 +01003004 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003005 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01003006 the machine's word size. Ranges can be partially defined. The
3007 higher bound can be omitted. In such case, it is replaced by
3008 the corresponding maximum value. If a proxy is bound to
3009 process numbers greater than the configured global.nbproc, it
3010 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02003011 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003012
3013 This keyword limits binding of certain instances to certain processes. This
3014 is useful in order not to have too many processes listening to the same
3015 ports. For instance, on a dual-core machine, it might make sense to set
3016 'nbproc 2' in the global section, then distributes the listeners among 'odd'
3017 and 'even' instances.
3018
Willy Tarreaua9db57e2013-01-18 11:29:29 +01003019 At the moment, it is not possible to reference more than 32 or 64 processes
3020 using this keyword, but this should be more than enough for most setups.
3021 Please note that 'all' really means all processes regardless of the machine's
3022 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003023
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003024 Each "bind" line may further be limited to a subset of the proxy's processes,
3025 please consult the "process" bind keyword in section 5.1.
3026
Willy Tarreaub369a042014-09-16 13:21:03 +02003027 When a frontend has no explicit "bind-process" line, it tries to bind to all
3028 the processes referenced by its "bind" lines. That means that frontends can
3029 easily adapt to their listeners' processes.
3030
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003031 If some backends are referenced by frontends bound to other processes, the
3032 backend automatically inherits the frontend's processes.
3033
3034 Example :
3035 listen app_ip1
3036 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003037 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003038
3039 listen app_ip2
3040 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003041 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003042
3043 listen management
3044 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02003045 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003046
Willy Tarreau110ecc12012-11-15 17:50:01 +01003047 listen management
3048 bind 10.0.0.4:80
3049 bind-process 1-4
3050
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02003051 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01003052
3053
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02003054block { if | unless } <condition> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003055 Block a layer 7 request if/unless a condition is matched
3056 May be used in sections : defaults | frontend | listen | backend
3057 no | yes | yes | yes
3058
3059 The HTTP request will be blocked very early in the layer 7 processing
3060 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003061 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02003062 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01003063 conditions are met or not met. There is no fixed limit to the number of
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03003064 "block" statements per instance. To block connections at layer 4 (without
3065 sending a 403 error) see "tcp-request connection reject" and
3066 "tcp-request content reject" rules.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003067
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02003068 This form is deprecated, do not use it in any new configuration, use the new
3069 "http-request deny" instead.
3070
Willy Tarreau0ba27502007-12-24 16:55:16 +01003071 Example:
3072 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
3073 acl invalid_src src_port 0:1023
3074 acl local_dst hdr(host) -i localhost
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +03003075 # block is deprecated. Use http-request deny instead:
3076 #block if invalid_src || local_dst
3077 http-request deny if invalid_src || local_dst
Willy Tarreau0ba27502007-12-24 16:55:16 +01003078
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03003079 See also : section 7 about ACL usage, "http-request deny",
3080 "http-response deny", "tcp-request connection reject" and
3081 "tcp-request content reject".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003082
3083capture cookie <name> len <length>
3084 Capture and log a cookie in the request and in the response.
3085 May be used in sections : defaults | frontend | listen | backend
3086 no | yes | yes | no
3087 Arguments :
3088 <name> is the beginning of the name of the cookie to capture. In order
3089 to match the exact name, simply suffix the name with an equal
3090 sign ('='). The full name will appear in the logs, which is
3091 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01003092 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003093
3094 <length> is the maximum number of characters to report in the logs, which
3095 include the cookie name, the equal sign and the value, all in the
3096 standard "name=value" form. The string will be truncated on the
3097 right if it exceeds <length>.
3098
3099 Only the first cookie is captured. Both the "cookie" request headers and the
3100 "set-cookie" response headers are monitored. This is particularly useful to
3101 check for application bugs causing session crossing or stealing between
3102 users, because generally the user's cookies can only change on a login page.
3103
3104 When the cookie was not presented by the client, the associated log column
3105 will report "-". When a request does not cause a cookie to be assigned by the
3106 server, a "-" is reported in the response column.
3107
3108 The capture is performed in the frontend only because it is necessary that
3109 the log format does not change for a given frontend depending on the
3110 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01003111 "capture cookie" statement in a frontend. The maximum capture length is set
3112 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
3113 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003114
3115 Example:
3116 capture cookie ASPSESSION len 32
3117
3118 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003119 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003120
3121
3122capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003123 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003124 May be used in sections : defaults | frontend | listen | backend
3125 no | yes | yes | no
3126 Arguments :
3127 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003128 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003129 appear in the requests, with the first letter of each word in
3130 upper case. The header name will not appear in the logs, only the
3131 value is reported, but the position in the logs is respected.
3132
3133 <length> is the maximum number of characters to extract from the value and
3134 report in the logs. The string will be truncated on the right if
3135 it exceeds <length>.
3136
Willy Tarreau4460d032012-11-21 23:37:37 +01003137 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003138 value will be added to the logs between braces ('{}'). If multiple headers
3139 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003140 in the same order they were declared in the configuration. Non-existent
3141 headers will be logged just as an empty string. Common uses for request
3142 header captures include the "Host" field in virtual hosting environments, the
3143 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003144 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003145 environments to find where the request came from.
3146
3147 Note that when capturing headers such as "User-agent", some spaces may be
3148 logged, making the log analysis more difficult. Thus be careful about what
3149 you log if you know your log parser is not smart enough to rely on the
3150 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003151
Willy Tarreau0900abb2012-11-22 00:21:46 +01003152 There is no limit to the number of captured request headers nor to their
3153 length, though it is wise to keep them low to limit memory usage per session.
3154 In order to keep log format consistent for a same frontend, header captures
3155 can only be declared in a frontend. It is not possible to specify a capture
3156 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003157
3158 Example:
3159 capture request header Host len 15
3160 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01003161 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01003162
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003163 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003164 about logging.
3165
3166
3167capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003168 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003169 May be used in sections : defaults | frontend | listen | backend
3170 no | yes | yes | no
3171 Arguments :
3172 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003173 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003174 appear in the response, with the first letter of each word in
3175 upper case. The header name will not appear in the logs, only the
3176 value is reported, but the position in the logs is respected.
3177
3178 <length> is the maximum number of characters to extract from the value and
3179 report in the logs. The string will be truncated on the right if
3180 it exceeds <length>.
3181
Willy Tarreau4460d032012-11-21 23:37:37 +01003182 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003183 result will be added to the logs between braces ('{}') after the captured
3184 request headers. If multiple headers are captured, they will be delimited by
3185 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003186 the configuration. Non-existent headers will be logged just as an empty
3187 string. Common uses for response header captures include the "Content-length"
3188 header which indicates how many bytes are expected to be returned, the
3189 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003190
Willy Tarreau0900abb2012-11-22 00:21:46 +01003191 There is no limit to the number of captured response headers nor to their
3192 length, though it is wise to keep them low to limit memory usage per session.
3193 In order to keep log format consistent for a same frontend, header captures
3194 can only be declared in a frontend. It is not possible to specify a capture
3195 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003196
3197 Example:
3198 capture response header Content-length len 9
3199 capture response header Location len 15
3200
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003201 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003202 about logging.
3203
3204
Cyril Bontéf0c60612010-02-06 14:44:47 +01003205clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003206 Set the maximum inactivity time on the client side.
3207 May be used in sections : defaults | frontend | listen | backend
3208 yes | yes | yes | no
3209 Arguments :
3210 <timeout> is the timeout value is specified in milliseconds by default, but
3211 can be in any other unit if the number is suffixed by the unit,
3212 as explained at the top of this document.
3213
3214 The inactivity timeout applies when the client is expected to acknowledge or
3215 send data. In HTTP mode, this timeout is particularly important to consider
3216 during the first phase, when the client sends the request, and during the
3217 response while it is reading data sent by the server. The value is specified
3218 in milliseconds by default, but can be in any other unit if the number is
3219 suffixed by the unit, as specified at the top of this document. In TCP mode
3220 (and to a lesser extent, in HTTP mode), it is highly recommended that the
3221 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003222 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01003223 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +01003224 (e.g. 4 or 5 seconds).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003225
3226 This parameter is specific to frontends, but can be specified once for all in
3227 "defaults" sections. This is in fact one of the easiest solutions not to
3228 forget about it. An unspecified timeout results in an infinite timeout, which
3229 is not recommended. Such a usage is accepted and works but reports a warning
3230 during startup because it may results in accumulation of expired sessions in
3231 the system if the system's timeouts are not configured either.
3232
3233 This parameter is provided for compatibility but is currently deprecated.
3234 Please use "timeout client" instead.
3235
Willy Tarreau036fae02008-01-06 13:24:40 +01003236 See also : "timeout client", "timeout http-request", "timeout server", and
3237 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003238
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003239compression algo <algorithm> ...
3240compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02003241compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02003242 Enable HTTP compression.
3243 May be used in sections : defaults | frontend | listen | backend
3244 yes | yes | yes | yes
3245 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003246 algo is followed by the list of supported compression algorithms.
3247 type is followed by the list of MIME types that will be compressed.
3248 offload makes haproxy work as a compression offloader only (see notes).
3249
3250 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01003251 identity this is mostly for debugging, and it was useful for developing
3252 the compression feature. Identity does not apply any change on
3253 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003254
Willy Tarreauc91840a2015-03-28 17:00:39 +01003255 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01003256 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01003257
3258 deflate same as "gzip", but with deflate algorithm and zlib format.
3259 Note that this algorithm has ambiguous support on many
3260 browsers and no support at all from recent ones. It is
3261 strongly recommended not to use it for anything else than
3262 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01003263 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003264
Willy Tarreauc91840a2015-03-28 17:00:39 +01003265 raw-deflate same as "deflate" without the zlib wrapper, and used as an
3266 alternative when the browser wants "deflate". All major
3267 browsers understand it and despite violating the standards,
3268 it is known to work better than "deflate", at least on MSIE
3269 and some versions of Safari. Do not use it in conjunction
3270 with "deflate", use either one or the other since both react
3271 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01003272 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003273
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04003274 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003275 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003276 If backend servers support HTTP compression, these directives
3277 will be no-op: haproxy will see the compressed response and will not
3278 compress again. If backend servers do not support HTTP compression and
3279 there is Accept-Encoding header in request, haproxy will compress the
3280 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02003281
3282 The "offload" setting makes haproxy remove the Accept-Encoding header to
3283 prevent backend servers from compressing responses. It is strongly
3284 recommended not to do this because this means that all the compression work
3285 will be done on the single point where haproxy is located. However in some
3286 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003287 with broken HTTP compression implementation which can't be turned off.
3288 In that case haproxy can be used to prevent that gateway from emitting
3289 invalid payloads. In this case, simply removing the header in the
3290 configuration does not work because it applies before the header is parsed,
3291 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02003292 then be used for such scenarios. Note: for now, the "offload" setting is
3293 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02003294
William Lallemand05097442012-11-20 12:14:28 +01003295 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003296 * the request does not advertise a supported compression algorithm in the
3297 "Accept-Encoding" header
3298 * the response message is not HTTP/1.1
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003299 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003300 * response contain neither a "Content-Length" header nor a
3301 "Transfer-Encoding" whose last value is "chunked"
3302 * response contains a "Content-Type" header whose first value starts with
3303 "multipart"
3304 * the response contains the "no-transform" value in the "Cache-control"
3305 header
3306 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
3307 and later
3308 * The response contains a "Content-Encoding" header, indicating that the
3309 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003310 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01003311
Tim Duesterhusb229f012019-01-29 16:38:56 +01003312 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01003313
William Lallemand82fe75c2012-10-23 10:25:10 +02003314 Examples :
3315 compression algo gzip
3316 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01003317
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003318
Cyril Bontéf0c60612010-02-06 14:44:47 +01003319contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003320 Set the maximum time to wait for a connection attempt to a server to succeed.
3321 May be used in sections : defaults | frontend | listen | backend
3322 yes | no | yes | yes
3323 Arguments :
3324 <timeout> is the timeout value is specified in milliseconds by default, but
3325 can be in any other unit if the number is suffixed by the unit,
3326 as explained at the top of this document.
3327
3328 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003329 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01003330 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +01003331 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003332 connect timeout also presets the queue timeout to the same value if this one
3333 has not been specified. Historically, the contimeout was also used to set the
3334 tarpit timeout in a listen section, which is not possible in a pure frontend.
3335
3336 This parameter is specific to backends, but can be specified once for all in
3337 "defaults" sections. This is in fact one of the easiest solutions not to
3338 forget about it. An unspecified timeout results in an infinite timeout, which
3339 is not recommended. Such a usage is accepted and works but reports a warning
3340 during startup because it may results in accumulation of failed sessions in
3341 the system if the system's timeouts are not configured either.
3342
3343 This parameter is provided for backwards compatibility but is currently
3344 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
3345 instead.
3346
3347 See also : "timeout connect", "timeout queue", "timeout tarpit",
3348 "timeout server", "contimeout".
3349
3350
Willy Tarreau55165fe2009-05-10 12:02:55 +02003351cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003352 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3353 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Christopher Fauletdb2cdbb2020-01-21 11:06:48 +01003354 [ dynamic ] [ attr <value> ]*
Willy Tarreau0ba27502007-12-24 16:55:16 +01003355 Enable cookie-based persistence in a backend.
3356 May be used in sections : defaults | frontend | listen | backend
3357 yes | no | yes | yes
3358 Arguments :
3359 <name> is the name of the cookie which will be monitored, modified or
3360 inserted in order to bring persistence. This cookie is sent to
3361 the client via a "Set-Cookie" header in the response, and is
3362 brought back by the client in a "Cookie" header in all requests.
3363 Special care should be taken to choose a name which does not
3364 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003365 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003366 HTTP/HTTPS), care should be taken to use different cookie names
3367 between all backends if persistence between them is not desired.
3368
3369 rewrite This keyword indicates that the cookie will be provided by the
3370 server and that haproxy will have to modify its value to set the
3371 server's identifier in it. This mode is handy when the management
3372 of complex combinations of "Set-cookie" and "Cache-control"
3373 headers is left to the application. The application can then
3374 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003375 cookie. Since all responses should be monitored, this mode
3376 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003377 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003378 start with this mode for new deployments. This keyword is
3379 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003380
3381 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003382 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003383
Willy Tarreaua79094d2010-08-31 22:54:15 +02003384 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003385 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02003386 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003387 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003388 existing configurations running in the "rewrite" mode. The cookie
3389 will only be a session cookie and will not be stored on the
3390 client's disk. By default, unless the "indirect" option is added,
3391 the server will see the cookies emitted by the client. Due to
3392 caching effects, it is generally wise to add the "nocache" or
3393 "postonly" keywords (see below). The "insert" keyword is not
3394 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003395
3396 prefix This keyword indicates that instead of relying on a dedicated
3397 cookie for the persistence, an existing one will be completed.
3398 This may be needed in some specific environments where the client
3399 does not support more than one single cookie and the application
3400 already needs it. In this case, whenever the server sets a cookie
3401 named <name>, it will be prefixed with the server's identifier
3402 and a delimiter. The prefix will be removed from all client
3403 requests so that the server still finds the cookie it emitted.
3404 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003405 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003406 not compatible with "rewrite" and "insert". Note: it is highly
3407 recommended not to use "indirect" with "prefix", otherwise server
3408 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003409
Willy Tarreaua79094d2010-08-31 22:54:15 +02003410 indirect When this option is specified, no cookie will be emitted to a
3411 client which already has a valid one for the server which has
3412 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003413 it will be removed, unless the "preserve" option is also set. In
3414 "insert" mode, this will additionally remove cookies from the
3415 requests transmitted to the server, making the persistence
3416 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003417 Note: it is highly recommended not to use "indirect" with
3418 "prefix", otherwise server cookie updates would not be sent to
3419 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003420
3421 nocache This option is recommended in conjunction with the insert mode
3422 when there is a cache between the client and HAProxy, as it
3423 ensures that a cacheable response will be tagged non-cacheable if
3424 a cookie needs to be inserted. This is important because if all
3425 persistence cookies are added on a cacheable home page for
3426 instance, then all customers will then fetch the page from an
3427 outer cache and will all share the same persistence cookie,
3428 leading to one server receiving much more traffic than others.
3429 See also the "insert" and "postonly" options.
3430
3431 postonly This option ensures that cookie insertion will only be performed
3432 on responses to POST requests. It is an alternative to the
3433 "nocache" option, because POST responses are not cacheable, so
3434 this ensures that the persistence cookie will never get cached.
3435 Since most sites do not need any sort of persistence before the
3436 first POST which generally is a login request, this is a very
3437 efficient method to optimize caching without risking to find a
3438 persistence cookie in the cache.
3439 See also the "insert" and "nocache" options.
3440
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003441 preserve This option may only be used with "insert" and/or "indirect". It
3442 allows the server to emit the persistence cookie itself. In this
3443 case, if a cookie is found in the response, haproxy will leave it
3444 untouched. This is useful in order to end persistence after a
3445 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003446 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003447 the past. By combining this mechanism with the "disable-on-404"
3448 check option, it is possible to perform a completely graceful
3449 shutdown because users will definitely leave the server after
3450 they logout.
3451
Willy Tarreau4992dd22012-05-31 21:02:17 +02003452 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3453 when a cookie is inserted. This attribute is used so that a
3454 user agent doesn't share the cookie with non-HTTP components.
3455 Please check RFC6265 for more information on this attribute.
3456
3457 secure This option tells haproxy to add a "Secure" cookie attribute when
3458 a cookie is inserted. This attribute is used so that a user agent
3459 never emits this cookie over non-secure channels, which means
3460 that a cookie learned with this flag will be presented only over
3461 SSL/TLS connections. Please check RFC6265 for more information on
3462 this attribute.
3463
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003464 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003465 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003466 name. If the domain begins with a dot, the browser is allowed to
3467 use it for any host ending with that name. It is also possible to
3468 specify several domain names by invoking this option multiple
3469 times. Some browsers might have small limits on the number of
3470 domains, so be careful when doing that. For the record, sending
3471 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003472
Willy Tarreau996a92c2010-10-13 19:30:47 +02003473 maxidle This option allows inserted cookies to be ignored after some idle
3474 time. It only works with insert-mode cookies. When a cookie is
3475 sent to the client, the date this cookie was emitted is sent too.
3476 Upon further presentations of this cookie, if the date is older
3477 than the delay indicated by the parameter (in seconds), it will
3478 be ignored. Otherwise, it will be refreshed if needed when the
3479 response is sent to the client. This is particularly useful to
3480 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003481 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003482 this option is set and a cookie has no date, it is always
3483 accepted, but gets refreshed in the response. This maintains the
3484 ability for admins to access their sites. Cookies that have a
3485 date in the future further than 24 hours are ignored. Doing so
3486 lets admins fix timezone issues without risking kicking users off
3487 the site.
3488
3489 maxlife This option allows inserted cookies to be ignored after some life
3490 time, whether they're in use or not. It only works with insert
3491 mode cookies. When a cookie is first sent to the client, the date
3492 this cookie was emitted is sent too. Upon further presentations
3493 of this cookie, if the date is older than the delay indicated by
3494 the parameter (in seconds), it will be ignored. If the cookie in
3495 the request has no date, it is accepted and a date will be set.
3496 Cookies that have a date in the future further than 24 hours are
3497 ignored. Doing so lets admins fix timezone issues without risking
3498 kicking users off the site. Contrary to maxidle, this value is
3499 not refreshed, only the first visit date counts. Both maxidle and
3500 maxlife may be used at the time. This is particularly useful to
3501 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003502 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003503 is stronger than the maxidle method in that it forces a
3504 redispatch after some absolute delay.
3505
Olivier Houchard4e694042017-03-14 20:01:29 +01003506 dynamic Activate dynamic cookies. When used, a session cookie is
3507 dynamically created for each server, based on the IP and port
3508 of the server, and a secret key, specified in the
3509 "dynamic-cookie-key" backend directive.
3510 The cookie will be regenerated each time the IP address change,
3511 and is only generated for IPv4/IPv6.
3512
Christopher Fauletdb2cdbb2020-01-21 11:06:48 +01003513 attr This option tells haproxy to add an extra attribute when a
3514 cookie is inserted. The attribute value can contain any
3515 characters except control ones or ";". This option may be
3516 repeated.
3517
Willy Tarreau0ba27502007-12-24 16:55:16 +01003518 There can be only one persistence cookie per HTTP backend, and it can be
3519 declared in a defaults section. The value of the cookie will be the value
3520 indicated after the "cookie" keyword in a "server" statement. If no cookie
3521 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003522
Willy Tarreau0ba27502007-12-24 16:55:16 +01003523 Examples :
3524 cookie JSESSIONID prefix
3525 cookie SRV insert indirect nocache
3526 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003527 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003528
Willy Tarreau294d0f02015-08-10 19:40:12 +02003529 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003530
Willy Tarreau983e01e2010-01-11 18:42:06 +01003531
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003532declare capture [ request | response ] len <length>
3533 Declares a capture slot.
3534 May be used in sections : defaults | frontend | listen | backend
3535 no | yes | yes | no
3536 Arguments:
3537 <length> is the length allowed for the capture.
3538
3539 This declaration is only available in the frontend or listen section, but the
3540 reserved slot can be used in the backends. The "request" keyword allocates a
3541 capture slot for use in the request, and "response" allocates a capture slot
3542 for use in the response.
3543
3544 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003545 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003546 "http-request capture" and "http-response capture".
3547
3548
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003549default-server [param*]
3550 Change default options for a server in a backend
3551 May be used in sections : defaults | frontend | listen | backend
3552 yes | no | yes | yes
3553 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003554 <param*> is a list of parameters for this server. The "default-server"
3555 keyword accepts an important number of options and has a complete
3556 section dedicated to it. Please refer to section 5 for more
3557 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003558
Willy Tarreau983e01e2010-01-11 18:42:06 +01003559 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003560 default-server inter 1000 weight 13
3561
3562 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003563
Willy Tarreau983e01e2010-01-11 18:42:06 +01003564
Willy Tarreau0ba27502007-12-24 16:55:16 +01003565default_backend <backend>
3566 Specify the backend to use when no "use_backend" rule has been matched.
3567 May be used in sections : defaults | frontend | listen | backend
3568 yes | yes | yes | no
3569 Arguments :
3570 <backend> is the name of the backend to use.
3571
3572 When doing content-switching between frontend and backends using the
3573 "use_backend" keyword, it is often useful to indicate which backend will be
3574 used when no rule has matched. It generally is the dynamic backend which
3575 will catch all undetermined requests.
3576
Willy Tarreau0ba27502007-12-24 16:55:16 +01003577 Example :
3578
3579 use_backend dynamic if url_dyn
3580 use_backend static if url_css url_img extension_img
3581 default_backend dynamic
3582
Willy Tarreau98d04852015-05-26 12:18:29 +02003583 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003584
Willy Tarreau0ba27502007-12-24 16:55:16 +01003585
Baptiste Assmann27f51342013-10-09 06:51:49 +02003586description <string>
3587 Describe a listen, frontend or backend.
3588 May be used in sections : defaults | frontend | listen | backend
3589 no | yes | yes | yes
3590 Arguments : string
3591
3592 Allows to add a sentence to describe the related object in the HAProxy HTML
3593 stats page. The description will be printed on the right of the object name
3594 it describes.
3595 No need to backslash spaces in the <string> arguments.
3596
3597
Willy Tarreau0ba27502007-12-24 16:55:16 +01003598disabled
3599 Disable a proxy, frontend or backend.
3600 May be used in sections : defaults | frontend | listen | backend
3601 yes | yes | yes | yes
3602 Arguments : none
3603
3604 The "disabled" keyword is used to disable an instance, mainly in order to
3605 liberate a listening port or to temporarily disable a service. The instance
3606 will still be created and its configuration will be checked, but it will be
3607 created in the "stopped" state and will appear as such in the statistics. It
3608 will not receive any traffic nor will it send any health-checks or logs. It
3609 is possible to disable many instances at once by adding the "disabled"
3610 keyword in a "defaults" section.
3611
3612 See also : "enabled"
3613
3614
Willy Tarreau5ce94572010-06-07 14:35:41 +02003615dispatch <address>:<port>
3616 Set a default server address
3617 May be used in sections : defaults | frontend | listen | backend
3618 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003619 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003620
3621 <address> is the IPv4 address of the default server. Alternatively, a
3622 resolvable hostname is supported, but this name will be resolved
3623 during start-up.
3624
3625 <ports> is a mandatory port specification. All connections will be sent
3626 to this port, and it is not permitted to use port offsets as is
3627 possible with normal servers.
3628
Willy Tarreau787aed52011-04-15 06:45:37 +02003629 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003630 server can take the connection. In the past it was used to forward non
3631 persistent connections to an auxiliary load balancer. Due to its simple
3632 syntax, it has also been used for simple TCP relays. It is recommended not to
3633 use it for more clarity, and to use the "server" directive instead.
3634
3635 See also : "server"
3636
Olivier Houchard4e694042017-03-14 20:01:29 +01003637
3638dynamic-cookie-key <string>
3639 Set the dynamic cookie secret key for a backend.
3640 May be used in sections : defaults | frontend | listen | backend
3641 yes | no | yes | yes
3642 Arguments : The secret key to be used.
3643
3644 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003645 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003646 specified on the "server" line), using a hash of the IP address of the
3647 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003648 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003649 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003650
Willy Tarreau0ba27502007-12-24 16:55:16 +01003651enabled
3652 Enable a proxy, frontend or backend.
3653 May be used in sections : defaults | frontend | listen | backend
3654 yes | yes | yes | yes
3655 Arguments : none
3656
3657 The "enabled" keyword is used to explicitly enable an instance, when the
3658 defaults has been set to "disabled". This is very rarely used.
3659
3660 See also : "disabled"
3661
3662
3663errorfile <code> <file>
3664 Return a file contents instead of errors generated by HAProxy
3665 May be used in sections : defaults | frontend | listen | backend
3666 yes | yes | yes | yes
3667 Arguments :
3668 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003669 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3670 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003671
3672 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003673 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003674 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003675 error pages, and to use absolute paths, since files are read
3676 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003677
3678 It is important to understand that this keyword is not meant to rewrite
3679 errors returned by the server, but errors detected and returned by HAProxy.
3680 This is why the list of supported errors is limited to a small set.
3681
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003682 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3683
Willy Tarreau0ba27502007-12-24 16:55:16 +01003684 The files are returned verbatim on the TCP socket. This allows any trick such
3685 as redirections to another URL or site, as well as tricks to clean cookies,
3686 force enable or disable caching, etc... The package provides default error
3687 files returning the same contents as default errors.
3688
Willy Tarreau59140a22009-02-22 12:02:30 +01003689 The files should not exceed the configured buffer size (BUFSIZE), which
3690 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003691 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003692 loops between the client and HAProxy when all servers are down, causing an
3693 error to be returned instead of an image. For better HTTP compliance, it is
3694 recommended that all header lines end with CR-LF and not LF alone.
3695
Willy Tarreau0ba27502007-12-24 16:55:16 +01003696 The files are read at the same time as the configuration and kept in memory.
3697 For this reason, the errors continue to be returned even when the process is
3698 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003699 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003700 403 status code and interrogating a blocked URL.
3701
3702 See also : "errorloc", "errorloc302", "errorloc303"
3703
Willy Tarreau59140a22009-02-22 12:02:30 +01003704 Example :
3705 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003706 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003707 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3708 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3709
Willy Tarreau2769aa02007-12-27 18:26:09 +01003710
3711errorloc <code> <url>
3712errorloc302 <code> <url>
3713 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3714 May be used in sections : defaults | frontend | listen | backend
3715 yes | yes | yes | yes
3716 Arguments :
3717 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003718 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3719 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003720
3721 <url> it is the exact contents of the "Location" header. It may contain
3722 either a relative URI to an error page hosted on the same site,
3723 or an absolute URI designating an error page on another site.
3724 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003725 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003726
3727 It is important to understand that this keyword is not meant to rewrite
3728 errors returned by the server, but errors detected and returned by HAProxy.
3729 This is why the list of supported errors is limited to a small set.
3730
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003731 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3732
Willy Tarreau2769aa02007-12-27 18:26:09 +01003733 Note that both keyword return the HTTP 302 status code, which tells the
3734 client to fetch the designated URL using the same HTTP method. This can be
3735 quite problematic in case of non-GET methods such as POST, because the URL
3736 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003737 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003738 status code, indicating to the client that the URL must be fetched with a GET
3739 request.
3740
3741 See also : "errorfile", "errorloc303"
3742
3743
3744errorloc303 <code> <url>
3745 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3746 May be used in sections : defaults | frontend | listen | backend
3747 yes | yes | yes | yes
3748 Arguments :
3749 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003750 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3751 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003752
3753 <url> it is the exact contents of the "Location" header. It may contain
3754 either a relative URI to an error page hosted on the same site,
3755 or an absolute URI designating an error page on another site.
3756 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003757 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003758
3759 It is important to understand that this keyword is not meant to rewrite
3760 errors returned by the server, but errors detected and returned by HAProxy.
3761 This is why the list of supported errors is limited to a small set.
3762
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003763 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3764
Willy Tarreau2769aa02007-12-27 18:26:09 +01003765 Note that both keyword return the HTTP 303 status code, which tells the
3766 client to fetch the designated URL using the same HTTP GET method. This
3767 solves the usual problems associated with "errorloc" and the 302 code. It is
3768 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003769 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003770
3771 See also : "errorfile", "errorloc", "errorloc302"
3772
3773
Simon Horman51a1cf62015-02-03 13:00:44 +09003774email-alert from <emailaddr>
3775 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003776 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003777 May be used in sections: defaults | frontend | listen | backend
3778 yes | yes | yes | yes
3779
3780 Arguments :
3781
3782 <emailaddr> is the from email address to use when sending email alerts
3783
3784 Also requires "email-alert mailers" and "email-alert to" to be set
3785 and if so sending email alerts is enabled for the proxy.
3786
Simon Horman64e34162015-02-06 11:11:57 +09003787 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003788 "email-alert myhostname", "email-alert to", section 3.6 about
3789 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003790
3791
3792email-alert level <level>
3793 Declare the maximum log level of messages for which email alerts will be
3794 sent. This acts as a filter on the sending of email alerts.
3795 May be used in sections: defaults | frontend | listen | backend
3796 yes | yes | yes | yes
3797
3798 Arguments :
3799
3800 <level> One of the 8 syslog levels:
3801 emerg alert crit err warning notice info debug
3802 The above syslog levels are ordered from lowest to highest.
3803
3804 By default level is alert
3805
3806 Also requires "email-alert from", "email-alert mailers" and
3807 "email-alert to" to be set and if so sending email alerts is enabled
3808 for the proxy.
3809
Simon Horman1421e212015-04-30 13:10:35 +09003810 Alerts are sent when :
3811
3812 * An un-paused server is marked as down and <level> is alert or lower
3813 * A paused server is marked as down and <level> is notice or lower
3814 * A server is marked as up or enters the drain state and <level>
3815 is notice or lower
3816 * "option log-health-checks" is enabled, <level> is info or lower,
3817 and a health check status update occurs
3818
Simon Horman64e34162015-02-06 11:11:57 +09003819 See also : "email-alert from", "email-alert mailers",
3820 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003821 section 3.6 about mailers.
3822
3823
3824email-alert mailers <mailersect>
3825 Declare the mailers to be used when sending email alerts
3826 May be used in sections: defaults | frontend | listen | backend
3827 yes | yes | yes | yes
3828
3829 Arguments :
3830
3831 <mailersect> is the name of the mailers section to send email alerts.
3832
3833 Also requires "email-alert from" and "email-alert to" to be set
3834 and if so sending email alerts is enabled for the proxy.
3835
Simon Horman64e34162015-02-06 11:11:57 +09003836 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3837 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003838
3839
3840email-alert myhostname <hostname>
3841 Declare the to hostname address to be used when communicating with
3842 mailers.
3843 May be used in sections: defaults | frontend | listen | backend
3844 yes | yes | yes | yes
3845
3846 Arguments :
3847
Baptiste Assmann738bad92015-12-21 15:27:53 +01003848 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003849
3850 By default the systems hostname is used.
3851
3852 Also requires "email-alert from", "email-alert mailers" and
3853 "email-alert to" to be set and if so sending email alerts is enabled
3854 for the proxy.
3855
Simon Horman64e34162015-02-06 11:11:57 +09003856 See also : "email-alert from", "email-alert level", "email-alert mailers",
3857 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003858
3859
3860email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003861 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003862 header of email alerts. This is the address that email alerts are sent to.
3863 May be used in sections: defaults | frontend | listen | backend
3864 yes | yes | yes | yes
3865
3866 Arguments :
3867
3868 <emailaddr> is the to email address to use when sending email alerts
3869
3870 Also requires "email-alert mailers" and "email-alert to" to be set
3871 and if so sending email alerts is enabled for the proxy.
3872
Simon Horman64e34162015-02-06 11:11:57 +09003873 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003874 "email-alert myhostname", section 3.6 about mailers.
3875
3876
Willy Tarreau4de91492010-01-22 19:10:05 +01003877force-persist { if | unless } <condition>
3878 Declare a condition to force persistence on down servers
3879 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003880 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003881
3882 By default, requests are not dispatched to down servers. It is possible to
3883 force this using "option persist", but it is unconditional and redispatches
3884 to a valid server if "option redispatch" is set. That leaves with very little
3885 possibilities to force some requests to reach a server which is artificially
3886 marked down for maintenance operations.
3887
3888 The "force-persist" statement allows one to declare various ACL-based
3889 conditions which, when met, will cause a request to ignore the down status of
3890 a server and still try to connect to it. That makes it possible to start a
3891 server, still replying an error to the health checks, and run a specially
3892 configured browser to test the service. Among the handy methods, one could
3893 use a specific source IP address, or a specific cookie. The cookie also has
3894 the advantage that it can easily be added/removed on the browser from a test
3895 page. Once the service is validated, it is then possible to open the service
3896 to the world by returning a valid response to health checks.
3897
3898 The forced persistence is enabled when an "if" condition is met, or unless an
3899 "unless" condition is met. The final redispatch is always disabled when this
3900 is used.
3901
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003902 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003903 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003904
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003905
3906filter <name> [param*]
3907 Add the filter <name> in the filter list attached to the proxy.
3908 May be used in sections : defaults | frontend | listen | backend
3909 no | yes | yes | yes
3910 Arguments :
3911 <name> is the name of the filter. Officially supported filters are
3912 referenced in section 9.
3913
Tim Düsterhus4896c442016-11-29 02:15:19 +01003914 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003915 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01003916 filter. Please refer to the documentation of the corresponding
3917 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003918
3919 Multiple occurrences of the filter line can be used for the same proxy. The
3920 same filter can be referenced many times if needed.
3921
3922 Example:
3923 listen
3924 bind *:80
3925
3926 filter trace name BEFORE-HTTP-COMP
3927 filter compression
3928 filter trace name AFTER-HTTP-COMP
3929
3930 compression algo gzip
3931 compression offload
3932
3933 server srv1 192.168.0.1:80
3934
3935 See also : section 9.
3936
Willy Tarreau4de91492010-01-22 19:10:05 +01003937
Willy Tarreau2769aa02007-12-27 18:26:09 +01003938fullconn <conns>
3939 Specify at what backend load the servers will reach their maxconn
3940 May be used in sections : defaults | frontend | listen | backend
3941 yes | no | yes | yes
3942 Arguments :
3943 <conns> is the number of connections on the backend which will make the
3944 servers use the maximal number of connections.
3945
Willy Tarreau198a7442008-01-17 12:05:32 +01003946 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003947 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003948 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003949 load. The server will then always accept at least <minconn> connections,
3950 never more than <maxconn>, and the limit will be on the ramp between both
3951 values when the backend has less than <conns> concurrent connections. This
3952 makes it possible to limit the load on the servers during normal loads, but
3953 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003954 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003955
Willy Tarreaufbb78422011-06-05 15:38:35 +02003956 Since it's hard to get this value right, haproxy automatically sets it to
3957 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003958 backend (based on "use_backend" and "default_backend" rules). That way it's
3959 safe to leave it unset. However, "use_backend" involving dynamic names are
3960 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003961
Willy Tarreau2769aa02007-12-27 18:26:09 +01003962 Example :
3963 # The servers will accept between 100 and 1000 concurrent connections each
3964 # and the maximum of 1000 will be reached when the backend reaches 10000
3965 # connections.
3966 backend dynamic
3967 fullconn 10000
3968 server srv1 dyn1:80 minconn 100 maxconn 1000
3969 server srv2 dyn2:80 minconn 100 maxconn 1000
3970
3971 See also : "maxconn", "server"
3972
3973
3974grace <time>
3975 Maintain a proxy operational for some time after a soft stop
3976 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003977 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003978 Arguments :
3979 <time> is the time (by default in milliseconds) for which the instance
3980 will remain operational with the frontend sockets still listening
3981 when a soft-stop is received via the SIGUSR1 signal.
3982
3983 This may be used to ensure that the services disappear in a certain order.
3984 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003985 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003986 needed by the equipment to detect the failure.
3987
3988 Note that currently, there is very little benefit in using this parameter,
3989 and it may in fact complicate the soft-reconfiguration process more than
3990 simplify it.
3991
Willy Tarreau0ba27502007-12-24 16:55:16 +01003992
Andrew Rodland17be45e2016-10-25 17:04:12 -04003993hash-balance-factor <factor>
3994 Specify the balancing factor for bounded-load consistent hashing
3995 May be used in sections : defaults | frontend | listen | backend
3996 yes | no | no | yes
3997 Arguments :
3998 <factor> is the control for the maximum number of concurrent requests to
3999 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01004000 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04004001
4002 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
4003 enables an algorithm that prevents any one server from getting too many
4004 requests at once, even if some hash buckets receive many more requests than
4005 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
4006 <factor> is a percentage greater than 100. For example, if <factor> is 150,
4007 then no server will be allowed to have a load more than 1.5 times the average.
4008 If server weights are used, they will be respected.
4009
4010 If the first-choice server is disqualified, the algorithm will choose another
4011 server based on the request hash, until a server with additional capacity is
4012 found. A higher <factor> allows more imbalance between the servers, while a
4013 lower <factor> means that more servers will be checked on average, affecting
4014 performance. Reasonable values are from 125 to 200.
4015
Willy Tarreau760e81d2018-05-03 07:20:40 +02004016 This setting is also used by "balance random" which internally relies on the
4017 consistent hashing mechanism.
4018
Andrew Rodland17be45e2016-10-25 17:04:12 -04004019 See also : "balance" and "hash-type".
4020
4021
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004022hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004023 Specify a method to use for mapping hashes to servers
4024 May be used in sections : defaults | frontend | listen | backend
4025 yes | no | yes | yes
4026 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04004027 <method> is the method used to select a server from the hash computed by
4028 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004029
Bhaskar98634f02013-10-29 23:30:51 -04004030 map-based the hash table is a static array containing all alive servers.
4031 The hashes will be very smooth, will consider weights, but
4032 will be static in that weight changes while a server is up
4033 will be ignored. This means that there will be no slow start.
4034 Also, since a server is selected by its position in the array,
4035 most mappings are changed when the server count changes. This
4036 means that when a server goes up or down, or when a server is
4037 added to a farm, most connections will be redistributed to
4038 different servers. This can be inconvenient with caches for
4039 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01004040
Bhaskar98634f02013-10-29 23:30:51 -04004041 consistent the hash table is a tree filled with many occurrences of each
4042 server. The hash key is looked up in the tree and the closest
4043 server is chosen. This hash is dynamic, it supports changing
4044 weights while the servers are up, so it is compatible with the
4045 slow start feature. It has the advantage that when a server
4046 goes up or down, only its associations are moved. When a
4047 server is added to the farm, only a few part of the mappings
4048 are redistributed, making it an ideal method for caches.
4049 However, due to its principle, the distribution will never be
4050 very smooth and it may sometimes be necessary to adjust a
4051 server's weight or its ID to get a more balanced distribution.
4052 In order to get the same distribution on multiple load
4053 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004054 same IDs. Note: consistent hash uses sdbm and avalanche if no
4055 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04004056
4057 <function> is the hash function to be used :
4058
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004059 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04004060 reimplementation of ndbm) database library. It was found to do
4061 well in scrambling bits, causing better distribution of the keys
4062 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004063 function with good distribution, unless the total server weight
4064 is a multiple of 64, in which case applying the avalanche
4065 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04004066
4067 djb2 this function was first proposed by Dan Bernstein many years ago
4068 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004069 function provides a better distribution than sdbm. It generally
4070 works well with text-based inputs though it can perform extremely
4071 poorly with numeric-only input or when the total server weight is
4072 a multiple of 33, unless the avalanche modifier is also used.
4073
Willy Tarreaua0f42712013-11-14 14:30:35 +01004074 wt6 this function was designed for haproxy while testing other
4075 functions in the past. It is not as smooth as the other ones, but
4076 is much less sensible to the input data set or to the number of
4077 servers. It can make sense as an alternative to sdbm+avalanche or
4078 djb2+avalanche for consistent hashing or when hashing on numeric
4079 data such as a source IP address or a visitor identifier in a URL
4080 parameter.
4081
Willy Tarreau324f07f2015-01-20 19:44:50 +01004082 crc32 this is the most common CRC32 implementation as used in Ethernet,
4083 gzip, PNG, etc. It is slower than the other ones but may provide
4084 a better distribution or less predictable results especially when
4085 used on strings.
4086
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004087 <modifier> indicates an optional method applied after hashing the key :
4088
4089 avalanche This directive indicates that the result from the hash
4090 function above should not be used in its raw form but that
4091 a 4-byte full avalanche hash must be applied first. The
4092 purpose of this step is to mix the resulting bits from the
4093 previous hash in order to avoid any undesired effect when
4094 the input contains some limited values or when the number of
4095 servers is a multiple of one of the hash's components (64
4096 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
4097 result less predictable, but it's also not as smooth as when
4098 using the original function. Some testing might be needed
4099 with some workloads. This hash is one of the many proposed
4100 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004101
Bhaskar98634f02013-10-29 23:30:51 -04004102 The default hash type is "map-based" and is recommended for most usages. The
4103 default function is "sdbm", the selection of a function should be based on
4104 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004105
Andrew Rodland17be45e2016-10-25 17:04:12 -04004106 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004107
4108
Willy Tarreau0ba27502007-12-24 16:55:16 +01004109http-check disable-on-404
4110 Enable a maintenance mode upon HTTP/404 response to health-checks
4111 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01004112 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01004113 Arguments : none
4114
4115 When this option is set, a server which returns an HTTP code 404 will be
4116 excluded from further load-balancing, but will still receive persistent
4117 connections. This provides a very convenient method for Web administrators
4118 to perform a graceful shutdown of their servers. It is also important to note
4119 that a server which is detected as failed while it was in this mode will not
4120 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
4121 will immediately be reinserted into the farm. The status on the stats page
4122 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01004123 option only works in conjunction with the "httpchk" option. If this option
4124 is used with "http-check expect", then it has precedence over it so that 404
4125 responses will still be considered as soft-stop.
4126
4127 See also : "option httpchk", "http-check expect"
4128
4129
4130http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004131 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01004132 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02004133 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01004134 Arguments :
4135 <match> is a keyword indicating how to look for a specific pattern in the
4136 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004137 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01004138 exclamation mark ("!") to negate the match. Spaces are allowed
4139 between the exclamation mark and the keyword. See below for more
4140 details on the supported keywords.
4141
4142 <pattern> is the pattern to look for. It may be a string or a regular
4143 expression. If the pattern contains spaces, they must be escaped
4144 with the usual backslash ('\').
4145
4146 By default, "option httpchk" considers that response statuses 2xx and 3xx
4147 are valid, and that others are invalid. When "http-check expect" is used,
4148 it defines what is considered valid or invalid. Only one "http-check"
4149 statement is supported in a backend. If a server fails to respond or times
4150 out, the check obviously fails. The available matches are :
4151
4152 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004153 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004154 response's status code is exactly this string. If the
4155 "status" keyword is prefixed with "!", then the response
4156 will be considered invalid if the status code matches.
4157
4158 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004159 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004160 response's status code matches the expression. If the
4161 "rstatus" keyword is prefixed with "!", then the response
4162 will be considered invalid if the status code matches.
4163 This is mostly used to check for multiple codes.
4164
4165 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004166 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004167 response's body contains this exact string. If the
4168 "string" keyword is prefixed with "!", then the response
4169 will be considered invalid if the body contains this
4170 string. This can be used to look for a mandatory word at
4171 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004172 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01004173 trace).
4174
4175 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004176 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004177 response's body matches this expression. If the "rstring"
4178 keyword is prefixed with "!", then the response will be
4179 considered invalid if the body matches the expression.
4180 This can be used to look for a mandatory word at the end
4181 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01004182 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01004183
4184 It is important to note that the responses will be limited to a certain size
4185 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
4186 Thus, too large responses may not contain the mandatory pattern when using
4187 "string" or "rstring". If a large response is absolutely required, it is
4188 possible to change the default max size by setting the global variable.
4189 However, it is worth keeping in mind that parsing very large responses can
4190 waste some CPU cycles, especially when regular expressions are used, and that
4191 it is always better to focus the checks on smaller resources.
4192
Cyril Bonté32602d22015-01-30 00:07:07 +01004193 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
4194 will automatically append a "Connection: close" header, meaning that this
4195 header should not be present in the request provided by "option httpchk".
4196
Willy Tarreaubd741542010-03-16 18:46:54 +01004197 Last, if "http-check expect" is combined with "http-check disable-on-404",
4198 then this last one has precedence when the server responds with 404.
4199
4200 Examples :
4201 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004202 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01004203
4204 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004205 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01004206
4207 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004208 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01004209
4210 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004211 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01004212
Willy Tarreaubd741542010-03-16 18:46:54 +01004213 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004214
4215
Willy Tarreauef781042010-01-27 11:53:01 +01004216http-check send-state
4217 Enable emission of a state header with HTTP health checks
4218 May be used in sections : defaults | frontend | listen | backend
4219 yes | no | yes | yes
4220 Arguments : none
4221
4222 When this option is set, haproxy will systematically send a special header
4223 "X-Haproxy-Server-State" with a list of parameters indicating to each server
4224 how they are seen by haproxy. This can be used for instance when a server is
4225 manipulated without access to haproxy and the operator needs to know whether
4226 haproxy still sees it up or not, or if the server is the last one in a farm.
4227
4228 The header is composed of fields delimited by semi-colons, the first of which
4229 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
4230 checks on the total number before transition, just as appears in the stats
4231 interface. Next headers are in the form "<variable>=<value>", indicating in
4232 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08004233 - a variable "address", containing the address of the backend server.
4234 This corresponds to the <address> field in the server declaration. For
4235 unix domain sockets, it will read "unix".
4236
4237 - a variable "port", containing the port of the backend server. This
4238 corresponds to the <port> field in the server declaration. For unix
4239 domain sockets, it will read "unix".
4240
Willy Tarreauef781042010-01-27 11:53:01 +01004241 - a variable "name", containing the name of the backend followed by a slash
4242 ("/") then the name of the server. This can be used when a server is
4243 checked in multiple backends.
4244
4245 - a variable "node" containing the name of the haproxy node, as set in the
4246 global "node" variable, otherwise the system's hostname if unspecified.
4247
4248 - a variable "weight" indicating the weight of the server, a slash ("/")
4249 and the total weight of the farm (just counting usable servers). This
4250 helps to know if other servers are available to handle the load when this
4251 one fails.
4252
4253 - a variable "scur" indicating the current number of concurrent connections
4254 on the server, followed by a slash ("/") then the total number of
4255 connections on all servers of the same backend.
4256
4257 - a variable "qcur" indicating the current number of requests in the
4258 server's queue.
4259
4260 Example of a header received by the application server :
4261 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
4262 scur=13/22; qcur=0
4263
4264 See also : "option httpchk", "http-check disable-on-404"
4265
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004266
4267http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004268 Access control for Layer 7 requests
4269
4270 May be used in sections: defaults | frontend | listen | backend
4271 no | yes | yes | yes
4272
Willy Tarreau20b0de52012-12-24 15:45:22 +01004273 The http-request statement defines a set of rules which apply to layer 7
4274 processing. The rules are evaluated in their declaration order when they are
4275 met in a frontend, listen or backend section. Any rule may optionally be
4276 followed by an ACL-based condition, in which case it will only be evaluated
4277 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004278
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004279 The first keyword is the rule's action. The supported actions are described
4280 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004281
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004282 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004283
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004284 It is important to know that http-request rules are processed very early in
4285 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
4286 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
4287 visible by almost all further ACL rules.
Willy Tarreau53275e82017-11-24 07:52:01 +01004288
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004289 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
4290 in newer versions (>= 1.5). But if you need to use regular expression to
4291 delete headers, you can still use "reqdel". Also please use
4292 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Willy Tarreauccbcc372012-12-27 12:37:57 +01004293
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004294 Example:
4295 acl nagios src 192.168.129.3
4296 acl local_net src 192.168.0.0/16
4297 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01004298
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004299 http-request allow if nagios
4300 http-request allow if local_net auth_ok
4301 http-request auth realm Gimme if local_net auth_ok
4302 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01004303
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004304 Example:
4305 acl key req.hdr(X-Add-Acl-Key) -m found
4306 acl add path /addacl
4307 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01004308
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004309 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01004310
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004311 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4312 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004313
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004314 Example:
4315 acl value req.hdr(X-Value) -m found
4316 acl setmap path /setmap
4317 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004318
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004319 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004320
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004321 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4322 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004323
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004324 See also : "stats http-request", section 3.4 about userlists and section 7
4325 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004326
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004327http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004328
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004329 This is used to add a new entry into an ACL. The ACL must be loaded from a
4330 file (even a dummy empty file). The file name of the ACL to be updated is
4331 passed between parentheses. It takes one argument: <key fmt>, which follows
4332 log-format rules, to collect content of the new entry. It performs a lookup
4333 in the ACL before insertion, to avoid duplicated (or more) values. This
4334 lookup is done by a linear search and can be expensive with large lists!
4335 It is the equivalent of the "add acl" command from the stats socket, but can
4336 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06004337
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004338http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004339
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004340 This appends an HTTP header field whose name is specified in <name> and
4341 whose value is defined by <fmt> which follows the log-format rules (see
4342 Custom Log Format in section 8.2.4). This is particularly useful to pass
4343 connection-specific information to the server (e.g. the client's SSL
4344 certificate), or to combine several headers into one. This rule is not
4345 final, so it is possible to add other similar rules. Note that header
4346 addition is performed immediately, so one rule might reuse the resulting
4347 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004348
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004349http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004350
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004351 This stops the evaluation of the rules and lets the request pass the check.
4352 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06004353
Sasha Pachev218f0642014-06-16 12:05:59 -06004354
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004355http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004356
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004357 This stops the evaluation of the rules and immediately responds with an
4358 HTTP 401 or 407 error code to invite the user to present a valid user name
4359 and password. No further "http-request" rules are evaluated. An optional
4360 "realm" parameter is supported, it sets the authentication realm that is
4361 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06004362
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004363 Example:
4364 acl auth_ok http_auth_group(L1) G1
4365 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06004366
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004367http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004368
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004369 See section 10.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004370
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004371http-request capture <sample> [ len <length> | id <id> ]
4372 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004373
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004374 This captures sample expression <sample> from the request buffer, and
4375 converts it to a string of at most <len> characters. The resulting string is
4376 stored into the next request "capture" slot, so it will possibly appear next
4377 to some captured HTTP headers. It will then automatically appear in the logs,
4378 and it will be possible to extract it using sample fetch rules to feed it
4379 into headers or anything. The length should be limited given that this size
4380 will be allocated for each capture during the whole session life.
4381 Please check section 7.3 (Fetching samples) and "capture request header" for
4382 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004383
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004384 If the keyword "id" is used instead of "len", the action tries to store the
4385 captured string in a previously declared capture slot. This is useful to run
4386 captures in backends. The slot id can be declared by a previous directive
Baptiste Assmann63b220d2020-01-16 14:34:22 +01004387 "http-request capture" or with the "declare capture" keyword.
4388
4389 When using this action in a backend, double check that the relevant
4390 frontend(s) have the required capture slots otherwise, this rule will be
4391 ignored at run time. This can't be detected at configuration parsing time
4392 due to HAProxy's ability to dynamically resolve backend name at runtime.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004393
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004394http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004395
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004396 This is used to delete an entry from an ACL. The ACL must be loaded from a
4397 file (even a dummy empty file). The file name of the ACL to be updated is
4398 passed between parentheses. It takes one argument: <key fmt>, which follows
4399 log-format rules, to collect content of the entry to delete.
4400 It is the equivalent of the "del acl" command from the stats socket, but can
4401 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004402
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004403http-request del-header <name> [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004404
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004405 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004406
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004407http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004408
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004409 This is used to delete an entry from a MAP. The MAP must be loaded from a
4410 file (even a dummy empty file). The file name of the MAP to be updated is
4411 passed between parentheses. It takes one argument: <key fmt>, which follows
4412 log-format rules, to collect content of the entry to delete.
4413 It takes one argument: "file name" It is the equivalent of the "del map"
4414 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02004415
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004416http-request deny [deny_status <status>] [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04004417
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004418 This stops the evaluation of the rules and immediately rejects the request
4419 and emits an HTTP 403 error, or optionally the status code specified as an
4420 argument to "deny_status". The list of permitted status codes is limited to
4421 those that can be overridden by the "errorfile" directive.
4422 No further "http-request" rules are evaluated.
Patrick Hemmer268a7072018-05-11 12:52:31 -04004423
Olivier Houchard602bf7d2019-05-10 13:59:15 +02004424http-request disable-l7-retry [ { if | unless } <condition> ]
4425 This disables any attempt to retry the request if it fails for any other
4426 reason than a connection failure. This can be useful for example to make
4427 sure POST requests aren't retried on failure.
4428
Baptiste Assmann333939c2019-01-21 08:34:50 +01004429http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
4430
4431 This action performs a DNS resolution of the output of <expr> and stores
4432 the result in the variable <var>. It uses the DNS resolvers section
4433 pointed by <resolvers>.
4434 It is possible to choose a resolution preference using the optional
4435 arguments 'ipv4' or 'ipv6'.
4436 When performing the DNS resolution, the client side connection is on
4437 pause waiting till the end of the resolution.
4438 If an IP address can be found, it is stored into <var>. If any kind of
4439 error occurs, then <var> is not set.
4440 One can use this action to discover a server IP address at run time and
4441 based on information found in the request (IE a Host header).
4442 If this action is used to find the server's IP address (using the
4443 "set-dst" action), then the server IP address in the backend must be set
4444 to 0.0.0.0.
4445
4446 Example:
4447 resolvers mydns
4448 nameserver local 127.0.0.53:53
4449 nameserver google 8.8.8.8:53
4450 timeout retry 1s
4451 hold valid 10s
4452 hold nx 3s
4453 hold other 3s
4454 hold obsolete 0s
4455 accepted_payload_size 8192
4456
4457 frontend fe
4458 bind 10.42.0.1:80
4459 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
4460 http-request capture var(txn.myip) len 40
4461
4462 # return 503 when the variable is not set,
4463 # which mean DNS resolution error
4464 use_backend b_503 unless { var(txn.myip) -m found }
4465
4466 default_backend be
4467
4468 backend b_503
4469 # dummy backend used to return 503.
4470 # one can use the errorfile directive to send a nice
4471 # 503 error page to end users
4472
4473 backend be
4474 # rule to prevent HAProxy from reconnecting to services
4475 # on the local network (forged DNS name used to scan the network)
4476 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
4477 http-request set-dst var(txn.myip)
4478 server clear 0.0.0.0:0
4479
4480 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
4481 be used to scan the network or worst won't loop over itself...
4482
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004483http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
4484
4485 This is used to build an HTTP 103 Early Hints response prior to any other one.
4486 This appends an HTTP header field to this response whose name is specified in
4487 <name> and whose value is defined by <fmt> which follows the log-format rules
4488 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01004489 to the client some Link headers to preload resources required to render the
4490 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004491
4492 See RFC 8297 for more information.
4493
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004494http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004495
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004496 This performs an HTTP redirection based on a redirect rule. This is exactly
4497 the same as the "redirect" statement except that it inserts a redirect rule
4498 which can be processed in the middle of other "http-request" rules and that
4499 these rules use the "log-format" strings. See the "redirect" keyword for the
4500 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004501
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004502http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004503
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004504 This stops the evaluation of the rules and immediately closes the connection
4505 without sending any response. It acts similarly to the
4506 "tcp-request content reject" rules. It can be useful to force an immediate
4507 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004508
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004509http-request replace-header <name> <match-regex> <replace-fmt>
4510 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02004511
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004512 This matches the value of all occurences of header field <name> against
4513 <match-regex>. Matching is performed case-sensitively. Matching values are
4514 completely replaced by <replace-fmt>. Format characters are allowed in
4515 <replace-fmt> and work like <fmt> arguments in "http-request add-header".
4516 Standard back-references using the backslash ('\') followed by a number are
4517 supported.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004518
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004519 This action acts on whole header lines, regardless of the number of values
4520 they may contain. Thus it is well-suited to process headers naturally
4521 containing commas in their value, such as If-Modified-Since. Headers that
4522 contain a comma-separated list of values, such as Accept, should be processed
4523 using "http-request replace-value".
William Lallemand86d0df02017-11-24 21:36:45 +01004524
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004525 Example:
4526 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
4527
4528 # applied to:
4529 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4530
4531 # outputs:
4532 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
4533
4534 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02004535
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004536 http-request replace-header User-Agent curl foo
4537
4538 # applied to:
4539 User-Agent: curl/7.47.0
Willy Tarreau09448f72014-06-25 18:12:15 +02004540
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004541 # outputs:
4542 User-Agent: foo
Willy Tarreau09448f72014-06-25 18:12:15 +02004543
Willy Tarreaudfc85772019-12-17 06:52:51 +01004544http-request replace-path <match-regex> <replace-fmt>
4545 [ { if | unless } <condition> ]
4546
4547 This works like "replace-header" except that it works on the request's path
4548 component instead of a header. The path component starts at the first '/'
4549 after an optional scheme+authority. It does contain the query string if any
4550 is present. The replacement does not modify the scheme nor authority.
4551
4552 It is worth noting that regular expressions may be more expensive to evaluate
4553 than certain ACLs, so rare replacements may benefit from a condition to avoid
4554 performing the evaluation at all if it does not match.
4555
4556 Example:
4557 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4558 http-request replace-path (.*) /foo\1
4559
4560 # suffix /foo : turn /bar?q=1 into /bar/foo?q=1 :
4561 http-request replace-path ([^?]*)(\?(.*))? \1/foo\2
4562
4563 # strip /foo : turn /foo/bar?q=1 into /bar?q=1
4564 http-request replace-path /foo/(.*) /\1
4565 # or more efficient if only some requests match :
4566 http-request replace-path /foo/(.*) /\1 if { url_beg /foo/ }
4567
Willy Tarreau33810222019-06-12 17:44:02 +02004568http-request replace-uri <match-regex> <replace-fmt>
4569 [ { if | unless } <condition> ]
4570
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004571 This works like "replace-header" except that it works on the request's URI part
4572 instead of a header. The URI part may contain an optional scheme, authority or
4573 query string. These are considered to be part of the value that is matched
4574 against.
4575
4576 It is worth noting that regular expressions may be more expensive to evaluate
4577 than certain ACLs, so rare replacements may benefit from a condition to avoid
4578 performing the evaluation at all if it does not match.
Willy Tarreau33810222019-06-12 17:44:02 +02004579
Willy Tarreaud41821d2019-12-17 06:51:20 +01004580 IMPORTANT NOTE: historically in HTTP/1.x, the vast majority of requests sent
4581 by browsers use the "origin form", which differs from the "absolute form" in
4582 that they do not contain a scheme nor authority in the URI portion. Mostly
4583 only requests sent to proxies, those forged by hand and some emitted by
4584 certain applications use the absolute form. As such, "replace-uri" usually
4585 works fine most of the time in HTTP/1.x with rules starting with a "/". But
4586 with HTTP/2, clients are encouraged to send absolute URIs only, which look
4587 like the ones HTTP/1 clients use to talk to proxies. Such partial replace-uri
4588 rules may then fail in HTTP/2 when they work in HTTP/1. Either the rules need
Willy Tarreaudfc85772019-12-17 06:52:51 +01004589 to be adapted to optionally match a scheme and authority, or replace-path
4590 should be used.
Willy Tarreau33810222019-06-12 17:44:02 +02004591
Willy Tarreaud41821d2019-12-17 06:51:20 +01004592 Example:
4593 # rewrite all "http" absolute requests to "https":
4594 http-request replace-uri ^http://(.*) https://\1
Willy Tarreau33810222019-06-12 17:44:02 +02004595
Willy Tarreaud41821d2019-12-17 06:51:20 +01004596 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4597 http-request replace-uri ([^/:]*://[^/]*)?(.*) \1/foo\2
Willy Tarreau33810222019-06-12 17:44:02 +02004598
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004599http-request replace-value <name> <match-regex> <replace-fmt>
4600 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02004601
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004602 This works like "replace-header" except that it matches the regex against
4603 every comma-delimited value of the header field <name> instead of the
4604 entire header. This is suited for all headers which are allowed to carry
4605 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02004606
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004607 Example:
4608 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004609
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004610 # applied to:
4611 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004612
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01004613 # outputs:
4614 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004615
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004616http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4617http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004618
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004619 This actions increments the GPC0 or GPC1 counter according with the sticky
4620 counter designated by <sc-id>. If an error occurs, this action silently fails
4621 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004622
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004623http-request sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004624
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004625 This action sets the GPT0 tag according to the sticky counter designated by
4626 <sc-id> and the value of <int>. The expected result is a boolean. If an error
4627 occurs, this action silently fails and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004628
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004629http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004630
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004631 This is used to set the destination IP address to the value of specified
4632 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
4633 but provides the correct IP in a HTTP header; or you want to mask the IP for
4634 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
4635 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004636
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004637 Arguments:
4638 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4639 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004640
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004641 Example:
4642 http-request set-dst hdr(x-dst)
4643 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01004644
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004645 When possible, set-dst preserves the original destination port as long as the
4646 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004647
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004648http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004649
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004650 This is used to set the destination port address to the value of specified
4651 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
4652 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004653
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004654 Arguments:
4655 <expr> Is a standard HAProxy expression formed by a sample-fetch
4656 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004657
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004658 Example:
4659 http-request set-dst-port hdr(x-port)
4660 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004661
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004662 When possible, set-dst-port preserves the original destination address as
4663 long as the address family supports a port, otherwise it forces the
4664 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004665
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004666http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004667
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004668 This does the same as "http-request add-header" except that the header name
4669 is first removed if it existed. This is useful when passing security
4670 information to the server, where the header must not be manipulated by
4671 external users. Note that the new value is computed before the removal so it
4672 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02004673
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004674 Example:
4675 http-request set-header X-Haproxy-Current-Date %T
4676 http-request set-header X-SSL %[ssl_fc]
4677 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
4678 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4679 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4680 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4681 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4682 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4683 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02004684
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004685http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004686
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004687 This is used to change the log level of the current request when a certain
4688 condition is met. Valid levels are the 8 syslog levels (see the "log"
4689 keyword) plus the special level "silent" which disables logging for this
4690 request. This rule is not final so the last matching rule wins. This rule
4691 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004692
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004693http-request set-map(<file-name>) <key fmt> <value fmt>
4694 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004695
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004696 This is used to add a new entry into a MAP. The MAP must be loaded from a
4697 file (even a dummy empty file). The file name of the MAP to be updated is
4698 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
4699 log-format rules, used to collect MAP key, and <value fmt>, which follows
4700 log-format rules, used to collect content for the new entry.
4701 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4702 more) values. This lookup is done by a linear search and can be expensive
4703 with large lists! It is the equivalent of the "set map" command from the
4704 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004705
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004706http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004707
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004708 This is used to set the Netfilter MARK on all packets sent to the client to
4709 the value passed in <mark> on platforms which support it. This value is an
4710 unsigned 32 bit value which can be matched by netfilter and by the routing
4711 table. It can be expressed both in decimal or hexadecimal format (prefixed by
4712 "0x"). This can be useful to force certain packets to take a different route
4713 (for example a cheaper network path for bulk downloads). This works on Linux
4714 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02004715
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004716http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004717
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004718 This rewrites the request method with the result of the evaluation of format
4719 string <fmt>. There should be very few valid reasons for having to do so as
4720 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004721
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004722http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004723
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004724 This sets the "nice" factor of the current request being processed. It only
4725 has effect against the other requests being processed at the same time.
4726 The default value is 0, unless altered by the "nice" setting on the "bind"
4727 line. The accepted range is -1024..1024. The higher the value, the nicest
4728 the request will be. Lower values will make the request more important than
4729 other ones. This can be useful to improve the speed of some requests, or
4730 lower the priority of non-important requests. Using this setting without
4731 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004732
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004733http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02004734
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004735 This rewrites the request path with the result of the evaluation of format
4736 string <fmt>. The query string, if any, is left intact. If a scheme and
4737 authority is found before the path, they are left intact as well. If the
4738 request doesn't have a path ("*"), this one is replaced with the format.
4739 This can be used to prepend a directory component in front of a path for
4740 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02004741
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004742 Example :
4743 # prepend the host name before the path
4744 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004745
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004746http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02004747
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004748 This is used to set the queue priority class of the current request.
4749 The value must be a sample expression which converts to an integer in the
4750 range -2047..2047. Results outside this range will be truncated.
4751 The priority class determines the order in which queued requests are
4752 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004753
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004754http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004755
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004756 This is used to set the queue priority timestamp offset of the current
4757 request. The value must be a sample expression which converts to an integer
4758 in the range -524287..524287. Results outside this range will be truncated.
4759 When a request is queued, it is ordered first by the priority class, then by
4760 the current timestamp adjusted by the given offset in milliseconds. Lower
4761 values have higher priority.
4762 Note that the resulting timestamp is is only tracked with enough precision
4763 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
4764 adjusted timestamp exceeds this value, it will be misidentified as highest
4765 priority. Thus it is important to set "timeout queue" to a value, where when
4766 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004767
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004768http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004769
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004770 This rewrites the request's query string which appears after the first
4771 question mark ("?") with the result of the evaluation of format string <fmt>.
4772 The part prior to the question mark is left intact. If the request doesn't
4773 contain a question mark and the new value is not empty, then one is added at
4774 the end of the URI, followed by the new value. If a question mark was
4775 present, it will never be removed even if the value is empty. This can be
4776 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004777
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004778 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004779
4780 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004781 # replace "%3D" with "=" in the query string
4782 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004783
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004784http-request set-src <expr> [ { if | unless } <condition> ]
4785 This is used to set the source IP address to the value of specified
4786 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
4787 provides the correct IP in a HTTP header; or you want to mask source IP for
4788 privacy.
4789
4790 Arguments :
4791 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4792 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004793
Cyril Bonté78caf842010-03-10 22:41:43 +01004794 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004795 http-request set-src hdr(x-forwarded-for)
4796 http-request set-src src,ipmask(24)
4797
4798 When possible, set-src preserves the original source port as long as the
4799 address family allows it, otherwise the source port is set to 0.
4800
4801http-request set-src-port <expr> [ { if | unless } <condition> ]
4802
4803 This is used to set the source port address to the value of specified
4804 expression.
4805
4806 Arguments:
4807 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4808 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004809
Willy Tarreau20b0de52012-12-24 15:45:22 +01004810 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004811 http-request set-src-port hdr(x-port)
4812 http-request set-src-port int(4000)
4813
4814 When possible, set-src-port preserves the original source address as long as
4815 the address family supports a port, otherwise it forces the source address to
4816 IPv4 "0.0.0.0" before rewriting the port.
4817
4818http-request set-tos <tos> [ { if | unless } <condition> ]
4819
4820 This is used to set the TOS or DSCP field value of packets sent to the client
4821 to the value passed in <tos> on platforms which support this. This value
4822 represents the whole 8 bits of the IP TOS field, and can be expressed both in
4823 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
4824 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
4825 be used to adjust some routing behavior on border routers based on some
4826 information from the request.
4827
4828 See RFC 2474, 2597, 3260 and 4594 for more information.
4829
4830http-request set-uri <fmt> [ { if | unless } <condition> ]
4831
4832 This rewrites the request URI with the result of the evaluation of format
4833 string <fmt>. The scheme, authority, path and query string are all replaced
4834 at once. This can be used to rewrite hosts in front of proxies, or to
4835 perform complex modifications to the URI such as moving parts between the
4836 path and the query string.
4837 See also "http-request set-path" and "http-request set-query".
4838
4839http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
4840
4841 This is used to set the contents of a variable. The variable is declared
4842 inline.
4843
4844 Arguments:
4845 <var-name> The name of the variable starts with an indication about its
4846 scope. The scopes allowed are:
4847 "proc" : the variable is shared with the whole process
4848 "sess" : the variable is shared with the whole session
4849 "txn" : the variable is shared with the transaction
4850 (request and response)
4851 "req" : the variable is shared only during request
4852 processing
4853 "res" : the variable is shared only during response
4854 processing
4855 This prefix is followed by a name. The separator is a '.'.
4856 The name may only contain characters 'a-z', 'A-Z', '0-9'
4857 and '_'.
4858
4859 <expr> Is a standard HAProxy expression formed by a sample-fetch
4860 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004861
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004862 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004863 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004864
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004865http-request send-spoe-group <engine-name> <group-name>
4866 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004867
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004868 This action is used to trigger sending of a group of SPOE messages. To do so,
4869 the SPOE engine used to send messages must be defined, as well as the SPOE
4870 group to send. Of course, the SPOE engine must refer to an existing SPOE
4871 filter. If not engine name is provided on the SPOE filter line, the SPOE
4872 agent name must be used.
4873
4874 Arguments:
4875 <engine-name> The SPOE engine name.
4876
4877 <group-name> The SPOE group name as specified in the engine
4878 configuration.
4879
4880http-request silent-drop [ { if | unless } <condition> ]
4881
4882 This stops the evaluation of the rules and makes the client-facing connection
4883 suddenly disappear using a system-dependent way that tries to prevent the
4884 client from being notified. The effect it then that the client still sees an
4885 established connection while there's none on HAProxy. The purpose is to
4886 achieve a comparable effect to "tarpit" except that it doesn't use any local
4887 resource at all on the machine running HAProxy. It can resist much higher
4888 loads than "tarpit", and slow down stronger attackers. It is important to
4889 understand the impact of using this mechanism. All stateful equipment placed
4890 between the client and HAProxy (firewalls, proxies, load balancers) will also
4891 keep the established connection for a long time and may suffer from this
4892 action.
4893 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
4894 option is used to block the emission of a TCP reset. On other systems, the
4895 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
4896 router, though it's still delivered to local networks. Do not use it unless
4897 you fully understand how it works.
4898
4899http-request tarpit [deny_status <status>] [ { if | unless } <condition> ]
4900
4901 This stops the evaluation of the rules and immediately blocks the request
4902 without responding for a delay specified by "timeout tarpit" or
4903 "timeout connect" if the former is not set. After that delay, if the client
4904 is still connected, an HTTP error 500 (or optionally the status code
4905 specified as an argument to "deny_status") is returned so that the client
4906 does not suspect it has been tarpitted. Logs will report the flags "PT".
4907 The goal of the tarpit rule is to slow down robots during an attack when
4908 they're limited on the number of concurrent requests. It can be very
4909 efficient against very dumb robots, and will significantly reduce the load
4910 on firewalls compared to a "deny" rule. But when facing "correctly"
4911 developed robots, it can make things worse by forcing haproxy and the front
4912 firewall to support insane number of concurrent connections.
4913 See also the "silent-drop" action.
4914
4915http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
4916http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
4917http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
4918
4919 This enables tracking of sticky counters from current request. These rules do
4920 not stop evaluation and do not change default action. The number of counters
4921 that may be simultaneously tracked by the same connection is set in
4922 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
4923 so the track-sc number is between 0 and (MAX_SESS_STCKTR-1). The first
4924 "track-sc0" rule executed enables tracking of the counters of the specified
4925 table as the first set. The first "track-sc1" rule executed enables tracking
4926 of the counters of the specified table as the second set. The first
4927 "track-sc2" rule executed enables tracking of the counters of the specified
4928 table as the third set. It is a recommended practice to use the first set of
4929 counters for the per-frontend counters and the second set for the per-backend
4930 ones. But this is just a guideline, all may be used everywhere.
4931
4932 Arguments :
4933 <key> is mandatory, and is a sample expression rule as described in
4934 section 7.3. It describes what elements of the incoming request or
4935 connection will be analyzed, extracted, combined, and used to
4936 select which table entry to update the counters.
4937
4938 <table> is an optional table to be used instead of the default one, which
4939 is the stick-table declared in the current proxy. All the counters
4940 for the matches and updates for the key will then be performed in
4941 that table until the session ends.
4942
4943 Once a "track-sc*" rule is executed, the key is looked up in the table and if
4944 it is not found, an entry is allocated for it. Then a pointer to that entry
4945 is kept during all the session's life, and this entry's counters are updated
4946 as often as possible, every time the session's counters are updated, and also
4947 systematically when the session ends. Counters are only updated for events
4948 that happen after the tracking has been started. As an exception, connection
4949 counters and request counters are systematically updated so that they reflect
4950 useful information.
4951
4952 If the entry tracks concurrent connection counters, one connection is counted
4953 for as long as the entry is tracked, and the entry will not expire during
4954 that time. Tracking counters also provides a performance advantage over just
4955 checking the keys, because only one table lookup is performed for all ACL
4956 checks that make use of it.
4957
4958http-request unset-var(<var-name>) [ { if | unless } <condition> ]
4959
4960 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004961
4962 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004963 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004964
Christopher Faulet6bd406e2019-11-22 15:34:17 +01004965http-request use-service <service-name> [ { if | unless } <condition> ]
4966
4967 This directive executes the configured HTTP service to reply to the request
4968 and stops the evaluation of the rules. An HTTP service may choose to reply by
4969 sending any valid HTTP response or it may immediately close the connection
4970 without sending any response. Outside natives services, for instance the
4971 Prometheus exporter, it is possible to write your own services in Lua. No
4972 further "http-request" rules are evaluated.
4973
4974 Arguments :
4975 <service-name> is mandatory. It is the service to call
4976
4977 Example:
4978 http-request use-service prometheus-exporter if { path /metrics }
4979
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004980http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004981
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004982 This will delay the processing of the request until the SSL handshake
4983 happened. This is mostly useful to delay processing early data until we're
4984 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004985
Willy Tarreauef781042010-01-27 11:53:01 +01004986
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004987http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004988 Access control for Layer 7 responses
4989
4990 May be used in sections: defaults | frontend | listen | backend
4991 no | yes | yes | yes
4992
4993 The http-response statement defines a set of rules which apply to layer 7
4994 processing. The rules are evaluated in their declaration order when they are
4995 met in a frontend, listen or backend section. Any rule may optionally be
4996 followed by an ACL-based condition, in which case it will only be evaluated
4997 if the condition is true. Since these rules apply on responses, the backend
4998 rules are applied first, followed by the frontend's rules.
4999
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005000 The first keyword is the rule's action. The supported actions are described
5001 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005002
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005003 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005004
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005005 It is important to know that http-response rules are processed very early in
5006 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
5007 headers added by "add-header"/"set-header" are visible by almost all further
5008 ACL rules.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005009
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005010 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
5011 in newer versions (>= 1.5). But if you need to use regular expression to
5012 delete headers, you can still use "rspdel". Also please use
5013 "http-response deny" instead of "rspdeny".
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005014
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005015 Example:
5016 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02005017
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005018 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06005019
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005020 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
5021 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06005022
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005023 Example:
5024 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06005025
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005026 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06005027
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005028 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
5029 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06005030
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005031 See also : "http-request", section 3.4 about userlists and section 7 about
5032 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06005033
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005034http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005035
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005036 This is used to add a new entry into an ACL. The ACL must be loaded from a
5037 file (even a dummy empty file). The file name of the ACL to be updated is
5038 passed between parentheses. It takes one argument: <key fmt>, which follows
5039 log-format rules, to collect content of the new entry. It performs a lookup
5040 in the ACL before insertion, to avoid duplicated (or more) values.
5041 This lookup is done by a linear search and can be expensive with large lists!
5042 It is the equivalent of the "add acl" command from the stats socket, but can
5043 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06005044
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005045http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005046
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005047 This appends an HTTP header field whose name is specified in <name> and whose
5048 value is defined by <fmt> which follows the log-format rules (see Custom Log
5049 Format in section 8.2.4). This may be used to send a cookie to a client for
5050 example, or to pass some internal information.
5051 This rule is not final, so it is possible to add other similar rules.
5052 Note that header addition is performed immediately, so one rule might reuse
5053 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06005054
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005055http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005056
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005057 This stops the evaluation of the rules and lets the response pass the check.
5058 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06005059
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02005060http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005061
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005062 See section 10.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06005063
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005064http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06005065
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005066 This captures sample expression <sample> from the response buffer, and
5067 converts it to a string. The resulting string is stored into the next request
5068 "capture" slot, so it will possibly appear next to some captured HTTP
5069 headers. It will then automatically appear in the logs, and it will be
5070 possible to extract it using sample fetch rules to feed it into headers or
5071 anything. Please check section 7.3 (Fetching samples) and
5072 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005073
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005074 The keyword "id" is the id of the capture slot which is used for storing the
5075 string. The capture slot must be defined in an associated frontend.
5076 This is useful to run captures in backends. The slot id can be declared by a
5077 previous directive "http-response capture" or with the "declare capture"
5078 keyword.
Baptiste Assmann63b220d2020-01-16 14:34:22 +01005079
5080 When using this action in a backend, double check that the relevant
5081 frontend(s) have the required capture slots otherwise, this rule will be
5082 ignored at run time. This can't be detected at configuration parsing time
5083 due to HAProxy's ability to dynamically resolve backend name at runtime.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005084
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005085http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02005086
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005087 This is used to delete an entry from an ACL. The ACL must be loaded from a
5088 file (even a dummy empty file). The file name of the ACL to be updated is
5089 passed between parentheses. It takes one argument: <key fmt>, which follows
5090 log-format rules, to collect content of the entry to delete.
5091 It is the equivalent of the "del acl" command from the stats socket, but can
5092 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02005093
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005094http-response del-header <name> [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02005095
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005096 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02005097
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005098http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02005099
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005100 This is used to delete an entry from a MAP. The MAP must be loaded from a
5101 file (even a dummy empty file). The file name of the MAP to be updated is
5102 passed between parentheses. It takes one argument: <key fmt>, which follows
5103 log-format rules, to collect content of the entry to delete.
5104 It takes one argument: "file name" It is the equivalent of the "del map"
5105 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005106
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005107http-response deny [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005108
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005109 This stops the evaluation of the rules and immediately rejects the response
5110 and emits an HTTP 502 error. No further "http-response" rules are evaluated.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005111
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005112http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005113
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005114 This performs an HTTP redirection based on a redirect rule.
5115 This supports a format string similarly to "http-request redirect" rules,
5116 with the exception that only the "location" type of redirect is possible on
5117 the response. See the "redirect" keyword for the rule's syntax. When a
5118 redirect rule is applied during a response, connections to the server are
5119 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005120
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005121http-response replace-header <name> <regex-match> <replace-fmt>
5122 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02005123
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01005124 This works like "http-request replace-header" except that it works on the
5125 server's response instead of the client's request.
William Lallemand86d0df02017-11-24 21:36:45 +01005126
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005127 Example:
5128 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02005129
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005130 # applied to:
5131 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005132
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005133 # outputs:
5134 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 +02005135
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005136 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005137
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005138http-response replace-value <name> <regex-match> <replace-fmt>
5139 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005140
Tim Duesterhus2f7045c2019-10-29 00:05:13 +01005141 This works like "http-response replace-value" except that it works on the
5142 server's response instead of the client's request.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005143
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005144 Example:
5145 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005146
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005147 # applied to:
5148 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01005149
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005150 # outputs:
5151 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005152
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005153http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
5154http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08005155
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005156 This action increments the GPC0 or GPC1 counter according with the sticky
5157 counter designated by <sc-id>. If an error occurs, this action silently fails
5158 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02005159
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005160http-response sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02005161
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005162 This action sets the GPT0 tag according to the sticky counter designated by
5163 <sc-id> and the value of <int>. The expected result is a boolean. If an error
5164 occurs, this action silently fails and the actions evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01005165
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005166http-response send-spoe-group [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02005167
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005168 This action is used to trigger sending of a group of SPOE messages. To do so,
5169 the SPOE engine used to send messages must be defined, as well as the SPOE
5170 group to send. Of course, the SPOE engine must refer to an existing SPOE
5171 filter. If not engine name is provided on the SPOE filter line, the SPOE
5172 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005173
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005174 Arguments:
5175 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005176
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005177 <group-name> The SPOE group name as specified in the engine
5178 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005179
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005180http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005181
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005182 This does the same as "add-header" except that the header name is first
5183 removed if it existed. This is useful when passing security information to
5184 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005185
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005186http-response set-log-level <level> [ { if | unless } <condition> ]
5187
5188 This is used to change the log level of the current request when a certain
5189 condition is met. Valid levels are the 8 syslog levels (see the "log"
5190 keyword) plus the special level "silent" which disables logging for this
5191 request. This rule is not final so the last matching rule wins. This rule can
5192 be useful to disable health checks coming from another equipment.
5193
5194http-response set-map(<file-name>) <key fmt> <value fmt>
5195
5196 This is used to add a new entry into a MAP. The MAP must be loaded from a
5197 file (even a dummy empty file). The file name of the MAP to be updated is
5198 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5199 log-format rules, used to collect MAP key, and <value fmt>, which follows
5200 log-format rules, used to collect content for the new entry. It performs a
5201 lookup in the MAP before insertion, to avoid duplicated (or more) values.
5202 This lookup is done by a linear search and can be expensive with large lists!
5203 It is the equivalent of the "set map" command from the stats socket, but can
5204 be triggered by an HTTP response.
5205
5206http-response set-mark <mark> [ { if | unless } <condition> ]
5207
5208 This is used to set the Netfilter MARK on all packets sent to the client to
5209 the value passed in <mark> on platforms which support it. This value is an
5210 unsigned 32 bit value which can be matched by netfilter and by the routing
5211 table. It can be expressed both in decimal or hexadecimal format (prefixed
5212 by "0x"). This can be useful to force certain packets to take a different
5213 route (for example a cheaper network path for bulk downloads). This works on
5214 Linux kernels 2.6.32 and above and requires admin privileges.
5215
5216http-response set-nice <nice> [ { if | unless } <condition> ]
5217
5218 This sets the "nice" factor of the current request being processed.
5219 It only has effect against the other requests being processed at the same
5220 time. The default value is 0, unless altered by the "nice" setting on the
5221 "bind" line. The accepted range is -1024..1024. The higher the value, the
5222 nicest the request will be. Lower values will make the request more important
5223 than other ones. This can be useful to improve the speed of some requests, or
5224 lower the priority of non-important requests. Using this setting without
5225 prior experimentation can cause some major slowdown.
5226
5227http-response set-status <status> [reason <str>]
5228 [ { if | unless } <condition> ]
5229
5230 This replaces the response status code with <status> which must be an integer
5231 between 100 and 999. Optionally, a custom reason text can be provided defined
5232 by <str>, or the default reason for the specified code will be used as a
5233 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005234
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005235 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005236 # return "431 Request Header Fields Too Large"
5237 http-response set-status 431
5238 # return "503 Slow Down", custom reason
5239 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005240
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005241http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005242
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005243 This is used to set the TOS or DSCP field value of packets sent to the client
5244 to the value passed in <tos> on platforms which support this.
5245 This value represents the whole 8 bits of the IP TOS field, and can be
5246 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
5247 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
5248 always 0. This can be used to adjust some routing behavior on border routers
5249 based on some information from the request.
5250
5251 See RFC 2474, 2597, 3260 and 4594 for more information.
5252
5253http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5254
5255 This is used to set the contents of a variable. The variable is declared
5256 inline.
5257
5258 Arguments:
5259 <var-name> The name of the variable starts with an indication about its
5260 scope. The scopes allowed are:
5261 "proc" : the variable is shared with the whole process
5262 "sess" : the variable is shared with the whole session
5263 "txn" : the variable is shared with the transaction
5264 (request and response)
5265 "req" : the variable is shared only during request
5266 processing
5267 "res" : the variable is shared only during response
5268 processing
5269 This prefix is followed by a name. The separator is a '.'.
5270 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5271 and '_'.
5272
5273 <expr> Is a standard HAProxy expression formed by a sample-fetch
5274 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005275
5276 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005277 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005278
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005279http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005280
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005281 This stops the evaluation of the rules and makes the client-facing connection
5282 suddenly disappear using a system-dependent way that tries to prevent the
5283 client from being notified. The effect it then that the client still sees an
5284 established connection while there's none on HAProxy. The purpose is to
5285 achieve a comparable effect to "tarpit" except that it doesn't use any local
5286 resource at all on the machine running HAProxy. It can resist much higher
5287 loads than "tarpit", and slow down stronger attackers. It is important to
5288 understand the impact of using this mechanism. All stateful equipment placed
5289 between the client and HAProxy (firewalls, proxies, load balancers) will also
5290 keep the established connection for a long time and may suffer from this
5291 action.
5292 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5293 option is used to block the emission of a TCP reset. On other systems, the
5294 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5295 router, though it's still delivered to local networks. Do not use it unless
5296 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005297
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005298http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5299http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5300http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005301
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005302 This enables tracking of sticky counters from current response. Please refer
5303 to "http-request track-sc" for a complete description. The only difference
5304 from "http-request track-sc" is the <key> sample expression can only make use
5305 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
5306 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
5307 supported, haproxy will fail and warn while parsing the config.
5308
5309http-response unset-var(<var-name>) [ { if | unless } <condition> ]
5310
5311 This is used to unset a variable. See "http-response set-var" for details
5312 about <var-name>.
5313
5314 Example:
5315 http-response unset-var(sess.last_redir)
5316
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02005317
Willy Tarreau30631952015-08-06 15:05:24 +02005318http-reuse { never | safe | aggressive | always }
5319 Declare how idle HTTP connections may be shared between requests
5320
5321 May be used in sections: defaults | frontend | listen | backend
5322 yes | no | yes | yes
5323
5324 By default, a connection established between haproxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01005325 which is considered safe for reuse is moved back to the server's idle
5326 connections pool so that any other request can make use of it. This is the
5327 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02005328
5329 The argument indicates the desired connection reuse strategy :
5330
Olivier Houchard86006a52018-12-14 19:37:49 +01005331 - "never" : idle connections are never shared between sessions. This mode
5332 may be enforced to cancel a different strategy inherited from
5333 a defaults section or for troubleshooting. For example, if an
5334 old bogus application considers that multiple requests over
5335 the same connection come from the same client and it is not
5336 possible to fix the application, it may be desirable to
5337 disable connection sharing in a single backend. An example of
5338 such an application could be an old haproxy using cookie
5339 insertion in tunnel mode and not checking any request past the
5340 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02005341
Olivier Houchard86006a52018-12-14 19:37:49 +01005342 - "safe" : this is the default and the recommended strategy. The first
5343 request of a session is always sent over its own connection,
5344 and only subsequent requests may be dispatched over other
5345 existing connections. This ensures that in case the server
5346 closes the connection when the request is being sent, the
5347 browser can decide to silently retry it. Since it is exactly
5348 equivalent to regular keep-alive, there should be no side
5349 effects.
Willy Tarreau30631952015-08-06 15:05:24 +02005350
5351 - "aggressive" : this mode may be useful in webservices environments where
5352 all servers are not necessarily known and where it would be
5353 appreciable to deliver most first requests over existing
5354 connections. In this case, first requests are only delivered
5355 over existing connections that have been reused at least once,
5356 proving that the server correctly supports connection reuse.
5357 It should only be used when it's sure that the client can
5358 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02005359 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02005360 downsides of rare connection failures.
5361
5362 - "always" : this mode is only recommended when the path to the server is
5363 known for never breaking existing connections quickly after
5364 releasing them. It allows the first request of a session to be
5365 sent to an existing connection. This can provide a significant
5366 performance increase over the "safe" strategy when the backend
5367 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005368 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02005369 sharing. It is recommended that the "http-keep-alive" timeout
5370 remains low in this mode so that no dead connections remain
5371 usable. In most cases, this will lead to the same performance
5372 gains as "aggressive" but with more risks. It should only be
5373 used when it improves the situation over "aggressive".
5374
5375 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005376 connection properties and compatibility. Specifically :
5377 - connections made with "usesrc" followed by a client-dependent value
5378 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005379
5380 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01005381 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005382
Lukas Tribusfd9b68c2018-10-27 20:06:59 +02005383 - connections with certain bogus authentication schemes (relying on the
5384 connection) like NTLM are detected, marked private and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005385
Lukas Tribus79a56932019-11-06 11:50:25 +01005386 A connection pool is involved and configurable with "pool-max-conn".
Willy Tarreau30631952015-08-06 15:05:24 +02005387
5388 Note: connection reuse improves the accuracy of the "server maxconn" setting,
5389 because almost no new connection will be established while idle connections
5390 remain available. This is particularly true with the "always" strategy.
5391
5392 See also : "option http-keep-alive", "server maxconn"
5393
5394
Mark Lamourinec2247f02012-01-04 13:02:01 -05005395http-send-name-header [<header>]
5396 Add the server name to a request. Use the header string given by <header>
Mark Lamourinec2247f02012-01-04 13:02:01 -05005397 May be used in sections: defaults | frontend | listen | backend
5398 yes | no | yes | yes
Mark Lamourinec2247f02012-01-04 13:02:01 -05005399 Arguments :
Mark Lamourinec2247f02012-01-04 13:02:01 -05005400 <header> The header string to use to send the server name
5401
Willy Tarreaue0e32792019-10-07 14:58:02 +02005402 The "http-send-name-header" statement causes the header field named <header>
5403 to be set to the name of the target server at the moment the request is about
5404 to be sent on the wire. Any existing occurrences of this header are removed.
5405 Upon retries and redispatches, the header field is updated to always reflect
5406 the server being attempted to connect to. Given that this header is modified
5407 very late in the connection setup, it may have unexpected effects on already
5408 modified headers. For example using it with transport-level header such as
5409 connection, content-length, transfer-encoding and so on will likely result in
5410 invalid requests being sent to the server. Additionally it has been reported
5411 that this directive is currently being used as a way to overwrite the Host
5412 header field in outgoing requests; while this trick has been known to work
5413 as a side effect of the feature for some time, it is not officially supported
5414 and might possibly not work anymore in a future version depending on the
5415 technical difficulties this feature induces. A long-term solution instead
5416 consists in fixing the application which required this trick so that it binds
5417 to the correct host name.
Mark Lamourinec2247f02012-01-04 13:02:01 -05005418
5419 See also : "server"
5420
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005421id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02005422 Set a persistent ID to a proxy.
5423 May be used in sections : defaults | frontend | listen | backend
5424 no | yes | yes | yes
5425 Arguments : none
5426
5427 Set a persistent ID for the proxy. This ID must be unique and positive.
5428 An unused ID will automatically be assigned if unset. The first assigned
5429 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005430
5431
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005432ignore-persist { if | unless } <condition>
5433 Declare a condition to ignore persistence
5434 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005435 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005436
5437 By default, when cookie persistence is enabled, every requests containing
5438 the cookie are unconditionally persistent (assuming the target server is up
5439 and running).
5440
5441 The "ignore-persist" statement allows one to declare various ACL-based
5442 conditions which, when met, will cause a request to ignore persistence.
5443 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005444 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005445 persistence for a specific User-Agent (for example, some web crawler bots).
5446
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005447 The persistence is ignored when an "if" condition is met, or unless an
5448 "unless" condition is met.
5449
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005450 Example:
5451 acl url_static path_beg /static /images /img /css
5452 acl url_static path_end .gif .png .jpg .css .js
5453 ignore-persist if url_static
5454
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005455 See also : "force-persist", "cookie", and section 7 about ACL usage.
5456
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005457load-server-state-from-file { global | local | none }
5458 Allow seamless reload of HAProxy
5459 May be used in sections: defaults | frontend | listen | backend
5460 yes | no | yes | yes
5461
5462 This directive points HAProxy to a file where server state from previous
5463 running process has been saved. That way, when starting up, before handling
5464 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01005465 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005466 to tell haproxy which file to use. For now, only 2 arguments to either prevent
5467 loading state or load states from a file containing all backends and servers.
5468 The state file can be generated by running the command "show servers state"
5469 over the stats socket and redirect output.
5470
Davor Ocelice9ed2812017-12-25 17:49:28 +01005471 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005472 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02005473 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005474
5475 Arguments:
5476 global load the content of the file pointed by the global directive
5477 named "server-state-file".
5478
5479 local load the content of the file pointed by the directive
5480 "server-state-file-name" if set. If not set, then the backend
5481 name is used as a file name.
5482
5483 none don't load any stat for this backend
5484
5485 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01005486 - server's IP address is preserved across reloads by default, but the
5487 order can be changed thanks to the server's "init-addr" setting. This
5488 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01005489 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01005490 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005491
5492 - server's weight is applied from previous running process unless it has
5493 has changed between previous and new configuration files.
5494
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005495 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005496
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005497 global
5498 stats socket /tmp/socket
5499 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005500
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005501 defaults
5502 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005503
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005504 backend bk
5505 server s1 127.0.0.1:22 check weight 11
5506 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005507
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005508
5509 Then one can run :
5510
5511 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
5512
5513 Content of the file /tmp/server_state would be like this:
5514
5515 1
5516 # <field names skipped for the doc example>
5517 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5518 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5519
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005520 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005521
5522 global
5523 stats socket /tmp/socket
5524 server-state-base /etc/haproxy/states
5525
5526 defaults
5527 load-server-state-from-file local
5528
5529 backend bk
5530 server s1 127.0.0.1:22 check weight 11
5531 server s2 127.0.0.1:22 check weight 12
5532
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005533
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005534 Then one can run :
5535
5536 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
5537
5538 Content of the file /etc/haproxy/states/bk would be like this:
5539
5540 1
5541 # <field names skipped for the doc example>
5542 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5543 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5544
5545 See also: "server-state-file", "server-state-file-name", and
5546 "show servers state"
5547
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005548
Willy Tarreau2769aa02007-12-27 18:26:09 +01005549log global
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005550log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
5551 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02005552no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01005553 Enable per-instance logging of events and traffic.
5554 May be used in sections : defaults | frontend | listen | backend
5555 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02005556
5557 Prefix :
5558 no should be used when the logger list must be flushed. For example,
5559 if you don't want to inherit from the default logger list. This
5560 prefix does not allow arguments.
5561
Willy Tarreau2769aa02007-12-27 18:26:09 +01005562 Arguments :
5563 global should be used when the instance's logging parameters are the
5564 same as the global ones. This is the most common usage. "global"
5565 replaces <address>, <facility> and <level> with those of the log
5566 entries found in the "global" section. Only one "log global"
5567 statement may be used per instance, and this form takes no other
5568 parameter.
5569
5570 <address> indicates where to send the logs. It takes the same format as
5571 for the "global" section's logs, and can be one of :
5572
5573 - An IPv4 address optionally followed by a colon (':') and a UDP
5574 port. If no port is specified, 514 is used by default (the
5575 standard syslog port).
5576
David du Colombier24bb5f52011-03-17 10:40:23 +01005577 - An IPv6 address followed by a colon (':') and optionally a UDP
5578 port. If no port is specified, 514 is used by default (the
5579 standard syslog port).
5580
Willy Tarreau2769aa02007-12-27 18:26:09 +01005581 - A filesystem path to a UNIX domain socket, keeping in mind
5582 considerations for chroot (be sure the path is accessible
5583 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01005584 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005585
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005586 - A file descriptor number in the form "fd@<number>", which may
5587 point to a pipe, terminal, or socket. In this case unbuffered
5588 logs are used and one writev() call per log is performed. This
5589 is a bit expensive but acceptable for most workloads. Messages
5590 sent this way will not be truncated but may be dropped, in
5591 which case the DroppedLogs counter will be incremented. The
5592 writev() call is atomic even on pipes for messages up to
5593 PIPE_BUF size, which POSIX recommends to be at least 512 and
5594 which is 4096 bytes on most modern operating systems. Any
5595 larger message may be interleaved with messages from other
5596 processes. Exceptionally for debugging purposes the file
5597 descriptor may also be directed to a file, but doing so will
5598 significantly slow haproxy down as non-blocking calls will be
5599 ignored. Also there will be no way to purge nor rotate this
5600 file without restarting the process. Note that the configured
5601 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01005602 with a TCP syslog server. See also the "short" and "raw"
5603 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005604
5605 - "stdout" / "stderr", which are respectively aliases for "fd@1"
5606 and "fd@2", see above.
5607
5608 You may want to reference some environment variables in the
5609 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01005610
Willy Tarreau18324f52014-06-27 18:10:07 +02005611 <length> is an optional maximum line length. Log lines larger than this
5612 value will be truncated before being sent. The reason is that
5613 syslog servers act differently on log line length. All servers
5614 support the default value of 1024, but some servers simply drop
5615 larger lines while others do log them. If a server supports long
5616 lines, it may make sense to set this value here in order to avoid
5617 truncating long lines. Similarly, if a server drops long lines,
5618 it is preferable to truncate them before sending them. Accepted
5619 values are 80 to 65535 inclusive. The default value of 1024 is
5620 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01005621 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02005622
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005623 <ranges> A list of comma-separated ranges to identify the logs to sample.
5624 This is used to balance the load of the logs to send to the log
5625 server. The limits of the ranges cannot be null. They are numbered
5626 from 1. The size or period (in number of logs) of the sample must
5627 be set with <sample_size> parameter.
5628
5629 <sample_size>
5630 The size of the sample in number of logs to consider when balancing
5631 their logging loads. It is used to balance the load of the logs to
5632 send to the syslog server. This size must be greater or equal to the
5633 maximum of the high limits of the ranges.
5634 (see also <ranges> parameter).
5635
Willy Tarreauadb345d2018-11-12 07:56:13 +01005636 <format> is the log format used when generating syslog messages. It may be
5637 one of the following :
5638
5639 rfc3164 The RFC3164 syslog message format. This is the default.
5640 (https://tools.ietf.org/html/rfc3164)
5641
5642 rfc5424 The RFC5424 syslog message format.
5643 (https://tools.ietf.org/html/rfc5424)
5644
Willy Tarreaue8746a02018-11-12 08:45:00 +01005645 short A message containing only a level between angle brackets such as
5646 '<3>', followed by the text. The PID, date, time, process name
5647 and system name are omitted. This is designed to be used with a
5648 local log server. This format is compatible with what the
5649 systemd logger consumes.
5650
Willy Tarreauc1b06452018-11-12 11:57:56 +01005651 raw A message containing only the text. The level, PID, date, time,
5652 process name and system name are omitted. This is designed to
5653 be used in containers or during development, where the severity
5654 only depends on the file descriptor used (stdout/stderr).
5655
Willy Tarreau2769aa02007-12-27 18:26:09 +01005656 <facility> must be one of the 24 standard syslog facilities :
5657
Willy Tarreaue8746a02018-11-12 08:45:00 +01005658 kern user mail daemon auth syslog lpr news
5659 uucp cron auth2 ftp ntp audit alert cron2
5660 local0 local1 local2 local3 local4 local5 local6 local7
5661
Willy Tarreauc1b06452018-11-12 11:57:56 +01005662 Note that the facility is ignored for the "short" and "raw"
5663 formats, but still required as a positional field. It is
5664 recommended to use "daemon" in this case to make it clear that
5665 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005666
5667 <level> is optional and can be specified to filter outgoing messages. By
5668 default, all messages are sent. If a level is specified, only
5669 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005670 will be sent. An optional minimum level can be specified. If it
5671 is set, logs emitted with a more severe level than this one will
5672 be capped to this level. This is used to avoid sending "emerg"
5673 messages on all terminals on some default syslog configurations.
5674 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005675
5676 emerg alert crit err warning notice info debug
5677
William Lallemand0f99e342011-10-12 17:50:54 +02005678 It is important to keep in mind that it is the frontend which decides what to
5679 log from a connection, and that in case of content switching, the log entries
5680 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005681
5682 However, backend log declaration define how and where servers status changes
5683 will be logged. Level "notice" will be used to indicate a server going up,
5684 "warning" will be used for termination signals and definitive service
5685 termination, and "alert" will be used for when a server goes down.
5686
5687 Note : According to RFC3164, messages are truncated to 1024 bytes before
5688 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005689
5690 Example :
5691 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01005692 log stdout format short daemon # send log to systemd
5693 log stdout format raw daemon # send everything to stdout
5694 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005695 log 127.0.0.1:514 local0 notice # only send important events
5696 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02005697 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01005698
Willy Tarreau2769aa02007-12-27 18:26:09 +01005699
William Lallemand48940402012-01-30 16:47:22 +01005700log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005701 Specifies the log format string to use for traffic logs
5702 May be used in sections: defaults | frontend | listen | backend
5703 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01005704
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005705 This directive specifies the log format string that will be used for all logs
5706 resulting from traffic passing through the frontend using this line. If the
5707 directive is used in a defaults section, all subsequent frontends will use
5708 the same log format. Please see section 8.2.4 which covers the log format
5709 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01005710
Guillaume de Lafond29f45602017-03-31 19:52:15 +02005711 "log-format" directive overrides previous "option tcplog", "log-format" and
5712 "option httplog" directives.
5713
Dragan Dosen7ad31542015-09-28 17:16:47 +02005714log-format-sd <string>
5715 Specifies the RFC5424 structured-data log format string
5716 May be used in sections: defaults | frontend | listen | backend
5717 yes | yes | yes | no
5718
5719 This directive specifies the RFC5424 structured-data log format string that
5720 will be used for all logs resulting from traffic passing through the frontend
5721 using this line. If the directive is used in a defaults section, all
5722 subsequent frontends will use the same log format. Please see section 8.2.4
5723 which covers the log format string in depth.
5724
5725 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
5726 about the RFC5424 structured-data part.
5727
5728 Note : This log format string will be used only for loggers that have set
5729 log format to "rfc5424".
5730
5731 Example :
5732 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
5733
5734
Willy Tarreau094af4e2015-01-07 15:03:42 +01005735log-tag <string>
5736 Specifies the log tag to use for all outgoing logs
5737 May be used in sections: defaults | frontend | listen | backend
5738 yes | yes | yes | yes
5739
5740 Sets the tag field in the syslog header to this string. It defaults to the
5741 log-tag set in the global section, otherwise the program name as launched
5742 from the command line, which usually is "haproxy". Sometimes it can be useful
5743 to differentiate between multiple processes running on the same host, or to
5744 differentiate customer instances running in the same process. In the backend,
5745 logs about servers up/down will use this tag. As a hint, it can be convenient
5746 to set a log-tag related to a hosted customer in a defaults section then put
5747 all the frontends and backends for that customer, then start another customer
5748 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005749
Willy Tarreauc35362a2014-04-25 13:58:37 +02005750max-keep-alive-queue <value>
5751 Set the maximum server queue size for maintaining keep-alive connections
5752 May be used in sections: defaults | frontend | listen | backend
5753 yes | no | yes | yes
5754
5755 HTTP keep-alive tries to reuse the same server connection whenever possible,
5756 but sometimes it can be counter-productive, for example if a server has a lot
5757 of connections while other ones are idle. This is especially true for static
5758 servers.
5759
5760 The purpose of this setting is to set a threshold on the number of queued
5761 connections at which haproxy stops trying to reuse the same server and prefers
5762 to find another one. The default value, -1, means there is no limit. A value
5763 of zero means that keep-alive requests will never be queued. For very close
5764 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01005765 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02005766 use a value of 10 or less). For remote servers suffering from a high latency,
5767 higher values might be needed to cover for the latency and/or the cost of
5768 picking a different server.
5769
5770 Note that this has no impact on responses which are maintained to the same
5771 server consecutively to a 401 response. They will still go to the same server
5772 even if they have to be queued.
5773
5774 See also : "option http-server-close", "option prefer-last-server", server
5775 "maxconn" and cookie persistence.
5776
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01005777max-session-srv-conns <nb>
5778 Set the maximum number of outgoing connections we can keep idling for a given
5779 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
5780 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02005781
Willy Tarreau2769aa02007-12-27 18:26:09 +01005782maxconn <conns>
5783 Fix the maximum number of concurrent connections on a frontend
5784 May be used in sections : defaults | frontend | listen | backend
5785 yes | yes | yes | no
5786 Arguments :
5787 <conns> is the maximum number of concurrent connections the frontend will
5788 accept to serve. Excess connections will be queued by the system
5789 in the socket's listen queue and will be served once a connection
5790 closes.
5791
5792 If the system supports it, it can be useful on big sites to raise this limit
5793 very high so that haproxy manages connection queues, instead of leaving the
5794 clients with unanswered connection attempts. This value should not exceed the
5795 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01005796 of tune.bufsize (16kB by default) each, as well as some other data resulting
5797 in about 33 kB of RAM being consumed per established connection. That means
5798 that a medium system equipped with 1GB of RAM can withstand around
5799 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005800
5801 Also, when <conns> is set to large values, it is possible that the servers
5802 are not sized to accept such loads, and for this reason it is generally wise
5803 to assign them some reasonable connection limits.
5804
Willy Tarreauc8d5b952019-02-27 17:25:52 +01005805 When this value is set to zero, which is the default, the global "maxconn"
5806 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02005807
Willy Tarreau2769aa02007-12-27 18:26:09 +01005808 See also : "server", global section's "maxconn", "fullconn"
5809
5810
5811mode { tcp|http|health }
5812 Set the running mode or protocol of the instance
5813 May be used in sections : defaults | frontend | listen | backend
5814 yes | yes | yes | yes
5815 Arguments :
5816 tcp The instance will work in pure TCP mode. A full-duplex connection
5817 will be established between clients and servers, and no layer 7
5818 examination will be performed. This is the default mode. It
5819 should be used for SSL, SSH, SMTP, ...
5820
5821 http The instance will work in HTTP mode. The client request will be
5822 analyzed in depth before connecting to any server. Any request
5823 which is not RFC-compliant will be rejected. Layer 7 filtering,
5824 processing and switching will be possible. This is the mode which
5825 brings HAProxy most of its value.
5826
5827 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02005828 to incoming connections and close the connection. Alternatively,
5829 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
5830 instead. Nothing will be logged in either case. This mode is used
5831 to reply to external components health checks. This mode is
5832 deprecated and should not be used anymore as it is possible to do
5833 the same and even better by combining TCP or HTTP modes with the
5834 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005835
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005836 When doing content switching, it is mandatory that the frontend and the
5837 backend are in the same mode (generally HTTP), otherwise the configuration
5838 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005839
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005840 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005841 defaults http_instances
5842 mode http
5843
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005844 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005845
Willy Tarreau0ba27502007-12-24 16:55:16 +01005846
Cyril Bontéf0c60612010-02-06 14:44:47 +01005847monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01005848 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005849 May be used in sections : defaults | frontend | listen | backend
5850 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01005851 Arguments :
5852 if <cond> the monitor request will fail if the condition is satisfied,
5853 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005854 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01005855 are met, for instance a low number of servers both in a
5856 backend and its backup.
5857
5858 unless <cond> the monitor request will succeed only if the condition is
5859 satisfied, and will fail otherwise. Such a condition may be
5860 based on a test on the presence of a minimum number of active
5861 servers in a list of backends.
5862
5863 This statement adds a condition which can force the response to a monitor
5864 request to report a failure. By default, when an external component queries
5865 the URI dedicated to monitoring, a 200 response is returned. When one of the
5866 conditions above is met, haproxy will return 503 instead of 200. This is
5867 very useful to report a site failure to an external component which may base
5868 routing advertisements between multiple sites on the availability reported by
5869 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005870 criterion. Note that "monitor fail" only works in HTTP mode. Both status
5871 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005872
5873 Example:
5874 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01005875 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01005876 acl site_dead nbsrv(dynamic) lt 2
5877 acl site_dead nbsrv(static) lt 2
5878 monitor-uri /site_alive
5879 monitor fail if site_dead
5880
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005881 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005882
5883
5884monitor-net <source>
5885 Declare a source network which is limited to monitor requests
5886 May be used in sections : defaults | frontend | listen | backend
5887 yes | yes | yes | no
5888 Arguments :
5889 <source> is the source IPv4 address or network which will only be able to
5890 get monitor responses to any request. It can be either an IPv4
5891 address, a host name, or an address followed by a slash ('/')
5892 followed by a mask.
5893
5894 In TCP mode, any connection coming from a source matching <source> will cause
5895 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005896 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01005897 forwarding the connection to a remote server.
5898
5899 In HTTP mode, a connection coming from a source matching <source> will be
5900 accepted, the following response will be sent without waiting for a request,
5901 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
5902 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02005903 running without forwarding the request to a backend server. Note that this
5904 response is sent in raw format, without any transformation. This is important
5905 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005906
Willy Tarreau82569f92012-09-27 23:48:56 +02005907 Monitor requests are processed very early, just after tcp-request connection
5908 ACLs which are the only ones able to block them. These connections are short
5909 lived and never wait for any data from the client. They cannot be logged, and
5910 it is the intended purpose. They are only used to report HAProxy's health to
5911 an upper component, nothing more. Please note that "monitor fail" rules do
5912 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005913
Willy Tarreau95cd2832010-03-04 23:36:33 +01005914 Last, please note that only one "monitor-net" statement can be specified in
5915 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005916
Willy Tarreau2769aa02007-12-27 18:26:09 +01005917 Example :
5918 # addresses .252 and .253 are just probing us.
5919 frontend www
5920 monitor-net 192.168.0.252/31
5921
5922 See also : "monitor fail", "monitor-uri"
5923
5924
5925monitor-uri <uri>
5926 Intercept a URI used by external components' monitor requests
5927 May be used in sections : defaults | frontend | listen | backend
5928 yes | yes | yes | no
5929 Arguments :
5930 <uri> is the exact URI which we want to intercept to return HAProxy's
5931 health status instead of forwarding the request.
5932
5933 When an HTTP request referencing <uri> will be received on a frontend,
5934 HAProxy will not forward it nor log it, but instead will return either
5935 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
5936 conditions defined with "monitor fail". This is normally enough for any
5937 front-end HTTP probe to detect that the service is UP and running without
5938 forwarding the request to a backend server. Note that the HTTP method, the
5939 version and all headers are ignored, but the request must at least be valid
5940 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
5941
Willy Tarreau721d8e02017-12-01 18:25:08 +01005942 Monitor requests are processed very early, just after the request is parsed
5943 and even before any "http-request" or "block" rulesets. The only rulesets
5944 applied before are the tcp-request ones. They cannot be logged either, and it
5945 is the intended purpose. They are only used to report HAProxy's health to an
5946 upper component, nothing more. However, it is possible to add any number of
5947 conditions using "monitor fail" and ACLs so that the result can be adjusted
5948 to whatever check can be imagined (most often the number of available servers
5949 in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005950
5951 Example :
5952 # Use /haproxy_test to report haproxy's status
5953 frontend www
5954 mode http
5955 monitor-uri /haproxy_test
5956
5957 See also : "monitor fail", "monitor-net"
5958
Willy Tarreau0ba27502007-12-24 16:55:16 +01005959
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005960option abortonclose
5961no option abortonclose
5962 Enable or disable early dropping of aborted requests pending in queues.
5963 May be used in sections : defaults | frontend | listen | backend
5964 yes | no | yes | yes
5965 Arguments : none
5966
5967 In presence of very high loads, the servers will take some time to respond.
5968 The per-instance connection queue will inflate, and the response time will
5969 increase respective to the size of the queue times the average per-session
5970 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01005971 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005972 the queue, and slowing down other users, and the servers as well, because the
5973 request will eventually be served, then aborted at the first error
5974 encountered while delivering the response.
5975
5976 As there is no way to distinguish between a full STOP and a simple output
5977 close on the client side, HTTP agents should be conservative and consider
5978 that the client might only have closed its output channel while waiting for
5979 the response. However, this introduces risks of congestion when lots of users
5980 do the same, and is completely useless nowadays because probably no client at
5981 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01005982 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005983 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01005984 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005985 of being the single component to break rare but valid traffic is extremely
5986 low, which adds to the temptation to be able to abort a session early while
5987 still not served and not pollute the servers.
5988
Davor Ocelice9ed2812017-12-25 17:49:28 +01005989 In HAProxy, the user can choose the desired behavior using the option
5990 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005991 compliant and aborted requests will be served. But when the option is
5992 specified, a session with an incoming channel closed will be aborted while
5993 it is still possible, either pending in the queue for a connection slot, or
5994 during the connection establishment if the server has not yet acknowledged
5995 the connection request. This considerably reduces the queue size and the load
5996 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005997 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005998
5999 If this option has been enabled in a "defaults" section, it can be disabled
6000 in a specific instance by prepending the "no" keyword before it.
6001
6002 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
6003
6004
Willy Tarreau4076a152009-04-02 15:18:36 +02006005option accept-invalid-http-request
6006no option accept-invalid-http-request
6007 Enable or disable relaxing of HTTP request parsing
6008 May be used in sections : defaults | frontend | listen | backend
6009 yes | yes | yes | no
6010 Arguments : none
6011
Willy Tarreau91852eb2015-05-01 13:26:00 +02006012 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02006013 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01006014 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02006015 forbidden characters are essentially used to build attacks exploiting server
6016 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
6017 server will emit invalid header names for whatever reason (configuration,
6018 implementation) and the issue will not be immediately fixed. In such a case,
6019 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01006020 even if that does not make sense, by specifying this option. Similarly, the
6021 list of characters allowed to appear in a URI is well defined by RFC3986, and
6022 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
6023 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01006024 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02006025 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02006026 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
6027 to pass through (no version specified) and multiple digits for both the major
6028 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02006029
6030 This option should never be enabled by default as it hides application bugs
6031 and open security breaches. It should only be deployed after a problem has
6032 been confirmed.
6033
6034 When this option is enabled, erroneous header names will still be accepted in
6035 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01006036 analysis using the "show errors" request on the UNIX stats socket. Similarly,
6037 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02006038 also helps confirming that the issue has been solved.
6039
6040 If this option has been enabled in a "defaults" section, it can be disabled
6041 in a specific instance by prepending the "no" keyword before it.
6042
6043 See also : "option accept-invalid-http-response" and "show errors" on the
6044 stats socket.
6045
6046
6047option accept-invalid-http-response
6048no option accept-invalid-http-response
6049 Enable or disable relaxing of HTTP response parsing
6050 May be used in sections : defaults | frontend | listen | backend
6051 yes | no | yes | yes
6052 Arguments : none
6053
Willy Tarreau91852eb2015-05-01 13:26:00 +02006054 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02006055 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01006056 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02006057 forbidden characters are essentially used to build attacks exploiting server
6058 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
6059 server will emit invalid header names for whatever reason (configuration,
6060 implementation) and the issue will not be immediately fixed. In such a case,
6061 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02006062 even if that does not make sense, by specifying this option. This option also
6063 relaxes the test on the HTTP version format, it allows multiple digits for
6064 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02006065
6066 This option should never be enabled by default as it hides application bugs
6067 and open security breaches. It should only be deployed after a problem has
6068 been confirmed.
6069
6070 When this option is enabled, erroneous header names will still be accepted in
6071 responses, but the complete response will be captured in order to permit
6072 later analysis using the "show errors" request on the UNIX stats socket.
6073 Doing this also helps confirming that the issue has been solved.
6074
6075 If this option has been enabled in a "defaults" section, it can be disabled
6076 in a specific instance by prepending the "no" keyword before it.
6077
6078 See also : "option accept-invalid-http-request" and "show errors" on the
6079 stats socket.
6080
6081
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006082option allbackups
6083no option allbackups
6084 Use either all backup servers at a time or only the first one
6085 May be used in sections : defaults | frontend | listen | backend
6086 yes | no | yes | yes
6087 Arguments : none
6088
6089 By default, the first operational backup server gets all traffic when normal
6090 servers are all down. Sometimes, it may be preferred to use multiple backups
6091 at once, because one will not be enough. When "option allbackups" is enabled,
6092 the load balancing will be performed among all backup servers when all normal
6093 ones are unavailable. The same load balancing algorithm will be used and the
6094 servers' weights will be respected. Thus, there will not be any priority
6095 order between the backup servers anymore.
6096
6097 This option is mostly used with static server farms dedicated to return a
6098 "sorry" page when an application is completely offline.
6099
6100 If this option has been enabled in a "defaults" section, it can be disabled
6101 in a specific instance by prepending the "no" keyword before it.
6102
6103
6104option checkcache
6105no option checkcache
Godbach7056a352013-12-11 20:01:07 +08006106 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006107 May be used in sections : defaults | frontend | listen | backend
6108 yes | no | yes | yes
6109 Arguments : none
6110
6111 Some high-level frameworks set application cookies everywhere and do not
6112 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006113 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006114 high risk of session crossing or stealing between users traversing the same
6115 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006116 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006117
6118 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006119 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01006120 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006121 response to check if there's a risk of caching a cookie on a client-side
6122 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01006123 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01006124 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01006125 - all those with a return code other than 200, 203, 204, 206, 300, 301,
6126 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01006127 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01006128 - all those that result from a request using a method other than GET, HEAD,
6129 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01006130 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006131 - those with a 'Pragma: no-cache' header
6132 - those with a 'Cache-control: private' header
6133 - those with a 'Cache-control: no-store' header
6134 - those with a 'Cache-control: max-age=0' header
6135 - those with a 'Cache-control: s-maxage=0' header
6136 - those with a 'Cache-control: no-cache' header
6137 - those with a 'Cache-control: no-cache="set-cookie"' header
6138 - those with a 'Cache-control: no-cache="set-cookie,' header
6139 (allowing other fields after set-cookie)
6140
6141 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01006142 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006143 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006144 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006145 that admins are informed that there's something to be fixed.
6146
6147 Due to the high impact on the application, the application should be tested
6148 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006149 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01006150 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006151
6152 If this option has been enabled in a "defaults" section, it can be disabled
6153 in a specific instance by prepending the "no" keyword before it.
6154
6155
6156option clitcpka
6157no option clitcpka
6158 Enable or disable the sending of TCP keepalive packets on the client side
6159 May be used in sections : defaults | frontend | listen | backend
6160 yes | yes | yes | no
6161 Arguments : none
6162
6163 When there is a firewall or any session-aware component between a client and
6164 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006165 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006166 components decides to expire a session which has remained idle for too long.
6167
6168 Enabling socket-level TCP keep-alives makes the system regularly send packets
6169 to the other end of the connection, leaving it active. The delay between
6170 keep-alive probes is controlled by the system only and depends both on the
6171 operating system and its tuning parameters.
6172
6173 It is important to understand that keep-alive packets are neither emitted nor
6174 received at the application level. It is only the network stacks which sees
6175 them. For this reason, even if one side of the proxy already uses keep-alives
6176 to maintain its connection alive, those keep-alive packets will not be
6177 forwarded to the other side of the proxy.
6178
6179 Please note that this has nothing to do with HTTP keep-alive.
6180
6181 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
6182 client side of a connection, which should help when session expirations are
6183 noticed between HAProxy and a client.
6184
6185 If this option has been enabled in a "defaults" section, it can be disabled
6186 in a specific instance by prepending the "no" keyword before it.
6187
6188 See also : "option srvtcpka", "option tcpka"
6189
6190
Willy Tarreau0ba27502007-12-24 16:55:16 +01006191option contstats
6192 Enable continuous traffic statistics updates
6193 May be used in sections : defaults | frontend | listen | backend
6194 yes | yes | yes | no
6195 Arguments : none
6196
6197 By default, counters used for statistics calculation are incremented
6198 only when a session finishes. It works quite well when serving small
6199 objects, but with big ones (for example large images or archives) or
6200 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01006201 a hedgehog. With this option enabled counters get incremented frequently
6202 along the session, typically every 5 seconds, which is often enough to
6203 produce clean graphs. Recounting touches a hotpath directly so it is not
6204 not enabled by default, as it can cause a lot of wakeups for very large
6205 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006206
6207
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006208option dontlog-normal
6209no option dontlog-normal
6210 Enable or disable logging of normal, successful connections
6211 May be used in sections : defaults | frontend | listen | backend
6212 yes | yes | yes | no
6213 Arguments : none
6214
6215 There are large sites dealing with several thousand connections per second
6216 and for which logging is a major pain. Some of them are even forced to turn
6217 logs off and cannot debug production issues. Setting this option ensures that
6218 normal connections, those which experience no error, no timeout, no retry nor
6219 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
6220 mode, the response status code is checked and return codes 5xx will still be
6221 logged.
6222
6223 It is strongly discouraged to use this option as most of the time, the key to
6224 complex issues is in the normal logs which will not be logged here. If you
6225 need to separate logs, see the "log-separate-errors" option instead.
6226
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006227 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006228 logging.
6229
6230
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006231option dontlognull
6232no option dontlognull
6233 Enable or disable logging of null connections
6234 May be used in sections : defaults | frontend | listen | backend
6235 yes | yes | yes | no
6236 Arguments : none
6237
6238 In certain environments, there are components which will regularly connect to
6239 various systems to ensure that they are still alive. It can be the case from
6240 another load balancer as well as from monitoring systems. By default, even a
6241 simple port probe or scan will produce a log. If those connections pollute
6242 the logs too much, it is possible to enable option "dontlognull" to indicate
6243 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02006244 which typically corresponds to those probes. Note that errors will still be
6245 returned to the client and accounted for in the stats. If this is not what is
6246 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006247
6248 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01006249 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006250 would not be logged.
6251
6252 If this option has been enabled in a "defaults" section, it can be disabled
6253 in a specific instance by prepending the "no" keyword before it.
6254
Willy Tarreau0f228a02015-05-01 15:37:53 +02006255 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
6256 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006257
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006258
Willy Tarreau87cf5142011-08-19 22:57:24 +02006259option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006260 Enable insertion of the X-Forwarded-For header to requests sent to servers
6261 May be used in sections : defaults | frontend | listen | backend
6262 yes | yes | yes | yes
6263 Arguments :
6264 <network> is an optional argument used to disable this option for sources
6265 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02006266 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01006267 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006268
6269 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
6270 their client address. This is sometimes annoying when the client's IP address
6271 is expected in server logs. To solve this problem, the well-known HTTP header
6272 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
6273 This header contains a value representing the client's IP address. Since this
6274 header is always appended at the end of the existing header list, the server
6275 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02006276 the server's manual to find how to enable use of this standard header. Note
6277 that only the last occurrence of the header must be used, since it is really
6278 possible that the client has already brought one.
6279
Willy Tarreaud72758d2010-01-12 10:42:19 +01006280 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02006281 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01006282 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01006283 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006284 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02006285 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01006286
6287 Sometimes, a same HAProxy instance may be shared between a direct client
6288 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6289 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6290 header for a known source address or network by adding the "except" keyword
6291 followed by the network address. In this case, any source IP matching the
6292 network will not cause an addition of this header. Most common uses are with
6293 private networks or 127.0.0.1.
6294
Willy Tarreau87cf5142011-08-19 22:57:24 +02006295 Alternatively, the keyword "if-none" states that the header will only be
6296 added if it is not present. This should only be used in perfectly trusted
6297 environment, as this might cause a security issue if headers reaching haproxy
6298 are under the control of the end-user.
6299
Willy Tarreauc27debf2008-01-06 08:57:02 +01006300 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02006301 least one of them uses it, the header will be added. Note that the backend's
6302 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02006303 both are defined. In the case of the "if-none" argument, if at least one of
6304 the frontend or the backend does not specify it, it wants the addition to be
6305 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006306
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02006307 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01006308 # Public HTTP address also used by stunnel on the same machine
6309 frontend www
6310 mode http
6311 option forwardfor except 127.0.0.1 # stunnel already adds the header
6312
Ross Westaf72a1d2008-08-03 10:51:45 +02006313 # Those servers want the IP Address in X-Client
6314 backend www
6315 mode http
6316 option forwardfor header X-Client
6317
Willy Tarreau87cf5142011-08-19 22:57:24 +02006318 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006319 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01006320
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006321
Christopher Fauleta99ff4d2019-07-22 16:18:24 +02006322option h1-case-adjust-bogus-client
6323no option h1-case-adjust-bogus-client
6324 Enable or disable the case adjustment of HTTP/1 headers sent to bogus clients
6325 May be used in sections : defaults | frontend | listen | backend
6326 yes | yes | yes | no
6327 Arguments : none
6328
6329 There is no standard case for header names because, as stated in RFC7230,
6330 they are case-insensitive. So applications must handle them in a case-
6331 insensitive manner. But some bogus applications violate the standards and
6332 erroneously rely on the cases most commonly used by browsers. This problem
6333 becomes critical with HTTP/2 because all header names must be exchanged in
6334 lower case, and HAProxy follows the same convention. All header names are
6335 sent in lower case to clients and servers, regardless of the HTTP version.
6336
6337 When HAProxy receives an HTTP/1 response, its header names are converted to
6338 lower case and manipulated and sent this way to the clients. If a client is
6339 known to violate the HTTP standards and to fail to process a response coming
6340 from HAProxy, it is possible to transform the lower case header names to a
6341 different format when the response is formatted and sent to the client, by
6342 enabling this option and specifying the list of headers to be reformatted
6343 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6344 must only be a temporary workaround for the time it takes the client to be
6345 fixed, because clients which require such workarounds might be vulnerable to
6346 content smuggling attacks and must absolutely be fixed.
6347
6348 Please note that this option will not affect standards-compliant clients.
6349
6350 If this option has been enabled in a "defaults" section, it can be disabled
6351 in a specific instance by prepending the "no" keyword before it.
6352
6353 See also: "option h1-case-adjust-bogus-server", "h1-case-adjust",
6354 "h1-case-adjust-file".
6355
6356
6357option h1-case-adjust-bogus-server
6358no option h1-case-adjust-bogus-server
6359 Enable or disable the case adjustment of HTTP/1 headers sent to bogus servers
6360 May be used in sections : defaults | frontend | listen | backend
6361 yes | no | yes | yes
6362 Arguments : none
6363
6364 There is no standard case for header names because, as stated in RFC7230,
6365 they are case-insensitive. So applications must handle them in a case-
6366 insensitive manner. But some bogus applications violate the standards and
6367 erroneously rely on the cases most commonly used by browsers. This problem
6368 becomes critical with HTTP/2 because all header names must be exchanged in
6369 lower case, and HAProxy follows the same convention. All header names are
6370 sent in lower case to clients and servers, regardless of the HTTP version.
6371
6372 When HAProxy receives an HTTP/1 request, its header names are converted to
6373 lower case and manipulated and sent this way to the servers. If a server is
6374 known to violate the HTTP standards and to fail to process a request coming
6375 from HAProxy, it is possible to transform the lower case header names to a
6376 different format when the request is formatted and sent to the server, by
6377 enabling this option and specifying the list of headers to be reformatted
6378 using the global directives "h1-case-adjust" or "h1-case-adjust-file". This
6379 must only be a temporary workaround for the time it takes the server to be
6380 fixed, because servers which require such workarounds might be vulnerable to
6381 content smuggling attacks and must absolutely be fixed.
6382
6383 Please note that this option will not affect standards-compliant servers.
6384
6385 If this option has been enabled in a "defaults" section, it can be disabled
6386 in a specific instance by prepending the "no" keyword before it.
6387
6388 See also: "option h1-case-adjust-bogus-client", "h1-case-adjust",
6389 "h1-case-adjust-file".
6390
6391
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006392option http-buffer-request
6393no option http-buffer-request
6394 Enable or disable waiting for whole HTTP request body before proceeding
6395 May be used in sections : defaults | frontend | listen | backend
6396 yes | yes | yes | yes
6397 Arguments : none
6398
6399 It is sometimes desirable to wait for the body of an HTTP request before
6400 taking a decision. This is what is being done by "balance url_param" for
6401 example. The first use case is to buffer requests from slow clients before
6402 connecting to the server. Another use case consists in taking the routing
6403 decision based on the request body's contents. This option placed in a
6404 frontend or backend forces the HTTP processing to wait until either the whole
6405 body is received, or the request buffer is full, or the first chunk is
6406 complete in case of chunked encoding. It can have undesired side effects with
Tim Düsterhus4896c442016-11-29 02:15:19 +01006407 some applications abusing HTTP by expecting unbuffered transmissions between
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006408 the frontend and the backend, so this should definitely not be used by
6409 default.
6410
Baptiste Assmanneccdf432015-10-28 13:49:01 +01006411 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006412
6413
Willy Tarreau0f228a02015-05-01 15:37:53 +02006414option http-ignore-probes
6415no option http-ignore-probes
6416 Enable or disable logging of null connections and request timeouts
6417 May be used in sections : defaults | frontend | listen | backend
6418 yes | yes | yes | no
6419 Arguments : none
6420
6421 Recently some browsers started to implement a "pre-connect" feature
6422 consisting in speculatively connecting to some recently visited web sites
6423 just in case the user would like to visit them. This results in many
6424 connections being established to web sites, which end up in 408 Request
6425 Timeout if the timeout strikes first, or 400 Bad Request when the browser
6426 decides to close them first. These ones pollute the log and feed the error
6427 counters. There was already "option dontlognull" but it's insufficient in
6428 this case. Instead, this option does the following things :
6429 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01006430 was received over a connection before it was closed;
6431 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02006432 - prevent any error counter from being incremented
6433
6434 That way the empty connection is silently ignored. Note that it is better
6435 not to use this unless it is clear that it is needed, because it will hide
6436 real problems. The most common reason for not receiving a request and seeing
6437 a 408 is due to an MTU inconsistency between the client and an intermediary
6438 element such as a VPN, which blocks too large packets. These issues are
6439 generally seen with POST requests as well as GET with large cookies. The logs
6440 are often the only way to detect them.
6441
6442 If this option has been enabled in a "defaults" section, it can be disabled
6443 in a specific instance by prepending the "no" keyword before it.
6444
6445 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
6446
6447
Willy Tarreau16bfb022010-01-16 19:48:41 +01006448option http-keep-alive
6449no option http-keep-alive
6450 Enable or disable HTTP keep-alive from client to server
6451 May be used in sections : defaults | frontend | listen | backend
6452 yes | yes | yes | yes
6453 Arguments : none
6454
Willy Tarreau70dffda2014-01-30 03:07:23 +01006455 By default HAProxy operates in keep-alive mode with regards to persistent
6456 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02006457 leaves the connection idle on both sides between the end of a response and
6458 the start of a new request. This mode may be changed by several options such
6459 as "option http-server-close", "option httpclose" or "option http-tunnel".
6460 This option allows to set back the keep-alive mode, which can be useful when
6461 another mode was used in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01006462
6463 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
6464 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01006465 network) and the fastest session reuse on the server side at the expense
6466 of maintaining idle connections to the servers. In general, it is possible
6467 with this option to achieve approximately twice the request rate that the
6468 "http-server-close" option achieves on small objects. There are mainly two
6469 situations where this option may be useful :
6470
6471 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01006472 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01006473
6474 - when the cost of establishing the connection to the server is significant
6475 compared to the cost of retrieving the associated object from the server.
6476
6477 This last case can happen when the server is a fast static server of cache.
6478 In this case, the server will need to be properly tuned to support high enough
6479 connection counts because connections will last until the client sends another
6480 request.
6481
6482 If the client request has to go to another backend or another server due to
6483 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01006484 immediately be closed and a new one re-opened. Option "prefer-last-server" is
6485 available to try optimize server selection so that if the server currently
6486 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01006487
Willy Tarreau16bfb022010-01-16 19:48:41 +01006488 At the moment, logs will not indicate whether requests came from the same
6489 session or not. The accept date reported in the logs corresponds to the end
6490 of the previous request, and the request time corresponds to the time spent
6491 waiting for a new request. The keep-alive request time is still bound to the
6492 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6493 not set.
6494
Cyril Bonté653dcd62014-02-20 00:13:15 +01006495 This option disables and replaces any previous "option httpclose", "option
Christopher Faulet315b39c2018-09-21 16:26:19 +02006496 http-server-close" or "option http-tunnel". When backend and frontend options
6497 differ, all of these 4 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006498
Christopher Faulet315b39c2018-09-21 16:26:19 +02006499 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01006500 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01006501 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006502
6503
Willy Tarreau96e31212011-05-30 18:10:30 +02006504option http-no-delay
6505no option http-no-delay
6506 Instruct the system to favor low interactive delays over performance in HTTP
6507 May be used in sections : defaults | frontend | listen | backend
6508 yes | yes | yes | yes
6509 Arguments : none
6510
6511 In HTTP, each payload is unidirectional and has no notion of interactivity.
6512 Any agent is expected to queue data somewhat for a reasonably low delay.
6513 There are some very rare server-to-server applications that abuse the HTTP
6514 protocol and expect the payload phase to be highly interactive, with many
6515 interleaved data chunks in both directions within a single request. This is
6516 absolutely not supported by the HTTP specification and will not work across
6517 most proxies or servers. When such applications attempt to do this through
6518 haproxy, it works but they will experience high delays due to the network
6519 optimizations which favor performance by instructing the system to wait for
6520 enough data to be available in order to only send full packets. Typical
6521 delays are around 200 ms per round trip. Note that this only happens with
6522 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
6523 affected.
6524
6525 When "option http-no-delay" is present in either the frontend or the backend
6526 used by a connection, all such optimizations will be disabled in order to
6527 make the exchanges as fast as possible. Of course this offers no guarantee on
6528 the functionality, as it may break at any other place. But if it works via
6529 HAProxy, it will work as fast as possible. This option should never be used
6530 by default, and should never be used at all unless such a buggy application
6531 is discovered. The impact of using this option is an increase of bandwidth
6532 usage and CPU usage, which may significantly lower performance in high
6533 latency environments.
6534
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006535 See also : "option http-buffer-request"
6536
Willy Tarreau96e31212011-05-30 18:10:30 +02006537
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006538option http-pretend-keepalive
6539no option http-pretend-keepalive
6540 Define whether haproxy will announce keepalive to the server or not
6541 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02006542 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006543 Arguments : none
6544
Christopher Faulet315b39c2018-09-21 16:26:19 +02006545 When running with "option http-server-close" or "option httpclose", haproxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006546 adds a "Connection: close" header to the request forwarded to the server.
6547 Unfortunately, when some servers see this header, they automatically refrain
6548 from using the chunked encoding for responses of unknown length, while this
6549 is totally unrelated. The immediate effect is that this prevents haproxy from
6550 maintaining the client connection alive. A second effect is that a client or
6551 a cache could receive an incomplete response without being aware of it, and
6552 consider the response complete.
6553
6554 By setting "option http-pretend-keepalive", haproxy will make the server
6555 believe it will keep the connection alive. The server will then not fall back
6556 to the abnormal undesired above. When haproxy gets the whole response, it
6557 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02006558 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006559 connection is correctly closed on the server side.
6560
6561 It is recommended not to enable this option by default, because most servers
6562 will more efficiently close the connection themselves after the last packet,
6563 and release its buffers slightly earlier. Also, the added packet on the
6564 network could slightly reduce the overall peak performance. However it is
6565 worth noting that when this option is enabled, haproxy will have slightly
6566 less work to do. So if haproxy is the bottleneck on the whole architecture,
6567 enabling this option might save a few CPU cycles.
6568
Christopher Faulet98db9762018-09-21 10:25:19 +02006569 This option may be set in backend and listen sections. Using it in a frontend
6570 section will be ignored and a warning will be reported during startup. It is
6571 a backend related option, so there is no real reason to set it on a
6572 frontend. This option may be combined with "option httpclose", which will
6573 cause keepalive to be announced to the server and close to be announced to
6574 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006575
6576 If this option has been enabled in a "defaults" section, it can be disabled
6577 in a specific instance by prepending the "no" keyword before it.
6578
Christopher Faulet315b39c2018-09-21 16:26:19 +02006579 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01006580 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006581
Willy Tarreauc27debf2008-01-06 08:57:02 +01006582
Willy Tarreaub608feb2010-01-02 22:47:18 +01006583option http-server-close
6584no option http-server-close
6585 Enable or disable HTTP connection closing on the server side
6586 May be used in sections : defaults | frontend | listen | backend
6587 yes | yes | yes | yes
6588 Arguments : none
6589
Willy Tarreau70dffda2014-01-30 03:07:23 +01006590 By default HAProxy operates in keep-alive mode with regards to persistent
6591 connections: for each connection it processes each request and response, and
6592 leaves the connection idle on both sides between the end of a response and
6593 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006594 as "option http-server-close", "option httpclose" or "option http-tunnel".
6595 Setting "option http-server-close" enables HTTP connection-close mode on the
6596 server side while keeping the ability to support HTTP keep-alive and
6597 pipelining on the client side. This provides the lowest latency on the client
6598 side (slow network) and the fastest session reuse on the server side to save
6599 server resources, similarly to "option httpclose". It also permits
6600 non-keepalive capable servers to be served in keep-alive mode to the clients
6601 if they conform to the requirements of RFC7230. Please note that some servers
6602 do not always conform to those requirements when they see "Connection: close"
6603 in the request. The effect will be that keep-alive will never be used. A
6604 workaround consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006605
6606 At the moment, logs will not indicate whether requests came from the same
6607 session or not. The accept date reported in the logs corresponds to the end
6608 of the previous request, and the request time corresponds to the time spent
6609 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01006610 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6611 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006612
6613 This option may be set both in a frontend and in a backend. It is enabled if
6614 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet315b39c2018-09-21 16:26:19 +02006615 It disables and replaces any previous "option httpclose", "option http-tunnel"
6616 or "option http-keep-alive". Please check section 4 ("Proxies") to see how
6617 this option combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006618
6619 If this option has been enabled in a "defaults" section, it can be disabled
6620 in a specific instance by prepending the "no" keyword before it.
6621
Christopher Faulet315b39c2018-09-21 16:26:19 +02006622 See also : "option httpclose", "option http-pretend-keepalive",
6623 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006624
6625
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006626option http-tunnel (deprecated)
6627no option http-tunnel (deprecated)
6628 Disable or enable HTTP connection processing after first transaction.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006629 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet4212a302018-09-21 10:42:19 +02006630 yes | yes | yes | no
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006631 Arguments : none
6632
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006633 Warning : Because it cannot work in HTTP/2, this option is deprecated and it
6634 is only supported on legacy HTTP frontends. In HTX, it is ignored and a
6635 warning is emitted during HAProxy startup.
6636
Willy Tarreau70dffda2014-01-30 03:07:23 +01006637 By default HAProxy operates in keep-alive mode with regards to persistent
6638 connections: for each connection it processes each request and response, and
6639 leaves the connection idle on both sides between the end of a response and
6640 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006641 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006642
6643 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006644 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01006645 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
6646 is normally not needed anymore unless in very specific cases such as when
6647 using an in-house protocol that looks like HTTP but is not compatible, or
6648 just to log one request per client in order to reduce log size. Note that
6649 everything which works at the HTTP level, including header parsing/addition,
6650 cookie processing or content switching will only work for the first request
6651 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006652
Christopher Faulet4212a302018-09-21 10:42:19 +02006653 This option may be set on frontend and listen sections. Using it on a backend
6654 section will be ignored and a warning will be reported during the startup. It
6655 is a frontend related option, so there is no real reason to set it on a
6656 backend.
6657
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006658 If this option has been enabled in a "defaults" section, it can be disabled
6659 in a specific instance by prepending the "no" keyword before it.
6660
Christopher Faulet315b39c2018-09-21 16:26:19 +02006661 See also : "option httpclose", "option http-server-close",
6662 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006663
6664
Willy Tarreau88d349d2010-01-25 12:15:43 +01006665option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01006666no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01006667 Make use of non-standard Proxy-Connection header instead of Connection
6668 May be used in sections : defaults | frontend | listen | backend
6669 yes | yes | yes | no
6670 Arguments : none
6671
Lukas Tribus23953682017-04-28 13:24:30 +00006672 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01006673 Connection header to indicate their wish of persistent or non-persistent
6674 connections, both browsers and proxies ignore this header for proxied
6675 connections and make use of the undocumented, non-standard Proxy-Connection
6676 header instead. The issue begins when trying to put a load balancer between
6677 browsers and such proxies, because there will be a difference between what
6678 haproxy understands and what the client and the proxy agree on.
6679
6680 By setting this option in a frontend, haproxy can automatically switch to use
6681 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006682 defined here as one where the URI begins with neither a '/' nor a '*'. This
6683 is incompatible with the HTTP tunnel mode. Note that this option can only be
6684 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01006685
Willy Tarreau844a7e72010-01-31 21:46:18 +01006686 Also, when this option is set, a request which requires authentication will
6687 automatically switch to use proxy authentication headers if it is itself a
6688 proxied request. That makes it possible to check or enforce authentication in
6689 front of an existing proxy.
6690
Willy Tarreau88d349d2010-01-25 12:15:43 +01006691 This option should normally never be used, except in front of a proxy.
6692
Christopher Faulet315b39c2018-09-21 16:26:19 +02006693 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01006694
6695
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006696option http-use-htx
6697no option http-use-htx
6698 Switch to the new HTX internal representation for HTTP protocol elements
6699 May be used in sections : defaults | frontend | listen | backend
6700 yes | yes | yes | yes
6701 Arguments : none
6702
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006703 Historically, the HTTP protocol is processed as-is. Inserting, deleting, or
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006704 modifying a header field requires to rewrite the affected part in the buffer
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006705 and to move the buffer's tail accordingly. This mode is known as the legacy
6706 HTTP mode. Since this principle has deep roots in haproxy, the HTTP/2
6707 protocol is converted to HTTP/1.1 before being processed this way. It also
6708 results in the inability to establish HTTP/2 connections to servers because
6709 of the loss of HTTP/2 semantics in the HTTP/1 representation.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006710
6711 HTX is the name of a totally new native internal representation for the HTTP
6712 protocol, that is agnostic to the version and aims at preserving semantics
6713 all along the chain. It relies on a fast parsing, tokenizing and indexing of
6714 the protocol elements so that no more memory moves are necessary and that
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006715 most elements are directly accessed. It supports using either HTTP/1 or
6716 HTTP/2 on any side regardless of the other side's version. It also supports
6717 upgrades from TCP to HTTP and implicit ones from HTTP/1 to HTTP/2 (matching
6718 the HTTP/2 preface).
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006719
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006720 This option indicates that HTX needs to be used. Since the version 2.0-dev3,
6721 the HTX is the default mode. To switch back on the legacy HTTP mode, the
6722 option must be explicitly disabled using the "no" prefix. For prior versions,
6723 the feature has incomplete functional coverage, so it is not enabled by
6724 default.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006725
6726 See also : "mode http"
6727
6728
Willy Tarreaud63335a2010-02-26 12:56:52 +01006729option httpchk
6730option httpchk <uri>
6731option httpchk <method> <uri>
6732option httpchk <method> <uri> <version>
6733 Enable HTTP protocol to check on the servers health
6734 May be used in sections : defaults | frontend | listen | backend
6735 yes | no | yes | yes
6736 Arguments :
6737 <method> is the optional HTTP method used with the requests. When not set,
6738 the "OPTIONS" method is used, as it generally requires low server
6739 processing and is easy to filter out from the logs. Any method
6740 may be used, though it is not recommended to invent non-standard
6741 ones.
6742
6743 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
6744 which is accessible by default on almost any server, but may be
6745 changed to any other URI. Query strings are permitted.
6746
6747 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
6748 but some servers might behave incorrectly in HTTP 1.0, so turning
6749 it to HTTP/1.1 may sometimes help. Note that the Host field is
6750 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
6751 after "\r\n" following the version string.
6752
6753 By default, server health checks only consist in trying to establish a TCP
6754 connection. When "option httpchk" is specified, a complete HTTP request is
6755 sent once the TCP connection is established, and responses 2xx and 3xx are
6756 considered valid, while all other ones indicate a server failure, including
6757 the lack of any response.
6758
6759 The port and interval are specified in the server configuration.
6760
6761 This option does not necessarily require an HTTP backend, it also works with
6762 plain TCP backends. This is particularly useful to check simple scripts bound
6763 to some dedicated ports using the inetd daemon.
6764
6765 Examples :
6766 # Relay HTTPS traffic to Apache instance and check service availability
6767 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
6768 backend https_relay
6769 mode tcp
6770 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
6771 server apache1 192.168.1.1:443 check port 80
6772
Simon Hormanafc47ee2013-11-25 10:46:35 +09006773 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
6774 "option pgsql-check", "http-check" and the "check", "port" and
6775 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006776
6777
Willy Tarreauc27debf2008-01-06 08:57:02 +01006778option httpclose
6779no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02006780 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01006781 May be used in sections : defaults | frontend | listen | backend
6782 yes | yes | yes | yes
6783 Arguments : none
6784
Willy Tarreau70dffda2014-01-30 03:07:23 +01006785 By default HAProxy operates in keep-alive mode with regards to persistent
6786 connections: for each connection it processes each request and response, and
6787 leaves the connection idle on both sides between the end of a response and
6788 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006789 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006790
Christopher Faulet315b39c2018-09-21 16:26:19 +02006791 If "option httpclose" is set, HAProxy will close connections with the server
6792 and the client as soon as the request and the response are received. It will
John Roesler27c754c2019-07-10 15:45:51 -05006793 also check if a "Connection: close" header is already set in each direction,
Christopher Faulet315b39c2018-09-21 16:26:19 +02006794 and will add one if missing. Any "Connection" header different from "close"
6795 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006796
Christopher Faulet315b39c2018-09-21 16:26:19 +02006797 This option may also be combined with "option http-pretend-keepalive", which
6798 will disable sending of the "Connection: close" header, but will still cause
6799 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006800
6801 This option may be set both in a frontend and in a backend. It is enabled if
6802 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01006803 It disables and replaces any previous "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006804 "option http-keep-alive" or "option http-tunnel". Please check section 4
6805 ("Proxies") to see how this option combines with others when frontend and
6806 backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006807
6808 If this option has been enabled in a "defaults" section, it can be disabled
6809 in a specific instance by prepending the "no" keyword before it.
6810
Christopher Faulet315b39c2018-09-21 16:26:19 +02006811 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01006812
6813
Emeric Brun3a058f32009-06-30 18:26:00 +02006814option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006815 Enable logging of HTTP request, session state and timers
6816 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006817 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02006818 Arguments :
6819 clf if the "clf" argument is added, then the output format will be
6820 the CLF format instead of HAProxy's default HTTP format. You can
6821 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01006822 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02006823 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006824
6825 By default, the log output format is very poor, as it only contains the
6826 source and destination addresses, and the instance name. By specifying
6827 "option httplog", each log line turns into a much richer format including,
6828 but not limited to, the HTTP request, the connection timers, the session
6829 status, the connections numbers, the captured headers and cookies, the
6830 frontend, backend and server name, and of course the source address and
6831 ports.
6832
PiBa-NLbd556bf2014-12-11 21:31:54 +01006833 Specifying only "option httplog" will automatically clear the 'clf' mode
6834 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02006835
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006836 "option httplog" overrides any previous "log-format" directive.
6837
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006838 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006839
Willy Tarreau55165fe2009-05-10 12:02:55 +02006840
6841option http_proxy
6842no option http_proxy
6843 Enable or disable plain HTTP proxy mode
6844 May be used in sections : defaults | frontend | listen | backend
6845 yes | yes | yes | yes
6846 Arguments : none
6847
6848 It sometimes happens that people need a pure HTTP proxy which understands
6849 basic proxy requests without caching nor any fancy feature. In this case,
6850 it may be worth setting up an HAProxy instance with the "option http_proxy"
6851 set. In this mode, no server is declared, and the connection is forwarded to
6852 the IP address and port found in the URL after the "http://" scheme.
6853
6854 No host address resolution is performed, so this only works when pure IP
6855 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006856 it will probably be used only by experts who know they need exactly it. This
6857 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02006858
6859 If this option has been enabled in a "defaults" section, it can be disabled
6860 in a specific instance by prepending the "no" keyword before it.
6861
6862 Example :
6863 # this backend understands HTTP proxy requests and forwards them directly.
6864 backend direct_forward
6865 option httpclose
6866 option http_proxy
6867
6868 See also : "option httpclose"
6869
Willy Tarreau211ad242009-10-03 21:45:07 +02006870
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006871option independent-streams
6872no option independent-streams
6873 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006874 May be used in sections : defaults | frontend | listen | backend
6875 yes | yes | yes | yes
6876 Arguments : none
6877
6878 By default, when data is sent over a socket, both the write timeout and the
6879 read timeout for that socket are refreshed, because we consider that there is
6880 activity on that socket, and we have no other means of guessing if we should
6881 receive data or not.
6882
Davor Ocelice9ed2812017-12-25 17:49:28 +01006883 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006884 exists a situation where it is desirable to disable it, and only refresh the
6885 read timeout if there are incoming data. This happens on sessions with large
6886 timeouts and low amounts of exchanged data such as telnet session. If the
6887 server suddenly disappears, the output data accumulates in the system's
6888 socket buffers, both timeouts are correctly refreshed, and there is no way
6889 to know the server does not receive them, so we don't timeout. However, when
6890 the underlying protocol always echoes sent data, it would be enough by itself
6891 to detect the issue using the read timeout. Note that this problem does not
6892 happen with more verbose protocols because data won't accumulate long in the
6893 socket buffers.
6894
6895 When this option is set on the frontend, it will disable read timeout updates
6896 on data sent to the client. There probably is little use of this case. When
6897 the option is set on the backend, it will disable read timeout updates on
6898 data sent to the server. Doing so will typically break large HTTP posts from
6899 slow lines, so use it with caution.
6900
Lukas Tribus745f15e2018-11-08 12:41:42 +01006901 Note: older versions used to call this setting "option independant-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006902 with a spelling mistake. This spelling is still supported but
6903 deprecated.
6904
Willy Tarreauce887fd2012-05-12 12:50:00 +02006905 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006906
6907
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02006908option ldap-check
6909 Use LDAPv3 health checks for server testing
6910 May be used in sections : defaults | frontend | listen | backend
6911 yes | no | yes | yes
6912 Arguments : none
6913
6914 It is possible to test that the server correctly talks LDAPv3 instead of just
6915 testing that it accepts the TCP connection. When this option is set, an
6916 LDAPv3 anonymous simple bind message is sent to the server, and the response
6917 is analyzed to find an LDAPv3 bind response message.
6918
6919 The server is considered valid only when the LDAP response contains success
6920 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
6921
6922 Logging of bind requests is server dependent see your documentation how to
6923 configure it.
6924
6925 Example :
6926 option ldap-check
6927
6928 See also : "option httpchk"
6929
6930
Simon Horman98637e52014-06-20 12:30:16 +09006931option external-check
6932 Use external processes for server health checks
6933 May be used in sections : defaults | frontend | listen | backend
6934 yes | no | yes | yes
6935
6936 It is possible to test the health of a server using an external command.
6937 This is achieved by running the executable set using "external-check
6938 command".
6939
6940 Requires the "external-check" global to be set.
6941
6942 See also : "external-check", "external-check command", "external-check path"
6943
6944
Willy Tarreau211ad242009-10-03 21:45:07 +02006945option log-health-checks
6946no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02006947 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02006948 May be used in sections : defaults | frontend | listen | backend
6949 yes | no | yes | yes
6950 Arguments : none
6951
Willy Tarreaubef1b322014-05-13 21:01:39 +02006952 By default, failed health check are logged if server is UP and successful
6953 health checks are logged if server is DOWN, so the amount of additional
6954 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02006955
Willy Tarreaubef1b322014-05-13 21:01:39 +02006956 When this option is enabled, any change of the health check status or to
6957 the server's health will be logged, so that it becomes possible to know
6958 that a server was failing occasional checks before crashing, or exactly when
6959 it failed to respond a valid HTTP status, then when the port started to
6960 reject connections, then when the server stopped responding at all.
6961
Davor Ocelice9ed2812017-12-25 17:49:28 +01006962 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02006963 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02006964
Willy Tarreaubef1b322014-05-13 21:01:39 +02006965 See also: "option httpchk", "option ldap-check", "option mysql-check",
6966 "option pgsql-check", "option redis-check", "option smtpchk",
6967 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02006968
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006969
6970option log-separate-errors
6971no option log-separate-errors
6972 Change log level for non-completely successful connections
6973 May be used in sections : defaults | frontend | listen | backend
6974 yes | yes | yes | no
6975 Arguments : none
6976
6977 Sometimes looking for errors in logs is not easy. This option makes haproxy
6978 raise the level of logs containing potentially interesting information such
6979 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
6980 level changes from "info" to "err". This makes it possible to log them
6981 separately to a different file with most syslog daemons. Be careful not to
6982 remove them from the original file, otherwise you would lose ordering which
6983 provides very important information.
6984
6985 Using this option, large sites dealing with several thousand connections per
6986 second may log normal traffic to a rotating buffer and only archive smaller
6987 error logs.
6988
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006989 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006990 logging.
6991
Willy Tarreauc27debf2008-01-06 08:57:02 +01006992
6993option logasap
6994no option logasap
6995 Enable or disable early logging of HTTP requests
6996 May be used in sections : defaults | frontend | listen | backend
6997 yes | yes | yes | no
6998 Arguments : none
6999
7000 By default, HTTP requests are logged upon termination so that the total
7001 transfer time and the number of bytes appear in the logs. When large objects
7002 are being transferred, it may take a while before the request appears in the
7003 logs. Using "option logasap", the request gets logged as soon as the server
7004 sends the complete headers. The only missing information in the logs will be
7005 the total number of bytes which will indicate everything except the amount
7006 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007007 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01007008 "Content-Length" response header so that the logs at least indicate how many
7009 bytes are expected to be transferred.
7010
Willy Tarreaucc6c8912009-02-22 10:53:55 +01007011 Examples :
7012 listen http_proxy 0.0.0.0:80
7013 mode http
7014 option httplog
7015 option logasap
7016 log 192.168.2.200 local3
7017
7018 >>> Feb 6 12:14:14 localhost \
7019 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
7020 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
7021 "GET /image.iso HTTP/1.0"
7022
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007023 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01007024 logging.
7025
7026
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02007027option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007028 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007029 May be used in sections : defaults | frontend | listen | backend
7030 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007031 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007032 <username> This is the username which will be used when connecting to MySQL
7033 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02007034 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007035
7036 If you specify a username, the check consists of sending two MySQL packet,
7037 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01007038 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007039 Error packet. It is a basic but useful test which does not produce error nor
7040 aborted connect on the server. However, it requires adding an authorization
7041 in the MySQL table, like this :
7042
7043 USE mysql;
7044 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
7045 FLUSH PRIVILEGES;
7046
7047 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007048 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02007049 Error packet, we don't send anything in this mode. It was reported that it
7050 can generate lockout if check is too frequent and/or if there is not enough
7051 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
7052 value as if a connection is established successfully within fewer than MySQL
7053 "max_connect_errors" attempts after a previous connection was interrupted,
7054 the error count for the host is cleared to zero. If HAProxy's server get
7055 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
7056
7057 Remember that this does not check database presence nor database consistency.
7058 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007059
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02007060 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007061
7062 Most often, an incoming MySQL server needs to see the client's IP address for
7063 various purposes, including IP privilege matching and connection logging.
7064 When possible, it is often wise to masquerade the client's IP address when
7065 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007066 which requires the transparent proxy feature to be compiled in, and the MySQL
7067 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01007068
7069 See also: "option httpchk"
7070
7071
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007072option nolinger
7073no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007074 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007075 May be used in sections: defaults | frontend | listen | backend
7076 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007077 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007078
Davor Ocelice9ed2812017-12-25 17:49:28 +01007079 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007080 physically disconnected), the session timeouts triggers and the session is
7081 closed. But it will remain in FIN_WAIT1 state for some time in the system,
7082 using some resources and possibly limiting the ability to establish newer
7083 connections.
7084
7085 When this happens, it is possible to activate "option nolinger" which forces
7086 the system to immediately remove any socket's pending data on close. Thus,
7087 the session is instantly purged from the system's tables. This usually has
7088 side effects such as increased number of TCP resets due to old retransmits
7089 getting immediately rejected. Some firewalls may sometimes complain about
7090 this too.
7091
7092 For this reason, it is not recommended to use this option when not absolutely
7093 needed. You know that you need it when you have thousands of FIN_WAIT1
7094 sessions on your system (TIME_WAIT ones do not count).
7095
7096 This option may be used both on frontends and backends, depending on the side
7097 where it is required. Use it on the frontend for clients, and on the backend
7098 for servers.
7099
7100 If this option has been enabled in a "defaults" section, it can be disabled
7101 in a specific instance by prepending the "no" keyword before it.
7102
7103
Willy Tarreau55165fe2009-05-10 12:02:55 +02007104option originalto [ except <network> ] [ header <name> ]
7105 Enable insertion of the X-Original-To header to requests sent to servers
7106 May be used in sections : defaults | frontend | listen | backend
7107 yes | yes | yes | yes
7108 Arguments :
7109 <network> is an optional argument used to disable this option for sources
7110 matching <network>
7111 <name> an optional argument to specify a different "X-Original-To"
7112 header name.
7113
7114 Since HAProxy can work in transparent mode, every request from a client can
7115 be redirected to the proxy and HAProxy itself can proxy every request to a
7116 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
7117 be lost. This is annoying when you want access rules based on destination ip
7118 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
7119 added by HAProxy to all requests sent to the server. This header contains a
7120 value representing the original destination IP address. Since this must be
7121 configured to always use the last occurrence of this header only. Note that
7122 only the last occurrence of the header must be used, since it is really
7123 possible that the client has already brought one.
7124
7125 The keyword "header" may be used to supply a different header name to replace
7126 the default "X-Original-To". This can be useful where you might already
7127 have a "X-Original-To" header from a different application, and you need
7128 preserve it. Also if your backend server doesn't use the "X-Original-To"
7129 header and requires different one.
7130
7131 Sometimes, a same HAProxy instance may be shared between a direct client
7132 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
7133 used to decrypt HTTPS traffic). It is possible to disable the addition of the
7134 header for a known source address or network by adding the "except" keyword
7135 followed by the network address. In this case, any source IP matching the
7136 network will not cause an addition of this header. Most common uses are with
7137 private networks or 127.0.0.1.
7138
7139 This option may be specified either in the frontend or in the backend. If at
7140 least one of them uses it, the header will be added. Note that the backend's
7141 setting of the header subargument takes precedence over the frontend's if
7142 both are defined.
7143
Willy Tarreau55165fe2009-05-10 12:02:55 +02007144 Examples :
7145 # Original Destination address
7146 frontend www
7147 mode http
7148 option originalto except 127.0.0.1
7149
7150 # Those servers want the IP Address in X-Client-Dst
7151 backend www
7152 mode http
7153 option originalto header X-Client-Dst
7154
Christopher Faulet315b39c2018-09-21 16:26:19 +02007155 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02007156
7157
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007158option persist
7159no option persist
7160 Enable or disable forced persistence on down servers
7161 May be used in sections: defaults | frontend | listen | backend
7162 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007163 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007164
7165 When an HTTP request reaches a backend with a cookie which references a dead
7166 server, by default it is redispatched to another server. It is possible to
7167 force the request to be sent to the dead server first using "option persist"
7168 if absolutely needed. A common use case is when servers are under extreme
7169 load and spend their time flapping. In this case, the users would still be
7170 directed to the server they opened the session on, in the hope they would be
7171 correctly served. It is recommended to use "option redispatch" in conjunction
7172 with this option so that in the event it would not be possible to connect to
7173 the server at all (server definitely dead), the client would finally be
7174 redirected to another valid server.
7175
7176 If this option has been enabled in a "defaults" section, it can be disabled
7177 in a specific instance by prepending the "no" keyword before it.
7178
Willy Tarreau4de91492010-01-22 19:10:05 +01007179 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007180
7181
Willy Tarreau0c122822013-12-15 18:49:01 +01007182option pgsql-check [ user <username> ]
7183 Use PostgreSQL health checks for server testing
7184 May be used in sections : defaults | frontend | listen | backend
7185 yes | no | yes | yes
7186 Arguments :
7187 <username> This is the username which will be used when connecting to
7188 PostgreSQL server.
7189
7190 The check sends a PostgreSQL StartupMessage and waits for either
7191 Authentication request or ErrorResponse message. It is a basic but useful
7192 test which does not produce error nor aborted connect on the server.
7193 This check is identical with the "mysql-check".
7194
7195 See also: "option httpchk"
7196
7197
Willy Tarreau9420b122013-12-15 18:58:25 +01007198option prefer-last-server
7199no option prefer-last-server
7200 Allow multiple load balanced requests to remain on the same server
7201 May be used in sections: defaults | frontend | listen | backend
7202 yes | no | yes | yes
7203 Arguments : none
7204
7205 When the load balancing algorithm in use is not deterministic, and a previous
7206 request was sent to a server to which haproxy still holds a connection, it is
7207 sometimes desirable that subsequent requests on a same session go to the same
7208 server as much as possible. Note that this is different from persistence, as
7209 we only indicate a preference which haproxy tries to apply without any form
7210 of warranty. The real use is for keep-alive connections sent to servers. When
7211 this option is used, haproxy will try to reuse the same connection that is
7212 attached to the server instead of rebalancing to another server, causing a
7213 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01007214 not make much sense to use this in combination with hashing algorithms. Note,
7215 haproxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02007216 to a proxy which sends a 407 (authentication required), when the load
7217 balancing algorithm is not deterministic. This is mandatory for use with the
7218 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01007219 troubleshooting some faulty applications. Option prefer-last-server might be
7220 desirable in these environments as well, to avoid redistributing the traffic
7221 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01007222
7223 If this option has been enabled in a "defaults" section, it can be disabled
7224 in a specific instance by prepending the "no" keyword before it.
7225
7226 See also: "option http-keep-alive"
7227
7228
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007229option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07007230option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007231no option redispatch
7232 Enable or disable session redistribution in case of connection failure
7233 May be used in sections: defaults | frontend | listen | backend
7234 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07007235 Arguments :
7236 <interval> The optional integer value that controls how often redispatches
7237 occur when retrying connections. Positive value P indicates a
7238 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01007239 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07007240 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007241 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07007242 positive value of 1 would indicate a redispatch on every retry,
7243 and a positive value of 3 would indicate a redispatch on every
7244 third retry. You can disable redispatches with a value of 0.
7245
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007246
7247 In HTTP mode, if a server designated by a cookie is down, clients may
7248 definitely stick to it because they cannot flush the cookie, so they will not
7249 be able to access the service anymore.
7250
Willy Tarreau59884a62019-01-02 14:48:31 +01007251 Specifying "option redispatch" will allow the proxy to break cookie or
7252 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007253
Joseph Lynch726ab712015-05-11 23:25:34 -07007254 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007255 connection failures. Of course, it requires having "retries" set to a nonzero
7256 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007257
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007258 This form is the preferred form, which replaces both the "redispatch" and
7259 "redisp" keywords.
7260
7261 If this option has been enabled in a "defaults" section, it can be disabled
7262 in a specific instance by prepending the "no" keyword before it.
7263
Willy Tarreau4de91492010-01-22 19:10:05 +01007264 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007265
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007266
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007267option redis-check
7268 Use redis health checks for server testing
7269 May be used in sections : defaults | frontend | listen | backend
7270 yes | no | yes | yes
7271 Arguments : none
7272
7273 It is possible to test that the server correctly talks REDIS protocol instead
7274 of just testing that it accepts the TCP connection. When this option is set,
7275 a PING redis command is sent to the server, and the response is analyzed to
7276 find the "+PONG" response message.
7277
7278 Example :
7279 option redis-check
7280
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007281 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007282
7283
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007284option smtpchk
7285option smtpchk <hello> <domain>
7286 Use SMTP health checks for server testing
7287 May be used in sections : defaults | frontend | listen | backend
7288 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01007289 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007290 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02007291 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007292 values will be turned into the default command ("HELO").
7293
7294 <domain> is the domain name to present to the server. It may only be
7295 specified (and is mandatory) if the hello command has been
7296 specified. By default, "localhost" is used.
7297
7298 When "option smtpchk" is set, the health checks will consist in TCP
7299 connections followed by an SMTP command. By default, this command is
7300 "HELO localhost". The server's return code is analyzed and only return codes
7301 starting with a "2" will be considered as valid. All other responses,
7302 including a lack of response will constitute an error and will indicate a
7303 dead server.
7304
7305 This test is meant to be used with SMTP servers or relays. Depending on the
7306 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01007307 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007308 25 is often easier than adjusting the configuration.
7309
7310 Most often, an incoming SMTP server needs to see the client's IP address for
7311 various purposes, including spam filtering, anti-spoofing and logging. When
7312 possible, it is often wise to masquerade the client's IP address when
7313 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007314 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007315
7316 Example :
7317 option smtpchk HELO mydomain.org
7318
7319 See also : "option httpchk", "source"
7320
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007321
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02007322option socket-stats
7323no option socket-stats
7324
7325 Enable or disable collecting & providing separate statistics for each socket.
7326 May be used in sections : defaults | frontend | listen | backend
7327 yes | yes | yes | no
7328
7329 Arguments : none
7330
7331
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007332option splice-auto
7333no option splice-auto
7334 Enable or disable automatic kernel acceleration on sockets in both directions
7335 May be used in sections : defaults | frontend | listen | backend
7336 yes | yes | yes | yes
7337 Arguments : none
7338
7339 When this option is enabled either on a frontend or on a backend, haproxy
7340 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007341 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007342 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007343 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007344 are not much aggressive in order to limit excessive use of splicing. This
7345 option requires splicing to be enabled at compile time, and may be globally
7346 disabled with the global option "nosplice". Since splice uses pipes, using it
7347 requires that there are enough spare pipes.
7348
7349 Important note: kernel-based TCP splicing is a Linux-specific feature which
7350 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
7351 transfer data between sockets without copying these data to user-space, thus
7352 providing noticeable performance gains and CPU cycles savings. Since many
7353 early implementations are buggy, corrupt data and/or are inefficient, this
7354 feature is not enabled by default, and it should be used with extreme care.
7355 While it is not possible to detect the correctness of an implementation,
7356 2.6.29 is the first version offering a properly working implementation. In
7357 case of doubt, splicing may be globally disabled using the global "nosplice"
7358 keyword.
7359
7360 Example :
7361 option splice-auto
7362
7363 If this option has been enabled in a "defaults" section, it can be disabled
7364 in a specific instance by prepending the "no" keyword before it.
7365
7366 See also : "option splice-request", "option splice-response", and global
7367 options "nosplice" and "maxpipes"
7368
7369
7370option splice-request
7371no option splice-request
7372 Enable or disable automatic kernel acceleration on sockets for requests
7373 May be used in sections : defaults | frontend | listen | backend
7374 yes | yes | yes | yes
7375 Arguments : none
7376
7377 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007378 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007379 the client to the server. It might still use the recv/send scheme if there
7380 are no spare pipes left. This option requires splicing to be enabled at
7381 compile time, and may be globally disabled with the global option "nosplice".
7382 Since splice uses pipes, using it requires that there are enough spare pipes.
7383
7384 Important note: see "option splice-auto" for usage limitations.
7385
7386 Example :
7387 option splice-request
7388
7389 If this option has been enabled in a "defaults" section, it can be disabled
7390 in a specific instance by prepending the "no" keyword before it.
7391
7392 See also : "option splice-auto", "option splice-response", and global options
7393 "nosplice" and "maxpipes"
7394
7395
7396option splice-response
7397no option splice-response
7398 Enable or disable automatic kernel acceleration on sockets for responses
7399 May be used in sections : defaults | frontend | listen | backend
7400 yes | yes | yes | yes
7401 Arguments : none
7402
7403 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007404 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007405 the server to the client. It might still use the recv/send scheme if there
7406 are no spare pipes left. This option requires splicing to be enabled at
7407 compile time, and may be globally disabled with the global option "nosplice".
7408 Since splice uses pipes, using it requires that there are enough spare pipes.
7409
7410 Important note: see "option splice-auto" for usage limitations.
7411
7412 Example :
7413 option splice-response
7414
7415 If this option has been enabled in a "defaults" section, it can be disabled
7416 in a specific instance by prepending the "no" keyword before it.
7417
7418 See also : "option splice-auto", "option splice-request", and global options
7419 "nosplice" and "maxpipes"
7420
7421
Christopher Fauletba7bc162016-11-07 21:07:38 +01007422option spop-check
7423 Use SPOP health checks for server testing
7424 May be used in sections : defaults | frontend | listen | backend
7425 no | no | no | yes
7426 Arguments : none
7427
7428 It is possible to test that the server correctly talks SPOP protocol instead
7429 of just testing that it accepts the TCP connection. When this option is set,
7430 a HELLO handshake is performed between HAProxy and the server, and the
7431 response is analyzed to check no error is reported.
7432
7433 Example :
7434 option spop-check
7435
7436 See also : "option httpchk"
7437
7438
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007439option srvtcpka
7440no option srvtcpka
7441 Enable or disable the sending of TCP keepalive packets on the server side
7442 May be used in sections : defaults | frontend | listen | backend
7443 yes | no | yes | yes
7444 Arguments : none
7445
7446 When there is a firewall or any session-aware component between a client and
7447 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007448 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007449 components decides to expire a session which has remained idle for too long.
7450
7451 Enabling socket-level TCP keep-alives makes the system regularly send packets
7452 to the other end of the connection, leaving it active. The delay between
7453 keep-alive probes is controlled by the system only and depends both on the
7454 operating system and its tuning parameters.
7455
7456 It is important to understand that keep-alive packets are neither emitted nor
7457 received at the application level. It is only the network stacks which sees
7458 them. For this reason, even if one side of the proxy already uses keep-alives
7459 to maintain its connection alive, those keep-alive packets will not be
7460 forwarded to the other side of the proxy.
7461
7462 Please note that this has nothing to do with HTTP keep-alive.
7463
7464 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
7465 server side of a connection, which should help when session expirations are
7466 noticed between HAProxy and a server.
7467
7468 If this option has been enabled in a "defaults" section, it can be disabled
7469 in a specific instance by prepending the "no" keyword before it.
7470
7471 See also : "option clitcpka", "option tcpka"
7472
7473
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007474option ssl-hello-chk
7475 Use SSLv3 client hello health checks for server testing
7476 May be used in sections : defaults | frontend | listen | backend
7477 yes | no | yes | yes
7478 Arguments : none
7479
7480 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
7481 possible to test that the server correctly talks SSL instead of just testing
7482 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
7483 SSLv3 client hello messages are sent once the connection is established to
7484 the server, and the response is analyzed to find an SSL server hello message.
7485 The server is considered valid only when the response contains this server
7486 hello message.
7487
7488 All servers tested till there correctly reply to SSLv3 client hello messages,
7489 and most servers tested do not even log the requests containing only hello
7490 messages, which is appreciable.
7491
Willy Tarreau763a95b2012-10-04 23:15:39 +02007492 Note that this check works even when SSL support was not built into haproxy
7493 because it forges the SSL message. When SSL support is available, it is best
7494 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007495
Willy Tarreau763a95b2012-10-04 23:15:39 +02007496 See also: "option httpchk", "check-ssl"
7497
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007498
Willy Tarreaued179852013-12-16 01:07:00 +01007499option tcp-check
7500 Perform health checks using tcp-check send/expect sequences
7501 May be used in sections: defaults | frontend | listen | backend
7502 yes | no | yes | yes
7503
7504 This health check method is intended to be combined with "tcp-check" command
7505 lists in order to support send/expect types of health check sequences.
7506
7507 TCP checks currently support 4 modes of operations :
7508 - no "tcp-check" directive : the health check only consists in a connection
7509 attempt, which remains the default mode.
7510
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007511 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01007512 used to send a string along with a connection opening. With some
7513 protocols, it helps sending a "QUIT" message for example that prevents
7514 the server from logging a connection error for each health check. The
7515 check result will still be based on the ability to open the connection
7516 only.
7517
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007518 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01007519 The connection is opened and haproxy waits for the server to present some
7520 contents which must validate some rules. The check result will be based
7521 on the matching between the contents and the rules. This is suited for
7522 POP, IMAP, SMTP, FTP, SSH, TELNET.
7523
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007524 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007525 used to test a hello-type protocol. HAProxy sends a message, the server
7526 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007527 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01007528 suited for protocols which require a binding or a request/response model.
7529 LDAP, MySQL, Redis and SSL are example of such protocols, though they
7530 already all have their dedicated checks with a deeper understanding of
7531 the respective protocols.
7532 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01007533 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01007534
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007535 A fifth mode can be used to insert comments in different steps of the
7536 script.
7537
7538 For each tcp-check rule you create, you can add a "comment" directive,
7539 followed by a string. This string will be reported in the log and stderr
7540 in debug mode. It is useful to make user-friendly error reporting.
7541 The "comment" is of course optional.
7542
7543
Willy Tarreaued179852013-12-16 01:07:00 +01007544 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007545 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007546 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007547 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007548
Davor Ocelice9ed2812017-12-25 17:49:28 +01007549 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007550 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007551 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007552
7553 # look for the redis master server after ensuring it speaks well
7554 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01007555 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01007556 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007557 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007558 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02007559 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007560 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01007561 tcp-check send info\ replication\r\n
7562 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007563 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007564 tcp-check send QUIT\r\n
7565 tcp-check expect string +OK
7566
Davor Ocelice9ed2812017-12-25 17:49:28 +01007567 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01007568 (send many headers before analyzing)
7569 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007570 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01007571 tcp-check send HEAD\ /\ HTTP/1.1\r\n
7572 tcp-check send Host:\ www.mydomain.com\r\n
7573 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
7574 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007575 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01007576
7577
7578 See also : "tcp-check expect", "tcp-check send"
7579
7580
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007581option tcp-smart-accept
7582no option tcp-smart-accept
7583 Enable or disable the saving of one ACK packet during the accept sequence
7584 May be used in sections : defaults | frontend | listen | backend
7585 yes | yes | yes | no
7586 Arguments : none
7587
7588 When an HTTP connection request comes in, the system acknowledges it on
7589 behalf of HAProxy, then the client immediately sends its request, and the
7590 system acknowledges it too while it is notifying HAProxy about the new
7591 connection. HAProxy then reads the request and responds. This means that we
7592 have one TCP ACK sent by the system for nothing, because the request could
7593 very well be acknowledged by HAProxy when it sends its response.
7594
7595 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
7596 sending this useless ACK on platforms which support it (currently at least
7597 Linux). It must not cause any problem, because the system will send it anyway
7598 after 40 ms if the response takes more time than expected to come.
7599
7600 During complex network debugging sessions, it may be desirable to disable
7601 this optimization because delayed ACKs can make troubleshooting more complex
7602 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007603 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007604
7605 It is also possible to force it for non-HTTP proxies by simply specifying
7606 "option tcp-smart-accept". For instance, it can make sense with some services
7607 such as SMTP where the server speaks first.
7608
7609 It is recommended to avoid forcing this option in a defaults section. In case
7610 of doubt, consider setting it back to automatic values by prepending the
7611 "default" keyword before it, or disabling it using the "no" keyword.
7612
Willy Tarreaud88edf22009-06-14 15:48:17 +02007613 See also : "option tcp-smart-connect"
7614
7615
7616option tcp-smart-connect
7617no option tcp-smart-connect
7618 Enable or disable the saving of one ACK packet during the connect sequence
7619 May be used in sections : defaults | frontend | listen | backend
7620 yes | no | yes | yes
7621 Arguments : none
7622
7623 On certain systems (at least Linux), HAProxy can ask the kernel not to
7624 immediately send an empty ACK upon a connection request, but to directly
7625 send the buffer request instead. This saves one packet on the network and
7626 thus boosts performance. It can also be useful for some servers, because they
7627 immediately get the request along with the incoming connection.
7628
7629 This feature is enabled when "option tcp-smart-connect" is set in a backend.
7630 It is not enabled by default because it makes network troubleshooting more
7631 complex.
7632
7633 It only makes sense to enable it with protocols where the client speaks first
7634 such as HTTP. In other situations, if there is no data to send in place of
7635 the ACK, a normal ACK is sent.
7636
7637 If this option has been enabled in a "defaults" section, it can be disabled
7638 in a specific instance by prepending the "no" keyword before it.
7639
7640 See also : "option tcp-smart-accept"
7641
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007642
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007643option tcpka
7644 Enable or disable the sending of TCP keepalive packets on both sides
7645 May be used in sections : defaults | frontend | listen | backend
7646 yes | yes | yes | yes
7647 Arguments : none
7648
7649 When there is a firewall or any session-aware component between a client and
7650 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007651 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007652 components decides to expire a session which has remained idle for too long.
7653
7654 Enabling socket-level TCP keep-alives makes the system regularly send packets
7655 to the other end of the connection, leaving it active. The delay between
7656 keep-alive probes is controlled by the system only and depends both on the
7657 operating system and its tuning parameters.
7658
7659 It is important to understand that keep-alive packets are neither emitted nor
7660 received at the application level. It is only the network stacks which sees
7661 them. For this reason, even if one side of the proxy already uses keep-alives
7662 to maintain its connection alive, those keep-alive packets will not be
7663 forwarded to the other side of the proxy.
7664
7665 Please note that this has nothing to do with HTTP keep-alive.
7666
7667 Using option "tcpka" enables the emission of TCP keep-alive probes on both
7668 the client and server sides of a connection. Note that this is meaningful
7669 only in "defaults" or "listen" sections. If this option is used in a
7670 frontend, only the client side will get keep-alives, and if this option is
7671 used in a backend, only the server side will get keep-alives. For this
7672 reason, it is strongly recommended to explicitly use "option clitcpka" and
7673 "option srvtcpka" when the configuration is split between frontends and
7674 backends.
7675
7676 See also : "option clitcpka", "option srvtcpka"
7677
Willy Tarreau844e3c52008-01-11 16:28:18 +01007678
7679option tcplog
7680 Enable advanced logging of TCP connections with session state and timers
7681 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01007682 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01007683 Arguments : none
7684
7685 By default, the log output format is very poor, as it only contains the
7686 source and destination addresses, and the instance name. By specifying
7687 "option tcplog", each log line turns into a much richer format including, but
7688 not limited to, the connection timers, the session status, the connections
7689 numbers, the frontend, backend and server name, and of course the source
7690 address and ports. This option is useful for pure TCP proxies in order to
7691 find which of the client or server disconnects or times out. For normal HTTP
7692 proxies, it's better to use "option httplog" which is even more complete.
7693
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007694 "option tcplog" overrides any previous "log-format" directive.
7695
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007696 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007697
7698
Willy Tarreau844e3c52008-01-11 16:28:18 +01007699option transparent
7700no option transparent
7701 Enable client-side transparent proxying
7702 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007703 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007704 Arguments : none
7705
7706 This option was introduced in order to provide layer 7 persistence to layer 3
7707 load balancers. The idea is to use the OS's ability to redirect an incoming
7708 connection for a remote address to a local process (here HAProxy), and let
7709 this process know what address was initially requested. When this option is
7710 used, sessions without cookies will be forwarded to the original destination
7711 IP address of the incoming request (which should match that of another
7712 equipment), while requests with cookies will still be forwarded to the
7713 appropriate server.
7714
7715 Note that contrary to a common belief, this option does NOT make HAProxy
7716 present the client's IP to the server when establishing the connection.
7717
Willy Tarreaua1146052011-03-01 09:51:54 +01007718 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007719 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007720
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007721
Simon Horman98637e52014-06-20 12:30:16 +09007722external-check command <command>
7723 Executable to run when performing an external-check
7724 May be used in sections : defaults | frontend | listen | backend
7725 yes | no | yes | yes
7726
7727 Arguments :
7728 <command> is the external command to run
7729
Simon Horman98637e52014-06-20 12:30:16 +09007730 The arguments passed to the to the command are:
7731
Cyril Bonté777be862014-12-02 21:21:35 +01007732 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09007733
Cyril Bonté777be862014-12-02 21:21:35 +01007734 The <proxy_address> and <proxy_port> are derived from the first listener
7735 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
7736 listener the proxy_address will be the path of the socket and the
7737 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
7738 possible to determine a listener, and both <proxy_address> and <proxy_port>
7739 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09007740
Cyril Bonté72cda2a2014-12-27 22:28:39 +01007741 Some values are also provided through environment variables.
7742
7743 Environment variables :
7744 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
7745 applicable, for example in a "backend" section).
7746
7747 HAPROXY_PROXY_ID The backend id.
7748
7749 HAPROXY_PROXY_NAME The backend name.
7750
7751 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
7752 applicable, for example in a "backend" section or
7753 for a UNIX socket).
7754
7755 HAPROXY_SERVER_ADDR The server address.
7756
7757 HAPROXY_SERVER_CURCONN The current number of connections on the server.
7758
7759 HAPROXY_SERVER_ID The server id.
7760
7761 HAPROXY_SERVER_MAXCONN The server max connections.
7762
7763 HAPROXY_SERVER_NAME The server name.
7764
7765 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
7766 socket).
7767
7768 PATH The PATH environment variable used when executing
7769 the command may be set using "external-check path".
7770
William Lallemand4d03e432019-06-14 15:35:37 +02007771 See also "2.3. Environment variables" for other variables.
7772
Simon Horman98637e52014-06-20 12:30:16 +09007773 If the command executed and exits with a zero status then the check is
7774 considered to have passed, otherwise the check is considered to have
7775 failed.
7776
7777 Example :
7778 external-check command /bin/true
7779
7780 See also : "external-check", "option external-check", "external-check path"
7781
7782
7783external-check path <path>
7784 The value of the PATH environment variable used when running an external-check
7785 May be used in sections : defaults | frontend | listen | backend
7786 yes | no | yes | yes
7787
7788 Arguments :
7789 <path> is the path used when executing external command to run
7790
7791 The default path is "".
7792
7793 Example :
7794 external-check path "/usr/bin:/bin"
7795
7796 See also : "external-check", "option external-check",
7797 "external-check command"
7798
7799
Emeric Brun647caf12009-06-30 17:57:00 +02007800persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02007801persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02007802 Enable RDP cookie-based persistence
7803 May be used in sections : defaults | frontend | listen | backend
7804 yes | no | yes | yes
7805 Arguments :
7806 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02007807 default cookie name "msts" will be used. There currently is no
7808 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02007809
7810 This statement enables persistence based on an RDP cookie. The RDP cookie
7811 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01007812 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02007813 and if an RDP cookie is found, it is decoded. If it matches a known server
7814 which is still UP (or if "option persist" is set), then the connection is
7815 forwarded to this server.
7816
7817 Note that this only makes sense in a TCP backend, but for this to work, the
7818 frontend must have waited long enough to ensure that an RDP cookie is present
7819 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007820 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02007821 a single "listen" section.
7822
Willy Tarreau61e28f22010-05-16 22:31:05 +02007823 Also, it is important to understand that the terminal server will emit this
7824 RDP cookie only if it is configured for "token redirection mode", which means
7825 that the "IP address redirection" option is disabled.
7826
Emeric Brun647caf12009-06-30 17:57:00 +02007827 Example :
7828 listen tse-farm
7829 bind :3389
7830 # wait up to 5s for an RDP cookie in the request
7831 tcp-request inspect-delay 5s
7832 tcp-request content accept if RDP_COOKIE
7833 # apply RDP cookie persistence
7834 persist rdp-cookie
7835 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02007836 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02007837 balance rdp-cookie
7838 server srv1 1.1.1.1:3389
7839 server srv2 1.1.1.2:3389
7840
Simon Hormanab814e02011-06-24 14:50:20 +09007841 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
7842 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02007843
7844
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007845rate-limit sessions <rate>
7846 Set a limit on the number of new sessions accepted per second on a frontend
7847 May be used in sections : defaults | frontend | listen | backend
7848 yes | yes | yes | no
7849 Arguments :
7850 <rate> The <rate> parameter is an integer designating the maximum number
7851 of new sessions per second to accept on the frontend.
7852
7853 When the frontend reaches the specified number of new sessions per second, it
7854 stops accepting new connections until the rate drops below the limit again.
7855 During this time, the pending sessions will be kept in the socket's backlog
7856 (in system buffers) and haproxy will not even be aware that sessions are
7857 pending. When applying very low limit on a highly loaded service, it may make
7858 sense to increase the socket's backlog using the "backlog" keyword.
7859
7860 This feature is particularly efficient at blocking connection-based attacks
7861 or service abuse on fragile servers. Since the session rate is measured every
7862 millisecond, it is extremely accurate. Also, the limit applies immediately,
7863 no delay is needed at all to detect the threshold.
7864
7865 Example : limit the connection rate on SMTP to 10 per second max
7866 listen smtp
7867 mode tcp
7868 bind :25
7869 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02007870 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007871
Willy Tarreaua17c2d92011-07-25 08:16:20 +02007872 Note : when the maximum rate is reached, the frontend's status is not changed
7873 but its sockets appear as "WAITING" in the statistics if the
7874 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007875
7876 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
7877
7878
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007879redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
7880redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
7881redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007882 Return an HTTP redirection if/unless a condition is matched
7883 May be used in sections : defaults | frontend | listen | backend
7884 no | yes | yes | yes
7885
7886 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01007887 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007888
Willy Tarreau0140f252008-11-19 21:07:09 +01007889 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007890 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007891 the HTTP "Location" header. When used in an "http-request" rule,
7892 <loc> value follows the log-format rules and can include some
7893 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007894
7895 <pfx> With "redirect prefix", the "Location" header is built from the
7896 concatenation of <pfx> and the complete URI path, including the
7897 query string, unless the "drop-query" option is specified (see
7898 below). As a special case, if <pfx> equals exactly "/", then
7899 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007900 redirect to the same URL (for instance, to insert a cookie). When
7901 used in an "http-request" rule, <pfx> value follows the log-format
7902 rules and can include some dynamic values (see Custom Log Format
7903 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007904
7905 <sch> With "redirect scheme", then the "Location" header is built by
7906 concatenating <sch> with "://" then the first occurrence of the
7907 "Host" header, and then the URI path, including the query string
7908 unless the "drop-query" option is specified (see below). If no
7909 path is found or if the path is "*", then "/" is used instead. If
7910 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007911 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007912 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007913 HTTPS. When used in an "http-request" rule, <sch> value follows
7914 the log-format rules and can include some dynamic values (see
7915 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01007916
7917 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007918 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
7919 with 302 used by default if no code is specified. 301 means
7920 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02007921 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007922 cache the redirection. 303 is equivalent to 302 except that the
7923 browser will fetch the location with a GET method. 307 is just
7924 like 302 but makes it clear that the same method must be reused.
7925 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01007926
7927 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007928 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01007929
7930 - "drop-query"
7931 When this keyword is used in a prefix-based redirection, then the
7932 location will be set without any possible query-string, which is useful
7933 for directing users to a non-secure page for instance. It has no effect
7934 with a location-type redirect.
7935
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007936 - "append-slash"
7937 This keyword may be used in conjunction with "drop-query" to redirect
7938 users who use a URL not ending with a '/' to the same one with the '/'.
7939 It can be useful to ensure that search engines will only see one URL.
7940 For this, a return code 301 is preferred.
7941
Willy Tarreau0140f252008-11-19 21:07:09 +01007942 - "set-cookie NAME[=value]"
7943 A "Set-Cookie" header will be added with NAME (and optionally "=value")
7944 to the response. This is sometimes used to indicate that a user has
7945 been seen, for instance to protect against some types of DoS. No other
7946 cookie option is added, so the cookie will be a session cookie. Note
7947 that for a browser, a sole cookie name without an equal sign is
7948 different from a cookie with an equal sign.
7949
7950 - "clear-cookie NAME[=]"
7951 A "Set-Cookie" header will be added with NAME (and optionally "="), but
7952 with the "Max-Age" attribute set to zero. This will tell the browser to
7953 delete this cookie. It is useful for instance on logout pages. It is
7954 important to note that clearing the cookie "NAME" will not remove a
7955 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
7956 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007957
7958 Example: move the login URL only to HTTPS.
7959 acl clear dst_port 80
7960 acl secure dst_port 8080
7961 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01007962 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01007963 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01007964 acl cookie_set hdr_sub(cookie) SEEN=1
7965
7966 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01007967 redirect prefix https://mysite.com if login_page !secure
7968 redirect prefix http://mysite.com drop-query if login_page !uid_given
7969 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01007970 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007971
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007972 Example: send redirects for request for articles without a '/'.
7973 acl missing_slash path_reg ^/article/[^/]*$
7974 redirect code 301 prefix / drop-query append-slash if missing_slash
7975
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007976 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01007977 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007978
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007979 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02007980 http-request redirect code 301 location \
7981 http://www.%[hdr(host)]%[capture.req.uri] \
7982 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007983
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007984 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007985
7986
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007987redisp (deprecated)
7988redispatch (deprecated)
7989 Enable or disable session redistribution in case of connection failure
7990 May be used in sections: defaults | frontend | listen | backend
7991 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007992 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007993
7994 In HTTP mode, if a server designated by a cookie is down, clients may
7995 definitely stick to it because they cannot flush the cookie, so they will not
7996 be able to access the service anymore.
7997
7998 Specifying "redispatch" will allow the proxy to break their persistence and
7999 redistribute them to a working server.
8000
8001 It also allows to retry last connection to another server in case of multiple
8002 connection failures. Of course, it requires having "retries" set to a nonzero
8003 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01008004
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01008005 This form is deprecated, do not use it in any new configuration, use the new
8006 "option redispatch" instead.
8007
8008 See also : "option redispatch"
8009
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008010
Willy Tarreau96d51952019-05-22 20:34:35 +02008011reqadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008012 Add a header at the end of the HTTP request
8013 May be used in sections : defaults | frontend | listen | backend
8014 no | yes | yes | yes
8015 Arguments :
8016 <string> is the complete line to be added. Any space or known delimiter
8017 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008018 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008019
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01008020 <cond> is an optional matching condition built from ACLs. It makes it
8021 possible to ignore this rule when other conditions are not met.
8022
Willy Tarreau303c0352008-01-17 19:01:39 +01008023 A new line consisting in <string> followed by a line feed will be added after
8024 the last header of an HTTP request.
8025
8026 Header transformations only apply to traffic which passes through HAProxy,
8027 and not to traffic generated by HAProxy, such as health-checks or error
8028 responses.
8029
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01008030 Example : add "X-Proto: SSL" to requests coming via port 81
8031 acl is-ssl dst_port 81
8032 reqadd X-Proto:\ SSL if is-ssl
8033
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008034 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
8035 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008036
8037
Willy Tarreau96d51952019-05-22 20:34:35 +02008038reqallow <search> [{if | unless} <cond>] (deprecated)
8039reqiallow <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008040 Definitely allow an HTTP request if a line matches a regular expression
8041 May be used in sections : defaults | frontend | listen | backend
8042 no | yes | yes | yes
8043 Arguments :
8044 <search> is the regular expression applied to HTTP headers and to the
8045 request line. This is an extended regular expression. Parenthesis
8046 grouping is supported and no preliminary backslash is required.
8047 Any space or known delimiter must be escaped using a backslash
8048 ('\'). The pattern applies to a full line at a time. The
8049 "reqallow" keyword strictly matches case while "reqiallow"
8050 ignores case.
8051
Willy Tarreau5321c422010-01-28 20:35:13 +01008052 <cond> is an optional matching condition built from ACLs. It makes it
8053 possible to ignore this rule when other conditions are not met.
8054
Willy Tarreau303c0352008-01-17 19:01:39 +01008055 A request containing any line which matches extended regular expression
8056 <search> will mark the request as allowed, even if any later test would
8057 result in a deny. The test applies both to the request line and to request
8058 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01008059 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01008060
8061 It is easier, faster and more powerful to use ACLs to write access policies.
8062 Reqdeny, reqallow and reqpass should be avoided in new designs.
8063
8064 Example :
8065 # allow www.* but refuse *.local
8066 reqiallow ^Host:\ www\.
8067 reqideny ^Host:\ .*\.local
8068
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008069 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
8070 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008071
8072
Willy Tarreau96d51952019-05-22 20:34:35 +02008073reqdel <search> [{if | unless} <cond>] (deprecated)
8074reqidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008075 Delete all headers matching a regular expression in an HTTP request
8076 May be used in sections : defaults | frontend | listen | backend
8077 no | yes | yes | yes
8078 Arguments :
8079 <search> is the regular expression applied to HTTP headers and to the
8080 request line. This is an extended regular expression. Parenthesis
8081 grouping is supported and no preliminary backslash is required.
8082 Any space or known delimiter must be escaped using a backslash
8083 ('\'). The pattern applies to a full line at a time. The "reqdel"
8084 keyword strictly matches case while "reqidel" ignores case.
8085
Willy Tarreau5321c422010-01-28 20:35:13 +01008086 <cond> is an optional matching condition built from ACLs. It makes it
8087 possible to ignore this rule when other conditions are not met.
8088
Willy Tarreau303c0352008-01-17 19:01:39 +01008089 Any header line matching extended regular expression <search> in the request
8090 will be completely deleted. Most common use of this is to remove unwanted
8091 and/or dangerous headers or cookies from a request before passing it to the
8092 next servers.
8093
8094 Header transformations only apply to traffic which passes through HAProxy,
8095 and not to traffic generated by HAProxy, such as health-checks or error
8096 responses. Keep in mind that header names are not case-sensitive.
8097
8098 Example :
8099 # remove X-Forwarded-For header and SERVER cookie
8100 reqidel ^X-Forwarded-For:.*
8101 reqidel ^Cookie:.*SERVER=
8102
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008103 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
8104 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008105
8106
Willy Tarreau96d51952019-05-22 20:34:35 +02008107reqdeny <search> [{if | unless} <cond>] (deprecated)
8108reqideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008109 Deny an HTTP request if a line matches a regular expression
8110 May be used in sections : defaults | frontend | listen | backend
8111 no | yes | yes | yes
8112 Arguments :
8113 <search> is the regular expression applied to HTTP headers and to the
8114 request line. This is an extended regular expression. Parenthesis
8115 grouping is supported and no preliminary backslash is required.
8116 Any space or known delimiter must be escaped using a backslash
8117 ('\'). The pattern applies to a full line at a time. The
8118 "reqdeny" keyword strictly matches case while "reqideny" ignores
8119 case.
8120
Willy Tarreau5321c422010-01-28 20:35:13 +01008121 <cond> is an optional matching condition built from ACLs. It makes it
8122 possible to ignore this rule when other conditions are not met.
8123
Willy Tarreau303c0352008-01-17 19:01:39 +01008124 A request containing any line which matches extended regular expression
8125 <search> will mark the request as denied, even if any later test would
8126 result in an allow. The test applies both to the request line and to request
8127 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01008128 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01008129
Willy Tarreauced27012008-01-17 20:35:34 +01008130 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008131 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01008132 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01008133
Willy Tarreau303c0352008-01-17 19:01:39 +01008134 It is easier, faster and more powerful to use ACLs to write access policies.
8135 Reqdeny, reqallow and reqpass should be avoided in new designs.
8136
8137 Example :
8138 # refuse *.local, then allow www.*
8139 reqideny ^Host:\ .*\.local
8140 reqiallow ^Host:\ www\.
8141
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008142 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
8143 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008144
8145
Willy Tarreau96d51952019-05-22 20:34:35 +02008146reqpass <search> [{if | unless} <cond>] (deprecated)
8147reqipass <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008148 Ignore any HTTP request line matching a regular expression in next rules
8149 May be used in sections : defaults | frontend | listen | backend
8150 no | yes | yes | yes
8151 Arguments :
8152 <search> is the regular expression applied to HTTP headers and to the
8153 request line. This is an extended regular expression. Parenthesis
8154 grouping is supported and no preliminary backslash is required.
8155 Any space or known delimiter must be escaped using a backslash
8156 ('\'). The pattern applies to a full line at a time. The
8157 "reqpass" keyword strictly matches case while "reqipass" ignores
8158 case.
8159
Willy Tarreau5321c422010-01-28 20:35:13 +01008160 <cond> is an optional matching condition built from ACLs. It makes it
8161 possible to ignore this rule when other conditions are not met.
8162
Willy Tarreau303c0352008-01-17 19:01:39 +01008163 A request containing any line which matches extended regular expression
8164 <search> will skip next rules, without assigning any deny or allow verdict.
8165 The test applies both to the request line and to request headers. Keep in
8166 mind that URLs in request line are case-sensitive while header names are not.
8167
8168 It is easier, faster and more powerful to use ACLs to write access policies.
8169 Reqdeny, reqallow and reqpass should be avoided in new designs.
8170
8171 Example :
8172 # refuse *.local, then allow www.*, but ignore "www.private.local"
8173 reqipass ^Host:\ www.private\.local
8174 reqideny ^Host:\ .*\.local
8175 reqiallow ^Host:\ www\.
8176
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008177 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
8178 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008179
8180
Willy Tarreau96d51952019-05-22 20:34:35 +02008181reqrep <search> <string> [{if | unless} <cond>] (deprecated)
8182reqirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008183 Replace a regular expression with a string in an HTTP request line
8184 May be used in sections : defaults | frontend | listen | backend
8185 no | yes | yes | yes
8186 Arguments :
8187 <search> is the regular expression applied to HTTP headers and to the
8188 request line. This is an extended regular expression. Parenthesis
8189 grouping is supported and no preliminary backslash is required.
8190 Any space or known delimiter must be escaped using a backslash
8191 ('\'). The pattern applies to a full line at a time. The "reqrep"
8192 keyword strictly matches case while "reqirep" ignores case.
8193
8194 <string> is the complete line to be added. Any space or known delimiter
8195 must be escaped using a backslash ('\'). References to matched
8196 pattern groups are possible using the common \N form, with N
8197 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008198 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008199
Willy Tarreau5321c422010-01-28 20:35:13 +01008200 <cond> is an optional matching condition built from ACLs. It makes it
8201 possible to ignore this rule when other conditions are not met.
8202
Willy Tarreau303c0352008-01-17 19:01:39 +01008203 Any line matching extended regular expression <search> in the request (both
8204 the request line and header lines) will be completely replaced with <string>.
8205 Most common use of this is to rewrite URLs or domain names in "Host" headers.
8206
8207 Header transformations only apply to traffic which passes through HAProxy,
8208 and not to traffic generated by HAProxy, such as health-checks or error
8209 responses. Note that for increased readability, it is suggested to add enough
8210 spaces between the request and the response. Keep in mind that URLs in
8211 request line are case-sensitive while header names are not.
8212
8213 Example :
8214 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04008215 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01008216 # replace "www.mydomain.com" with "www" in the host name.
8217 reqirep ^Host:\ www.mydomain.com Host:\ www
8218
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008219 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
8220 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008221
8222
Willy Tarreau96d51952019-05-22 20:34:35 +02008223reqtarpit <search> [{if | unless} <cond>] (deprecated)
8224reqitarpit <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008225 Tarpit an HTTP request containing a line matching a regular expression
8226 May be used in sections : defaults | frontend | listen | backend
8227 no | yes | yes | yes
8228 Arguments :
8229 <search> is the regular expression applied to HTTP headers and to the
8230 request line. This is an extended regular expression. Parenthesis
8231 grouping is supported and no preliminary backslash is required.
8232 Any space or known delimiter must be escaped using a backslash
8233 ('\'). The pattern applies to a full line at a time. The
8234 "reqtarpit" keyword strictly matches case while "reqitarpit"
8235 ignores case.
8236
Willy Tarreau5321c422010-01-28 20:35:13 +01008237 <cond> is an optional matching condition built from ACLs. It makes it
8238 possible to ignore this rule when other conditions are not met.
8239
Willy Tarreau303c0352008-01-17 19:01:39 +01008240 A request containing any line which matches extended regular expression
8241 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01008242 be kept open for a pre-defined time, then will return an HTTP error 500 so
8243 that the attacker does not suspect it has been tarpitted. The status 500 will
8244 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01008245 delay is defined by "timeout tarpit", or "timeout connect" if the former is
8246 not set.
8247
8248 The goal of the tarpit is to slow down robots attacking servers with
8249 identifiable requests. Many robots limit their outgoing number of connections
8250 and stay connected waiting for a reply which can take several minutes to
8251 come. Depending on the environment and attack, it may be particularly
8252 efficient at reducing the load on the network and firewalls.
8253
Willy Tarreau5321c422010-01-28 20:35:13 +01008254 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01008255 # ignore user-agents reporting any flavor of "Mozilla" or "MSIE", but
Willy Tarreau303c0352008-01-17 19:01:39 +01008256 # block all others.
8257 reqipass ^User-Agent:\.*(Mozilla|MSIE)
8258 reqitarpit ^User-Agent:
8259
Willy Tarreau5321c422010-01-28 20:35:13 +01008260 # block bad guys
8261 acl badguys src 10.1.0.3 172.16.13.20/28
8262 reqitarpit . if badguys
8263
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008264 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
8265 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008266
8267
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008268retries <value>
8269 Set the number of retries to perform on a server after a connection failure
8270 May be used in sections: defaults | frontend | listen | backend
8271 yes | no | yes | yes
8272 Arguments :
8273 <value> is the number of times a connection attempt should be retried on
8274 a server when a connection either is refused or times out. The
8275 default value is 3.
8276
8277 It is important to understand that this value applies to the number of
8278 connection attempts, not full requests. When a connection has effectively
8279 been established to a server, there will be no more retry.
8280
8281 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07008282 a turn-around timer of min("timeout connect", one second) is applied before
8283 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008284
8285 When "option redispatch" is set, the last retry may be performed on another
8286 server even if a cookie references a different server.
8287
8288 See also : "option redispatch"
8289
8290
Olivier Houcharda254a372019-04-05 15:30:12 +02008291retry-on [list of keywords]
8292 Specify when to attempt to automatically retry a failed request
8293 May be used in sections: defaults | frontend | listen | backend
8294 yes | no | yes | yes
8295 Arguments :
8296 <keywords> is a list of keywords or HTTP status codes, each representing a
8297 type of failure event on which an attempt to retry the request
8298 is desired. Please read the notes at the bottom before changing
8299 this setting. The following keywords are supported :
8300
8301 none never retry
8302
8303 conn-failure retry when the connection or the SSL handshake failed
8304 and the request could not be sent. This is the default.
8305
8306 empty-response retry when the server connection was closed after part
8307 of the request was sent, and nothing was received from
8308 the server. This type of failure may be caused by the
8309 request timeout on the server side, poor network
8310 condition, or a server crash or restart while
8311 processing the request.
8312
Olivier Houcharde3249a92019-05-03 23:01:47 +02008313 junk-response retry when the server returned something not looking
8314 like a complete HTTP response. This includes partial
8315 responses headers as well as non-HTTP contents. It
8316 usually is a bad idea to retry on such events, which
8317 may be caused a configuration issue (wrong server port)
8318 or by the request being harmful to the server (buffer
8319 overflow attack for example).
8320
Olivier Houcharda254a372019-04-05 15:30:12 +02008321 response-timeout the server timeout stroke while waiting for the server
8322 to respond to the request. This may be caused by poor
8323 network condition, the reuse of an idle connection
8324 which has expired on the path, or by the request being
8325 extremely expensive to process. It generally is a bad
8326 idea to retry on such events on servers dealing with
8327 heavy database processing (full scans, etc) as it may
8328 amplify denial of service attacks.
8329
Olivier Houchard865d8392019-05-03 22:46:27 +02008330 0rtt-rejected retry requests which were sent over early data and were
8331 rejected by the server. These requests are generally
8332 considered to be safe to retry.
8333
Olivier Houcharda254a372019-04-05 15:30:12 +02008334 <status> any HTTP status code among "404" (Not Found), "408"
8335 (Request Timeout), "425" (Too Early), "500" (Server
8336 Error), "501" (Not Implemented), "502" (Bad Gateway),
8337 "503" (Service Unavailable), "504" (Gateway Timeout).
8338
Olivier Houchardddf0e032019-05-10 18:05:40 +02008339 all-retryable-errors
8340 retry request for any error that are considered
8341 retryable. This currently activates "conn-failure",
8342 "empty-response", "junk-response", "response-timeout",
8343 "0rtt-rejected", "500", "502", "503", and "504".
8344
Olivier Houcharda254a372019-04-05 15:30:12 +02008345 Using this directive replaces any previous settings with the new ones; it is
8346 not cumulative.
8347
8348 Please note that using anything other than "none" and "conn-failure" requires
8349 to allocate a buffer and copy the whole request into it, so it has memory and
8350 performance impacts. Requests not fitting in a single buffer will never be
8351 retried (see the global tune.bufsize setting).
8352
8353 You have to make sure the application has a replay protection mechanism built
8354 in such as a unique transaction IDs passed in requests, or that replaying the
8355 same request has no consequence, or it is very dangerous to use any retry-on
8356 value beside "conn-failure" and "none". Static file servers and caches are
8357 generally considered safe against any type of retry. Using a status code can
8358 be useful to quickly leave a server showing an abnormal behavior (out of
8359 memory, file system issues, etc), but in this case it may be a good idea to
8360 immediately redispatch the connection to another server (please see "option
8361 redispatch" for this). Last, it is important to understand that most causes
8362 of failures are the requests themselves and that retrying a request causing a
8363 server to misbehave will often make the situation even worse for this server,
8364 or for the whole service in case of redispatch.
8365
8366 Unless you know exactly how the application deals with replayed requests, you
8367 should not use this directive.
8368
8369 The default is "conn-failure".
8370
8371 See also: "retries", "option redispatch", "tune.bufsize"
8372
Willy Tarreau96d51952019-05-22 20:34:35 +02008373rspadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008374 Add a header at the end of the HTTP response
8375 May be used in sections : defaults | frontend | listen | backend
8376 no | yes | yes | yes
8377 Arguments :
8378 <string> is the complete line to be added. Any space or known delimiter
8379 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008380 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008381
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008382 <cond> is an optional matching condition built from ACLs. It makes it
8383 possible to ignore this rule when other conditions are not met.
8384
Willy Tarreau303c0352008-01-17 19:01:39 +01008385 A new line consisting in <string> followed by a line feed will be added after
8386 the last header of an HTTP response.
8387
8388 Header transformations only apply to traffic which passes through HAProxy,
8389 and not to traffic generated by HAProxy, such as health-checks or error
8390 responses.
8391
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008392 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
8393 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008394
8395
Willy Tarreau96d51952019-05-22 20:34:35 +02008396rspdel <search> [{if | unless} <cond>] (deprecated)
8397rspidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008398 Delete all headers matching a regular expression in an HTTP response
8399 May be used in sections : defaults | frontend | listen | backend
8400 no | yes | yes | yes
8401 Arguments :
8402 <search> is the regular expression applied to HTTP headers and to the
8403 response line. This is an extended regular expression, so
8404 parenthesis grouping is supported and no preliminary backslash
8405 is required. Any space or known delimiter must be escaped using
8406 a backslash ('\'). The pattern applies to a full line at a time.
8407 The "rspdel" keyword strictly matches case while "rspidel"
8408 ignores case.
8409
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008410 <cond> is an optional matching condition built from ACLs. It makes it
8411 possible to ignore this rule when other conditions are not met.
8412
Willy Tarreau303c0352008-01-17 19:01:39 +01008413 Any header line matching extended regular expression <search> in the response
8414 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008415 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01008416 client.
8417
8418 Header transformations only apply to traffic which passes through HAProxy,
8419 and not to traffic generated by HAProxy, such as health-checks or error
8420 responses. Keep in mind that header names are not case-sensitive.
8421
8422 Example :
8423 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02008424 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01008425
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008426 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
8427 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008428
8429
Willy Tarreau96d51952019-05-22 20:34:35 +02008430rspdeny <search> [{if | unless} <cond>] (deprecated)
8431rspideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008432 Block an HTTP response if a line matches a regular expression
8433 May be used in sections : defaults | frontend | listen | backend
8434 no | yes | yes | yes
8435 Arguments :
8436 <search> is the regular expression applied to HTTP headers and to the
8437 response line. This is an extended regular expression, so
8438 parenthesis grouping is supported and no preliminary backslash
8439 is required. Any space or known delimiter must be escaped using
8440 a backslash ('\'). The pattern applies to a full line at a time.
8441 The "rspdeny" keyword strictly matches case while "rspideny"
8442 ignores case.
8443
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008444 <cond> is an optional matching condition built from ACLs. It makes it
8445 possible to ignore this rule when other conditions are not met.
8446
Willy Tarreau303c0352008-01-17 19:01:39 +01008447 A response containing any line which matches extended regular expression
8448 <search> will mark the request as denied. The test applies both to the
8449 response line and to response headers. Keep in mind that header names are not
8450 case-sensitive.
8451
8452 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01008453 block the response before it reaches the client. If a response is denied, it
8454 will be replaced with an HTTP 502 error so that the client never retrieves
8455 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01008456
8457 It is easier, faster and more powerful to use ACLs to write access policies.
8458 Rspdeny should be avoided in new designs.
8459
8460 Example :
8461 # Ensure that no content type matching ms-word will leak
8462 rspideny ^Content-type:\.*/ms-word
8463
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008464 See also: "reqdeny", "acl", "block", "http-response", section 6 about
8465 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008466
8467
Willy Tarreau96d51952019-05-22 20:34:35 +02008468rsprep <search> <string> [{if | unless} <cond>] (deprecated)
8469rspirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008470 Replace a regular expression with a string in an HTTP response line
8471 May be used in sections : defaults | frontend | listen | backend
8472 no | yes | yes | yes
8473 Arguments :
8474 <search> is the regular expression applied to HTTP headers and to the
8475 response line. This is an extended regular expression, so
8476 parenthesis grouping is supported and no preliminary backslash
8477 is required. Any space or known delimiter must be escaped using
8478 a backslash ('\'). The pattern applies to a full line at a time.
8479 The "rsprep" keyword strictly matches case while "rspirep"
8480 ignores case.
8481
8482 <string> is the complete line to be added. Any space or known delimiter
8483 must be escaped using a backslash ('\'). References to matched
8484 pattern groups are possible using the common \N form, with N
8485 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008486 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008487
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008488 <cond> is an optional matching condition built from ACLs. It makes it
8489 possible to ignore this rule when other conditions are not met.
8490
Willy Tarreau303c0352008-01-17 19:01:39 +01008491 Any line matching extended regular expression <search> in the response (both
8492 the response line and header lines) will be completely replaced with
8493 <string>. Most common use of this is to rewrite Location headers.
8494
8495 Header transformations only apply to traffic which passes through HAProxy,
8496 and not to traffic generated by HAProxy, such as health-checks or error
8497 responses. Note that for increased readability, it is suggested to add enough
8498 spaces between the request and the response. Keep in mind that header names
8499 are not case-sensitive.
8500
8501 Example :
8502 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
8503 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
8504
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008505 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
8506 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008507
8508
David du Colombier486df472011-03-17 10:40:26 +01008509server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008510 Declare a server in a backend
8511 May be used in sections : defaults | frontend | listen | backend
8512 no | no | yes | yes
8513 Arguments :
8514 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01008515 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05008516 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008517
David du Colombier486df472011-03-17 10:40:26 +01008518 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
8519 resolvable hostname is supported, but this name will be resolved
8520 during start-up. Address "0.0.0.0" or "*" has a special meaning.
8521 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02008522 address as the one from the client connection. This is useful in
8523 transparent proxy architectures where the client's connection is
8524 intercepted and haproxy must forward to the original destination
8525 address. This is more or less what the "transparent" keyword does
8526 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01008527 to report statistics. Optionally, an address family prefix may be
8528 used before the address to force the family regardless of the
8529 address format, which can be useful to specify a path to a unix
8530 socket with no slash ('/'). Currently supported prefixes are :
8531 - 'ipv4@' -> address is always IPv4
8532 - 'ipv6@' -> address is always IPv6
8533 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008534 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +02008535 - 'sockpair@' -> address is the FD of a connected unix
8536 socket or of a socketpair. During a connection, the
8537 backend creates a pair of connected sockets, and passes
8538 one of them over the FD. The bind part will use the
8539 received socket as the client FD. Should be used
8540 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02008541 You may want to reference some environment variables in the
8542 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01008543 variables. The "init-addr" setting can be used to modify the way
8544 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008545
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008546 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008547 be sent to this port. If unset, the same port the client
8548 connected to will be used. The port may also be prefixed by a "+"
8549 or a "-". In this case, the server's port will be determined by
8550 adding this value to the client's port.
8551
8552 <param*> is a list of parameters for this server. The "server" keywords
8553 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008554 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008555
8556 Examples :
8557 server first 10.1.1.1:1080 cookie first check inter 1000
8558 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01008559 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02008560 server backup "${SRV_BACKUP}:1080" backup
8561 server www1_dc1 "${LAN_DC1}.101:80"
8562 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008563
Willy Tarreau55dcaf62015-09-27 15:03:15 +02008564 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
8565 sun_path length is used for the address length. Some other programs
8566 such as socat use the string length only by default. Pass the option
8567 ",unix-tightsocklen=0" to any abstract socket definition in socat to
8568 make it compatible with HAProxy's.
8569
Mark Lamourinec2247f02012-01-04 13:02:01 -05008570 See also: "default-server", "http-send-name-header" and section 5 about
8571 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008572
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02008573server-state-file-name [<file>]
8574 Set the server state file to read, load and apply to servers available in
8575 this backend. It only applies when the directive "load-server-state-from-file"
8576 is set to "local". When <file> is not provided or if this directive is not
8577 set, then backend name is used. If <file> starts with a slash '/', then it is
8578 considered as an absolute path. Otherwise, <file> is concatenated to the
8579 global directive "server-state-file-base".
8580
8581 Example: the minimal configuration below would make HAProxy look for the
8582 state server file '/etc/haproxy/states/bk':
8583
8584 global
8585 server-state-file-base /etc/haproxy/states
8586
8587 backend bk
8588 load-server-state-from-file
8589
8590 See also: "server-state-file-base", "load-server-state-from-file", and
8591 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008592
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02008593server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
8594 Set a template to initialize servers with shared parameters.
8595 The names of these servers are built from <prefix> and <num | range> parameters.
8596 May be used in sections : defaults | frontend | listen | backend
8597 no | no | yes | yes
8598
8599 Arguments:
8600 <prefix> A prefix for the server names to be built.
8601
8602 <num | range>
8603 If <num> is provided, this template initializes <num> servers
8604 with 1 up to <num> as server name suffixes. A range of numbers
8605 <num_low>-<num_high> may also be used to use <num_low> up to
8606 <num_high> as server name suffixes.
8607
8608 <fqdn> A FQDN for all the servers this template initializes.
8609
8610 <port> Same meaning as "server" <port> argument (see "server" keyword).
8611
8612 <params*>
8613 Remaining server parameters among all those supported by "server"
8614 keyword.
8615
8616 Examples:
8617 # Initializes 3 servers with srv1, srv2 and srv3 as names,
8618 # google.com as FQDN, and health-check enabled.
8619 server-template srv 1-3 google.com:80 check
8620
8621 # or
8622 server-template srv 3 google.com:80 check
8623
8624 # would be equivalent to:
8625 server srv1 google.com:80 check
8626 server srv2 google.com:80 check
8627 server srv3 google.com:80 check
8628
8629
8630
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008631source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008632source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008633source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008634 Set the source address for outgoing connections
8635 May be used in sections : defaults | frontend | listen | backend
8636 yes | no | yes | yes
8637 Arguments :
8638 <addr> is the IPv4 address HAProxy will bind to before connecting to a
8639 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01008640
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008641 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01008642 the most appropriate address to reach its destination. Optionally
8643 an address family prefix may be used before the address to force
8644 the family regardless of the address format, which can be useful
8645 to specify a path to a unix socket with no slash ('/'). Currently
8646 supported prefixes are :
8647 - 'ipv4@' -> address is always IPv4
8648 - 'ipv6@' -> address is always IPv6
8649 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008650 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02008651 You may want to reference some environment variables in the
8652 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008653
8654 <port> is an optional port. It is normally not needed but may be useful
8655 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008656 the system will select a free port. Note that port ranges are not
8657 supported in the backend. If you want to force port ranges, you
8658 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008659
8660 <addr2> is the IP address to present to the server when connections are
8661 forwarded in full transparent proxy mode. This is currently only
8662 supported on some patched Linux kernels. When this address is
8663 specified, clients connecting to the server will be presented
8664 with this address, while health checks will still use the address
8665 <addr>.
8666
8667 <port2> is the optional port to present to the server when connections
8668 are forwarded in full transparent proxy mode (see <addr2> above).
8669 The default value of zero means the system will select a free
8670 port.
8671
Willy Tarreaubce70882009-09-07 11:51:47 +02008672 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
8673 This is the name of a comma-separated header list which can
8674 contain multiple IP addresses. By default, the last occurrence is
8675 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01008676 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02008677 by previous proxy, typically Stunnel. In order to use another
8678 occurrence from the last one, please see the <occ> parameter
8679 below. When the header (or occurrence) is not found, no binding
8680 is performed so that the proxy's default IP address is used. Also
8681 keep in mind that the header name is case insensitive, as for any
8682 HTTP header.
8683
8684 <occ> is the occurrence number of a value to be used in a multi-value
8685 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008686 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02008687 address. Positive values indicate a position from the first
8688 occurrence, 1 being the first one. Negative values indicate
8689 positions relative to the last one, -1 being the last one. This
8690 is helpful for situations where an X-Forwarded-For header is set
8691 at the entry point of an infrastructure and must be used several
8692 proxy layers away. When this value is not specified, -1 is
8693 assumed. Passing a zero here disables the feature.
8694
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008695 <name> is an optional interface name to which to bind to for outgoing
8696 traffic. On systems supporting this features (currently, only
8697 Linux), this allows one to bind all traffic to the server to
8698 this interface even if it is not the one the system would select
8699 based on routing tables. This should be used with extreme care.
8700 Note that using this option requires root privileges.
8701
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008702 The "source" keyword is useful in complex environments where a specific
8703 address only is allowed to connect to the servers. It may be needed when a
8704 private address must be used through a public gateway for instance, and it is
8705 known that the system cannot determine the adequate source address by itself.
8706
8707 An extension which is available on certain patched Linux kernels may be used
8708 through the "usesrc" optional keyword. It makes it possible to connect to the
8709 servers with an IP address which does not belong to the system itself. This
8710 is called "full transparent proxy mode". For this to work, the destination
8711 servers have to route their traffic back to this address through the machine
8712 running HAProxy, and IP forwarding must generally be enabled on this machine.
8713
8714 In this "full transparent proxy" mode, it is possible to force a specific IP
8715 address to be presented to the servers. This is not much used in fact. A more
8716 common use is to tell HAProxy to present the client's IP address. For this,
8717 there are two methods :
8718
8719 - present the client's IP and port addresses. This is the most transparent
8720 mode, but it can cause problems when IP connection tracking is enabled on
8721 the machine, because a same connection may be seen twice with different
8722 states. However, this solution presents the huge advantage of not
8723 limiting the system to the 64k outgoing address+port couples, because all
8724 of the client ranges may be used.
8725
8726 - present only the client's IP address and select a spare port. This
8727 solution is still quite elegant but slightly less transparent (downstream
8728 firewalls logs will not match upstream's). It also presents the downside
8729 of limiting the number of concurrent connections to the usual 64k ports.
8730 However, since the upstream and downstream ports are different, local IP
8731 connection tracking on the machine will not be upset by the reuse of the
8732 same session.
8733
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008734 This option sets the default source for all servers in the backend. It may
8735 also be specified in a "defaults" section. Finer source address specification
8736 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008737 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008738
Baptiste Assmann91bd3372015-07-17 21:59:42 +02008739 In order to work, "usesrc" requires root privileges.
8740
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008741 Examples :
8742 backend private
8743 # Connect to the servers using our 192.168.1.200 source address
8744 source 192.168.1.200
8745
8746 backend transparent_ssl1
8747 # Connect to the SSL farm from the client's source address
8748 source 192.168.1.200 usesrc clientip
8749
8750 backend transparent_ssl2
8751 # Connect to the SSL farm from the client's source address and port
8752 # not recommended if IP conntrack is present on the local machine.
8753 source 192.168.1.200 usesrc client
8754
8755 backend transparent_ssl3
8756 # Connect to the SSL farm from the client's source address. It
8757 # is more conntrack-friendly.
8758 source 192.168.1.200 usesrc clientip
8759
8760 backend transparent_smtp
8761 # Connect to the SMTP farm from the client's source address/port
8762 # with Tproxy version 4.
8763 source 0.0.0.0 usesrc clientip
8764
Willy Tarreaubce70882009-09-07 11:51:47 +02008765 backend transparent_http
8766 # Connect to the servers using the client's IP as seen by previous
8767 # proxy.
8768 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
8769
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008770 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008771 the Linux kernel on www.balabit.com, the "bind" keyword.
8772
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01008773
Willy Tarreau844e3c52008-01-11 16:28:18 +01008774srvtimeout <timeout> (deprecated)
8775 Set the maximum inactivity time on the server side.
8776 May be used in sections : defaults | frontend | listen | backend
8777 yes | no | yes | yes
8778 Arguments :
8779 <timeout> is the timeout value specified in milliseconds by default, but
8780 can be in any other unit if the number is suffixed by the unit,
8781 as explained at the top of this document.
8782
8783 The inactivity timeout applies when the server is expected to acknowledge or
8784 send data. In HTTP mode, this timeout is particularly important to consider
8785 during the first phase of the server's response, when it has to send the
8786 headers, as it directly represents the server's processing time for the
8787 request. To find out what value to put there, it's often good to start with
8788 what would be considered as unacceptable response times, then check the logs
8789 to observe the response time distribution, and adjust the value accordingly.
8790
8791 The value is specified in milliseconds by default, but can be in any other
8792 unit if the number is suffixed by the unit, as specified at the top of this
8793 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
8794 recommended that the client timeout remains equal to the server timeout in
8795 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008796 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01008797 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +01008798 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01008799
8800 This parameter is specific to backends, but can be specified once for all in
8801 "defaults" sections. This is in fact one of the easiest solutions not to
8802 forget about it. An unspecified timeout results in an infinite timeout, which
8803 is not recommended. Such a usage is accepted and works but reports a warning
8804 during startup because it may results in accumulation of expired sessions in
8805 the system if the system's timeouts are not configured either.
8806
8807 This parameter is provided for compatibility but is currently deprecated.
8808 Please use "timeout server" instead.
8809
Willy Tarreauce887fd2012-05-12 12:50:00 +02008810 See also : "timeout server", "timeout tunnel", "timeout client" and
8811 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008812
8813
Cyril Bonté66c327d2010-10-12 00:14:37 +02008814stats admin { if | unless } <cond>
8815 Enable statistics admin level if/unless a condition is matched
8816 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008817 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008818
8819 This statement enables the statistics admin level if/unless a condition is
8820 matched.
8821
8822 The admin level allows to enable/disable servers from the web interface. By
8823 default, statistics page is read-only for security reasons.
8824
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008825 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8826 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008827 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008828
Cyril Bonté23b39d92011-02-10 22:54:44 +01008829 Currently, the POST request is limited to the buffer size minus the reserved
8830 buffer space, which means that if the list of servers is too long, the
8831 request won't be processed. It is recommended to alter few servers at a
8832 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008833
8834 Example :
8835 # statistics admin level only for localhost
8836 backend stats_localhost
8837 stats enable
8838 stats admin if LOCALHOST
8839
8840 Example :
8841 # statistics admin level always enabled because of the authentication
8842 backend stats_auth
8843 stats enable
8844 stats auth admin:AdMiN123
8845 stats admin if TRUE
8846
8847 Example :
8848 # statistics admin level depends on the authenticated user
8849 userlist stats-auth
8850 group admin users admin
8851 user admin insecure-password AdMiN123
8852 group readonly users haproxy
8853 user haproxy insecure-password haproxy
8854
8855 backend stats_auth
8856 stats enable
8857 acl AUTH http_auth(stats-auth)
8858 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8859 stats http-request auth unless AUTH
8860 stats admin if AUTH_ADMIN
8861
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008862 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8863 "bind-process", section 3.4 about userlists and section 7 about
8864 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008865
8866
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008867stats auth <user>:<passwd>
8868 Enable statistics with authentication and grant access to an account
8869 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008870 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008871 Arguments :
8872 <user> is a user name to grant access to
8873
8874 <passwd> is the cleartext password associated to this user
8875
8876 This statement enables statistics with default settings, and restricts access
8877 to declared users only. It may be repeated as many times as necessary to
8878 allow as many users as desired. When a user tries to access the statistics
8879 without a valid account, a "401 Forbidden" response will be returned so that
8880 the browser asks the user to provide a valid user and password. The real
8881 which will be returned to the browser is configurable using "stats realm".
8882
8883 Since the authentication method is HTTP Basic Authentication, the passwords
8884 circulate in cleartext on the network. Thus, it was decided that the
8885 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008886 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008887
8888 It is also possible to reduce the scope of the proxies which appear in the
8889 report using "stats scope".
8890
8891 Though this statement alone is enough to enable statistics reporting, it is
8892 recommended to set all other settings in order to avoid relying on default
8893 unobvious parameters.
8894
8895 Example :
8896 # public access (limited to this backend only)
8897 backend public_www
8898 server srv1 192.168.0.1:80
8899 stats enable
8900 stats hide-version
8901 stats scope .
8902 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008903 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008904 stats auth admin1:AdMiN123
8905 stats auth admin2:AdMiN321
8906
8907 # internal monitoring access (unlimited)
8908 backend private_monitoring
8909 stats enable
8910 stats uri /admin?stats
8911 stats refresh 5s
8912
8913 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8914
8915
8916stats enable
8917 Enable statistics reporting with default settings
8918 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008919 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008920 Arguments : none
8921
8922 This statement enables statistics reporting with default settings defined
8923 at build time. Unless stated otherwise, these settings are used :
8924 - stats uri : /haproxy?stats
8925 - stats realm : "HAProxy Statistics"
8926 - stats auth : no authentication
8927 - stats scope : no restriction
8928
8929 Though this statement alone is enough to enable statistics reporting, it is
8930 recommended to set all other settings in order to avoid relying on default
8931 unobvious parameters.
8932
8933 Example :
8934 # public access (limited to this backend only)
8935 backend public_www
8936 server srv1 192.168.0.1:80
8937 stats enable
8938 stats hide-version
8939 stats scope .
8940 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008941 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008942 stats auth admin1:AdMiN123
8943 stats auth admin2:AdMiN321
8944
8945 # internal monitoring access (unlimited)
8946 backend private_monitoring
8947 stats enable
8948 stats uri /admin?stats
8949 stats refresh 5s
8950
8951 See also : "stats auth", "stats realm", "stats uri"
8952
8953
Willy Tarreaud63335a2010-02-26 12:56:52 +01008954stats hide-version
8955 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008956 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008957 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008958 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008959
Willy Tarreaud63335a2010-02-26 12:56:52 +01008960 By default, the stats page reports some useful status information along with
8961 the statistics. Among them is HAProxy's version. However, it is generally
8962 considered dangerous to report precise version to anyone, as it can help them
8963 target known weaknesses with specific attacks. The "stats hide-version"
8964 statement removes the version from the statistics report. This is recommended
8965 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008966
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008967 Though this statement alone is enough to enable statistics reporting, it is
8968 recommended to set all other settings in order to avoid relying on default
8969 unobvious parameters.
8970
Willy Tarreaud63335a2010-02-26 12:56:52 +01008971 Example :
8972 # public access (limited to this backend only)
8973 backend public_www
8974 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008975 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008976 stats hide-version
8977 stats scope .
8978 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008979 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008980 stats auth admin1:AdMiN123
8981 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008982
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008983 # internal monitoring access (unlimited)
8984 backend private_monitoring
8985 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008986 stats uri /admin?stats
8987 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008988
Willy Tarreaud63335a2010-02-26 12:56:52 +01008989 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008990
Willy Tarreau983e01e2010-01-11 18:42:06 +01008991
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008992stats http-request { allow | deny | auth [realm <realm>] }
8993 [ { if | unless } <condition> ]
8994 Access control for statistics
8995
8996 May be used in sections: defaults | frontend | listen | backend
8997 no | no | yes | yes
8998
8999 As "http-request", these set of options allow to fine control access to
9000 statistics. Each option may be followed by if/unless and acl.
9001 First option with matched condition (or option without condition) is final.
9002 For "deny" a 403 error will be returned, for "allow" normal processing is
9003 performed, for "auth" a 401/407 error code is returned so the client
9004 should be asked to enter a username and password.
9005
9006 There is no fixed limit to the number of http-request statements per
9007 instance.
9008
9009 See also : "http-request", section 3.4 about userlists and section 7
9010 about ACL usage.
9011
9012
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009013stats realm <realm>
9014 Enable statistics and set authentication realm
9015 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009016 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009017 Arguments :
9018 <realm> is the name of the HTTP Basic Authentication realm reported to
9019 the browser. The browser uses it to display it in the pop-up
9020 inviting the user to enter a valid username and password.
9021
9022 The realm is read as a single word, so any spaces in it should be escaped
9023 using a backslash ('\').
9024
9025 This statement is useful only in conjunction with "stats auth" since it is
9026 only related to authentication.
9027
9028 Though this statement alone is enough to enable statistics reporting, it is
9029 recommended to set all other settings in order to avoid relying on default
9030 unobvious parameters.
9031
9032 Example :
9033 # public access (limited to this backend only)
9034 backend public_www
9035 server srv1 192.168.0.1:80
9036 stats enable
9037 stats hide-version
9038 stats scope .
9039 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009040 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009041 stats auth admin1:AdMiN123
9042 stats auth admin2:AdMiN321
9043
9044 # internal monitoring access (unlimited)
9045 backend private_monitoring
9046 stats enable
9047 stats uri /admin?stats
9048 stats refresh 5s
9049
9050 See also : "stats auth", "stats enable", "stats uri"
9051
9052
9053stats refresh <delay>
9054 Enable statistics with automatic refresh
9055 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009056 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009057 Arguments :
9058 <delay> is the suggested refresh delay, specified in seconds, which will
9059 be returned to the browser consulting the report page. While the
9060 browser is free to apply any delay, it will generally respect it
9061 and refresh the page this every seconds. The refresh interval may
9062 be specified in any other non-default time unit, by suffixing the
9063 unit after the value, as explained at the top of this document.
9064
9065 This statement is useful on monitoring displays with a permanent page
9066 reporting the load balancer's activity. When set, the HTML report page will
9067 include a link "refresh"/"stop refresh" so that the user can select whether
9068 he wants automatic refresh of the page or not.
9069
9070 Though this statement alone is enough to enable statistics reporting, it is
9071 recommended to set all other settings in order to avoid relying on default
9072 unobvious parameters.
9073
9074 Example :
9075 # public access (limited to this backend only)
9076 backend public_www
9077 server srv1 192.168.0.1:80
9078 stats enable
9079 stats hide-version
9080 stats scope .
9081 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009082 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009083 stats auth admin1:AdMiN123
9084 stats auth admin2:AdMiN321
9085
9086 # internal monitoring access (unlimited)
9087 backend private_monitoring
9088 stats enable
9089 stats uri /admin?stats
9090 stats refresh 5s
9091
9092 See also : "stats auth", "stats enable", "stats realm", "stats uri"
9093
9094
9095stats scope { <name> | "." }
9096 Enable statistics and limit access scope
9097 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009098 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009099 Arguments :
9100 <name> is the name of a listen, frontend or backend section to be
9101 reported. The special name "." (a single dot) designates the
9102 section in which the statement appears.
9103
9104 When this statement is specified, only the sections enumerated with this
9105 statement will appear in the report. All other ones will be hidden. This
9106 statement may appear as many times as needed if multiple sections need to be
9107 reported. Please note that the name checking is performed as simple string
9108 comparisons, and that it is never checked that a give section name really
9109 exists.
9110
9111 Though this statement alone is enough to enable statistics reporting, it is
9112 recommended to set all other settings in order to avoid relying on default
9113 unobvious parameters.
9114
9115 Example :
9116 # public access (limited to this backend only)
9117 backend public_www
9118 server srv1 192.168.0.1:80
9119 stats enable
9120 stats hide-version
9121 stats scope .
9122 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009123 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009124 stats auth admin1:AdMiN123
9125 stats auth admin2:AdMiN321
9126
9127 # internal monitoring access (unlimited)
9128 backend private_monitoring
9129 stats enable
9130 stats uri /admin?stats
9131 stats refresh 5s
9132
9133 See also : "stats auth", "stats enable", "stats realm", "stats uri"
9134
Willy Tarreaud63335a2010-02-26 12:56:52 +01009135
Willy Tarreauc9705a12010-07-27 20:05:50 +02009136stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01009137 Enable reporting of a description on the statistics page.
9138 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009139 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01009140
Willy Tarreauc9705a12010-07-27 20:05:50 +02009141 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01009142 description from global section is automatically used instead.
9143
9144 This statement is useful for users that offer shared services to their
9145 customers, where node or description should be different for each customer.
9146
9147 Though this statement alone is enough to enable statistics reporting, it is
9148 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01009149 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009150
9151 Example :
9152 # internal monitoring access (unlimited)
9153 backend private_monitoring
9154 stats enable
9155 stats show-desc Master node for Europe, Asia, Africa
9156 stats uri /admin?stats
9157 stats refresh 5s
9158
9159 See also: "show-node", "stats enable", "stats uri" and "description" in
9160 global section.
9161
9162
9163stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02009164 Enable reporting additional information on the statistics page
9165 May be used in sections : defaults | frontend | listen | backend
9166 yes | yes | yes | yes
9167 Arguments : none
9168
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009169 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01009170 - cap: capabilities (proxy)
9171 - mode: one of tcp, http or health (proxy)
9172 - id: SNMP ID (proxy, socket, server)
9173 - IP (socket, server)
9174 - cookie (backend, server)
9175
9176 Though this statement alone is enough to enable statistics reporting, it is
9177 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01009178 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009179
9180 See also: "stats enable", "stats uri".
9181
9182
9183stats show-node [ <name> ]
9184 Enable reporting of a host name on the statistics page.
9185 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009186 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01009187 Arguments:
9188 <name> is an optional name to be reported. If unspecified, the
9189 node name from global section is automatically used instead.
9190
9191 This statement is useful for users that offer shared services to their
9192 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01009193 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009194
9195 Though this statement alone is enough to enable statistics reporting, it is
9196 recommended to set all other settings in order to avoid relying on default
9197 unobvious parameters.
9198
9199 Example:
9200 # internal monitoring access (unlimited)
9201 backend private_monitoring
9202 stats enable
9203 stats show-node Europe-1
9204 stats uri /admin?stats
9205 stats refresh 5s
9206
9207 See also: "show-desc", "stats enable", "stats uri", and "node" in global
9208 section.
9209
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009210
9211stats uri <prefix>
9212 Enable statistics and define the URI prefix to access them
9213 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009214 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009215 Arguments :
9216 <prefix> is the prefix of any URI which will be redirected to stats. This
9217 prefix may contain a question mark ('?') to indicate part of a
9218 query string.
9219
9220 The statistics URI is intercepted on the relayed traffic, so it appears as a
9221 page within the normal application. It is strongly advised to ensure that the
9222 selected URI will never appear in the application, otherwise it will never be
9223 possible to reach it in the application.
9224
9225 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009226 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009227 It is generally a good idea to include a question mark in the URI so that
9228 intermediate proxies refrain from caching the results. Also, since any string
9229 beginning with the prefix will be accepted as a stats request, the question
9230 mark helps ensuring that no valid URI will begin with the same words.
9231
9232 It is sometimes very convenient to use "/" as the URI prefix, and put that
9233 statement in a "listen" instance of its own. That makes it easy to dedicate
9234 an address or a port to statistics only.
9235
9236 Though this statement alone is enough to enable statistics reporting, it is
9237 recommended to set all other settings in order to avoid relying on default
9238 unobvious parameters.
9239
9240 Example :
9241 # public access (limited to this backend only)
9242 backend public_www
9243 server srv1 192.168.0.1:80
9244 stats enable
9245 stats hide-version
9246 stats scope .
9247 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009248 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009249 stats auth admin1:AdMiN123
9250 stats auth admin2:AdMiN321
9251
9252 # internal monitoring access (unlimited)
9253 backend private_monitoring
9254 stats enable
9255 stats uri /admin?stats
9256 stats refresh 5s
9257
9258 See also : "stats auth", "stats enable", "stats realm"
9259
9260
Willy Tarreaud63335a2010-02-26 12:56:52 +01009261stick match <pattern> [table <table>] [{if | unless} <cond>]
9262 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009263 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01009264 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009265
9266 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009267 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009268 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009269 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009270 stickiness table. This rule is mandatory.
9271
9272 <table> is an optional stickiness table name. If unspecified, the same
9273 backend's table is used. A stickiness table is declared using
9274 the "stick-table" statement.
9275
9276 <cond> is an optional matching condition. It makes it possible to match
9277 on a certain criterion only when other conditions are met (or
9278 not met). For instance, it could be used to match on a source IP
9279 address except when a request passes through a known proxy, in
9280 which case we'd match on a header containing that IP address.
9281
9282 Some protocols or applications require complex stickiness rules and cannot
9283 always simply rely on cookies nor hashing. The "stick match" statement
9284 describes a rule to extract the stickiness criterion from an incoming request
9285 or connection. See section 7 for a complete list of possible patterns and
9286 transformation rules.
9287
9288 The table has to be declared using the "stick-table" statement. It must be of
9289 a type compatible with the pattern. By default it is the one which is present
9290 in the same backend. It is possible to share a table with other backends by
9291 referencing it using the "table" keyword. If another table is referenced,
9292 the server's ID inside the backends are used. By default, all server IDs
9293 start at 1 in each backend, so the server ordering is enough. But in case of
9294 doubt, it is highly recommended to force server IDs using their "id" setting.
9295
9296 It is possible to restrict the conditions where a "stick match" statement
9297 will apply, using "if" or "unless" followed by a condition. See section 7 for
9298 ACL based conditions.
9299
9300 There is no limit on the number of "stick match" statements. The first that
9301 applies and matches will cause the request to be directed to the same server
9302 as was used for the request which created the entry. That way, multiple
9303 matches can be used as fallbacks.
9304
9305 The stick rules are checked after the persistence cookies, so they will not
9306 affect stickiness if a cookie has already been used to select a server. That
9307 way, it becomes very easy to insert cookies and match on IP addresses in
9308 order to maintain stickiness between HTTP and HTTPS.
9309
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009310 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9311 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009312 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009313
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009314 Example :
9315 # forward SMTP users to the same server they just used for POP in the
9316 # last 30 minutes
9317 backend pop
9318 mode tcp
9319 balance roundrobin
9320 stick store-request src
9321 stick-table type ip size 200k expire 30m
9322 server s1 192.168.1.1:110
9323 server s2 192.168.1.1:110
9324
9325 backend smtp
9326 mode tcp
9327 balance roundrobin
9328 stick match src table pop
9329 server s1 192.168.1.1:25
9330 server s2 192.168.1.1:25
9331
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009332 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009333 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009334
9335
9336stick on <pattern> [table <table>] [{if | unless} <condition>]
9337 Define a request pattern to associate a user to a server
9338 May be used in sections : defaults | frontend | listen | backend
9339 no | no | yes | yes
9340
9341 Note : This form is exactly equivalent to "stick match" followed by
9342 "stick store-request", all with the same arguments. Please refer
9343 to both keywords for details. It is only provided as a convenience
9344 for writing more maintainable configurations.
9345
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009346 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9347 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009348 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009349
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009350 Examples :
9351 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01009352 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009353
9354 # ...is strictly equivalent to this one :
9355 stick match src table pop if !localhost
9356 stick store-request src table pop if !localhost
9357
9358
9359 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
9360 # well as HTTP without cookie. Share the same table between both accesses.
9361 backend http
9362 mode http
9363 balance roundrobin
9364 stick on src table https
9365 cookie SRV insert indirect nocache
9366 server s1 192.168.1.1:80 cookie s1
9367 server s2 192.168.1.1:80 cookie s2
9368
9369 backend https
9370 mode tcp
9371 balance roundrobin
9372 stick-table type ip size 200k expire 30m
9373 stick on src
9374 server s1 192.168.1.1:443
9375 server s2 192.168.1.1:443
9376
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009377 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009378
9379
9380stick store-request <pattern> [table <table>] [{if | unless} <condition>]
9381 Define a request pattern used to create an entry in a stickiness table
9382 May be used in sections : defaults | frontend | listen | backend
9383 no | no | yes | yes
9384
9385 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009386 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009387 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009388 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009389 server is selected.
9390
9391 <table> is an optional stickiness table name. If unspecified, the same
9392 backend's table is used. A stickiness table is declared using
9393 the "stick-table" statement.
9394
9395 <cond> is an optional storage condition. It makes it possible to store
9396 certain criteria only when some conditions are met (or not met).
9397 For instance, it could be used to store the source IP address
9398 except when the request passes through a known proxy, in which
9399 case we'd store a converted form of a header containing that IP
9400 address.
9401
9402 Some protocols or applications require complex stickiness rules and cannot
9403 always simply rely on cookies nor hashing. The "stick store-request" statement
9404 describes a rule to decide what to extract from the request and when to do
9405 it, in order to store it into a stickiness table for further requests to
9406 match it using the "stick match" statement. Obviously the extracted part must
9407 make sense and have a chance to be matched in a further request. Storing a
9408 client's IP address for instance often makes sense. Storing an ID found in a
9409 URL parameter also makes sense. Storing a source port will almost never make
9410 any sense because it will be randomly matched. See section 7 for a complete
9411 list of possible patterns and transformation rules.
9412
9413 The table has to be declared using the "stick-table" statement. It must be of
9414 a type compatible with the pattern. By default it is the one which is present
9415 in the same backend. It is possible to share a table with other backends by
9416 referencing it using the "table" keyword. If another table is referenced,
9417 the server's ID inside the backends are used. By default, all server IDs
9418 start at 1 in each backend, so the server ordering is enough. But in case of
9419 doubt, it is highly recommended to force server IDs using their "id" setting.
9420
9421 It is possible to restrict the conditions where a "stick store-request"
9422 statement will apply, using "if" or "unless" followed by a condition. This
9423 condition will be evaluated while parsing the request, so any criteria can be
9424 used. See section 7 for ACL based conditions.
9425
9426 There is no limit on the number of "stick store-request" statements, but
9427 there is a limit of 8 simultaneous stores per request or response. This
9428 makes it possible to store up to 8 criteria, all extracted from either the
9429 request or the response, regardless of the number of rules. Only the 8 first
9430 ones which match will be kept. Using this, it is possible to feed multiple
9431 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009432 another protocol or access method. Using multiple store-request rules with
9433 the same table is possible and may be used to find the best criterion to rely
9434 on, by arranging the rules by decreasing preference order. Only the first
9435 extracted criterion for a given table will be stored. All subsequent store-
9436 request rules referencing the same table will be skipped and their ACLs will
9437 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009438
9439 The "store-request" rules are evaluated once the server connection has been
9440 established, so that the table will contain the real server that processed
9441 the request.
9442
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009443 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9444 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009445 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009446
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009447 Example :
9448 # forward SMTP users to the same server they just used for POP in the
9449 # last 30 minutes
9450 backend pop
9451 mode tcp
9452 balance roundrobin
9453 stick store-request src
9454 stick-table type ip size 200k expire 30m
9455 server s1 192.168.1.1:110
9456 server s2 192.168.1.1:110
9457
9458 backend smtp
9459 mode tcp
9460 balance roundrobin
9461 stick match src table pop
9462 server s1 192.168.1.1:25
9463 server s2 192.168.1.1:25
9464
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009465 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009466 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009467
9468
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009469stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02009470 size <size> [expire <expire>] [nopurge] [peers <peersect>]
9471 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08009472 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009473 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009474 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009475
9476 Arguments :
9477 ip a table declared with "type ip" will only store IPv4 addresses.
9478 This form is very compact (about 50 bytes per entry) and allows
9479 very fast entry lookup and stores with almost no overhead. This
9480 is mainly used to store client source IP addresses.
9481
David du Colombier9a6d3c92011-03-17 10:40:24 +01009482 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
9483 This form is very compact (about 60 bytes per entry) and allows
9484 very fast entry lookup and stores with almost no overhead. This
9485 is mainly used to store client source IP addresses.
9486
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009487 integer a table declared with "type integer" will store 32bit integers
9488 which can represent a client identifier found in a request for
9489 instance.
9490
9491 string a table declared with "type string" will store substrings of up
9492 to <len> characters. If the string provided by the pattern
9493 extractor is larger than <len>, it will be truncated before
9494 being stored. During matching, at most <len> characters will be
9495 compared between the string in the table and the extracted
9496 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009497 to 32 characters.
9498
9499 binary a table declared with "type binary" will store binary blocks
9500 of <len> bytes. If the block provided by the pattern
9501 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02009502 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009503 is shorter than <len>, it will be padded by 0. When not
9504 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009505
9506 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009507 "string" type table (See type "string" above). Or the number
9508 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009509 changing this parameter as memory usage will proportionally
9510 increase.
9511
9512 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01009513 value directly impacts memory usage. Count approximately
9514 50 bytes per entry, plus the size of a string if any. The size
9515 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009516
9517 [nopurge] indicates that we refuse to purge older entries when the table
9518 is full. When not specified and the table is full when haproxy
9519 wants to store an entry in it, it will flush a few of the oldest
9520 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009521 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009522 be desirable to refuse new entries instead of purging the older
9523 ones. That may be the case when the amount of data to store is
9524 far above the hardware limits and we prefer not to offer access
9525 to new clients than to reject the ones already connected. When
9526 using this parameter, be sure to properly set the "expire"
9527 parameter (see below).
9528
Emeric Brunf099e792010-09-27 12:05:28 +02009529 <peersect> is the name of the peers section to use for replication. Entries
9530 which associate keys to server IDs are kept synchronized with
9531 the remote peers declared in this section. All entries are also
9532 automatically learned from the local peer (old process) during a
9533 soft restart.
9534
Willy Tarreau1abc6732015-05-01 19:21:02 +02009535 NOTE : each peers section may be referenced only by tables
9536 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009537
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009538 <expire> defines the maximum duration of an entry in the table since it
9539 was last created, refreshed or matched. The expiration delay is
9540 defined using the standard time format, similarly as the various
9541 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009542 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02009543 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009544 be removed once full. Be sure not to use the "nopurge" parameter
9545 if not expiration delay is specified.
9546
Willy Tarreau08d5f982010-06-06 13:34:54 +02009547 <data_type> is used to store additional information in the stick-table. This
9548 may be used by ACLs in order to control various criteria related
9549 to the activity of the client matching the stick-table. For each
9550 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02009551 that the additional data can fit. Several data types may be
9552 stored with an entry. Multiple data types may be specified after
9553 the "store" keyword, as a comma-separated list. Alternatively,
9554 it is possible to repeat the "store" keyword followed by one or
9555 several data types. Except for the "server_id" type which is
9556 automatically detected and enabled, all data types must be
9557 explicitly declared to be stored. If an ACL references a data
9558 type which is not stored, the ACL will simply not match. Some
9559 data types require an argument which must be passed just after
9560 the type between parenthesis. See below for the supported data
9561 types and their arguments.
9562
9563 The data types that can be stored with an entry are the following :
9564 - server_id : this is an integer which holds the numeric ID of the server a
9565 request was assigned to. It is used by the "stick match", "stick store",
9566 and "stick on" rules. It is automatically enabled when referenced.
9567
9568 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
9569 integer which may be used for anything. Most of the time it will be used
9570 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01009571 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009572
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009573 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
9574 over a period. It is a positive 32-bit integer integer which may be used
9575 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01009576 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009577 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01009578 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009579
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009580 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
9581 integer which may be used for anything. Most of the time it will be used
9582 to put a special tag on some entries, for instance to note that a
9583 specific behavior was detected and must be known for future matches.
9584
9585 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
9586 over a period. It is a positive 32-bit integer integer which may be used
9587 for anything. Just like <gpc1>, it counts events, but instead of keeping
9588 a cumulative number, it maintains the rate at which the counter is
9589 incremented. Most of the time it will be used to measure the frequency of
9590 occurrence of certain events (e.g. requests to a specific URL).
9591
Willy Tarreauc9705a12010-07-27 20:05:50 +02009592 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
9593 the absolute number of connections received from clients which matched
9594 this entry. It does not mean the connections were accepted, just that
9595 they were received.
9596
9597 - conn_cur : Current Connections. It is a positive 32-bit integer which
9598 stores the concurrent connection counts for the entry. It is incremented
9599 once an incoming connection matches the entry, and decremented once the
9600 connection leaves. That way it is possible to know at any time the exact
9601 number of concurrent connections for an entry.
9602
9603 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9604 integer parameter <period> which indicates in milliseconds the length
9605 of the period over which the average is measured. It reports the average
9606 incoming connection rate over that period, in connections per period. The
9607 result is an integer which can be matched using ACLs.
9608
9609 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
9610 the absolute number of sessions received from clients which matched this
9611 entry. A session is a connection that was accepted by the layer 4 rules.
9612
9613 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9614 integer parameter <period> which indicates in milliseconds the length
9615 of the period over which the average is measured. It reports the average
9616 incoming session rate over that period, in sessions per period. The
9617 result is an integer which can be matched using ACLs.
9618
9619 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
9620 counts the absolute number of HTTP requests received from clients which
9621 matched this entry. It does not matter whether they are valid requests or
9622 not. Note that this is different from sessions when keep-alive is used on
9623 the client side.
9624
9625 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9626 integer parameter <period> which indicates in milliseconds the length
9627 of the period over which the average is measured. It reports the average
9628 HTTP request rate over that period, in requests per period. The result is
9629 an integer which can be matched using ACLs. It does not matter whether
9630 they are valid requests or not. Note that this is different from sessions
9631 when keep-alive is used on the client side.
9632
9633 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
9634 counts the absolute number of HTTP requests errors induced by clients
9635 which matched this entry. Errors are counted on invalid and truncated
9636 requests, as well as on denied or tarpitted requests, and on failed
9637 authentications. If the server responds with 4xx, then the request is
9638 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01009639 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02009640
9641 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9642 integer parameter <period> which indicates in milliseconds the length
9643 of the period over which the average is measured. It reports the average
9644 HTTP request error rate over that period, in requests per period (see
9645 http_err_cnt above for what is accounted as an error). The result is an
9646 integer which can be matched using ACLs.
9647
9648 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009649 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02009650 which matched this entry. Headers are included in the count. This may be
9651 used to limit abuse of upload features on photo or video servers.
9652
9653 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9654 integer parameter <period> which indicates in milliseconds the length
9655 of the period over which the average is measured. It reports the average
9656 incoming bytes rate over that period, in bytes per period. It may be used
9657 to detect users which upload too much and too fast. Warning: with large
9658 uploads, it is possible that the amount of uploaded data will be counted
9659 once upon termination, thus causing spikes in the average transfer speed
9660 instead of having a smooth one. This may partially be smoothed with
9661 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
9662 recommended for better fairness.
9663
9664 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009665 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02009666 matched this entry. Headers are included in the count. This may be used
9667 to limit abuse of bots sucking the whole site.
9668
9669 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
9670 an integer parameter <period> which indicates in milliseconds the length
9671 of the period over which the average is measured. It reports the average
9672 outgoing bytes rate over that period, in bytes per period. It may be used
9673 to detect users which download too much and too fast. Warning: with large
9674 transfers, it is possible that the amount of transferred data will be
9675 counted once upon termination, thus causing spikes in the average
9676 transfer speed instead of having a smooth one. This may partially be
9677 smoothed with "option contstats" though this is not perfect yet. Use of
9678 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02009679
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009680 There is only one stick-table per proxy. At the moment of writing this doc,
9681 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009682 to be required, simply create a dummy backend with a stick-table in it and
9683 reference it.
9684
9685 It is important to understand that stickiness based on learning information
9686 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01009687 lost upon restart unless peers are properly configured to transfer such
9688 information upon restart (recommended). In general it can be good as a
9689 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009690
Willy Tarreauc9705a12010-07-27 20:05:50 +02009691 Last, memory requirements may be important when storing many data types.
9692 Indeed, storing all indicators above at once in each entry requires 116 bytes
9693 per entry, or 116 MB for a 1-million entries table. This is definitely not
9694 something that can be ignored.
9695
9696 Example:
9697 # Keep track of counters of up to 1 million IP addresses over 5 minutes
9698 # and store a general purpose counter and the average connection rate
9699 # computed over a sliding window of 30 seconds.
9700 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
9701
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009702 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01009703 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009704
9705
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009706stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01009707 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009708 May be used in sections : defaults | frontend | listen | backend
9709 no | no | yes | yes
9710
9711 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009712 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009713 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01009714 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009715 server is selected.
9716
9717 <table> is an optional stickiness table name. If unspecified, the same
9718 backend's table is used. A stickiness table is declared using
9719 the "stick-table" statement.
9720
9721 <cond> is an optional storage condition. It makes it possible to store
9722 certain criteria only when some conditions are met (or not met).
9723 For instance, it could be used to store the SSL session ID only
9724 when the response is a SSL server hello.
9725
9726 Some protocols or applications require complex stickiness rules and cannot
9727 always simply rely on cookies nor hashing. The "stick store-response"
9728 statement describes a rule to decide what to extract from the response and
9729 when to do it, in order to store it into a stickiness table for further
9730 requests to match it using the "stick match" statement. Obviously the
9731 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009732 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009733 See section 7 for a complete list of possible patterns and transformation
9734 rules.
9735
9736 The table has to be declared using the "stick-table" statement. It must be of
9737 a type compatible with the pattern. By default it is the one which is present
9738 in the same backend. It is possible to share a table with other backends by
9739 referencing it using the "table" keyword. If another table is referenced,
9740 the server's ID inside the backends are used. By default, all server IDs
9741 start at 1 in each backend, so the server ordering is enough. But in case of
9742 doubt, it is highly recommended to force server IDs using their "id" setting.
9743
9744 It is possible to restrict the conditions where a "stick store-response"
9745 statement will apply, using "if" or "unless" followed by a condition. This
9746 condition will be evaluated while parsing the response, so any criteria can
9747 be used. See section 7 for ACL based conditions.
9748
9749 There is no limit on the number of "stick store-response" statements, but
9750 there is a limit of 8 simultaneous stores per request or response. This
9751 makes it possible to store up to 8 criteria, all extracted from either the
9752 request or the response, regardless of the number of rules. Only the 8 first
9753 ones which match will be kept. Using this, it is possible to feed multiple
9754 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009755 another protocol or access method. Using multiple store-response rules with
9756 the same table is possible and may be used to find the best criterion to rely
9757 on, by arranging the rules by decreasing preference order. Only the first
9758 extracted criterion for a given table will be stored. All subsequent store-
9759 response rules referencing the same table will be skipped and their ACLs will
9760 not be evaluated. However, even if a store-request rule references a table, a
9761 store-response rule may also use the same table. This means that each table
9762 may learn exactly one element from the request and one element from the
9763 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009764
9765 The table will contain the real server that processed the request.
9766
9767 Example :
9768 # Learn SSL session ID from both request and response and create affinity.
9769 backend https
9770 mode tcp
9771 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02009772 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009773 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009774
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009775 acl clienthello req_ssl_hello_type 1
9776 acl serverhello rep_ssl_hello_type 2
9777
9778 # use tcp content accepts to detects ssl client and server hello.
9779 tcp-request inspect-delay 5s
9780 tcp-request content accept if clienthello
9781
9782 # no timeout on response inspect delay by default.
9783 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009784
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009785 # SSL session ID (SSLID) may be present on a client or server hello.
9786 # Its length is coded on 1 byte at offset 43 and its value starts
9787 # at offset 44.
9788
9789 # Match and learn on request if client hello.
9790 stick on payload_lv(43,1) if clienthello
9791
9792 # Learn on response if server hello.
9793 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02009794
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009795 server s1 192.168.1.1:443
9796 server s2 192.168.1.1:443
9797
9798 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
9799 extraction.
9800
9801
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009802tcp-check connect [params*]
9803 Opens a new connection
9804 May be used in sections: defaults | frontend | listen | backend
9805 no | no | yes | yes
9806
9807 When an application lies on more than a single TCP port or when HAProxy
9808 load-balance many services in a single backend, it makes sense to probe all
9809 the services individually before considering a server as operational.
9810
9811 When there are no TCP port configured on the server line neither server port
9812 directive, then the 'tcp-check connect port <port>' must be the first step
9813 of the sequence.
9814
9815 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9816 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9817 do.
9818
9819 Parameters :
9820 They are optional and can be used to describe how HAProxy should open and
9821 use the TCP connection.
9822
9823 port if not set, check port or server port is used.
9824 It tells HAProxy where to open the connection to.
9825 <port> must be a valid TCP port source integer, from 1 to 65535.
9826
9827 send-proxy send a PROXY protocol string
9828
9829 ssl opens a ciphered connection
9830
9831 Examples:
9832 # check HTTP and HTTPs services on a server.
9833 # first open port 80 thanks to server line port directive, then
9834 # tcp-check opens port 443, ciphered and run a request on it:
9835 option tcp-check
9836 tcp-check connect
9837 tcp-check send GET\ /\ HTTP/1.0\r\n
9838 tcp-check send Host:\ haproxy.1wt.eu\r\n
9839 tcp-check send \r\n
9840 tcp-check expect rstring (2..|3..)
9841 tcp-check connect port 443 ssl
9842 tcp-check send GET\ /\ HTTP/1.0\r\n
9843 tcp-check send Host:\ haproxy.1wt.eu\r\n
9844 tcp-check send \r\n
9845 tcp-check expect rstring (2..|3..)
9846 server www 10.0.0.1 check port 80
9847
9848 # check both POP and IMAP from a single server:
9849 option tcp-check
9850 tcp-check connect port 110
9851 tcp-check expect string +OK\ POP3\ ready
9852 tcp-check connect port 143
9853 tcp-check expect string *\ OK\ IMAP4\ ready
9854 server mail 10.0.0.1 check
9855
9856 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9857
9858
9859tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009860 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009861 May be used in sections: defaults | frontend | listen | backend
9862 no | no | yes | yes
9863
9864 Arguments :
9865 <match> is a keyword indicating how to look for a specific pattern in the
9866 response. The keyword may be one of "string", "rstring" or
9867 binary.
9868 The keyword may be preceded by an exclamation mark ("!") to negate
9869 the match. Spaces are allowed between the exclamation mark and the
9870 keyword. See below for more details on the supported keywords.
9871
9872 <pattern> is the pattern to look for. It may be a string or a regular
9873 expression. If the pattern contains spaces, they must be escaped
9874 with the usual backslash ('\').
9875 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009876 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009877 two digits will represent a byte. The hexadecimal digits may be
9878 used upper or lower case.
9879
9880
9881 The available matches are intentionally similar to their http-check cousins :
9882
9883 string <string> : test the exact string matches in the response buffer.
9884 A health check response will be considered valid if the
9885 response's buffer contains this exact string. If the
9886 "string" keyword is prefixed with "!", then the response
9887 will be considered invalid if the body contains this
9888 string. This can be used to look for a mandatory pattern
9889 in a protocol response, or to detect a failure when a
9890 specific error appears in a protocol banner.
9891
9892 rstring <regex> : test a regular expression on the response buffer.
9893 A health check response will be considered valid if the
9894 response's buffer matches this expression. If the
9895 "rstring" keyword is prefixed with "!", then the response
9896 will be considered invalid if the body matches the
9897 expression.
9898
9899 binary <hexstring> : test the exact string in its hexadecimal form matches
9900 in the response buffer. A health check response will
9901 be considered valid if the response's buffer contains
9902 this exact hexadecimal string.
9903 Purpose is to match data on binary protocols.
9904
9905 It is important to note that the responses will be limited to a certain size
9906 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9907 Thus, too large responses may not contain the mandatory pattern when using
9908 "string", "rstring" or binary. If a large response is absolutely required, it
9909 is possible to change the default max size by setting the global variable.
9910 However, it is worth keeping in mind that parsing very large responses can
9911 waste some CPU cycles, especially when regular expressions are used, and that
9912 it is always better to focus the checks on smaller resources. Also, in its
9913 current state, the check will not find any string nor regex past a null
9914 character in the response. Similarly it is not possible to request matching
9915 the null character.
9916
9917 Examples :
9918 # perform a POP check
9919 option tcp-check
9920 tcp-check expect string +OK\ POP3\ ready
9921
9922 # perform an IMAP check
9923 option tcp-check
9924 tcp-check expect string *\ OK\ IMAP4\ ready
9925
9926 # look for the redis master server
9927 option tcp-check
9928 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009929 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009930 tcp-check send info\ replication\r\n
9931 tcp-check expect string role:master
9932 tcp-check send QUIT\r\n
9933 tcp-check expect string +OK
9934
9935
9936 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9937 "tcp-check send-binary", "http-check expect", tune.chksize
9938
9939
9940tcp-check send <data>
9941 Specify a string to be sent as a question during a generic health check
9942 May be used in sections: defaults | frontend | listen | backend
9943 no | no | yes | yes
9944
9945 <data> : the data to be sent as a question during a generic health check
9946 session. For now, <data> must be a string.
9947
9948 Examples :
9949 # look for the redis master server
9950 option tcp-check
9951 tcp-check send info\ replication\r\n
9952 tcp-check expect string role:master
9953
9954 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9955 "tcp-check send-binary", tune.chksize
9956
9957
Davor Ocelice9ed2812017-12-25 17:49:28 +01009958tcp-check send-binary <hexstring>
9959 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009960 tcp health check
9961 May be used in sections: defaults | frontend | listen | backend
9962 no | no | yes | yes
9963
9964 <data> : the data to be sent as a question during a generic health check
9965 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009966 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009967 response buffer. A health check response will be considered
9968 valid if the response's buffer contains this exact
9969 hexadecimal string.
9970 Purpose is to send binary data to ask on binary protocols.
9971
9972 Examples :
9973 # redis check in binary
9974 option tcp-check
9975 tcp-check send-binary 50494e470d0a # PING\r\n
9976 tcp-check expect binary 2b504F4e47 # +PONG
9977
9978
9979 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9980 "tcp-check send", tune.chksize
9981
9982
Willy Tarreaue9656522010-08-17 15:40:09 +02009983tcp-request connection <action> [{if | unless} <condition>]
9984 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009985 May be used in sections : defaults | frontend | listen | backend
9986 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009987 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009988 <action> defines the action to perform if the condition applies. See
9989 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009990
Willy Tarreaue9656522010-08-17 15:40:09 +02009991 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009992
9993 Immediately after acceptance of a new incoming connection, it is possible to
9994 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009995 or dropped or have its counters tracked. Those conditions cannot make use of
9996 any data contents because the connection has not been read from yet, and the
9997 buffers are not yet allocated. This is used to selectively and very quickly
9998 accept or drop connections from various sources with a very low overhead. If
9999 some contents need to be inspected in order to take the decision, the
10000 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010001
Willy Tarreaue9656522010-08-17 15:40:09 +020010002 The "tcp-request connection" rules are evaluated in their exact declaration
10003 order. If no rule matches or if there is no rule, the default action is to
10004 accept the incoming connection. There is no specific limit to the number of
10005 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010006
Willy Tarreaua9083d02015-05-08 15:27:59 +020010007 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +020010008 - accept :
10009 accepts the connection if the condition is true (when used with "if")
10010 or false (when used with "unless"). The first such rule executed ends
10011 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010012
Willy Tarreaue9656522010-08-17 15:40:09 +020010013 - reject :
10014 rejects the connection if the condition is true (when used with "if")
10015 or false (when used with "unless"). The first such rule executed ends
10016 the rules evaluation. Rejected connections do not even become a
10017 session, which is why they are accounted separately for in the stats,
10018 as "denied connections". They are not considered for the session
10019 rate-limit and are not logged either. The reason is that these rules
10020 should only be used to filter extremely high connection rates such as
10021 the ones encountered during a massive DDoS attack. Under these extreme
10022 conditions, the simple action of logging each event would make the
10023 system collapse and would considerably lower the filtering capacity. If
10024 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +020010025 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010026
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010027 - expect-proxy layer4 :
10028 configures the client-facing connection to receive a PROXY protocol
10029 header before any byte is read from the socket. This is equivalent to
10030 having the "accept-proxy" keyword on the "bind" line, except that using
10031 the TCP rule allows the PROXY protocol to be accepted only for certain
10032 IP address ranges using an ACL. This is convenient when multiple layers
10033 of load balancers are passed through by traffic coming from public
10034 hosts.
10035
Bertrand Jacquin90759682016-06-06 15:35:39 +010010036 - expect-netscaler-cip layer4 :
10037 configures the client-facing connection to receive a NetScaler Client
10038 IP insertion protocol header before any byte is read from the socket.
10039 This is equivalent to having the "accept-netscaler-cip" keyword on the
10040 "bind" line, except that using the TCP rule allows the PROXY protocol
10041 to be accepted only for certain IP address ranges using an ACL. This
10042 is convenient when multiple layers of load balancers are passed
10043 through by traffic coming from public hosts.
10044
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010045 - capture <sample> len <length> :
10046 This only applies to "tcp-request content" rules. It captures sample
10047 expression <sample> from the request buffer, and converts it to a
10048 string of at most <len> characters. The resulting string is stored into
10049 the next request "capture" slot, so it will possibly appear next to
10050 some captured HTTP headers. It will then automatically appear in the
10051 logs, and it will be possible to extract it using sample fetch rules to
10052 feed it into headers or anything. The length should be limited given
10053 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +020010054 session life. Please check section 7.3 (Fetching samples) and "capture
10055 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010056
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010057 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +020010058 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020010059 rules do not stop evaluation and do not change default action. The
10060 number of counters that may be simultaneously tracked by the same
10061 connection is set in MAX_SESS_STKCTR at build time (reported in
John Roesler27c754c2019-07-10 15:45:51 -050010062 haproxy -vv) which defaults to 3, so the track-sc number is between 0
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020010063 and (MAX_SESS_STCKTR-1). The first "track-sc0" rule executed enables
10064 tracking of the counters of the specified table as the first set. The
10065 first "track-sc1" rule executed enables tracking of the counters of the
10066 specified table as the second set. The first "track-sc2" rule executed
10067 enables tracking of the counters of the specified table as the third
10068 set. It is a recommended practice to use the first set of counters for
10069 the per-frontend counters and the second set for the per-backend ones.
10070 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010071
Willy Tarreaue9656522010-08-17 15:40:09 +020010072 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +020010073 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +020010074 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +010010075 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010076 and used to select which table entry to update the counters.
10077 Note that "tcp-request connection" cannot use content-based
10078 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010079
Willy Tarreaue9656522010-08-17 15:40:09 +020010080 <table> is an optional table to be used instead of the default one,
10081 which is the stick-table declared in the current proxy. All
10082 the counters for the matches and updates for the key will
10083 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010084
Willy Tarreaue9656522010-08-17 15:40:09 +020010085 Once a "track-sc*" rule is executed, the key is looked up in the table
10086 and if it is not found, an entry is allocated for it. Then a pointer to
10087 that entry is kept during all the session's life, and this entry's
10088 counters are updated as often as possible, every time the session's
10089 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010090 Counters are only updated for events that happen after the tracking has
10091 been started. For example, connection counters will not be updated when
10092 tracking layer 7 information, since the connection event happens before
10093 layer7 information is extracted.
10094
Willy Tarreaue9656522010-08-17 15:40:09 +020010095 If the entry tracks concurrent connection counters, one connection is
10096 counted for as long as the entry is tracked, and the entry will not
10097 expire during that time. Tracking counters also provides a performance
10098 advantage over just checking the keys, because only one table lookup is
10099 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010100
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010101 - sc-inc-gpc0(<sc-id>):
10102 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
10103 counter designated by <sc-id>. If an error occurs, this action silently
10104 fails and the actions evaluation continues.
10105
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010106 - sc-inc-gpc1(<sc-id>):
10107 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
10108 counter designated by <sc-id>. If an error occurs, this action silently
10109 fails and the actions evaluation continues.
10110
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010111 - sc-set-gpt0(<sc-id>) <int>:
10112 This action sets the GPT0 tag according to the sticky counter designated
10113 by <sc-id> and the value of <int>. The expected result is a boolean. If
10114 an error occurs, this action silently fails and the actions evaluation
10115 continues.
10116
William Lallemand2e785f22016-05-25 01:48:42 +020010117 - set-src <expr> :
10118 Is used to set the source IP address to the value of specified
10119 expression. Useful if you want to mask source IP for privacy.
10120 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010121 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +020010122
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010123 Arguments:
10124 <expr> Is a standard HAProxy expression formed by a sample-fetch
10125 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +020010126
10127 Example:
William Lallemand2e785f22016-05-25 01:48:42 +020010128 tcp-request connection set-src src,ipmask(24)
10129
Willy Tarreau0c630532016-10-21 17:52:58 +020010130 When possible, set-src preserves the original source port as long as the
10131 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +020010132
William Lallemand44be6402016-05-25 01:51:35 +020010133 - set-src-port <expr> :
10134 Is used to set the source port address to the value of specified
10135 expression.
10136
Cyril Bonté6c81d5f2018-10-17 00:14:51 +020010137 Arguments:
10138 <expr> Is a standard HAProxy expression formed by a sample-fetch
10139 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +020010140
10141 Example:
William Lallemand44be6402016-05-25 01:51:35 +020010142 tcp-request connection set-src-port int(4000)
10143
Willy Tarreau0c630532016-10-21 17:52:58 +020010144 When possible, set-src-port preserves the original source address as long
10145 as the address family supports a port, otherwise it forces the source
10146 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +020010147
William Lallemand13e9b0c2016-05-25 02:34:07 +020010148 - set-dst <expr> :
10149 Is used to set the destination IP address to the value of specified
10150 expression. Useful if you want to mask IP for privacy in log.
10151 If you want to provide an IP from a HTTP header use "http-request
10152 set-dst". If you want to connect to the new address/port, use
10153 '0.0.0.0:0' as a server address in the backend.
10154
10155 <expr> Is a standard HAProxy expression formed by a sample-fetch
10156 followed by some converters.
10157
10158 Example:
10159
10160 tcp-request connection set-dst dst,ipmask(24)
10161 tcp-request connection set-dst ipv4(10.0.0.1)
10162
Willy Tarreau0c630532016-10-21 17:52:58 +020010163 When possible, set-dst preserves the original destination port as long as
10164 the address family allows it, otherwise the destination port is set to 0.
10165
William Lallemand13e9b0c2016-05-25 02:34:07 +020010166 - set-dst-port <expr> :
10167 Is used to set the destination port address to the value of specified
10168 expression. If you want to connect to the new address/port, use
10169 '0.0.0.0:0' as a server address in the backend.
10170
10171
10172 <expr> Is a standard HAProxy expression formed by a sample-fetch
10173 followed by some converters.
10174
10175 Example:
10176
10177 tcp-request connection set-dst-port int(4000)
10178
Willy Tarreau0c630532016-10-21 17:52:58 +020010179 When possible, set-dst-port preserves the original destination address as
10180 long as the address family supports a port, otherwise it forces the
10181 destination address to IPv4 "0.0.0.0" before rewriting the port.
10182
Willy Tarreau2d392c22015-08-24 01:43:45 +020010183 - "silent-drop" :
10184 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010185 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010186 to prevent the client from being notified. The effect it then that the
10187 client still sees an established connection while there's none on
10188 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10189 except that it doesn't use any local resource at all on the machine
10190 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010191 slow down stronger attackers. It is important to understand the impact
10192 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010193 client and HAProxy (firewalls, proxies, load balancers) will also keep
10194 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010195 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010196 TCP_REPAIR socket option is used to block the emission of a TCP
10197 reset. On other systems, the socket's TTL is reduced to 1 so that the
10198 TCP reset doesn't pass the first router, though it's still delivered to
10199 local networks. Do not use it unless you fully understand how it works.
10200
Willy Tarreaue9656522010-08-17 15:40:09 +020010201 Note that the "if/unless" condition is optional. If no condition is set on
10202 the action, it is simply performed unconditionally. That can be useful for
10203 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010204
Willy Tarreaue9656522010-08-17 15:40:09 +020010205 Example: accept all connections from white-listed hosts, reject too fast
10206 connection without counting them, and track accepted connections.
10207 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010208
Willy Tarreaue9656522010-08-17 15:40:09 +020010209 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010210 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010211 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010212
Willy Tarreaue9656522010-08-17 15:40:09 +020010213 Example: accept all connections from white-listed hosts, count all other
10214 connections and reject too fast ones. This results in abusive ones
10215 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010216
Willy Tarreaue9656522010-08-17 15:40:09 +020010217 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010218 tcp-request connection track-sc0 src
10219 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010220
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010221 Example: enable the PROXY protocol for traffic coming from all known proxies.
10222
10223 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10224
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010225 See section 7 about ACL usage.
10226
Willy Tarreau4f614292016-10-21 17:49:36 +020010227 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010228
10229
Willy Tarreaue9656522010-08-17 15:40:09 +020010230tcp-request content <action> [{if | unless} <condition>]
10231 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +020010232 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010233 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +020010234 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010235 <action> defines the action to perform if the condition applies. See
10236 below.
Willy Tarreau62644772008-07-16 18:36:06 +020010237
Willy Tarreaue9656522010-08-17 15:40:09 +020010238 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +020010239
Davor Ocelice9ed2812017-12-25 17:49:28 +010010240 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +020010241 called "TCP content inspection". During this stage, ACL-based rules are
10242 evaluated every time the request contents are updated, until either an
10243 "accept" or a "reject" rule matches, or the TCP request inspection delay
10244 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +020010245
Willy Tarreaue9656522010-08-17 15:40:09 +020010246 The first difference between these rules and "tcp-request connection" rules
10247 is that "tcp-request content" rules can make use of contents to take a
10248 decision. Most often, these decisions will consider a protocol recognition or
10249 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +010010250 both frontends and backends. In case of HTTP keep-alive with the client, all
10251 tcp-request content rules are evaluated again, so haproxy keeps a record of
10252 what sticky counters were assigned by a "tcp-request connection" versus a
10253 "tcp-request content" rule, and flushes all the content-related ones after
10254 processing an HTTP request, so that they may be evaluated again by the rules
10255 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010256 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +010010257 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +020010258
Willy Tarreaue9656522010-08-17 15:40:09 +020010259 Content-based rules are evaluated in their exact declaration order. If no
10260 rule matches or if there is no rule, the default action is to accept the
10261 contents. There is no specific limit to the number of rules which may be
10262 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +020010263
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010264 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010265 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +010010266 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010267 - reject : the request is rejected and the connection is closed
10268 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -040010269 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010270 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010271 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010272 - sc-inc-gpc1(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +020010273 - sc-set-gpt0(<sc-id>) <int>
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010274 - set-dst <expr>
10275 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010276 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010277 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +020010278 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +010010279 - send-spoe-group <engine-name> <group-name>
Christopher Faulet6bd406e2019-11-22 15:34:17 +010010280 - use-service <service-name>
Willy Tarreau62644772008-07-16 18:36:06 +020010281
Willy Tarreaue9656522010-08-17 15:40:09 +020010282 They have the same meaning as their counter-parts in "tcp-request connection"
10283 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +010010284 For "do-resolve" action, please check the "http-request do-resolve"
10285 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +020010286
Willy Tarreauf3338342014-01-28 21:40:28 +010010287 While there is nothing mandatory about it, it is recommended to use the
10288 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
10289 content" rules in the frontend, and track-sc2 for "tcp-request content"
10290 rules in the backend, because that makes the configuration more readable
10291 and easier to troubleshoot, but this is just a guideline and all counters
10292 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +020010293
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010294 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +020010295 the action, it is simply performed unconditionally. That can be useful for
10296 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +020010297
Willy Tarreaue9656522010-08-17 15:40:09 +020010298 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +020010299 rules, since HTTP-specific ACL matches are able to preliminarily parse the
10300 contents of a buffer before extracting the required data. If the buffered
10301 contents do not parse as a valid HTTP message, then the ACL does not match.
10302 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +010010303 processing, so there is no risk of parsing something differently. In an HTTP
10304 backend connected to from an HTTP frontend, it is guaranteed that HTTP
10305 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +020010306
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010307 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010308 are present when the rule is processed. The rule processing engine is able to
10309 wait until the inspect delay expires when the data to be tracked is not yet
10310 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010311
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010312 The "set-dst" and "set-dst-port" are used to set respectively the destination
10313 IP and port. More information on how to use it at "http-request set-dst".
10314
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010315 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +020010316 declared inline. For "tcp-request session" rules, only session-level
10317 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010318
Daniel Schneller0b547052016-03-21 20:46:57 +010010319 <var-name> The name of the variable starts with an indication about
10320 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010321 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010322 "sess" : the variable is shared with the whole session
10323 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010324 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010325 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010326 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010327 "res" : the variable is shared only during response
10328 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010329 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010330 The name may only contain characters 'a-z', 'A-Z', '0-9',
10331 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010332
10333 <expr> Is a standard HAProxy expression formed by a sample-fetch
10334 followed by some converters.
10335
Christopher Faulet85d79c92016-11-09 16:54:56 +010010336 The "unset-var" is used to unset a variable. See above for details about
10337 <var-name>.
10338
Patrick Hemmer268a7072018-05-11 12:52:31 -040010339 The "set-priority-class" is used to set the queue priority class of the
10340 current request. The value must be a sample expression which converts to an
10341 integer in the range -2047..2047. Results outside this range will be
10342 truncated. The priority class determines the order in which queued requests
10343 are processed. Lower values have higher priority.
10344
10345 The "set-priority-offset" is used to set the queue priority timestamp offset
10346 of the current request. The value must be a sample expression which converts
10347 to an integer in the range -524287..524287. Results outside this range will be
10348 truncated. When a request is queued, it is ordered first by the priority
10349 class, then by the current timestamp adjusted by the given offset in
10350 milliseconds. Lower values have higher priority.
10351 Note that the resulting timestamp is is only tracked with enough precision for
10352 524,287ms (8m44s287ms). If the request is queued long enough to where the
10353 adjusted timestamp exceeds this value, it will be misidentified as highest
10354 priority. Thus it is important to set "timeout queue" to a value, where when
10355 combined with the offset, does not exceed this limit.
10356
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010357 The "send-spoe-group" is used to trigger sending of a group of SPOE
10358 messages. To do so, the SPOE engine used to send messages must be defined, as
10359 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10360 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10361 the SPOE agent name must be used.
10362
10363 <engine-name> The SPOE engine name.
10364
10365 <group-name> The SPOE group name as specified in the engine configuration.
10366
Christopher Faulet6bd406e2019-11-22 15:34:17 +010010367 The "use-service" is used to executes a TCP service which will reply to the
10368 request and stop the evaluation of the rules. This service may choose to
10369 reply by sending any valid response or it may immediately close the
10370 connection without sending anything. Outside natives services, it is possible
10371 to write your own services in Lua. No further "tcp-request" rules are
10372 evaluated.
10373
10374 Example:
10375 tcp-request content use-service lua.deny { src -f /etc/haproxy/blacklist.lst }
10376
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010377 Example:
10378
10379 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +010010380 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010381
Willy Tarreau62644772008-07-16 18:36:06 +020010382 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +020010383 # Accept HTTP requests containing a Host header saying "example.com"
10384 # and reject everything else.
10385 acl is_host_com hdr(Host) -i example.com
10386 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +020010387 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +020010388 tcp-request content reject
10389
10390 Example:
Willy Tarreau62644772008-07-16 18:36:06 +020010391 # reject SMTP connection if client speaks first
10392 tcp-request inspect-delay 30s
10393 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010394 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +020010395
10396 # Forward HTTPS connection only if client speaks
10397 tcp-request inspect-delay 30s
10398 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010399 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +020010400 tcp-request content reject
10401
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010402 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010403 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010404 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010405 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010406 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
10407 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010408
10409 Example:
10410 # track request counts per "base" (concatenation of Host+URL)
10411 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010412 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010413
Willy Tarreaue9656522010-08-17 15:40:09 +020010414 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010415 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +020010416
10417 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +010010418 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +020010419 # protecting all our sites
10420 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010421 tcp-request connection track-sc0 src
10422 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +020010423 ...
10424 use_backend http_dynamic if { path_end .php }
10425
10426 backend http_dynamic
10427 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010428 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +020010429 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010430 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010431 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010432 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +020010433 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +020010434
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010435 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +020010436
Jarno Huuskonen95b012b2017-04-06 13:59:14 +030010437 See also : "tcp-request connection", "tcp-request session",
10438 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +020010439
10440
10441tcp-request inspect-delay <timeout>
10442 Set the maximum allowed time to wait for data during content inspection
10443 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010444 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +020010445 Arguments :
10446 <timeout> is the timeout value specified in milliseconds by default, but
10447 can be in any other unit if the number is suffixed by the unit,
10448 as explained at the top of this document.
10449
10450 People using haproxy primarily as a TCP relay are often worried about the
10451 risk of passing any type of protocol to a server without any analysis. In
10452 order to be able to analyze the request contents, we must first withhold
10453 the data then analyze them. This statement simply enables withholding of
10454 data for at most the specified amount of time.
10455
Willy Tarreaufb356202010-08-03 14:02:05 +020010456 TCP content inspection applies very early when a connection reaches a
10457 frontend, then very early when the connection is forwarded to a backend. This
10458 means that a connection may experience a first delay in the frontend and a
10459 second delay in the backend if both have tcp-request rules.
10460
Willy Tarreau62644772008-07-16 18:36:06 +020010461 Note that when performing content inspection, haproxy will evaluate the whole
10462 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010463 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +020010464 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +010010465 contents are definitive. If no delay is set, haproxy will not wait at all
10466 and will immediately apply a verdict based on the available information.
10467 Obviously this is unlikely to be very useful and might even be racy, so such
10468 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020010469
10470 As soon as a rule matches, the request is released and continues as usual. If
10471 the timeout is reached and no rule matches, the default policy will be to let
10472 it pass through unaffected.
10473
10474 For most protocols, it is enough to set it to a few seconds, as most clients
10475 send the full request immediately upon connection. Add 3 or more seconds to
10476 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +010010477 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +010010478 before the server (e.g. SMTP), or to wait for a client to talk before passing
10479 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +020010480 least the inspection delay, otherwise it will expire first. If the client
10481 closes the connection or if the buffer is full, the delay immediately expires
10482 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +020010483
Willy Tarreau55165fe2009-05-10 12:02:55 +020010484 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +020010485 "timeout client".
10486
10487
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010488tcp-response content <action> [{if | unless} <condition>]
10489 Perform an action on a session response depending on a layer 4-7 condition
10490 May be used in sections : defaults | frontend | listen | backend
10491 no | no | yes | yes
10492 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010493 <action> defines the action to perform if the condition applies. See
10494 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010495
10496 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
10497
Davor Ocelice9ed2812017-12-25 17:49:28 +010010498 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010499 called "TCP content inspection". During this stage, ACL-based rules are
10500 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010501 "accept", "close" or a "reject" rule matches, or a TCP response inspection
10502 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010503
10504 Most often, these decisions will consider a protocol recognition or validity.
10505
10506 Content-based rules are evaluated in their exact declaration order. If no
10507 rule matches or if there is no rule, the default action is to accept the
10508 contents. There is no specific limit to the number of rules which may be
10509 inserted.
10510
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010511 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010512 - accept :
10513 accepts the response if the condition is true (when used with "if")
10514 or false (when used with "unless"). The first such rule executed ends
10515 the rules evaluation.
10516
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010517 - close :
10518 immediately closes the connection with the server if the condition is
10519 true (when used with "if"), or false (when used with "unless"). The
10520 first such rule executed ends the rules evaluation. The main purpose of
10521 this action is to force a connection to be finished between a client
10522 and a server after an exchange when the application protocol expects
10523 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010524 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010525 protocols.
10526
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010527 - reject :
10528 rejects the response if the condition is true (when used with "if")
10529 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010530 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010531
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010532 - set-var(<var-name>) <expr>
10533 Sets a variable.
10534
Christopher Faulet85d79c92016-11-09 16:54:56 +010010535 - unset-var(<var-name>)
10536 Unsets a variable.
10537
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010538 - sc-inc-gpc0(<sc-id>):
10539 This action increments the GPC0 counter according to the sticky
10540 counter designated by <sc-id>. If an error occurs, this action fails
10541 silently and the actions evaluation continues.
10542
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010543 - sc-inc-gpc1(<sc-id>):
10544 This action increments the GPC1 counter according to the sticky
10545 counter designated by <sc-id>. If an error occurs, this action fails
10546 silently and the actions evaluation continues.
10547
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010548 - sc-set-gpt0(<sc-id>) <int> :
10549 This action sets the GPT0 tag according to the sticky counter designated
10550 by <sc-id> and the value of <int>. The expected result is a boolean. If
10551 an error occurs, this action silently fails and the actions evaluation
10552 continues.
10553
Willy Tarreau2d392c22015-08-24 01:43:45 +020010554 - "silent-drop" :
10555 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010556 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010557 to prevent the client from being notified. The effect it then that the
10558 client still sees an established connection while there's none on
10559 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10560 except that it doesn't use any local resource at all on the machine
10561 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010562 slow down stronger attackers. It is important to understand the impact
10563 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010564 client and HAProxy (firewalls, proxies, load balancers) will also keep
10565 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010566 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010567 TCP_REPAIR socket option is used to block the emission of a TCP
10568 reset. On other systems, the socket's TTL is reduced to 1 so that the
10569 TCP reset doesn't pass the first router, though it's still delivered to
10570 local networks. Do not use it unless you fully understand how it works.
10571
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010572 - send-spoe-group <engine-name> <group-name>
10573 Send a group of SPOE messages.
10574
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010575 Note that the "if/unless" condition is optional. If no condition is set on
10576 the action, it is simply performed unconditionally. That can be useful for
10577 for changing the default action to a reject.
10578
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010579 It is perfectly possible to match layer 7 contents with "tcp-response
10580 content" rules, but then it is important to ensure that a full response has
10581 been buffered, otherwise no contents will match. In order to achieve this,
10582 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010583 period.
10584
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010585 The "set-var" is used to set the content of a variable. The variable is
10586 declared inline.
10587
Daniel Schneller0b547052016-03-21 20:46:57 +010010588 <var-name> The name of the variable starts with an indication about
10589 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010590 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010591 "sess" : the variable is shared with the whole session
10592 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010593 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010594 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010595 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010596 "res" : the variable is shared only during response
10597 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010598 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010599 The name may only contain characters 'a-z', 'A-Z', '0-9',
10600 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010601
10602 <expr> Is a standard HAProxy expression formed by a sample-fetch
10603 followed by some converters.
10604
10605 Example:
10606
10607 tcp-request content set-var(sess.my_var) src
10608
Christopher Faulet85d79c92016-11-09 16:54:56 +010010609 The "unset-var" is used to unset a variable. See above for details about
10610 <var-name>.
10611
10612 Example:
10613
10614 tcp-request content unset-var(sess.my_var)
10615
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010616 The "send-spoe-group" is used to trigger sending of a group of SPOE
10617 messages. To do so, the SPOE engine used to send messages must be defined, as
10618 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10619 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10620 the SPOE agent name must be used.
10621
10622 <engine-name> The SPOE engine name.
10623
10624 <group-name> The SPOE group name as specified in the engine configuration.
10625
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010626 See section 7 about ACL usage.
10627
10628 See also : "tcp-request content", "tcp-response inspect-delay"
10629
10630
Willy Tarreau4f614292016-10-21 17:49:36 +020010631tcp-request session <action> [{if | unless} <condition>]
10632 Perform an action on a validated session depending on a layer 5 condition
10633 May be used in sections : defaults | frontend | listen | backend
10634 no | yes | yes | no
10635 Arguments :
10636 <action> defines the action to perform if the condition applies. See
10637 below.
10638
10639 <condition> is a standard layer5-only ACL-based condition (see section 7).
10640
Davor Ocelice9ed2812017-12-25 17:49:28 +010010641 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020010642 it is possible to evaluate some conditions to decide whether this session
10643 must be accepted or dropped or have its counters tracked. Those conditions
10644 cannot make use of any data contents because no buffers are allocated yet and
10645 the processing cannot wait at this stage. The main use case it to copy some
10646 early information into variables (since variables are accessible in the
10647 session), or to keep track of some information collected after the handshake,
10648 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010010649 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020010650 extracted information can thus be copied to a variable or tracked using
10651 "track-sc" rules. Of course it is also possible to decide to accept/reject as
10652 with other rulesets. Most operations performed here could also be performed
10653 in "tcp-request content" rules, except that in HTTP these rules are evaluated
10654 for each new request, and that might not always be acceptable. For example a
10655 rule might increment a counter on each evaluation. It would also be possible
10656 that a country is resolved by geolocation from the source IP address,
10657 assigned to a session-wide variable, then the source address rewritten from
10658 an HTTP header for all requests. If some contents need to be inspected in
10659 order to take the decision, the "tcp-request content" statements must be used
10660 instead.
10661
10662 The "tcp-request session" rules are evaluated in their exact declaration
10663 order. If no rule matches or if there is no rule, the default action is to
10664 accept the incoming session. There is no specific limit to the number of
10665 rules which may be inserted.
10666
10667 Several types of actions are supported :
10668 - accept : the request is accepted
10669 - reject : the request is rejected and the connection is closed
10670 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
10671 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010672 - sc-inc-gpc1(<sc-id>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010673 - sc-set-gpt0(<sc-id>) <int>
10674 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010675 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010676 - silent-drop
10677
10678 These actions have the same meaning as their respective counter-parts in
10679 "tcp-request connection" and "tcp-request content", so please refer to these
10680 sections for a complete description.
10681
10682 Note that the "if/unless" condition is optional. If no condition is set on
10683 the action, it is simply performed unconditionally. That can be useful for
10684 "track-sc*" actions as well as for changing the default action to a reject.
10685
10686 Example: track the original source address by default, or the one advertised
10687 in the PROXY protocol header for connection coming from the local
10688 proxies. The first connection-level rule enables receipt of the
10689 PROXY protocol for these ones, the second rule tracks whatever
10690 address we decide to keep after optional decoding.
10691
10692 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10693 tcp-request session track-sc0 src
10694
10695 Example: accept all sessions from white-listed hosts, reject too fast
10696 sessions without counting them, and track accepted sessions.
10697 This results in session rate being capped from abusive sources.
10698
10699 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10700 tcp-request session reject if { src_sess_rate gt 10 }
10701 tcp-request session track-sc0 src
10702
10703 Example: accept all sessions from white-listed hosts, count all other
10704 sessions and reject too fast ones. This results in abusive ones
10705 being blocked as long as they don't slow down.
10706
10707 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10708 tcp-request session track-sc0 src
10709 tcp-request session reject if { sc0_sess_rate gt 10 }
10710
10711 See section 7 about ACL usage.
10712
10713 See also : "tcp-request connection", "tcp-request content", "stick-table"
10714
10715
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010716tcp-response inspect-delay <timeout>
10717 Set the maximum allowed time to wait for a response during content inspection
10718 May be used in sections : defaults | frontend | listen | backend
10719 no | no | yes | yes
10720 Arguments :
10721 <timeout> is the timeout value specified in milliseconds by default, but
10722 can be in any other unit if the number is suffixed by the unit,
10723 as explained at the top of this document.
10724
10725 See also : "tcp-response content", "tcp-request inspect-delay".
10726
10727
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010728timeout check <timeout>
10729 Set additional check timeout, but only after a connection has been already
10730 established.
10731
10732 May be used in sections: defaults | frontend | listen | backend
10733 yes | no | yes | yes
10734 Arguments:
10735 <timeout> is the timeout value specified in milliseconds by default, but
10736 can be in any other unit if the number is suffixed by the unit,
10737 as explained at the top of this document.
10738
10739 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
10740 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010010741 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010742 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010010743 (Please also note that there is no valid reason to have such long connect
10744 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
10745 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010746
10747 If "timeout check" is not set haproxy uses "inter" for complete check
10748 timeout (connect + read) exactly like all <1.3.15 version.
10749
10750 In most cases check request is much simpler and faster to handle than normal
10751 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010010752 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010753
10754 This parameter is specific to backends, but can be specified once for all in
10755 "defaults" sections. This is in fact one of the easiest solutions not to
10756 forget about it.
10757
Willy Tarreau41a340d2008-01-22 12:25:31 +010010758 See also: "timeout connect", "timeout queue", "timeout server",
10759 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010760
10761
Willy Tarreau0ba27502007-12-24 16:55:16 +010010762timeout client <timeout>
10763timeout clitimeout <timeout> (deprecated)
10764 Set the maximum inactivity time on the client side.
10765 May be used in sections : defaults | frontend | listen | backend
10766 yes | yes | yes | no
10767 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010768 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010769 can be in any other unit if the number is suffixed by the unit,
10770 as explained at the top of this document.
10771
10772 The inactivity timeout applies when the client is expected to acknowledge or
10773 send data. In HTTP mode, this timeout is particularly important to consider
10774 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010775 response while it is reading data sent by the server. That said, for the
10776 first phase, it is preferable to set the "timeout http-request" to better
10777 protect HAProxy from Slowloris like attacks. The value is specified in
10778 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010010779 suffixed by the unit, as specified at the top of this document. In TCP mode
10780 (and to a lesser extent, in HTTP mode), it is highly recommended that the
10781 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010782 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010010783 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010010784 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
10785 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020010786 which overrides "timeout client" and "timeout server" for tunnels, as well as
10787 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010788
10789 This parameter is specific to frontends, but can be specified once for all in
10790 "defaults" sections. This is in fact one of the easiest solutions not to
10791 forget about it. An unspecified timeout results in an infinite timeout, which
10792 is not recommended. Such a usage is accepted and works but reports a warning
John Roesler27c754c2019-07-10 15:45:51 -050010793 during startup because it may result in accumulation of expired sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010794 the system if the system's timeouts are not configured either.
10795
Willy Tarreau95c4e142017-11-26 12:18:55 +010010796 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010010797
Willy Tarreau0ba27502007-12-24 16:55:16 +010010798 This parameter replaces the old, deprecated "clitimeout". It is recommended
10799 to use it to write new configurations. The form "timeout clitimeout" is
10800 provided only by backwards compatibility but its use is strongly discouraged.
10801
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010802 See also : "clitimeout", "timeout server", "timeout tunnel",
10803 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010804
10805
Willy Tarreau05cdd962014-05-10 14:30:07 +020010806timeout client-fin <timeout>
10807 Set the inactivity timeout on the client side for half-closed connections.
10808 May be used in sections : defaults | frontend | listen | backend
10809 yes | yes | yes | no
10810 Arguments :
10811 <timeout> is the timeout value specified in milliseconds by default, but
10812 can be in any other unit if the number is suffixed by the unit,
10813 as explained at the top of this document.
10814
10815 The inactivity timeout applies when the client is expected to acknowledge or
10816 send data while one direction is already shut down. This timeout is different
10817 from "timeout client" in that it only applies to connections which are closed
10818 in one direction. This is particularly useful to avoid keeping connections in
10819 FIN_WAIT state for too long when clients do not disconnect cleanly. This
10820 problem is particularly common long connections such as RDP or WebSocket.
10821 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010010822 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
10823 frame was sent, often indicating an expectation that the connection quickly
10824 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020010825
10826 This parameter is specific to frontends, but can be specified once for all in
10827 "defaults" sections. By default it is not set, so half-closed connections
10828 will use the other timeouts (timeout.client or timeout.tunnel).
10829
10830 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
10831
10832
Willy Tarreau0ba27502007-12-24 16:55:16 +010010833timeout connect <timeout>
10834timeout contimeout <timeout> (deprecated)
10835 Set the maximum time to wait for a connection attempt to a server to succeed.
10836 May be used in sections : defaults | frontend | listen | backend
10837 yes | no | yes | yes
10838 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010839 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010840 can be in any other unit if the number is suffixed by the unit,
10841 as explained at the top of this document.
10842
10843 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010844 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010010845 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010010846 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010847 connect timeout also presets both queue and tarpit timeouts to the same value
10848 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010849
10850 This parameter is specific to backends, but can be specified once for all in
10851 "defaults" sections. This is in fact one of the easiest solutions not to
10852 forget about it. An unspecified timeout results in an infinite timeout, which
10853 is not recommended. Such a usage is accepted and works but reports a warning
John Roesler27c754c2019-07-10 15:45:51 -050010854 during startup because it may result in accumulation of failed sessions in
Willy Tarreau0ba27502007-12-24 16:55:16 +010010855 the system if the system's timeouts are not configured either.
10856
10857 This parameter replaces the old, deprecated "contimeout". It is recommended
10858 to use it to write new configurations. The form "timeout contimeout" is
10859 provided only by backwards compatibility but its use is strongly discouraged.
10860
Willy Tarreau41a340d2008-01-22 12:25:31 +010010861 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
10862 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010863
10864
Willy Tarreaub16a5742010-01-10 14:46:16 +010010865timeout http-keep-alive <timeout>
10866 Set the maximum allowed time to wait for a new HTTP request to appear
10867 May be used in sections : defaults | frontend | listen | backend
10868 yes | yes | yes | yes
10869 Arguments :
10870 <timeout> is the timeout value specified in milliseconds by default, but
10871 can be in any other unit if the number is suffixed by the unit,
10872 as explained at the top of this document.
10873
10874 By default, the time to wait for a new request in case of keep-alive is set
10875 by "timeout http-request". However this is not always convenient because some
10876 people want very short keep-alive timeouts in order to release connections
10877 faster, and others prefer to have larger ones but still have short timeouts
10878 once the request has started to present itself.
10879
10880 The "http-keep-alive" timeout covers these needs. It will define how long to
10881 wait for a new HTTP request to start coming after a response was sent. Once
10882 the first byte of request has been seen, the "http-request" timeout is used
10883 to wait for the complete request to come. Note that empty lines prior to a
10884 new request do not refresh the timeout and are not counted as a new request.
10885
10886 There is also another difference between the two timeouts : when a connection
10887 expires during timeout http-keep-alive, no error is returned, the connection
10888 just closes. If the connection expires in "http-request" while waiting for a
10889 connection to complete, a HTTP 408 error is returned.
10890
10891 In general it is optimal to set this value to a few tens to hundreds of
10892 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010893 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010894 1 millisecond) it will probably only accept pipelined requests but not the
10895 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010896 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010897
10898 If this parameter is not set, the "http-request" timeout applies, and if both
10899 are not set, "timeout client" still applies at the lower level. It should be
10900 set in the frontend to take effect, unless the frontend is in TCP mode, in
10901 which case the HTTP backend's timeout will be used.
10902
Willy Tarreau95c4e142017-11-26 12:18:55 +010010903 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10904 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010905 (where we only have one connection per client and a connection setup).
10906
Willy Tarreaub16a5742010-01-10 14:46:16 +010010907 See also : "timeout http-request", "timeout client".
10908
10909
Willy Tarreau036fae02008-01-06 13:24:40 +010010910timeout http-request <timeout>
10911 Set the maximum allowed time to wait for a complete HTTP request
10912 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010913 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010914 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010915 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010916 can be in any other unit if the number is suffixed by the unit,
10917 as explained at the top of this document.
10918
10919 In order to offer DoS protection, it may be required to lower the maximum
10920 accepted time to receive a complete HTTP request without affecting the client
10921 timeout. This helps protecting against established connections on which
10922 nothing is sent. The client timeout cannot offer a good protection against
10923 this abuse because it is an inactivity timeout, which means that if the
10924 attacker sends one character every now and then, the timeout will not
10925 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010926 types, the request will be aborted if it does not complete in time. When the
10927 timeout expires, an HTTP 408 response is sent to the client to inform it
10928 about the problem, and the connection is closed. The logs will report
10929 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010930 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010931 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10932 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010933
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010934 By default, this timeout only applies to the header part of the request,
10935 and not to any data. As soon as the empty line is received, this timeout is
10936 not used anymore. When combined with "option http-buffer-request", this
10937 timeout also applies to the body of the request..
10938 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010939 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010940
10941 Generally it is enough to set it to a few seconds, as most clients send the
10942 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010943 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010944 generally work on local networks as long as there are no packet losses. This
10945 will prevent people from sending bare HTTP requests using telnet.
10946
10947 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010948 chunk of the incoming request. It should be set in the frontend to take
10949 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10950 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010951
Willy Tarreau0f228a02015-05-01 15:37:53 +020010952 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010953 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010954
Willy Tarreau844e3c52008-01-11 16:28:18 +010010955
10956timeout queue <timeout>
10957 Set the maximum time to wait in the queue for a connection slot to be free
10958 May be used in sections : defaults | frontend | listen | backend
10959 yes | no | yes | yes
10960 Arguments :
10961 <timeout> is the timeout value specified in milliseconds by default, but
10962 can be in any other unit if the number is suffixed by the unit,
10963 as explained at the top of this document.
10964
10965 When a server's maxconn is reached, connections are left pending in a queue
10966 which may be server-specific or global to the backend. In order not to wait
10967 indefinitely, a timeout is applied to requests pending in the queue. If the
10968 timeout is reached, it is considered that the request will almost never be
10969 served, so it is dropped and a 503 error is returned to the client.
10970
10971 The "timeout queue" statement allows to fix the maximum time for a request to
10972 be left pending in a queue. If unspecified, the same value as the backend's
10973 connection timeout ("timeout connect") is used, for backwards compatibility
10974 with older versions with no "timeout queue" parameter.
10975
10976 See also : "timeout connect", "contimeout".
10977
10978
10979timeout server <timeout>
10980timeout srvtimeout <timeout> (deprecated)
10981 Set the maximum inactivity time on the server side.
10982 May be used in sections : defaults | frontend | listen | backend
10983 yes | no | yes | yes
10984 Arguments :
10985 <timeout> is the timeout value specified in milliseconds by default, but
10986 can be in any other unit if the number is suffixed by the unit,
10987 as explained at the top of this document.
10988
10989 The inactivity timeout applies when the server is expected to acknowledge or
10990 send data. In HTTP mode, this timeout is particularly important to consider
10991 during the first phase of the server's response, when it has to send the
10992 headers, as it directly represents the server's processing time for the
10993 request. To find out what value to put there, it's often good to start with
10994 what would be considered as unacceptable response times, then check the logs
10995 to observe the response time distribution, and adjust the value accordingly.
10996
10997 The value is specified in milliseconds by default, but can be in any other
10998 unit if the number is suffixed by the unit, as specified at the top of this
10999 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
11000 recommended that the client timeout remains equal to the server timeout in
11001 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010011002 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010011003 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010011004 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
11005 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020011006 "timeout tunnel", which overrides "timeout client" and "timeout server" for
11007 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011008
11009 This parameter is specific to backends, but can be specified once for all in
11010 "defaults" sections. This is in fact one of the easiest solutions not to
11011 forget about it. An unspecified timeout results in an infinite timeout, which
11012 is not recommended. Such a usage is accepted and works but reports a warning
John Roesler27c754c2019-07-10 15:45:51 -050011013 during startup because it may result in accumulation of expired sessions in
Willy Tarreau844e3c52008-01-11 16:28:18 +010011014 the system if the system's timeouts are not configured either.
11015
11016 This parameter replaces the old, deprecated "srvtimeout". It is recommended
11017 to use it to write new configurations. The form "timeout srvtimeout" is
11018 provided only by backwards compatibility but its use is strongly discouraged.
11019
Willy Tarreauce887fd2012-05-12 12:50:00 +020011020 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010011021
Willy Tarreau05cdd962014-05-10 14:30:07 +020011022
11023timeout server-fin <timeout>
11024 Set the inactivity timeout on the server side for half-closed connections.
11025 May be used in sections : defaults | frontend | listen | backend
11026 yes | no | yes | yes
11027 Arguments :
11028 <timeout> is the timeout value specified in milliseconds by default, but
11029 can be in any other unit if the number is suffixed by the unit,
11030 as explained at the top of this document.
11031
11032 The inactivity timeout applies when the server is expected to acknowledge or
11033 send data while one direction is already shut down. This timeout is different
11034 from "timeout server" in that it only applies to connections which are closed
11035 in one direction. This is particularly useful to avoid keeping connections in
11036 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
11037 This problem is particularly common long connections such as RDP or WebSocket.
11038 Note that this timeout can override "timeout tunnel" when a connection shuts
11039 down in one direction. This setting was provided for completeness, but in most
11040 situations, it should not be needed.
11041
11042 This parameter is specific to backends, but can be specified once for all in
11043 "defaults" sections. By default it is not set, so half-closed connections
11044 will use the other timeouts (timeout.server or timeout.tunnel).
11045
11046 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
11047
Willy Tarreau844e3c52008-01-11 16:28:18 +010011048
11049timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010011050 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010011051 May be used in sections : defaults | frontend | listen | backend
11052 yes | yes | yes | yes
11053 Arguments :
11054 <timeout> is the tarpit duration specified in milliseconds by default, but
11055 can be in any other unit if the number is suffixed by the unit,
11056 as explained at the top of this document.
11057
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030011058 When a connection is tarpitted using "http-request tarpit" or
11059 "reqtarpit", it is maintained open with no activity for a certain
11060 amount of time, then closed. "timeout tarpit" defines how long it will
11061 be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011062
11063 The value is specified in milliseconds by default, but can be in any other
11064 unit if the number is suffixed by the unit, as specified at the top of this
11065 document. If unspecified, the same value as the backend's connection timeout
11066 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010011067 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011068
11069 See also : "timeout connect", "contimeout".
11070
11071
Willy Tarreauce887fd2012-05-12 12:50:00 +020011072timeout tunnel <timeout>
11073 Set the maximum inactivity time on the client and server side for tunnels.
11074 May be used in sections : defaults | frontend | listen | backend
11075 yes | no | yes | yes
11076 Arguments :
11077 <timeout> is the timeout value specified in milliseconds by default, but
11078 can be in any other unit if the number is suffixed by the unit,
11079 as explained at the top of this document.
11080
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011081 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020011082 between a client and a server, and the connection remains inactive in both
11083 directions. This timeout supersedes both the client and server timeouts once
11084 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010011085 analyzer remains attached to either connection (e.g. tcp content rules are
11086 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020011087 when switching to the WebSocket protocol, or forwarding a CONNECT request
11088 to a proxy), or after the first response when no keepalive/close option is
11089 specified.
11090
Willy Tarreau05cdd962014-05-10 14:30:07 +020011091 Since this timeout is usually used in conjunction with long-lived connections,
11092 it usually is a good idea to also set "timeout client-fin" to handle the
11093 situation where a client suddenly disappears from the net and does not
11094 acknowledge a close, or sends a shutdown and does not acknowledge pending
11095 data anymore. This can happen in lossy networks where firewalls are present,
11096 and is detected by the presence of large amounts of sessions in a FIN_WAIT
11097 state.
11098
Willy Tarreauce887fd2012-05-12 12:50:00 +020011099 The value is specified in milliseconds by default, but can be in any other
11100 unit if the number is suffixed by the unit, as specified at the top of this
11101 document. Whatever the expected normal idle time, it is a good practice to
11102 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010011103 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020011104
11105 This parameter is specific to backends, but can be specified once for all in
11106 "defaults" sections. This is in fact one of the easiest solutions not to
11107 forget about it.
11108
11109 Example :
11110 defaults http
11111 option http-server-close
11112 timeout connect 5s
11113 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020011114 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020011115 timeout server 30s
11116 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
11117
Willy Tarreau05cdd962014-05-10 14:30:07 +020011118 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020011119
11120
Willy Tarreau844e3c52008-01-11 16:28:18 +010011121transparent (deprecated)
11122 Enable client-side transparent proxying
11123 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010011124 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010011125 Arguments : none
11126
11127 This keyword was introduced in order to provide layer 7 persistence to layer
11128 3 load balancers. The idea is to use the OS's ability to redirect an incoming
11129 connection for a remote address to a local process (here HAProxy), and let
11130 this process know what address was initially requested. When this option is
11131 used, sessions without cookies will be forwarded to the original destination
11132 IP address of the incoming request (which should match that of another
11133 equipment), while requests with cookies will still be forwarded to the
11134 appropriate server.
11135
11136 The "transparent" keyword is deprecated, use "option transparent" instead.
11137
11138 Note that contrary to a common belief, this option does NOT make HAProxy
11139 present the client's IP to the server when establishing the connection.
11140
Willy Tarreau844e3c52008-01-11 16:28:18 +010011141 See also: "option transparent"
11142
William Lallemanda73203e2012-03-12 12:48:57 +010011143unique-id-format <string>
11144 Generate a unique ID for each request.
11145 May be used in sections : defaults | frontend | listen | backend
11146 yes | yes | yes | no
11147 Arguments :
11148 <string> is a log-format string.
11149
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011150 This keyword creates a ID for each request using the custom log format. A
11151 unique ID is useful to trace a request passing through many components of
11152 a complex infrastructure. The newly created ID may also be logged using the
11153 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010011154
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011155 The format should be composed from elements that are guaranteed to be
11156 unique when combined together. For instance, if multiple haproxy instances
11157 are involved, it might be important to include the node name. It is often
11158 needed to log the incoming connection's source and destination addresses
11159 and ports. Note that since multiple requests may be performed over the same
11160 connection, including a request counter may help differentiate them.
11161 Similarly, a timestamp may protect against a rollover of the counter.
11162 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010011163
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011164 It is recommended to use hexadecimal notation for many fields since it
11165 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010011166
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011167 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010011168
Julien Vehentf21be322014-03-07 08:27:34 -050011169 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010011170
11171 will generate:
11172
11173 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
11174
11175 See also: "unique-id-header"
11176
11177unique-id-header <name>
11178 Add a unique ID header in the HTTP request.
11179 May be used in sections : defaults | frontend | listen | backend
11180 yes | yes | yes | no
11181 Arguments :
11182 <name> is the name of the header.
11183
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011184 Add a unique-id header in the HTTP request sent to the server, using the
11185 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010011186
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011187 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010011188
Julien Vehentf21be322014-03-07 08:27:34 -050011189 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010011190 unique-id-header X-Unique-ID
11191
11192 will generate:
11193
11194 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
11195
11196 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010011197
Willy Tarreauf51658d2014-04-23 01:21:56 +020011198use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011199 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011200 May be used in sections : defaults | frontend | listen | backend
11201 no | yes | yes | no
11202 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011203 <backend> is the name of a valid backend or "listen" section, or a
11204 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011205
Willy Tarreauf51658d2014-04-23 01:21:56 +020011206 <condition> is a condition composed of ACLs, as described in section 7. If
11207 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011208
11209 When doing content-switching, connections arrive on a frontend and are then
11210 dispatched to various backends depending on a number of conditions. The
11211 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011212 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011213 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011214 source address validation) or combined with a "tcp-request" rule to wait for
11215 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011216
11217 There may be as many "use_backend" rules as desired. All of these rules are
11218 evaluated in their declaration order, and the first one which matches will
11219 assign the backend.
11220
11221 In the first form, the backend will be used if the condition is met. In the
11222 second form, the backend will be used if the condition is not met. If no
11223 condition is valid, the backend defined with "default_backend" will be used.
11224 If no default backend is defined, either the servers in the same section are
11225 used (in case of a "listen" section) or, in case of a frontend, no server is
11226 used and a 503 service unavailable response is returned.
11227
Willy Tarreau51aecc72009-07-12 09:47:04 +020011228 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011229 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020011230 and backend processing will immediately follow, or the backend will wait for
11231 a complete HTTP request to get in. This feature is useful when a frontend
11232 must decode several protocols on a unique port, one of them being HTTP.
11233
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011234 When <backend> is a simple name, it is resolved at configuration time, and an
11235 error is reported if the specified backend does not exist. If <backend> is
11236 a log-format string instead, no check may be done at configuration time, so
11237 the backend name is resolved dynamically at run time. If the resulting
11238 backend name does not correspond to any valid backend, no other rule is
11239 evaluated, and the default_backend directive is applied instead. Note that
11240 when using dynamic backend names, it is highly recommended to use a prefix
11241 that no other backend uses in order to ensure that an unauthorized backend
11242 cannot be forced from the request.
11243
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011244 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011245 used to detect the association between frontends and backends to compute the
11246 backend's "fullconn" setting. This cannot be done for dynamic names.
11247
11248 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
11249 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010011250
Willy Tarreau036fae02008-01-06 13:24:40 +010011251
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011252use-server <server> if <condition>
11253use-server <server> unless <condition>
11254 Only use a specific server if/unless an ACL-based condition is matched.
11255 May be used in sections : defaults | frontend | listen | backend
11256 no | no | yes | yes
11257 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011258 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011259
11260 <condition> is a condition composed of ACLs, as described in section 7.
11261
11262 By default, connections which arrive to a backend are load-balanced across
11263 the available servers according to the configured algorithm, unless a
11264 persistence mechanism such as a cookie is used and found in the request.
11265
11266 Sometimes it is desirable to forward a particular request to a specific
11267 server without having to declare a dedicated backend for this server. This
11268 can be achieved using the "use-server" rules. These rules are evaluated after
11269 the "redirect" rules and before evaluating cookies, and they have precedence
11270 on them. There may be as many "use-server" rules as desired. All of these
11271 rules are evaluated in their declaration order, and the first one which
11272 matches will assign the server.
11273
11274 If a rule designates a server which is down, and "option persist" is not used
11275 and no force-persist rule was validated, it is ignored and evaluation goes on
11276 with the next rules until one matches.
11277
11278 In the first form, the server will be used if the condition is met. In the
11279 second form, the server will be used if the condition is not met. If no
11280 condition is valid, the processing continues and the server will be assigned
11281 according to other persistence mechanisms.
11282
11283 Note that even if a rule is matched, cookie processing is still performed but
11284 does not assign the server. This allows prefixed cookies to have their prefix
11285 stripped.
11286
11287 The "use-server" statement works both in HTTP and TCP mode. This makes it
11288 suitable for use with content-based inspection. For instance, a server could
11289 be selected in a farm according to the TLS SNI field. And if these servers
11290 have their weight set to zero, they will not be used for other traffic.
11291
11292 Example :
11293 # intercept incoming TLS requests based on the SNI field
11294 use-server www if { req_ssl_sni -i www.example.com }
11295 server www 192.168.0.1:443 weight 0
11296 use-server mail if { req_ssl_sni -i mail.example.com }
11297 server mail 192.168.0.1:587 weight 0
11298 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000011299 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011300 # all the rest is forwarded to this server
11301 server default 192.168.0.2:443 check
11302
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011303 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011304
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011305
Davor Ocelice9ed2812017-12-25 17:49:28 +0100113065. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011307--------------------------
11308
11309The "bind", "server" and "default-server" keywords support a number of settings
11310depending on some build options and on the system HAProxy was built on. These
11311settings generally each consist in one word sometimes followed by a value,
11312written on the same line as the "bind" or "server" line. All these options are
11313described in this section.
11314
11315
113165.1. Bind options
11317-----------------
11318
11319The "bind" keyword supports a certain number of settings which are all passed
11320as arguments on the same line. The order in which those arguments appear makes
11321no importance, provided that they appear after the bind address. All of these
11322parameters are optional. Some of them consist in a single words (booleans),
11323while other ones expect a value after them. In this case, the value must be
11324provided immediately after the setting name.
11325
11326The currently supported settings are the following ones.
11327
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011328accept-netscaler-cip <magic number>
11329 Enforces the use of the NetScaler Client IP insertion protocol over any
11330 connection accepted by any of the TCP sockets declared on the same line. The
11331 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
11332 the incoming connection to be used everywhere an address is used, with the
11333 only exception of "tcp-request connection" rules which will only see the
11334 real connection address. Logs will reflect the addresses indicated in the
11335 protocol, unless it is violated, in which case the real address will still
11336 be used. This keyword combined with support from external components can be
11337 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010011338 mechanism which is not always reliable and not even always usable. See also
11339 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
11340 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011341
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011342accept-proxy
11343 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020011344 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
11345 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011346 3/4 addresses of the incoming connection to be used everywhere an address is
11347 used, with the only exception of "tcp-request connection" rules which will
11348 only see the real connection address. Logs will reflect the addresses
11349 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010011350 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011351 components can be used as an efficient and reliable alternative to the
11352 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020011353 usable. See also "tcp-request connection expect-proxy" for a finer-grained
11354 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011355
Olivier Houchardc2aae742017-09-22 18:26:28 +020011356allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010011357 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010011358 due to security considerations. Because it is vulnerable to replay attacks,
John Roesler27c754c2019-07-10 15:45:51 -050011359 you should only allow if for requests that are safe to replay, i.e. requests
Olivier Houchard69752962019-01-08 15:35:32 +010011360 that are idempotent. You can use the "wait-for-handshake" action for any
11361 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020011362
Willy Tarreauab861d32013-04-02 02:30:41 +020011363alpn <protocols>
11364 This enables the TLS ALPN extension and advertises the specified protocol
11365 list as supported on top of ALPN. The protocol list consists in a comma-
11366 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roesler27c754c2019-07-10 15:45:51 -050011367 quotes). This requires that the SSL library is built with support for TLS
Willy Tarreauab861d32013-04-02 02:30:41 +020011368 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010011369 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
11370 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11371 now obsolete NPN extension. At the time of writing this, most browsers still
11372 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
11373 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
11374 are expected to be supported, both versions can be advertised, in order of
11375 preference, like below :
11376
11377 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020011378
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011379backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010011380 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011381 backlog is used instead, which generally defaults to the maxconn value.
11382
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010011383curves <curves>
11384 This setting is only available when support for OpenSSL was built in. It sets
11385 the string describing the list of elliptic curves algorithms ("curve suite")
11386 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
11387 string is a colon-delimited list of curve name.
11388 Example: "X25519:P-256" (without quote)
11389 When "curves" is set, "ecdhe" parameter is ignored.
11390
Emeric Brun7fb34422012-09-28 15:26:15 +020011391ecdhe <named curve>
11392 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010011393 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
11394 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020011395
Emeric Brunfd33a262012-10-11 16:28:27 +020011396ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011397 This setting is only available when support for OpenSSL was built in. It
11398 designates a PEM file from which to load CA certificates used to verify
11399 client's certificate.
11400
Emeric Brunb6dc9342012-09-28 17:55:37 +020011401ca-ignore-err [all|<errorID>,...]
11402 This setting is only available when support for OpenSSL was built in.
11403 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
11404 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
11405 error is ignored.
11406
Christopher Faulet31af49d2015-06-09 17:29:50 +020011407ca-sign-file <cafile>
11408 This setting is only available when support for OpenSSL was built in. It
11409 designates a PEM file containing both the CA certificate and the CA private
11410 key used to create and sign server's certificates. This is a mandatory
11411 setting when the dynamic generation of certificates is enabled. See
11412 'generate-certificates' for details.
11413
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000011414ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020011415 This setting is only available when support for OpenSSL was built in. It is
11416 the CA private key passphrase. This setting is optional and used only when
11417 the dynamic generation of certificates is enabled. See
11418 'generate-certificates' for details.
11419
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011420ciphers <ciphers>
11421 This setting is only available when support for OpenSSL was built in. It sets
11422 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000011423 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011424 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011425 information and recommendations see e.g.
11426 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11427 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11428 cipher configuration, please check the "ciphersuites" keyword.
11429
11430ciphersuites <ciphersuites>
11431 This setting is only available when support for OpenSSL was built in and
11432 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
11433 the list of cipher algorithms ("cipher suite") that are negotiated during the
11434 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011435 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
11436 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011437
Emeric Brunfd33a262012-10-11 16:28:27 +020011438crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011439 This setting is only available when support for OpenSSL was built in. It
11440 designates a PEM file from which to load certificate revocation list used
11441 to verify client's certificate.
11442
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011443crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000011444 This setting is only available when support for OpenSSL was built in. It
11445 designates a PEM file containing both the required certificates and any
11446 associated private keys. This file can be built by concatenating multiple
11447 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
11448 requires an intermediate certificate, this can also be concatenated into this
11449 file.
11450
11451 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
11452 are loaded.
11453
11454 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011455 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011456 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
11457 specified multiple times in order to load certificates from multiple files or
11458 directories. The certificates will be presented to clients who provide a
11459 valid TLS Server Name Indication field matching one of their CN or alt
Davor Ocelice9ed2812017-12-25 17:49:28 +010011460 subjects. Wildcards are supported, where a wildcard character '*' is used
11461 instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011462 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000011463
11464 If no SNI is provided by the client or if the SSL library does not support
11465 TLS extensions, or if the client provides an SNI hostname which does not
11466 match any certificate, then the first loaded certificate will be presented.
11467 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011468 recommended to load the default one first as a file or to ensure that it will
11469 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000011470
Emeric Brune032bfa2012-09-28 13:01:45 +020011471 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011472
Davor Ocelice9ed2812017-12-25 17:49:28 +010011473 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000011474 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010011475 choose a web server that the CA believes requires an intermediate CA (for
11476 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000011477 others, e.g. nginx, result in a wrong bundle that will not work for some
11478 clients).
11479
Emeric Brun4147b2e2014-06-16 18:36:30 +020011480 For each PEM file, haproxy checks for the presence of file at the same path
11481 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
11482 Status Request extension (also known as "OCSP stapling") is automatically
11483 enabled. The content of this file is optional. If not empty, it must contain
11484 a valid OCSP Response in DER format. In order to be valid an OCSP Response
11485 must comply with the following rules: it has to indicate a good status,
11486 it has to be a single response for the certificate of the PEM file, and it
11487 has to be valid at the moment of addition. If these rules are not respected
11488 the OCSP Response is ignored and a warning is emitted. In order to identify
11489 which certificate an OCSP Response applies to, the issuer's certificate is
11490 necessary. If the issuer's certificate is not found in the PEM file, it will
11491 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
11492 if it exists otherwise it will fail with an error.
11493
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011494 For each PEM file, haproxy also checks for the presence of file at the same
11495 path suffixed by ".sctl". If such file is found, support for Certificate
11496 Transparency (RFC6962) TLS extension is enabled. The file must contain a
11497 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
11498 to check basic syntax, but no signatures are verified.
11499
yanbzhu6c25e9e2016-01-05 12:52:02 -050011500 There are cases where it is desirable to support multiple key types, e.g. RSA
11501 and ECDSA in the cipher suites offered to the clients. This allows clients
11502 that support EC certificates to be able to use EC ciphers, while
11503 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050011504
11505 In order to provide this functionality, multiple PEM files, each with a
11506 different key type, are required. To associate these PEM files into a
11507 "cert bundle" that is recognized by haproxy, they must be named in the
11508 following way: All PEM files that are to be bundled must have the same base
11509 name, with a suffix indicating the key type. Currently, three suffixes are
11510 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
11511 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
11512 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
11513 suffix matters. To load this bundle into haproxy, specify the base name only:
11514
11515 Example : bind :8443 ssl crt example.pem
11516
yanbzhu6c25e9e2016-01-05 12:52:02 -050011517 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050011518 a cert bundle.
11519
Davor Ocelice9ed2812017-12-25 17:49:28 +010011520 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050011521 support multiple key types. PEM files are combined based on Common Name
11522 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
11523 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
11524 entries in the certificates in that bundle, haproxy will not be able to
11525 provide multi-cert support.
11526
11527 Assuming bundle in the example above contained the following:
11528
11529 Filename | CN | SAN
11530 -------------------+-----------------+-------------------
11531 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050011532 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050011533 example.pem.ecdsa | www.example.com | ecdsa.example.com
11534 -------------------+-----------------+-------------------
11535
11536 Users connecting with an SNI of "www.example.com" will be able
11537 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
11538 "rsa.example.com" will only be able to use RSA cipher suites, and users
11539 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011540 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
11541 no need to bundle certificates. ECDSA certificate will be preferred if client
11542 support it.
yanbzhud19630c2015-12-14 15:10:25 -050011543
11544 If a directory name is given as the <cert> argument, haproxy will
11545 automatically search and load bundled files in that directory.
11546
11547 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
11548 bundling. Each certificate can have its own .ocsp and .issuer file. At this
11549 time, sctl is not supported in multi-certificate bundling.
11550
Emeric Brunb6dc9342012-09-28 17:55:37 +020011551crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000011552 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010011553 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011554 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000011555 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020011556
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011557crt-list <file>
11558 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011559 designates a list of PEM file with an optional ssl configuration and a SNI
11560 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011561
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011562 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
11563
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011564 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
11565 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011566 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011567 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011568
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020011569 Wildcards are supported in the SNI filter. Negative filter are also supported,
11570 only useful in combination with a wildcard filter to exclude a particular SNI.
11571 The certificates will be presented to clients who provide a valid TLS Server
11572 Name Indication field matching one of the SNI filters. If no SNI filter is
11573 specified, the CN and alt subjects are used. This directive may be specified
11574 multiple times. See the "crt" option for more information. The default
11575 certificate is still needed to meet OpenSSL expectations. If it is not used,
11576 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011577
yanbzhu6c25e9e2016-01-05 12:52:02 -050011578 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020011579 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011580 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
11581 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
11582 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050011583
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011584 crt-list file example:
11585 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011586 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011587 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011588 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011589
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011590defer-accept
11591 Is an optional keyword which is supported only on certain Linux kernels. It
11592 states that a connection will only be accepted once some data arrive on it,
11593 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010011594 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011595 performance by ensuring that most of the request is already available when
11596 the connection is accepted. On the other hand, it will not be able to detect
11597 connections which don't talk. It is important to note that this option is
11598 broken in all kernels up to 2.6.31, as the connection is never accepted until
11599 the client talks. This can cause issues with front firewalls which would see
11600 an established connection while the proxy will only see it in SYN_RECV. This
11601 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
11602
William Lallemandf6975e92017-05-26 17:42:10 +020011603expose-fd listeners
11604 This option is only usable with the stats socket. It gives your stats socket
11605 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020011606 During a reload with the master-worker mode, the process is automatically
11607 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011608 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020011609
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011610force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011611 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011612 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011613 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011614 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011615
11616force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011617 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011618 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011619 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011620
11621force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011622 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011623 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011624 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011625
11626force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011627 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011628 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011629 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011630
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011631force-tlsv13
11632 This option enforces use of TLSv1.3 only on SSL connections instantiated from
11633 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011634 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011635
Christopher Faulet31af49d2015-06-09 17:29:50 +020011636generate-certificates
11637 This setting is only available when support for OpenSSL was built in. It
11638 enables the dynamic SSL certificates generation. A CA certificate and its
11639 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
11640 a transparent forward proxy, SSL requests generate errors because of a common
11641 name mismatch on the certificate presented to the client. With this option
11642 enabled, HAProxy will try to forge a certificate using the SNI hostname
11643 indicated by the client. This is done only if no certificate matches the SNI
11644 hostname (see 'crt-list'). If an error occurs, the default certificate is
11645 used, else the 'strict-sni' option is set.
11646 It can also be used when HAProxy is configured as a reverse proxy to ease the
11647 deployment of an architecture with many backends.
11648
11649 Creating a SSL certificate is an expensive operation, so a LRU cache is used
11650 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010011651 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020011652 certificate is used many times.
11653
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011654gid <gid>
11655 Sets the group of the UNIX sockets to the designated system gid. It can also
11656 be set by default in the global section's "unix-bind" statement. Note that
11657 some platforms simply ignore this. This setting is equivalent to the "group"
11658 setting except that the group ID is used instead of its name. This setting is
11659 ignored by non UNIX sockets.
11660
11661group <group>
11662 Sets the group of the UNIX sockets to the designated system group. It can
11663 also be set by default in the global section's "unix-bind" statement. Note
11664 that some platforms simply ignore this. This setting is equivalent to the
11665 "gid" setting except that the group name is used instead of its gid. This
11666 setting is ignored by non UNIX sockets.
11667
11668id <id>
11669 Fixes the socket ID. By default, socket IDs are automatically assigned, but
11670 sometimes it is more convenient to fix them to ease monitoring. This value
11671 must be strictly positive and unique within the listener/frontend. This
11672 option can only be used when defining only a single socket.
11673
11674interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010011675 Restricts the socket to a specific interface. When specified, only packets
11676 received from that particular interface are processed by the socket. This is
11677 currently only supported on Linux. The interface must be a primary system
11678 interface, not an aliased interface. It is also possible to bind multiple
11679 frontends to the same address if they are bound to different interfaces. Note
11680 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010011681 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
11682 uses the same interface as inbound traffic, and its associated routing table,
11683 even if there are explicit routes through different interfaces configured.
11684 This can prove useful to address asymmetric routing issues when the same
11685 client IP addresses need to be able to reach frontends hosted on different
11686 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011687
Willy Tarreauabb175f2012-09-24 12:43:26 +020011688level <level>
11689 This setting is used with the stats sockets only to restrict the nature of
11690 the commands that can be issued on the socket. It is ignored by other
11691 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010011692 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020011693 read, and no change is allowed. It would make sense on systems where it
11694 is not easy to restrict access to the socket.
11695 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011696 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020011697 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010011698 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020011699 all counters).
11700
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020011701severity-output <format>
11702 This setting is used with the stats sockets only to configure severity
11703 level output prepended to informational feedback messages. Severity
11704 level of messages can range between 0 and 7, conforming to syslog
11705 rfc5424. Valid and successful socket commands requesting data
11706 (i.e. "show map", "get acl foo" etc.) will never have a severity level
11707 prepended. It is ignored by other sockets. <format> can be one of :
11708 - "none" (default) no severity level is prepended to feedback messages.
11709 - "number" severity level is prepended as a number.
11710 - "string" severity level is prepended as a string following the
11711 rfc5424 convention.
11712
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011713maxconn <maxconn>
11714 Limits the sockets to this number of concurrent connections. Extraneous
11715 connections will remain in the system's backlog until a connection is
11716 released. If unspecified, the limit will be the same as the frontend's
11717 maxconn. Note that in case of port ranges or multiple addresses, the same
11718 value will be applied to each socket. This setting enables different
11719 limitations on expensive sockets, for instance SSL entries which may easily
11720 eat all memory.
11721
11722mode <mode>
11723 Sets the octal mode used to define access permissions on the UNIX socket. It
11724 can also be set by default in the global section's "unix-bind" statement.
11725 Note that some platforms simply ignore this. This setting is ignored by non
11726 UNIX sockets.
11727
11728mss <maxseg>
11729 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
11730 connections. This can be used to force a lower MSS for certain specific
11731 ports, for instance for connections passing through a VPN. Note that this
11732 relies on a kernel feature which is theoretically supported under Linux but
11733 was buggy in all versions prior to 2.6.28. It may or may not work on other
11734 operating systems. It may also not change the advertised value but change the
11735 effective size of outgoing segments. The commonly advertised value for TCPv4
11736 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
11737 positive, it will be used as the advertised MSS. If it is negative, it will
11738 indicate by how much to reduce the incoming connection's advertised MSS for
11739 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
11740
11741name <name>
11742 Sets an optional name for these sockets, which will be reported on the stats
11743 page.
11744
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011745namespace <name>
11746 On Linux, it is possible to specify which network namespace a socket will
11747 belong to. This directive makes it possible to explicitly bind a listener to
11748 a namespace different from the default one. Please refer to your operating
11749 system's documentation to find more details about network namespaces.
11750
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011751nice <nice>
11752 Sets the 'niceness' of connections initiated from the socket. Value must be
11753 in the range -1024..1024 inclusive, and defaults to zero. Positive values
11754 means that such connections are more friendly to others and easily offer
11755 their place in the scheduler. On the opposite, negative values mean that
11756 connections want to run with a higher priority than others. The difference
11757 only happens under high loads when the system is close to saturation.
11758 Negative values are appropriate for low-latency or administration services,
11759 and high values are generally recommended for CPU intensive tasks such as SSL
11760 processing or bulk transfers which are less sensible to latency. For example,
11761 it may make sense to use a positive value for an SMTP socket and a negative
11762 one for an RDP socket.
11763
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011764no-ca-names
11765 This setting is only available when support for OpenSSL was built in. It
11766 prevents from send CA names in server hello message when ca-file is used.
11767
Emeric Brun9b3009b2012-10-05 11:55:06 +020011768no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011769 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011770 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011771 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011772 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011773 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
11774 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011775
Emeric Brun90ad8722012-10-02 14:00:59 +020011776no-tls-tickets
11777 This setting is only available when support for OpenSSL was built in. It
11778 disables the stateless session resumption (RFC 5077 TLS Ticket
11779 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011780 session resumption is more expensive in CPU usage. This option is also
11781 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020011782
Emeric Brun9b3009b2012-10-05 11:55:06 +020011783no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011784 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011785 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011786 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011787 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011788 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11789 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011790
Emeric Brun9b3009b2012-10-05 11:55:06 +020011791no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020011792 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011793 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011794 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011795 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011796 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11797 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011798
Emeric Brun9b3009b2012-10-05 11:55:06 +020011799no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020011800 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011801 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011802 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011803 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011804 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11805 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011806
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011807no-tlsv13
11808 This setting is only available when support for OpenSSL was built in. It
11809 disables support for TLSv1.3 on any sockets instantiated from the listener
11810 when SSL is supported. Note that SSLv2 is forced disabled in the code and
11811 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011812 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11813 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011814
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011815npn <protocols>
11816 This enables the NPN TLS extension and advertises the specified protocol list
11817 as supported on top of NPN. The protocol list consists in a comma-delimited
11818 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roesler27c754c2019-07-10 15:45:51 -050011819 This requires that the SSL library is built with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020011820 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010011821 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11822 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
11823 version of OpenSSL, NPN might still be used as most clients still support it
11824 at the time of writing this. It is possible to enable both NPN and ALPN
11825 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011826
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011827prefer-client-ciphers
11828 Use the client's preference when selecting the cipher suite, by default
11829 the server's preference is enforced. This option is also available on
11830 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020011831 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
11832 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
11833 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011834
Christopher Fauletc644fa92017-11-23 22:44:11 +010011835process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010011836 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010011837 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010011838 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010011839 between the two is applied. If in the end the listener is not allowed to run
11840 on any remaining process, a warning is emitted, and the listener will either
11841 run on the first process of the listener if a single process was specified,
11842 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010011843 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010011844 connections for this listener, for the the process set. If multiple processes
11845 and threads are configured, a warning is emitted, as it either results from a
11846 configuration error or a misunderstanding of these models. For the unlikely
11847 case where several ranges are needed, this directive may be repeated.
11848 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010011849
11850 all | odd | even | number[-[number]]
11851
11852 Ranges can be partially defined. The higher bound can be omitted. In such
11853 case, it is replaced by the corresponding maximum value. The main purpose of
11854 this directive is to be used with the stats sockets and have one different
11855 socket per process. The second purpose is to have multiple bind lines sharing
11856 the same IP:port but not the same process in a listener, so that the system
11857 can distribute the incoming connections into multiple queues and allow a
11858 smoother inter-process load balancing. Currently Linux 3.9 and above is known
11859 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020011860
Christopher Fauleta717b992018-04-10 14:43:00 +020011861proto <name>
11862 Forces the multiplexer's protocol to use for the incoming connections. It
11863 must be compatible with the mode of the frontend (TCP or HTTP). It must also
11864 be usable on the frontend side. The list of available protocols is reported
11865 in haproxy -vv.
11866 Idea behind this optipon is to bypass the selection of the best multiplexer's
11867 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080011868 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020011869 h2" on the bind line.
11870
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011871ssl
11872 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011873 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011874 certificate is necessary (see "crt" above). All contents in the buffers will
11875 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011876 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11877 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011878
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011879ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11880 This option enforces use of <version> or lower on SSL connections instantiated
11881 from this listener. This option is also available on global statement
11882 "ssl-default-bind-options". See also "ssl-min-ver".
11883
11884ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11885 This option enforces use of <version> or upper on SSL connections instantiated
11886 from this listener. This option is also available on global statement
11887 "ssl-default-bind-options". See also "ssl-max-ver".
11888
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011889strict-sni
11890 This setting is only available when support for OpenSSL was built in. The
11891 SSL/TLS negotiation is allow only if the client provided an SNI which match
11892 a certificate. The default certificate is not used.
11893 See the "crt" option for more information.
11894
Willy Tarreau2af207a2015-02-04 00:45:58 +010011895tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011896 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011897 listening socket. This option is available on Linux since version 2.6.37. It
11898 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011899 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011900 useful on long-lived connections experiencing long idle periods such as
11901 remote terminals or database connection pools, where the client and server
11902 timeouts must remain high to allow a long period of idle, but where it is
11903 important to detect that the client has disappeared in order to release all
11904 resources associated with its connection (and the server's session). The
11905 argument is a delay expressed in milliseconds by default. This only works
11906 for regular TCP connections, and is ignored for other protocols.
11907
Willy Tarreau1c862c52012-10-05 16:21:00 +020011908tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011909 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011910 enables TCP Fast Open on the listening socket, which means that clients which
11911 support this feature will be able to send a request and receive a response
11912 during the 3-way handshake starting from second connection, thus saving one
11913 round-trip after the first connection. This only makes sense with protocols
11914 that use high connection rates and where each round trip matters. This can
11915 possibly cause issues with many firewalls which do not accept data on SYN
11916 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011917 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11918 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11919 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011920
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011921tls-ticket-keys <keyfile>
11922 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010011923 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
11924 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
11925 The first key determines the key length used for next keys: you can't mix
11926 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
11927 build option (default 3) and at least as many keys need to be present in
11928 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
11929 penultimate one for encryption. This enables easy key rotation by just
11930 appending new key to the file and reloading the process. Keys must be
11931 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
11932 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011933 storage such as hard drives (hint: use tmpfs and don't swap those files).
11934 Lifetime hint can be changed using tune.ssl.timeout.
11935
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011936transparent
11937 Is an optional keyword which is supported only on certain Linux kernels. It
11938 indicates that the addresses will be bound even if they do not belong to the
11939 local machine, and that packets targeting any of these addresses will be
11940 intercepted just as if the addresses were locally configured. This normally
11941 requires that IP forwarding is enabled. Caution! do not use this with the
11942 default address '*', as it would redirect any traffic for the specified port.
11943 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11944 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11945 kernel version. Some distribution kernels include backports of the feature,
11946 so check for support with your vendor.
11947
Willy Tarreau77e3af92012-11-24 15:07:23 +010011948v4v6
11949 Is an optional keyword which is supported only on most recent systems
11950 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11951 and IPv6 when it uses the default address. Doing so is sometimes necessary
11952 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011953 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011954
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011955v6only
11956 Is an optional keyword which is supported only on most recent systems
11957 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11958 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011959 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11960 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011961
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011962uid <uid>
11963 Sets the owner of the UNIX sockets to the designated system uid. It can also
11964 be set by default in the global section's "unix-bind" statement. Note that
11965 some platforms simply ignore this. This setting is equivalent to the "user"
11966 setting except that the user numeric ID is used instead of its name. This
11967 setting is ignored by non UNIX sockets.
11968
11969user <user>
11970 Sets the owner of the UNIX sockets to the designated system user. It can also
11971 be set by default in the global section's "unix-bind" statement. Note that
11972 some platforms simply ignore this. This setting is equivalent to the "uid"
11973 setting except that the user name is used instead of its uid. This setting is
11974 ignored by non UNIX sockets.
11975
Emeric Brun1a073b42012-09-28 17:07:34 +020011976verify [none|optional|required]
11977 This setting is only available when support for OpenSSL was built in. If set
11978 to 'none', client certificate is not requested. This is the default. In other
11979 cases, a client certificate is requested. If the client does not provide a
11980 certificate after the request and if 'verify' is set to 'required', then the
11981 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011982 certificate provided by the client is always verified using CAs from
11983 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11984 is aborted, regardless of the 'verify' option, unless the error code exactly
11985 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011986
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200119875.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011988------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011989
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011990The "server" and "default-server" keywords support a certain number of settings
11991which are all passed as arguments on the server line. The order in which those
11992arguments appear does not count, and they are all optional. Some of those
11993settings are single words (booleans) while others expect one or several values
11994after them. In this case, the values must immediately follow the setting name.
11995Except default-server, all those settings must be specified after the server's
11996address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011997
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011998 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011999 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020012000
Frédéric Lécailled2376272017-03-21 18:52:12 +010012001Note that all these settings are supported both by "server" and "default-server"
12002keywords, except "id" which is only supported by "server".
12003
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012004The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010012005
Willy Tarreauceb4ac92012-04-28 00:41:46 +020012006addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012007 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010012008 to send health-checks or to probe the agent-check. On some servers, it may be
12009 desirable to dedicate an IP address to specific component able to perform
12010 complex tests which are more suitable to health-checks than the application.
12011 This parameter is ignored if the "check" parameter is not set. See also the
12012 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012013
Simon Hormand60d6912013-11-25 10:46:36 +090012014agent-check
12015 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010012016 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010012017 to the port set by the "agent-port" parameter and reading an ASCII string
12018 terminated by the first '\r' or '\n' met. The string is made of a series of
12019 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090012020
Willy Tarreau81f5d942013-12-09 20:51:51 +010012021 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090012022 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020012023 weight of a server as configured when haproxy starts. Note that a zero
12024 weight is reported on the stats page as "DRAIN" since it has the same
12025 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090012026
Davor Ocelice9ed2812017-12-25 17:49:28 +010012027 - The string "maxconn:" followed by an integer (no space between). Values
12028 in this format will set the maxconn of a server. The maximum number of
12029 connections advertised needs to be multiplied by the number of load
12030 balancers and different backends that use this health check to get the
12031 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020012032
Willy Tarreau81f5d942013-12-09 20:51:51 +010012033 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012034 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090012035
Willy Tarreau81f5d942013-12-09 20:51:51 +010012036 - The word "drain". This will turn the server's administrative state to the
12037 DRAIN mode, thus it will not accept any new connections other than those
12038 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090012039
Willy Tarreau81f5d942013-12-09 20:51:51 +010012040 - The word "maint". This will turn the server's administrative state to the
12041 MAINT mode, thus it will not accept any new connections at all, and health
12042 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090012043
Willy Tarreau81f5d942013-12-09 20:51:51 +010012044 - The words "down", "failed", or "stopped", optionally followed by a
12045 description string after a sharp ('#'). All of these mark the server's
12046 operating state as DOWN, but since the word itself is reported on the stats
12047 page, the difference allows an administrator to know if the situation was
12048 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010012049 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010012050 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090012051
Willy Tarreau81f5d942013-12-09 20:51:51 +010012052 - The word "up" sets back the server's operating state as UP if health checks
12053 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090012054
Willy Tarreau81f5d942013-12-09 20:51:51 +010012055 Parameters which are not advertised by the agent are not changed. For
12056 example, an agent might be designed to monitor CPU usage and only report a
12057 relative weight and never interact with the operating status. Similarly, an
12058 agent could be designed as an end-user interface with 3 radio buttons
12059 allowing an administrator to change only the administrative state. However,
12060 it is important to consider that only the agent may revert its own actions,
12061 so if a server is set to DRAIN mode or to DOWN state using the agent, the
12062 agent must implement the other equivalent actions to bring the service into
12063 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090012064
Simon Horman2f1f9552013-11-25 10:46:37 +090012065 Failure to connect to the agent is not considered an error as connectivity
12066 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010012067 parameter. Warning though, it is not a good idea to stop an agent after it
12068 reports "down", since only an agent reporting "up" will be able to turn the
12069 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010012070 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090012071
Willy Tarreau81f5d942013-12-09 20:51:51 +010012072 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010012073 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090012074
James Brown55f9ff12015-10-21 18:19:05 -070012075agent-send <string>
12076 If this option is specified, haproxy will send the given string (verbatim)
12077 to the agent server upon connection. You could, for example, encode
12078 the backend name into this string, which would enable your agent to send
12079 different responses based on the backend. Make sure to include a '\n' if
12080 you want to terminate your request with a newline.
12081
Simon Hormand60d6912013-11-25 10:46:36 +090012082agent-inter <delay>
12083 The "agent-inter" parameter sets the interval between two agent checks
12084 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
12085
12086 Just as with every other time-based parameter, it may be entered in any
12087 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
12088 parameter also serves as a timeout for agent checks "timeout check" is
12089 not set. In order to reduce "resonance" effects when multiple servers are
12090 hosted on the same hardware, the agent and health checks of all servers
12091 are started with a small time offset between them. It is also possible to
12092 add some random noise in the agent and health checks interval using the
12093 global "spread-checks" keyword. This makes sense for instance when a lot
12094 of backends use the same servers.
12095
12096 See also the "agent-check" and "agent-port" parameters.
12097
Misiek768d8602017-01-09 09:52:43 +010012098agent-addr <addr>
12099 The "agent-addr" parameter sets address for agent check.
12100
12101 You can offload agent-check to another target, so you can make single place
12102 managing status and weights of servers defined in haproxy in case you can't
12103 make self-aware and self-managing services. You can specify both IP or
12104 hostname, it will be resolved.
12105
Simon Hormand60d6912013-11-25 10:46:36 +090012106agent-port <port>
12107 The "agent-port" parameter sets the TCP port used for agent checks.
12108
12109 See also the "agent-check" and "agent-inter" parameters.
12110
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020012111allow-0rtt
12112 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020012113 Note that early data will be sent only if the client used early data, or
12114 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020012115
Olivier Houchardc7566002018-11-20 23:33:50 +010012116alpn <protocols>
12117 This enables the TLS ALPN extension and advertises the specified protocol
12118 list as supported on top of ALPN. The protocol list consists in a comma-
12119 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
John Roesler27c754c2019-07-10 15:45:51 -050012120 quotes). This requires that the SSL library is built with support for TLS
Olivier Houchardc7566002018-11-20 23:33:50 +010012121 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
12122 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
12123 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
12124 now obsolete NPN extension.
12125 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
12126 be advertised, in order of preference, like below :
12127
12128 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
12129
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012130backup
12131 When "backup" is present on a server line, the server is only used in load
12132 balancing when all other non-backup servers are unavailable. Requests coming
12133 with a persistence cookie referencing the server will always be served
12134 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010012135 the "allbackups" option is set in the backend. See also the "no-backup" and
12136 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012137
Emeric Brunef42d922012-10-11 16:11:36 +020012138ca-file <cafile>
12139 This setting is only available when support for OpenSSL was built in. It
12140 designates a PEM file from which to load CA certificates used to verify
12141 server's certificate.
12142
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012143check
12144 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010012145 always considered available. If "check" is set, the server is available when
12146 accepting periodic TCP connections, to ensure that it is really able to serve
12147 requests. The default address and port to send the tests to are those of the
12148 server, and the default source is the same as the one defined in the
12149 backend. It is possible to change the address using the "addr" parameter, the
12150 port using the "port" parameter, the source address using the "source"
12151 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090012152 parameters. The request method is define in the backend using the "httpchk",
12153 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010012154 refer to those options and parameters for more information. See also
12155 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012156
Willy Tarreau6c16adc2012-10-05 00:04:16 +020012157check-send-proxy
12158 This option forces emission of a PROXY protocol line with outgoing health
12159 checks, regardless of whether the server uses send-proxy or not for the
12160 normal traffic. By default, the PROXY protocol is enabled for health checks
12161 if it is already enabled for normal traffic and if no "port" nor "addr"
12162 directive is present. However, if such a directive is present, the
12163 "check-send-proxy" option needs to be used to force the use of the
12164 protocol. See also the "send-proxy" option for more information.
12165
Olivier Houchard92150142018-12-21 19:47:01 +010012166check-alpn <protocols>
12167 Defines which protocols to advertise with ALPN. The protocol list consists in
12168 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
12169 (without quotes). If it is not set, the server ALPN is used.
12170
Jérôme Magninae9bb762018-12-09 16:08:26 +010012171check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020012172 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010012173 over SSL. It is only possible to use a string to set <sni>. If you want to
12174 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020012175
Willy Tarreau763a95b2012-10-04 23:15:39 +020012176check-ssl
12177 This option forces encryption of all health checks over SSL, regardless of
12178 whether the server uses SSL or not for the normal traffic. This is generally
12179 used when an explicit "port" or "addr" directive is specified and SSL health
12180 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012181 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020012182 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
12183 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012184 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010012185 See the "ssl" option for more information and "no-check-ssl" to disable
12186 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012187
Alexander Liu2a54bb72019-05-22 19:44:48 +080012188check-via-socks4
John Roesler27c754c2019-07-10 15:45:51 -050012189 This option enables outgoing health checks using upstream socks4 proxy. By
Alexander Liu2a54bb72019-05-22 19:44:48 +080012190 default, the health checks won't go through socks tunnel even it was enabled
12191 for normal traffic.
12192
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012193ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012194 This setting is only available when support for OpenSSL was built in. This
12195 option sets the string describing the list of cipher algorithms that is
12196 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000012197 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
12198 information and recommendations see e.g.
12199 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
12200 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
12201 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012202
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012203ciphersuites <ciphersuites>
12204 This setting is only available when support for OpenSSL was built in and
12205 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
12206 describing the list of cipher algorithms that is negotiated during the TLS
12207 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000012208 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
12209 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
12210 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012211
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012212cookie <value>
12213 The "cookie" parameter sets the cookie value assigned to the server to
12214 <value>. This value will be checked in incoming requests, and the first
12215 operational server possessing the same value will be selected. In return, in
12216 cookie insertion or rewrite modes, this value will be assigned to the cookie
12217 sent to the client. There is nothing wrong in having several servers sharing
12218 the same cookie value, and it is in fact somewhat common between normal and
12219 backup servers. See also the "cookie" keyword in backend section.
12220
Emeric Brunef42d922012-10-11 16:11:36 +020012221crl-file <crlfile>
12222 This setting is only available when support for OpenSSL was built in. It
12223 designates a PEM file from which to load certificate revocation list used
12224 to verify server's certificate.
12225
Emeric Bruna7aa3092012-10-26 12:58:00 +020012226crt <cert>
12227 This setting is only available when support for OpenSSL was built in.
12228 It designates a PEM file from which to load both a certificate and the
12229 associated private key. This file can be built by concatenating both PEM
12230 files into one. This certificate will be sent if the server send a client
12231 certificate request.
12232
Willy Tarreau96839092010-03-29 10:02:24 +020012233disabled
12234 The "disabled" keyword starts the server in the "disabled" state. That means
12235 that it is marked down in maintenance mode, and no connection other than the
12236 ones allowed by persist mode will reach it. It is very well suited to setup
12237 new servers, because normal traffic will never reach them, while it is still
12238 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012239 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012240
Frédéric Lécailled2376272017-03-21 18:52:12 +010012241enabled
12242 This option may be used as 'server' setting to reset any 'disabled'
12243 setting which would have been inherited from 'default-server' directive as
12244 default value.
12245 It may also be used as 'default-server' setting to reset any previous
12246 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012247
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012248error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010012249 If health observing is enabled, the "error-limit" parameter specifies the
12250 number of consecutive errors that triggers event selected by the "on-error"
12251 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012252
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012253 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012254
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012255fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012256 The "fall" parameter states that a server will be considered as dead after
12257 <count> consecutive unsuccessful health checks. This value defaults to 3 if
12258 unspecified. See also the "check", "inter" and "rise" parameters.
12259
Emeric Brun8694b9a2012-10-05 14:39:07 +020012260force-sslv3
12261 This option enforces use of SSLv3 only when SSL is used to communicate with
12262 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012263 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012264 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012265
12266force-tlsv10
12267 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012268 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012269 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012270
12271force-tlsv11
12272 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012273 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012274 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012275
12276force-tlsv12
12277 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012278 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012279 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012280
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012281force-tlsv13
12282 This option enforces use of TLSv1.3 only when SSL is used to communicate with
12283 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012284 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012285
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012286id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020012287 Set a persistent ID for the server. This ID must be positive and unique for
12288 the proxy. An unused ID will automatically be assigned if unset. The first
12289 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012290
Willy Tarreau6a031d12016-11-07 19:42:35 +010012291init-addr {last | libc | none | <ip>},[...]*
12292 Indicate in what order the server's address should be resolved upon startup
12293 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012294 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010012295 method which succeeds is used. If the end of the list is reached without
12296 finding a working method, an error is thrown. Method "last" suggests to pick
12297 the address which appears in the state file (see "server-state-file"). Method
12298 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
12299 depending on the operating system and build options). Method "none"
12300 specifically indicates that the server should start without any valid IP
12301 address in a down state. It can be useful to ignore some DNS issues upon
12302 startup, waiting for the situation to get fixed later. Finally, an IP address
12303 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012304 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010012305 server used to catch old sessions and present them with a decent error
12306 message for example. When the "first" load balancing algorithm is used, this
12307 IP address could point to a fake server used to trigger the creation of new
12308 instances on the fly. This option defaults to "last,libc" indicating that the
12309 previous address found in the state file (if any) is used first, otherwise
12310 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012311 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010012312
12313 Example:
12314 defaults
12315 # never fail on address resolution
12316 default-server init-addr last,libc,none
12317
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012318inter <delay>
12319fastinter <delay>
12320downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012321 The "inter" parameter sets the interval between two consecutive health checks
12322 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
12323 It is also possible to use "fastinter" and "downinter" to optimize delays
12324 between checks depending on the server state :
12325
Pieter Baauw44fc9df2015-09-17 21:30:46 +020012326 Server state | Interval used
12327 ----------------------------------------+----------------------------------
12328 UP 100% (non-transitional) | "inter"
12329 ----------------------------------------+----------------------------------
12330 Transitionally UP (going down "fall"), | "fastinter" if set,
12331 Transitionally DOWN (going up "rise"), | "inter" otherwise.
12332 or yet unchecked. |
12333 ----------------------------------------+----------------------------------
12334 DOWN 100% (non-transitional) | "downinter" if set,
12335 | "inter" otherwise.
12336 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010012337
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012338 Just as with every other time-based parameter, they can be entered in any
12339 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
12340 serves as a timeout for health checks sent to servers if "timeout check" is
12341 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090012342 hosted on the same hardware, the agent and health checks of all servers
12343 are started with a small time offset between them. It is also possible to
12344 add some random noise in the agent and health checks interval using the
12345 global "spread-checks" keyword. This makes sense for instance when a lot
12346 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012347
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012348maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012349 The "maxconn" parameter specifies the maximal number of concurrent
12350 connections that will be sent to this server. If the number of incoming
Tim Duesterhus50cfb312019-11-27 22:35:27 +010012351 concurrent connections goes higher than this value, they will be queued,
12352 waiting for a slot to be released. This parameter is very important as it can
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012353 save fragile servers from going down under extreme loads. If a "minconn"
12354 parameter is specified, the limit becomes dynamic. The default value is "0"
12355 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
12356 the backend's "fullconn" keyword.
12357
Tim Duesterhus50cfb312019-11-27 22:35:27 +010012358 In HTTP mode this parameter limits the number of concurrent requests instead
12359 of the number of connections. Multiple requests might be multiplexed over a
12360 single TCP connection to the server. As an example if you specify a maxconn
12361 of 50 you might see between 1 and 50 actual server connections, but no more
12362 than 50 concurrent requests.
12363
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012364maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012365 The "maxqueue" parameter specifies the maximal number of connections which
12366 will wait in the queue for this server. If this limit is reached, next
12367 requests will be redispatched to other servers instead of indefinitely
12368 waiting to be served. This will break persistence but may allow people to
12369 quickly re-log in when the server they try to connect to is dying. The
12370 default value is "0" which means the queue is unlimited. See also the
12371 "maxconn" and "minconn" parameters.
12372
Willy Tarreau9c538e02019-01-23 10:21:49 +010012373max-reuse <count>
12374 The "max-reuse" argument indicates the HTTP connection processors that they
12375 should not reuse a server connection more than this number of times to send
12376 new requests. Permitted values are -1 (the default), which disables this
12377 limit, or any positive value. Value zero will effectively disable keep-alive.
12378 This is only used to work around certain server bugs which cause them to leak
12379 resources over time. The argument is not necessarily respected by the lower
12380 layers as there might be technical limitations making it impossible to
12381 enforce. At least HTTP/2 connections to servers will respect it.
12382
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012383minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012384 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
12385 limit following the backend's load. The server will always accept at least
12386 <minconn> connections, never more than <maxconn>, and the limit will be on
12387 the ramp between both values when the backend has less than <fullconn>
12388 concurrent connections. This makes it possible to limit the load on the
12389 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012390 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012391 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012392
Willy Tarreaud72f0f32015-10-13 14:50:22 +020012393namespace <name>
12394 On Linux, it is possible to specify which network namespace a socket will
12395 belong to. This directive makes it possible to explicitly bind a server to
12396 a namespace different from the default one. Please refer to your operating
12397 system's documentation to find more details about network namespaces.
12398
Frédéric Lécailled2376272017-03-21 18:52:12 +010012399no-agent-check
12400 This option may be used as "server" setting to reset any "agent-check"
12401 setting which would have been inherited from "default-server" directive as
12402 default value.
12403 It may also be used as "default-server" setting to reset any previous
12404 "default-server" "agent-check" setting.
12405
12406no-backup
12407 This option may be used as "server" setting to reset any "backup"
12408 setting which would have been inherited from "default-server" directive as
12409 default value.
12410 It may also be used as "default-server" setting to reset any previous
12411 "default-server" "backup" setting.
12412
12413no-check
12414 This option may be used as "server" setting to reset any "check"
12415 setting which would have been inherited from "default-server" directive as
12416 default value.
12417 It may also be used as "default-server" setting to reset any previous
12418 "default-server" "check" setting.
12419
12420no-check-ssl
12421 This option may be used as "server" setting to reset any "check-ssl"
12422 setting which would have been inherited from "default-server" directive as
12423 default value.
12424 It may also be used as "default-server" setting to reset any previous
12425 "default-server" "check-ssl" setting.
12426
Frédéric Lécailled2376272017-03-21 18:52:12 +010012427no-send-proxy
12428 This option may be used as "server" setting to reset any "send-proxy"
12429 setting which would have been inherited from "default-server" directive as
12430 default value.
12431 It may also be used as "default-server" setting to reset any previous
12432 "default-server" "send-proxy" setting.
12433
12434no-send-proxy-v2
12435 This option may be used as "server" setting to reset any "send-proxy-v2"
12436 setting which would have been inherited from "default-server" directive as
12437 default value.
12438 It may also be used as "default-server" setting to reset any previous
12439 "default-server" "send-proxy-v2" setting.
12440
12441no-send-proxy-v2-ssl
12442 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
12443 setting which would have been inherited from "default-server" directive as
12444 default value.
12445 It may also be used as "default-server" setting to reset any previous
12446 "default-server" "send-proxy-v2-ssl" setting.
12447
12448no-send-proxy-v2-ssl-cn
12449 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
12450 setting which would have been inherited from "default-server" directive as
12451 default value.
12452 It may also be used as "default-server" setting to reset any previous
12453 "default-server" "send-proxy-v2-ssl-cn" setting.
12454
12455no-ssl
12456 This option may be used as "server" setting to reset any "ssl"
12457 setting which would have been inherited from "default-server" directive as
12458 default value.
12459 It may also be used as "default-server" setting to reset any previous
12460 "default-server" "ssl" setting.
12461
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010012462no-ssl-reuse
12463 This option disables SSL session reuse when SSL is used to communicate with
12464 the server. It will force the server to perform a full handshake for every
12465 new connection. It's probably only useful for benchmarking, troubleshooting,
12466 and for paranoid users.
12467
Emeric Brun9b3009b2012-10-05 11:55:06 +020012468no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012469 This option disables support for SSLv3 when SSL is used to communicate with
12470 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012471 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012472
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012473 Supported in default-server: No
12474
Emeric Brunf9c5c472012-10-11 15:28:34 +020012475no-tls-tickets
12476 This setting is only available when support for OpenSSL was built in. It
12477 disables the stateless session resumption (RFC 5077 TLS Ticket
12478 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012479 session resumption is more expensive in CPU usage for servers. This option
12480 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010012481 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020012482
Emeric Brun9b3009b2012-10-05 11:55:06 +020012483no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020012484 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012485 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12486 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012487 often makes sense to disable it when communicating with local servers. This
12488 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012489 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012490
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012491 Supported in default-server: No
12492
Emeric Brun9b3009b2012-10-05 11:55:06 +020012493no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020012494 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012495 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12496 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012497 often makes sense to disable it when communicating with local servers. This
12498 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012499 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012500
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012501 Supported in default-server: No
12502
Emeric Brun9b3009b2012-10-05 11:55:06 +020012503no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020012504 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012505 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12506 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012507 often makes sense to disable it when communicating with local servers. This
12508 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012509 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012510
12511 Supported in default-server: No
12512
12513no-tlsv13
12514 This option disables support for TLSv1.3 when SSL is used to communicate with
12515 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12516 using any configuration option. TLSv1 is more expensive than SSLv3 so it
12517 often makes sense to disable it when communicating with local servers. This
12518 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012519 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012520
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012521 Supported in default-server: No
12522
Frédéric Lécailled2376272017-03-21 18:52:12 +010012523no-verifyhost
12524 This option may be used as "server" setting to reset any "verifyhost"
12525 setting which would have been inherited from "default-server" directive as
12526 default value.
12527 It may also be used as "default-server" setting to reset any previous
12528 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012529
Frédéric Lécailleaeeb1c92019-07-04 14:19:06 +020012530no-tfo
12531 This option may be used as "server" setting to reset any "tfo"
12532 setting which would have been inherited from "default-server" directive as
12533 default value.
12534 It may also be used as "default-server" setting to reset any previous
12535 "default-server" "tfo" setting.
12536
Simon Hormanfa461682011-06-25 09:39:49 +090012537non-stick
12538 Never add connections allocated to this sever to a stick-table.
12539 This may be used in conjunction with backup to ensure that
12540 stick-table persistence is disabled for backup servers.
12541
Olivier Houchardc7566002018-11-20 23:33:50 +010012542npn <protocols>
12543 This enables the NPN TLS extension and advertises the specified protocol list
12544 as supported on top of NPN. The protocol list consists in a comma-delimited
12545 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
John Roesler27c754c2019-07-10 15:45:51 -050012546 This requires that the SSL library is built with support for TLS extensions
Olivier Houchardc7566002018-11-20 23:33:50 +010012547 enabled (check with haproxy -vv). Note that the NPN extension has been
12548 replaced with the ALPN extension (see the "alpn" keyword), though this one is
12549 only available starting with OpenSSL 1.0.2.
12550
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012551observe <mode>
12552 This option enables health adjusting based on observing communication with
12553 the server. By default this functionality is disabled and enabling it also
12554 requires to enable health checks. There are two supported modes: "layer4" and
12555 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
12556 significant. In layer7, which is only allowed for http proxies, responses
12557 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010012558 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012559
12560 See also the "check", "on-error" and "error-limit".
12561
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012562on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012563 Select what should happen when enough consecutive errors are detected.
12564 Currently, four modes are available:
12565 - fastinter: force fastinter
12566 - fail-check: simulate a failed check, also forces fastinter (default)
12567 - sudden-death: simulate a pre-fatal failed health check, one more failed
12568 check will mark a server down, forces fastinter
12569 - mark-down: mark the server immediately down and force fastinter
12570
12571 See also the "check", "observe" and "error-limit".
12572
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012573on-marked-down <action>
12574 Modify what occurs when a server is marked down.
12575 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012576 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
12577 all connections to the server are immediately terminated when the server
12578 goes down. It might be used if the health check detects more complex cases
12579 than a simple connection status, and long timeouts would cause the service
12580 to remain unresponsive for too long a time. For instance, a health check
12581 might detect that a database is stuck and that there's no chance to reuse
12582 existing connections anymore. Connections killed this way are logged with
12583 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012584
12585 Actions are disabled by default
12586
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012587on-marked-up <action>
12588 Modify what occurs when a server is marked up.
12589 Currently one action is available:
12590 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
12591 done only if the server is not in backup state and if it is not disabled
12592 (it must have an effective weight > 0). This can be used sometimes to force
12593 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012594 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
12595 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012596 with extreme care. Sessions killed because a server comes up are logged
12597 with an 'U' termination code (for "Up").
12598
12599 Actions are disabled by default
12600
Olivier Houchard006e3102018-12-10 18:30:32 +010012601pool-max-conn <max>
12602 Set the maximum number of idling connections for a server. -1 means unlimited
12603 connections, 0 means no idle connections. The default is -1. When idle
12604 connections are enabled, orphaned idle connections which do not belong to any
12605 client session anymore are moved to a dedicated pool so that they remain
12606 usable by future clients. This only applies to connections that can be shared
12607 according to the same principles as those applying to "http-reuse".
12608
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012609pool-purge-delay <delay>
12610 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010012611 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020012612 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012613
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012614port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012615 Using the "port" parameter, it becomes possible to use a different port to
12616 send health-checks. On some servers, it may be desirable to dedicate a port
12617 to a specific component able to perform complex tests which are more suitable
12618 to health-checks than the application. It is common to run a simple script in
12619 inetd for instance. This parameter is ignored if the "check" parameter is not
12620 set. See also the "addr" parameter.
12621
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020012622proto <name>
12623
12624 Forces the multiplexer's protocol to use for the outgoing connections to this
12625 server. It must be compatible with the mode of the backend (TCP or HTTP). It
12626 must also be usable on the backend side. The list of available protocols is
12627 reported in haproxy -vv.
12628 Idea behind this optipon is to bypass the selection of the best multiplexer's
12629 protocol for all connections established to this server.
12630
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012631redir <prefix>
12632 The "redir" parameter enables the redirection mode for all GET and HEAD
12633 requests addressing this server. This means that instead of having HAProxy
12634 forward the request to the server, it will send an "HTTP 302" response with
12635 the "Location" header composed of this prefix immediately followed by the
12636 requested URI beginning at the leading '/' of the path component. That means
12637 that no trailing slash should be used after <prefix>. All invalid requests
12638 will be rejected, and all non-GET or HEAD requests will be normally served by
12639 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012640 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012641 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012642 users to a remote location in case of local disaster. Main use consists in
12643 increasing bandwidth for static servers by having the clients directly
12644 connect to them. Note: never use a relative location here, it would cause a
12645 loop between the client and HAProxy!
12646
12647 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
12648
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012649rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012650 The "rise" parameter states that a server will be considered as operational
12651 after <count> consecutive successful health checks. This value defaults to 2
12652 if unspecified. See also the "check", "inter" and "fall" parameters.
12653
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012654resolve-opts <option>,<option>,...
12655 Comma separated list of options to apply to DNS resolution linked to this
12656 server.
12657
12658 Available options:
12659
12660 * allow-dup-ip
12661 By default, HAProxy prevents IP address duplication in a backend when DNS
12662 resolution at runtime is in operation.
12663 That said, for some cases, it makes sense that two servers (in the same
12664 backend, being resolved by the same FQDN) have the same IP address.
12665 For such case, simply enable this option.
12666 This is the opposite of prevent-dup-ip.
12667
12668 * prevent-dup-ip
12669 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
12670 an IP address already set to a server in the same backend and sharing the
12671 same fqdn.
12672 This is the opposite of allow-dup-ip.
12673
12674 Example:
12675 backend b_myapp
12676 default-server init-addr none resolvers dns
12677 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
12678 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
12679
12680 With the option allow-dup-ip set:
12681 * if the nameserver returns a single IP address, then both servers will use
12682 it
12683 * If the nameserver returns 2 IP addresses, then each server will pick up a
12684 different address
12685
12686 Default value: not set
12687
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012688resolve-prefer <family>
12689 When DNS resolution is enabled for a server and multiple IP addresses from
12690 different families are returned, HAProxy will prefer using an IP address
12691 from the family mentioned in the "resolve-prefer" parameter.
12692 Available families: "ipv4" and "ipv6"
12693
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020012694 Default value: ipv6
12695
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012696 Example:
12697
12698 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012699
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012700resolve-net <network>[,<network[,...]]
John Roesler27c754c2019-07-10 15:45:51 -050012701 This option prioritizes the choice of an ip address matching a network. This is
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012702 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010012703 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010012704 different datacenters. The latency between datacenter is not negligible, so
12705 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012706 configured network, another address is selected.
12707
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012708 Example:
12709
12710 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012711
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012712resolvers <id>
12713 Points to an existing "resolvers" section to resolve current server's
12714 hostname.
12715
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012716 Example:
12717
12718 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012719
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012720 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012721
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012722send-proxy
12723 The "send-proxy" parameter enforces use of the PROXY protocol over any
12724 connection established to this server. The PROXY protocol informs the other
12725 end about the layer 3/4 addresses of the incoming connection, so that it can
12726 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010012727 upper layer protocol. For connections accepted by an "accept-proxy" or
12728 "accept-netscaler-cip" listener, the advertised address will be used. Only
12729 TCPv4 and TCPv6 address families are supported. Other families such as
12730 Unix sockets, will report an UNKNOWN family. Servers using this option can
12731 fully be chained to another instance of haproxy listening with an
12732 "accept-proxy" setting. This setting must not be used if the server isn't
12733 aware of the protocol. When health checks are sent to the server, the PROXY
12734 protocol is automatically used when this option is set, unless there is an
12735 explicit "port" or "addr" directive, in which case an explicit
12736 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012737 See also the "no-send-proxy" option of this section and "accept-proxy" and
12738 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012739
David Safb76832014-05-08 23:42:08 -040012740send-proxy-v2
12741 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
12742 over any connection established to this server. The PROXY protocol informs
12743 the other end about the layer 3/4 addresses of the incoming connection, so
12744 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020012745 whatever the upper layer protocol. It also send ALPN information if an alpn
12746 have been negotiated. This setting must not be used if the server isn't aware
12747 of this version of the protocol. See also the "no-send-proxy-v2" option of
12748 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012749
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012750proxy-v2-options <option>[,<option>]*
12751 The "proxy-v2-options" parameter add option to send in PROXY protocol version
12752 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010012753 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
12754 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010012755 certificate, "cert-key": key algorithm of the used certificate), "authority":
12756 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010012757 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012758
David Safb76832014-05-08 23:42:08 -040012759send-proxy-v2-ssl
12760 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12761 2 over any connection established to this server. The PROXY protocol informs
12762 the other end about the layer 3/4 addresses of the incoming connection, so
12763 that it can know the client's address or the public address it accessed to,
12764 whatever the upper layer protocol. In addition, the SSL information extension
12765 of the PROXY protocol is added to the PROXY protocol header. This setting
12766 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012767 See also the "no-send-proxy-v2-ssl" option of this section and the
12768 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012769
12770send-proxy-v2-ssl-cn
12771 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12772 2 over any connection established to this server. The PROXY protocol informs
12773 the other end about the layer 3/4 addresses of the incoming connection, so
12774 that it can know the client's address or the public address it accessed to,
12775 whatever the upper layer protocol. In addition, the SSL information extension
12776 of the PROXY protocol, along along with the Common Name from the subject of
12777 the client certificate (if any), is added to the PROXY protocol header. This
12778 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012779 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
12780 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012781
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012782slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012783 The "slowstart" parameter for a server accepts a value in milliseconds which
12784 indicates after how long a server which has just come back up will run at
12785 full speed. Just as with every other time-based parameter, it can be entered
12786 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
12787 linearly from 0 to 100% during this time. The limitation applies to two
12788 parameters :
12789
12790 - maxconn: the number of connections accepted by the server will grow from 1
12791 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
12792
12793 - weight: when the backend uses a dynamic weighted algorithm, the weight
12794 grows linearly from 1 to 100%. In this case, the weight is updated at every
12795 health-check. For this reason, it is important that the "inter" parameter
12796 is smaller than the "slowstart", in order to maximize the number of steps.
12797
12798 The slowstart never applies when haproxy starts, otherwise it would cause
12799 trouble to running servers. It only applies when a server has been previously
12800 seen as failed.
12801
Willy Tarreau732eac42015-07-09 11:40:25 +020012802sni <expression>
12803 The "sni" parameter evaluates the sample fetch expression, converts it to a
12804 string and uses the result as the host name sent in the SNI TLS extension to
12805 the server. A typical use case is to send the SNI received from the client in
12806 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020012807 expression, though alternatives such as req.hdr(host) can also make sense. If
12808 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012809 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010012810 "verify" directive for more details. If you want to set a SNI for health
12811 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020012812
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012813source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020012814source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012815source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012816 The "source" parameter sets the source address which will be used when
12817 connecting to the server. It follows the exact same parameters and principle
12818 as the backend "source" keyword, except that it only applies to the server
12819 referencing it. Please consult the "source" keyword for details.
12820
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012821 Additionally, the "source" statement on a server line allows one to specify a
12822 source port range by indicating the lower and higher bounds delimited by a
12823 dash ('-'). Some operating systems might require a valid IP address when a
12824 source port range is specified. It is permitted to have the same IP/range for
12825 several servers. Doing so makes it possible to bypass the maximum of 64k
12826 total concurrent connections. The limit will then reach 64k connections per
12827 server.
12828
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000012829 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
12830 specifying the source address without port(s).
12831
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012832ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020012833 This option enables SSL ciphering on outgoing connections to the server. It
12834 is critical to verify server certificates using "verify" when using SSL to
12835 connect to servers, otherwise the communication is prone to trivial man in
12836 the-middle attacks rendering SSL useless. When this option is used, health
12837 checks are automatically sent in SSL too unless there is a "port" or an
12838 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012839 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
12840 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012841
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012842ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12843 This option enforces use of <version> or lower when SSL is used to communicate
12844 with the server. This option is also available on global statement
12845 "ssl-default-server-options". See also "ssl-min-ver".
12846
12847ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12848 This option enforces use of <version> or upper when SSL is used to communicate
12849 with the server. This option is also available on global statement
12850 "ssl-default-server-options". See also "ssl-max-ver".
12851
Frédéric Lécailled2376272017-03-21 18:52:12 +010012852ssl-reuse
12853 This option may be used as "server" setting to reset any "no-ssl-reuse"
12854 setting which would have been inherited from "default-server" directive as
12855 default value.
12856 It may also be used as "default-server" setting to reset any previous
12857 "default-server" "no-ssl-reuse" setting.
12858
12859stick
12860 This option may be used as "server" setting to reset any "non-stick"
12861 setting which would have been inherited from "default-server" directive as
12862 default value.
12863 It may also be used as "default-server" setting to reset any previous
12864 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012865
Alexander Liu2a54bb72019-05-22 19:44:48 +080012866socks4 <addr>:<port>
John Roesler27c754c2019-07-10 15:45:51 -050012867 This option enables upstream socks4 tunnel for outgoing connections to the
Alexander Liu2a54bb72019-05-22 19:44:48 +080012868 server. Using this option won't force the health check to go via socks4 by
12869 default. You will have to use the keyword "check-via-socks4" to enable it.
12870
Willy Tarreau163d4622015-10-13 16:16:41 +020012871tcp-ut <delay>
12872 Sets the TCP User Timeout for all outgoing connections to this server. This
12873 option is available on Linux since version 2.6.37. It allows haproxy to
12874 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012875 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020012876 long-lived connections experiencing long idle periods such as remote
12877 terminals or database connection pools, where the client and server timeouts
12878 must remain high to allow a long period of idle, but where it is important to
12879 detect that the server has disappeared in order to release all resources
12880 associated with its connection (and the client's session). One typical use
12881 case is also to force dead server connections to die when health checks are
12882 too slow or during a soft reload since health checks are then disabled. The
12883 argument is a delay expressed in milliseconds by default. This only works for
12884 regular TCP connections, and is ignored for other protocols.
12885
Willy Tarreau034c88c2017-01-23 23:36:45 +010012886tfo
12887 This option enables using TCP fast open when connecting to servers, on
12888 systems that support it (currently only the Linux kernel >= 4.11).
12889 See the "tfo" bind option for more information about TCP fast open.
12890 Please note that when using tfo, you should also use the "conn-failure",
12891 "empty-response" and "response-timeout" keywords for "retry-on", or haproxy
Frédéric Lécailleaeeb1c92019-07-04 14:19:06 +020012892 won't be able to retry the connection on failure. See also "no-tfo".
Willy Tarreau034c88c2017-01-23 23:36:45 +010012893
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012894track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020012895 This option enables ability to set the current state of the server by tracking
12896 another one. It is possible to track a server which itself tracks another
12897 server, provided that at the end of the chain, a server has health checks
12898 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012899 used, it has to be enabled on both proxies.
12900
Frédéric Lécailled2376272017-03-21 18:52:12 +010012901tls-tickets
12902 This option may be used as "server" setting to reset any "no-tls-tickets"
12903 setting which would have been inherited from "default-server" directive as
12904 default value.
12905 It may also be used as "default-server" setting to reset any previous
12906 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012907
Emeric Brunef42d922012-10-11 16:11:36 +020012908verify [none|required]
12909 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010012910 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012911 certificate provided by the server is verified using CAs from 'ca-file' and
12912 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012913 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012914 the name passed using the "sni" directive, or if not provided, the static
12915 host name passed using the "verifyhost" directive. When no name is found, the
12916 certificate's names are ignored. For this reason, without SNI it's important
12917 to use "verifyhost". On verification failure the handshake is aborted. It is
12918 critically important to verify server certificates when using SSL to connect
12919 to servers, otherwise the communication is prone to trivial man-in-the-middle
12920 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
12921 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020012922
Evan Broderbe554312013-06-27 00:05:25 -070012923verifyhost <hostname>
12924 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012925 only takes effect if 'verify required' is also specified. This directive sets
12926 a default static hostname to check the server's certificate against when no
12927 SNI was used to connect to the server. If SNI is not used, this is the only
12928 way to enable hostname verification. This static hostname, when set, will
12929 also be used for health checks (which cannot provide an SNI value). If none
12930 of the hostnames in the certificate match the specified hostname, the
12931 handshake is aborted. The hostnames in the server-provided certificate may
12932 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070012933
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012934weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012935 The "weight" parameter is used to adjust the server's weight relative to
12936 other servers. All servers will receive a load proportional to their weight
12937 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020012938 load. The default weight is 1, and the maximal value is 256. A value of 0
12939 means the server will not participate in load-balancing but will still accept
12940 persistent connections. If this parameter is used to distribute the load
12941 according to server's capacity, it is recommended to start with values which
12942 can both grow and shrink, for instance between 10 and 100 to leave enough
12943 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012944
12945
Cyril Bonté46175dd2015-07-02 22:45:32 +0200129465.3. Server IP address resolution using DNS
12947-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012948
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012949HAProxy allows using a host name on the server line to retrieve its IP address
12950using name servers. By default, HAProxy resolves the name when parsing the
12951configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012952This is not sufficient in some cases, such as in Amazon where a server's IP
12953can change after a reboot or an ELB Virtual IP can change based on current
12954workload.
12955This chapter describes how HAProxy can be configured to process server's name
12956resolution at run time.
12957Whether run time server name resolution has been enable or not, HAProxy will
12958carry on doing the first resolution when parsing the configuration.
12959
12960
Cyril Bonté46175dd2015-07-02 22:45:32 +0200129615.3.1. Global overview
12962----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012963
12964As we've seen in introduction, name resolution in HAProxy occurs at two
12965different steps of the process life:
12966
12967 1. when starting up, HAProxy parses the server line definition and matches a
12968 host name. It uses libc functions to get the host name resolved. This
12969 resolution relies on /etc/resolv.conf file.
12970
Christopher Faulet67957bd2017-09-27 11:00:59 +020012971 2. at run time, HAProxy performs periodically name resolutions for servers
12972 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012973
12974A few other events can trigger a name resolution at run time:
12975 - when a server's health check ends up in a connection timeout: this may be
12976 because the server has a new IP address. So we need to trigger a name
12977 resolution to know this new IP.
12978
Christopher Faulet67957bd2017-09-27 11:00:59 +020012979When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012980HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020012981SRV label is specified, then the corresponding SRV records will be retrieved
12982from the DNS server, and the provided hostnames will be used. The SRV label
12983will be checked periodically, and if any server are added or removed, haproxy
12984will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020012985
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012986A few things important to notice:
John Roesler27c754c2019-07-10 15:45:51 -050012987 - all the name servers are queried in the meantime. HAProxy will process the
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012988 first valid response.
12989
12990 - a resolution is considered as invalid (NX, timeout, refused), when all the
12991 servers return an error.
12992
12993
Cyril Bonté46175dd2015-07-02 22:45:32 +0200129945.3.2. The resolvers section
12995----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012996
12997This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020012998HAProxy. There can be as many as resolvers section as needed. Each section can
12999contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013000
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013001When multiple name servers are configured in a resolvers section, then HAProxy
13002uses the first valid response. In case of invalid responses, only the last one
13003is treated. Purpose is to give the chance to a slow server to deliver a valid
13004answer after a fast faulty or outdated server.
13005
13006When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020013007used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013008
Christopher Faulet67957bd2017-09-27 11:00:59 +020013009 1. HAProxy retries the same DNS query with a new query type. The A queries are
13010 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
13011 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013012
Christopher Faulet67957bd2017-09-27 11:00:59 +020013013 2. When the fallback on the query type was done (or not applicable), HAProxy
13014 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013015
Christopher Faulet67957bd2017-09-27 11:00:59 +020013016 3. HAProxy retries previous steps <resolve_retires> times. If no valid
13017 response is received after that, it stops the DNS resolution and reports
13018 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013019
Christopher Faulet67957bd2017-09-27 11:00:59 +020013020For example, with 2 name servers configured in a resolvers section, the
13021following scenarios are possible:
13022
13023 - First response is valid and is applied directly, second response is
13024 ignored
13025
13026 - First response is invalid and second one is valid, then second response is
13027 applied
13028
13029 - First response is a NX domain and second one a truncated response, then
13030 HAProxy retries the query with a new type
13031
13032 - First response is a NX domain and second one is a timeout, then HAProxy
13033 retries the query with a new type
13034
13035 - Query timed out for both name servers, then HAProxy retries it with the
13036 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013037
Olivier Houcharda8c6db82017-07-06 18:46:47 +020013038As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
13039a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020013040<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020013041
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013042
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013043resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010013044 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013045
13046A resolvers section accept the following parameters:
13047
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020013048accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010013049 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020013050 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020013051 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
13052 by RFC 6891)
13053
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020013054 Note: the maximum allowed value is 8192.
13055
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013056nameserver <id> <ip>:<port>
13057 DNS server description:
13058 <id> : label of the server, should be unique
13059 <ip> : IP address of the server
13060 <port> : port where the DNS service actually runs
13061
Ben Draut44e609b2018-05-29 15:40:08 -060013062parse-resolv-conf
13063 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
13064 list. Ordered as if each nameserver in /etc/resolv.conf was individually
13065 placed in the resolvers section in place of this directive.
13066
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013067hold <status> <period>
13068 Defines <period> during which the last name resolution should be kept based
13069 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010013070 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020013071 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013072 <period> : interval between two successive name resolution when the last
13073 answer was in <status>. It follows the HAProxy time format.
13074 <period> is in milliseconds by default.
13075
Baptiste Assmann686408b2017-08-18 10:15:42 +020013076 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013077
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013078resolve_retries <nb>
13079 Defines the number <nb> of queries to send to resolve a server name before
13080 giving up.
13081 Default value: 3
13082
Baptiste Assmann62b75b42015-09-09 01:11:36 +020013083 A retry occurs on name server timeout or when the full sequence of DNS query
13084 type failover is over and we need to start up from the default ANY query
13085 type.
13086
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013087timeout <event> <time>
13088 Defines timeouts related to name resolution
13089 <event> : the event on which the <time> timeout period applies to.
13090 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010013091 - resolve : default time to trigger name resolutions when no
13092 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020013093 Default value: 1s
13094 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010013095 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020013096 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013097 <time> : time related to the event. It follows the HAProxy time format.
13098 <time> is expressed in milliseconds.
13099
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020013100 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013101
13102 resolvers mydns
13103 nameserver dns1 10.0.0.1:53
13104 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060013105 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013106 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020013107 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013108 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010013109 hold other 30s
13110 hold refused 30s
13111 hold nx 30s
13112 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013113 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020013114 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020013115
13116
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200131176. HTTP header manipulation
13118---------------------------
13119
13120In HTTP mode, it is possible to rewrite, add or delete some of the request and
13121response headers based on regular expressions. It is also possible to block a
13122request or a response if a particular header matches a regular expression,
13123which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010013124against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013125
Willy Tarreau70dffda2014-01-30 03:07:23 +010013126If HAProxy encounters an "Informational Response" (status code 1xx), it is able
13127to process all rsp* rules which can allow, deny, rewrite or delete a header,
13128but it will refuse to add a header to any such messages as this is not
13129HTTP-compliant. The reason for still processing headers in such responses is to
13130stop and/or fix any possible information leak which may happen, for instance
13131because another downstream equipment would unconditionally add a header, or if
13132a server name appears there. When such messages are seen, normal processing
13133still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020013134
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013135This section covers common usage of the following keywords, described in detail
13136in section 4.2 :
13137
13138 - reqadd <string>
13139 - reqallow <search>
13140 - reqiallow <search>
13141 - reqdel <search>
13142 - reqidel <search>
13143 - reqdeny <search>
13144 - reqideny <search>
13145 - reqpass <search>
13146 - reqipass <search>
13147 - reqrep <search> <replace>
13148 - reqirep <search> <replace>
13149 - reqtarpit <search>
13150 - reqitarpit <search>
13151 - rspadd <string>
13152 - rspdel <search>
13153 - rspidel <search>
13154 - rspdeny <search>
13155 - rspideny <search>
13156 - rsprep <search> <replace>
13157 - rspirep <search> <replace>
13158
13159With all these keywords, the same conventions are used. The <search> parameter
13160is a POSIX extended regular expression (regex) which supports grouping through
13161parenthesis (without the backslash). Spaces and other delimiters must be
13162prefixed with a backslash ('\') to avoid confusion with a field delimiter.
13163Other characters may be prefixed with a backslash to change their meaning :
13164
13165 \t for a tab
13166 \r for a carriage return (CR)
13167 \n for a new line (LF)
13168 \ to mark a space and differentiate it from a delimiter
13169 \# to mark a sharp and differentiate it from a comment
13170 \\ to use a backslash in a regex
13171 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
13172 \xXX to write the ASCII hex code XX as in the C language
13173
13174The <replace> parameter contains the string to be used to replace the largest
13175portion of text matching the regex. It can make use of the special characters
13176above, and can reference a substring which is delimited by parenthesis in the
13177regex, by writing a backslash ('\') immediately followed by one digit from 0 to
131789 indicating the group position (0 designating the entire line). This practice
13179is very common to users of the "sed" program.
13180
13181The <string> parameter represents the string which will systematically be added
13182after the last header line. It can also use special character sequences above.
13183
13184Notes related to these keywords :
13185---------------------------------
13186 - these keywords are not always convenient to allow/deny based on header
13187 contents. It is strongly recommended to use ACLs with the "block" keyword
13188 instead, resulting in far more flexible and manageable rules.
13189
13190 - lines are always considered as a whole. It is not possible to reference
13191 a header name only or a value only. This is important because of the way
13192 headers are written (notably the number of spaces after the colon).
13193
13194 - the first line is always considered as a header, which makes it possible to
13195 rewrite or filter HTTP requests URIs or response codes, but in turn makes
13196 it harder to distinguish between headers and request line. The regex prefix
13197 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
13198 ^[^ \t:]*: matches any header name followed by a colon.
13199
13200 - for performances reasons, the number of characters added to a request or to
13201 a response is limited at build time to values between 1 and 4 kB. This
13202 should normally be far more than enough for most usages. If it is too short
13203 on occasional usages, it is possible to gain some space by removing some
13204 useless headers before adding new ones.
13205
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013206 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013207 without the 'i' letter except that they ignore case when matching patterns.
13208
13209 - when a request passes through a frontend then a backend, all req* rules
13210 from the frontend will be evaluated, then all req* rules from the backend
13211 will be evaluated. The reverse path is applied to responses.
13212
13213 - req* statements are applied after "block" statements, so that "block" is
13214 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010013215 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013216
13217
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132187. Using ACLs and fetching samples
13219----------------------------------
13220
Davor Ocelice9ed2812017-12-25 17:49:28 +010013221HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020013222client or server information, from tables, environmental information etc...
13223The action of extracting such data is called fetching a sample. Once retrieved,
13224these samples may be used for various purposes such as a key to a stick-table,
13225but most common usages consist in matching them against predefined constant
13226data called patterns.
13227
13228
132297.1. ACL basics
13230---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013231
13232The use of Access Control Lists (ACL) provides a flexible solution to perform
13233content switching and generally to take decisions based on content extracted
13234from the request, the response or any environmental status. The principle is
13235simple :
13236
Willy Tarreau74ca5042013-06-11 23:12:07 +020013237 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013238 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020013239 - apply one or multiple pattern matching methods on this sample
13240 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013241
Willy Tarreau74ca5042013-06-11 23:12:07 +020013242The actions generally consist in blocking a request, selecting a backend, or
13243adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013244
13245In order to define a test, the "acl" keyword is used. The syntax is :
13246
Willy Tarreau74ca5042013-06-11 23:12:07 +020013247 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013248
13249This creates a new ACL <aclname> or completes an existing one with new tests.
13250Those tests apply to the portion of request/response specified in <criterion>
13251and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013252an operator which may be specified before the set of values. Optionally some
13253conversion operators may be applied to the sample, and they will be specified
13254as a comma-delimited list of keywords just after the first keyword. The values
13255are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013256
13257ACL names must be formed from upper and lower case letters, digits, '-' (dash),
13258'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
13259which means that "my_acl" and "My_Acl" are two different ACLs.
13260
13261There is no enforced limit to the number of ACLs. The unused ones do not affect
13262performance, they just consume a small amount of memory.
13263
Willy Tarreau74ca5042013-06-11 23:12:07 +020013264The criterion generally is the name of a sample fetch method, or one of its ACL
13265specific declinations. The default test method is implied by the output type of
13266this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013267methods of a same sample fetch method. The sample fetch methods are the only
13268ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013269
13270Sample fetch methods return data which can be of the following types :
13271 - boolean
13272 - integer (signed or unsigned)
13273 - IPv4 or IPv6 address
13274 - string
13275 - data block
13276
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013277Converters transform any of these data into any of these. For example, some
13278converters might convert a string to a lower-case string while other ones
13279would turn a string to an IPv4 address, or apply a netmask to an IP address.
13280The resulting sample is of the type of the last converter applied to the list,
13281which defaults to the type of the sample fetch method.
13282
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013283Each sample or converter returns data of a specific type, specified with its
13284keyword in this documentation. When an ACL is declared using a standard sample
13285fetch method, certain types automatically involved a default matching method
13286which are summarized in the table below :
13287
13288 +---------------------+-----------------+
13289 | Sample or converter | Default |
13290 | output type | matching method |
13291 +---------------------+-----------------+
13292 | boolean | bool |
13293 +---------------------+-----------------+
13294 | integer | int |
13295 +---------------------+-----------------+
13296 | ip | ip |
13297 +---------------------+-----------------+
13298 | string | str |
13299 +---------------------+-----------------+
13300 | binary | none, use "-m" |
13301 +---------------------+-----------------+
13302
13303Note that in order to match a binary samples, it is mandatory to specify a
13304matching method, see below.
13305
Willy Tarreau74ca5042013-06-11 23:12:07 +020013306The ACL engine can match these types against patterns of the following types :
13307 - boolean
13308 - integer or integer range
13309 - IP address / network
13310 - string (exact, substring, suffix, prefix, subdir, domain)
13311 - regular expression
13312 - hex block
13313
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013314The following ACL flags are currently supported :
13315
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013316 -i : ignore case during matching of all subsequent patterns.
13317 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013318 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013319 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013320 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013321 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020013322 -- : force end of flags. Useful when a string looks like one of the flags.
13323
Willy Tarreau74ca5042013-06-11 23:12:07 +020013324The "-f" flag is followed by the name of a file from which all lines will be
13325read as individual values. It is even possible to pass multiple "-f" arguments
13326if the patterns are to be loaded from multiple files. Empty lines as well as
13327lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
13328will be stripped. If it is absolutely necessary to insert a valid pattern
13329beginning with a sharp, just prefix it with a space so that it is not taken for
13330a comment. Depending on the data type and match method, haproxy may load the
13331lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
13332exact string matching. In this case, duplicates will automatically be removed.
13333
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013334The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
13335parsed as two column file. The first column contains the patterns used by the
13336ACL, and the second column contain the samples. The sample can be used later by
13337a map. This can be useful in some rare cases where an ACL would just be used to
13338check for the existence of a pattern in a map before a mapping is applied.
13339
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013340The "-u" flag forces the unique id of the ACL. This unique id is used with the
13341socket interface to identify ACL and dynamically change its values. Note that a
13342file is always identified by its name even if an id is set.
13343
Willy Tarreau74ca5042013-06-11 23:12:07 +020013344Also, note that the "-i" flag applies to subsequent entries and not to entries
13345loaded from files preceding it. For instance :
13346
13347 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
13348
13349In this example, each line of "exact-ua.lst" will be exactly matched against
13350the "user-agent" header of the request. Then each line of "generic-ua" will be
13351case-insensitively matched. Then the word "test" will be insensitively matched
13352as well.
13353
13354The "-m" flag is used to select a specific pattern matching method on the input
13355sample. All ACL-specific criteria imply a pattern matching method and generally
13356do not need this flag. However, this flag is useful with generic sample fetch
13357methods to describe how they're going to be matched against the patterns. This
13358is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010013359obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013360followed by a pattern matching method name, this method is used instead of the
13361default one for the criterion. This makes it possible to match contents in ways
13362that were not initially planned, or with sample fetch methods which return a
13363string. The matching method also affects the way the patterns are parsed.
13364
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013365The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
13366By default, if the parser cannot parse ip address it considers that the parsed
13367string is maybe a domain name and try dns resolution. The flag "-n" disable this
13368resolution. It is useful for detecting malformed ip lists. Note that if the DNS
13369server is not reachable, the haproxy configuration parsing may last many minutes
John Roesler27c754c2019-07-10 15:45:51 -050013370waiting for the timeout. During this time no error messages are displayed. The
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013371flag "-n" disable this behavior. Note also that during the runtime, this
13372function is disabled for the dynamic acl modifications.
13373
Willy Tarreau74ca5042013-06-11 23:12:07 +020013374There are some restrictions however. Not all methods can be used with all
13375sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
13376be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020013377
13378 - "found" : only check if the requested sample could be found in the stream,
13379 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020013380 to pass any pattern to avoid confusion. This matching method is
13381 particularly useful to detect presence of certain contents such
13382 as headers, cookies, etc... even if they are empty and without
13383 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013384
13385 - "bool" : check the value as a boolean. It can only be applied to fetches
13386 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013387 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013388
13389 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013390 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013391
13392 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020013393 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013394
Davor Ocelice9ed2812017-12-25 17:49:28 +010013395 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020013396 binary sequence. This may be used with binary or string samples.
13397
13398 - "len" : match the sample's length as an integer. This may be used with
13399 binary or string samples.
13400
Willy Tarreau74ca5042013-06-11 23:12:07 +020013401 - "str" : exact match : match the contents against a string. This may be
13402 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013403
Willy Tarreau74ca5042013-06-11 23:12:07 +020013404 - "sub" : substring match : check that the contents contain at least one of
13405 the provided string patterns. This may be used with binary or
13406 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013407
Willy Tarreau74ca5042013-06-11 23:12:07 +020013408 - "reg" : regex match : match the contents against a list of regular
13409 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013410
Willy Tarreau74ca5042013-06-11 23:12:07 +020013411 - "beg" : prefix match : check that the contents begin like the provided
13412 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013413
Willy Tarreau74ca5042013-06-11 23:12:07 +020013414 - "end" : suffix match : check that the contents end like the provided
13415 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013416
Willy Tarreau74ca5042013-06-11 23:12:07 +020013417 - "dir" : subdir match : check that a slash-delimited portion of the
13418 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013419 This may be used with binary or string samples.
13420
Willy Tarreau74ca5042013-06-11 23:12:07 +020013421 - "dom" : domain match : check that a dot-delimited portion of the contents
13422 exactly match one of the provided string patterns. This may be
13423 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013424
13425For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
13426request, it is possible to do :
13427
13428 acl jsess_present cook(JSESSIONID) -m found
13429
13430In order to apply a regular expression on the 500 first bytes of data in the
13431buffer, one would use the following acl :
13432
13433 acl script_tag payload(0,500) -m reg -i <script>
13434
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013435On systems where the regex library is much slower when using "-i", it is
13436possible to convert the sample to lowercase before matching, like this :
13437
13438 acl script_tag payload(0,500),lower -m reg <script>
13439
Willy Tarreau74ca5042013-06-11 23:12:07 +020013440All ACL-specific criteria imply a default matching method. Most often, these
13441criteria are composed by concatenating the name of the original sample fetch
13442method and the matching method. For example, "hdr_beg" applies the "beg" match
13443to samples retrieved using the "hdr" fetch method. Since all ACL-specific
13444criteria rely on a sample fetch method, it is always possible instead to use
13445the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013446
Willy Tarreau74ca5042013-06-11 23:12:07 +020013447If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013448the matching method is simply applied to the underlying sample fetch method.
13449For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013450
Willy Tarreau74ca5042013-06-11 23:12:07 +020013451 acl short_form hdr_beg(host) www.
13452 acl alternate1 hdr_beg(host) -m beg www.
13453 acl alternate2 hdr_dom(host) -m beg www.
13454 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013455
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013456
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013457The table below summarizes the compatibility matrix between sample or converter
13458types and the pattern types to fetch against. It indicates for each compatible
13459combination the name of the matching method to be used, surrounded with angle
13460brackets ">" and "<" when the method is the default one and will work by
13461default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013462
Willy Tarreau74ca5042013-06-11 23:12:07 +020013463 +-------------------------------------------------+
13464 | Input sample type |
13465 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013466 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013467 +----------------------+---------+---------+---------+---------+---------+
13468 | none (presence only) | found | found | found | found | found |
13469 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013470 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013471 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013472 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013473 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013474 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013475 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013476 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013477 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013478 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013479 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013480 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013481 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013482 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013483 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013484 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013485 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013486 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013487 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013488 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013489 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013490 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013491 +----------------------+---------+---------+---------+---------+---------+
13492 | hex block | | | | bin | bin |
13493 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020013494
13495
Willy Tarreau74ca5042013-06-11 23:12:07 +0200134967.1.1. Matching booleans
13497------------------------
13498
13499In order to match a boolean, no value is needed and all values are ignored.
13500Boolean matching is used by default for all fetch methods of type "boolean".
13501When boolean matching is used, the fetched value is returned as-is, which means
13502that a boolean "true" will always match and a boolean "false" will never match.
13503
13504Boolean matching may also be enforced using "-m bool" on fetch methods which
13505return an integer value. Then, integer value 0 is converted to the boolean
13506"false" and all other values are converted to "true".
13507
Willy Tarreau6a06a402007-07-15 20:15:28 +020013508
Willy Tarreau74ca5042013-06-11 23:12:07 +0200135097.1.2. Matching integers
13510------------------------
13511
13512Integer matching applies by default to integer fetch methods. It can also be
13513enforced on boolean fetches using "-m int". In this case, "false" is converted
13514to the integer 0, and "true" is converted to the integer 1.
13515
13516Integer matching also supports integer ranges and operators. Note that integer
13517matching only applies to positive values. A range is a value expressed with a
13518lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013519
13520For instance, "1024:65535" is a valid range to represent a range of
13521unprivileged ports, and "1024:" would also work. "0:1023" is a valid
13522representation of privileged ports, and ":1023" would also work.
13523
Willy Tarreau62644772008-07-16 18:36:06 +020013524As a special case, some ACL functions support decimal numbers which are in fact
13525two integers separated by a dot. This is used with some version checks for
13526instance. All integer properties apply to those decimal numbers, including
13527ranges and operators.
13528
Willy Tarreau6a06a402007-07-15 20:15:28 +020013529For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010013530operators with ranges does not make much sense and is strongly discouraged.
13531Similarly, it does not make much sense to perform order comparisons with a set
13532of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013533
Willy Tarreau0ba27502007-12-24 16:55:16 +010013534Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013535
13536 eq : true if the tested value equals at least one value
13537 ge : true if the tested value is greater than or equal to at least one value
13538 gt : true if the tested value is greater than at least one value
13539 le : true if the tested value is less than or equal to at least one value
13540 lt : true if the tested value is less than at least one value
13541
Willy Tarreau0ba27502007-12-24 16:55:16 +010013542For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013543
13544 acl negative-length hdr_val(content-length) lt 0
13545
Willy Tarreau62644772008-07-16 18:36:06 +020013546This one matches SSL versions between 3.0 and 3.1 (inclusive) :
13547
13548 acl sslv3 req_ssl_ver 3:3.1
13549
Willy Tarreau6a06a402007-07-15 20:15:28 +020013550
Willy Tarreau74ca5042013-06-11 23:12:07 +0200135517.1.3. Matching strings
13552-----------------------
13553
13554String matching applies to string or binary fetch methods, and exists in 6
13555different forms :
13556
13557 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013558 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013559
13560 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013561 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013562
13563 - prefix match (-m beg) : the patterns are compared with the beginning of
13564 the extracted string, and the ACL matches if any of them matches.
13565
13566 - suffix match (-m end) : the patterns are compared with the end of the
13567 extracted string, and the ACL matches if any of them matches.
13568
Baptiste Assmann33db6002016-03-06 23:32:10 +010013569 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020013570 string, delimited with slashes ("/"), and the ACL matches if any of them
13571 matches.
13572
13573 - domain match (-m dom) : the patterns are looked up inside the extracted
13574 string, delimited with dots ("."), and the ACL matches if any of them
13575 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013576
13577String matching applies to verbatim strings as they are passed, with the
13578exception of the backslash ("\") which makes it possible to escape some
13579characters such as the space. If the "-i" flag is passed before the first
13580string, then the matching will be performed ignoring the case. In order
13581to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010013582before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013583
Mathias Weiersmuellerb2fe2232019-12-02 09:43:40 +010013584Do not use string matches for binary fetches which might contain null bytes
13585(0x00), as the comparison stops at the occurrence of the first null byte.
13586Instead, convert the binary fetch to a hex string with the hex converter first.
13587
13588Example:
13589 # matches if the string <tag> is present in the binary sample
13590 acl tag_found req.payload(0,0),hex -m sub 3C7461673E
13591
Willy Tarreau6a06a402007-07-15 20:15:28 +020013592
Willy Tarreau74ca5042013-06-11 23:12:07 +0200135937.1.4. Matching regular expressions (regexes)
13594---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013595
13596Just like with string matching, regex matching applies to verbatim strings as
13597they are passed, with the exception of the backslash ("\") which makes it
13598possible to escape some characters such as the space. If the "-i" flag is
13599passed before the first regex, then the matching will be performed ignoring
13600the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010013601the "--" flag before the first string. Same principle applies of course to
13602match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013603
13604
Willy Tarreau74ca5042013-06-11 23:12:07 +0200136057.1.5. Matching arbitrary data blocks
13606-------------------------------------
13607
13608It is possible to match some extracted samples against a binary block which may
13609not safely be represented as a string. For this, the patterns must be passed as
13610a series of hexadecimal digits in an even number, when the match method is set
13611to binary. Each sequence of two digits will represent a byte. The hexadecimal
13612digits may be used upper or lower case.
13613
13614Example :
13615 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
13616 acl hello payload(0,6) -m bin 48656c6c6f0a
13617
13618
136197.1.6. Matching IPv4 and IPv6 addresses
13620---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013621
13622IPv4 addresses values can be specified either as plain addresses or with a
13623netmask appended, in which case the IPv4 address matches whenever it is
13624within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010013625host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010013626difficult to read and debug configurations. If hostnames are used, you should
13627at least ensure that they are present in /etc/hosts so that the configuration
13628does not depend on any random DNS match at the moment the configuration is
13629parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013630
Daniel Schnellereba56342016-04-13 00:26:52 +020013631The dotted IPv4 address notation is supported in both regular as well as the
13632abbreviated form with all-0-octets omitted:
13633
13634 +------------------+------------------+------------------+
13635 | Example 1 | Example 2 | Example 3 |
13636 +------------------+------------------+------------------+
13637 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
13638 | 192.168.1 | 10.12 | 127.1 |
13639 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
13640 | 192.168.1/22 | 10.12/8 | 127.1/8 |
13641 +------------------+------------------+------------------+
13642
13643Notice that this is different from RFC 4632 CIDR address notation in which
13644192.168.42/24 would be equivalent to 192.168.42.0/24.
13645
Willy Tarreauceb4ac92012-04-28 00:41:46 +020013646IPv6 may be entered in their usual form, with or without a netmask appended.
13647Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
13648trouble with randomly resolved IP addresses, host names are never allowed in
13649IPv6 patterns.
13650
13651HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
13652following situations :
13653 - tested address is IPv4, pattern address is IPv4, the match applies
13654 in IPv4 using the supplied mask if any.
13655 - tested address is IPv6, pattern address is IPv6, the match applies
13656 in IPv6 using the supplied mask if any.
13657 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
13658 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
13659 ::IPV4 or ::ffff:IPV4, otherwise it fails.
13660 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
13661 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
13662 applied in IPv6 using the supplied IPv6 mask.
13663
Willy Tarreau74ca5042013-06-11 23:12:07 +020013664
136657.2. Using ACLs to form conditions
13666----------------------------------
13667
13668Some actions are only performed upon a valid condition. A condition is a
13669combination of ACLs with operators. 3 operators are supported :
13670
13671 - AND (implicit)
13672 - OR (explicit with the "or" keyword or the "||" operator)
13673 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020013674
Willy Tarreau74ca5042013-06-11 23:12:07 +020013675A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020013676
Willy Tarreau74ca5042013-06-11 23:12:07 +020013677 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020013678
Willy Tarreau74ca5042013-06-11 23:12:07 +020013679Such conditions are generally used after an "if" or "unless" statement,
13680indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020013681
Willy Tarreau74ca5042013-06-11 23:12:07 +020013682For instance, to block HTTP requests to the "*" URL with methods other than
13683"OPTIONS", as well as POST requests without content-length, and GET or HEAD
13684requests with a content-length greater than 0, and finally every request which
13685is not either GET/HEAD/POST/OPTIONS !
13686
13687 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013688 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
13689 http-request deny if METH_GET HTTP_CONTENT
13690 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020013691
13692To select a different backend for requests to static contents on the "www" site
13693and to every request on the "img", "video", "download" and "ftp" hosts :
13694
13695 acl url_static path_beg /static /images /img /css
13696 acl url_static path_end .gif .png .jpg .css .js
13697 acl host_www hdr_beg(host) -i www
13698 acl host_static hdr_beg(host) -i img. video. download. ftp.
13699
Davor Ocelice9ed2812017-12-25 17:49:28 +010013700 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020013701 # of host "www". Use backend "www" for the rest.
13702 use_backend static if host_static or host_www url_static
13703 use_backend www if host_www
13704
13705It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
13706expressions that are built on the fly without needing to be declared. They must
13707be enclosed between braces, with a space before and after each brace (because
13708the braces must be seen as independent words). Example :
13709
13710 The following rule :
13711
13712 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013713 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020013714
13715 Can also be written that way :
13716
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013717 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020013718
13719It is generally not recommended to use this construct because it's a lot easier
13720to leave errors in the configuration when written that way. However, for very
13721simple rules matching only one source IP address for instance, it can make more
13722sense to use them than to declare ACLs with random names. Another example of
13723good use is the following :
13724
13725 With named ACLs :
13726
13727 acl site_dead nbsrv(dynamic) lt 2
13728 acl site_dead nbsrv(static) lt 2
13729 monitor fail if site_dead
13730
13731 With anonymous ACLs :
13732
13733 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
13734
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013735See section 4.2 for detailed help on the "http-request deny" and "use_backend"
13736keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013737
13738
137397.3. Fetching samples
13740---------------------
13741
13742Historically, sample fetch methods were only used to retrieve data to match
13743against patterns using ACLs. With the arrival of stick-tables, a new class of
13744sample fetch methods was created, most often sharing the same syntax as their
13745ACL counterpart. These sample fetch methods are also known as "fetches". As
13746of now, ACLs and fetches have converged. All ACL fetch methods have been made
13747available as fetch methods, and ACLs may use any sample fetch method as well.
13748
13749This section details all available sample fetch methods and their output type.
13750Some sample fetch methods have deprecated aliases that are used to maintain
13751compatibility with existing configurations. They are then explicitly marked as
13752deprecated and should not be used in new setups.
13753
13754The ACL derivatives are also indicated when available, with their respective
13755matching methods. These ones all have a well defined default pattern matching
13756method, so it is never necessary (though allowed) to pass the "-m" option to
13757indicate how the sample will be matched using ACLs.
13758
13759As indicated in the sample type versus matching compatibility matrix above,
13760when using a generic sample fetch method in an ACL, the "-m" option is
13761mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
13762the same keyword exists as an ACL keyword and as a standard fetch method, the
13763ACL engine will automatically pick the ACL-only one by default.
13764
13765Some of these keywords support one or multiple mandatory arguments, and one or
13766multiple optional arguments. These arguments are strongly typed and are checked
13767when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010013768incorrect argument (e.g. an unresolved backend name). Fetch function arguments
13769are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020013770is optional, it will be indicated below between square brackets ('[ ]'). When
13771all arguments are optional, the parenthesis may be omitted.
13772
13773Thus, the syntax of a standard sample fetch method is one of the following :
13774 - name
13775 - name(arg1)
13776 - name(arg1,arg2)
13777
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013778
137797.3.1. Converters
13780-----------------
13781
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013782Sample fetch methods may be combined with transformations to be applied on top
13783of the fetched sample (also called "converters"). These combinations form what
13784is called "sample expressions" and the result is a "sample". Initially this
13785was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010013786has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013787unique-id-format, add-header, ...).
13788
13789These transformations are enumerated as a series of specific keywords after the
13790sample fetch method. These keywords may equally be appended immediately after
13791the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010013792support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013793
Willy Tarreau97707872015-01-27 15:12:13 +010013794A certain category of converters are bitwise and arithmetic operators which
13795support performing basic operations on integers. Some bitwise operations are
13796supported (and, or, xor, cpl) and some arithmetic operations are supported
13797(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
13798bool) which make it possible to report a match without having to write an ACL.
13799
Willy Tarreau74ca5042013-06-11 23:12:07 +020013800The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010013801
Ben Shillitof25e8e52016-12-02 14:25:37 +00001380251d.single(<prop>[,<prop>*])
13803 Returns values for the properties requested as a string, where values are
13804 separated by the delimiter specified with "51degrees-property-separator".
13805 The device is identified using the User-Agent header passed to the
13806 converter. The function can be passed up to five property names, and if a
13807 property name can't be found, the value "NoData" is returned.
13808
13809 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013810 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
13811 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000013812 # User-Agent passed to the converter.
13813 frontend http-in
13814 bind *:8081
13815 default_backend servers
13816 http-request set-header X-51D-DeviceTypeMobileTablet \
13817 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
13818
Willy Tarreau97707872015-01-27 15:12:13 +010013819add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013820 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013821 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013822 name. The name of the variable starts with an indication about its scope. The
13823 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013824 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013825 "sess" : the variable is shared with the whole session
13826 "txn" : the variable is shared with the transaction (request and response)
13827 "req" : the variable is shared only during request processing
13828 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013829 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013830 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013831
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010013832aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
13833 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
13834 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
13835 need to be base64 encoded and the returned result is in raw byte format.
13836 If the <aead_tag> validation fails, the converter doesn't return any data.
13837 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
13838 converter requires at least OpenSSL 1.0.1.
13839
13840 Example:
13841 http-response set-header X-Decrypted-Text %[var(txn.enc),\
13842 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
13843
Willy Tarreau97707872015-01-27 15:12:13 +010013844and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013845 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013846 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013847 numeric value or a variable name. The name of the variable starts with an
13848 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013849 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013850 "sess" : the variable is shared with the whole session
13851 "txn" : the variable is shared with the transaction (request and response)
13852 "req" : the variable is shared only during request processing
13853 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013854 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013855 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013856
Holger Just1bfc24b2017-05-06 00:56:53 +020013857b64dec
13858 Converts (decodes) a base64 encoded input string to its binary
13859 representation. It performs the inverse operation of base64().
13860
Emeric Brun53d1a982014-04-30 18:21:37 +020013861base64
13862 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010013863 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020013864 an SSL ID can be copied in a header).
13865
Willy Tarreau97707872015-01-27 15:12:13 +010013866bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013867 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013868 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013869 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013870 presence of a flag).
13871
Emeric Brun54c4ac82014-11-03 15:32:43 +010013872bytes(<offset>[,<length>])
13873 Extracts some bytes from an input binary sample. The result is a binary
13874 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010013875 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010013876
Willy Tarreau280f42b2018-02-19 15:34:12 +010013877concat([<start>],[<var>],[<end>])
13878 Concatenates up to 3 fields after the current sample which is then turned to
13879 a string. The first one, <start>, is a constant string, that will be appended
13880 immediately after the existing sample. It may be omitted if not used. The
13881 second one, <var>, is a variable name. The variable will be looked up, its
13882 contents converted to a string, and it will be appended immediately after the
13883 <first> part. If the variable is not found, nothing is appended. It may be
13884 omitted as well. The third field, <end> is a constant string that will be
13885 appended after the variable. It may also be omitted. Together, these elements
13886 allow to concatenate variables with delimiters to an existing set of
13887 variables. This can be used to build new variables made of a succession of
John Roesler27c754c2019-07-10 15:45:51 -050013888 other variables, such as colon-delimited values. Note that due to the config
Willy Tarreau280f42b2018-02-19 15:34:12 +010013889 parser, it is not possible to use a comma nor a closing parenthesis as
John Roesler27c754c2019-07-10 15:45:51 -050013890 delimiters.
Willy Tarreau280f42b2018-02-19 15:34:12 +010013891
13892 Example:
13893 tcp-request session set-var(sess.src) src
13894 tcp-request session set-var(sess.dn) ssl_c_s_dn
13895 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
13896 http-request set-header x-hap-sig %[var(txn.sig)]
13897
Willy Tarreau97707872015-01-27 15:12:13 +010013898cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013899 Takes the input value of type signed integer, applies a ones-complement
13900 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010013901
Willy Tarreau80599772015-01-20 19:35:24 +010013902crc32([<avalanche>])
13903 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
13904 hash function. Optionally, it is possible to apply a full avalanche hash
13905 function to the output if the optional <avalanche> argument equals 1. This
13906 converter uses the same functions as used by the various hash-based load
13907 balancing algorithms, so it will provide exactly the same results. It is
13908 provided for compatibility with other software which want a CRC32 to be
13909 computed on some input keys, so it follows the most common implementation as
13910 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
13911 but may provide a better or at least less predictable distribution. It must
13912 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013913 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
13914
13915crc32c([<avalanche>])
13916 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
13917 hash function. Optionally, it is possible to apply a full avalanche hash
13918 function to the output if the optional <avalanche> argument equals 1. This
13919 converter uses the same functions as described in RFC4960, Appendix B [8].
13920 It is provided for compatibility with other software which want a CRC32C to be
13921 computed on some input keys. It is slower than the other algorithms and it must
13922 not be used for security purposes as a 32-bit hash is trivial to break. See
13923 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010013924
David Carlier29b3ca32015-09-25 14:09:21 +010013925da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020013926 Asks the DeviceAtlas converter to identify the User Agent string passed on
13927 input, and to emit a string made of the concatenation of the properties
13928 enumerated in argument, delimited by the separator defined by the global
13929 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000013930 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020013931 configuration language.
13932
13933 Example:
13934 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013935 bind *:8881
13936 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000013937 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 +020013938
Thierry FOURNIER9687c772015-05-07 15:46:29 +020013939debug
13940 This converter is used as debug tool. It dumps on screen the content and the
13941 type of the input sample. The sample is returned as is on its output. This
13942 converter only exists when haproxy was built with debugging enabled.
13943
Willy Tarreau97707872015-01-27 15:12:13 +010013944div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013945 Divides the input value of type signed integer by <value>, and returns the
13946 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013947 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010013948 variable name. The name of the variable starts with an indication about its
13949 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013950 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013951 "sess" : the variable is shared with the whole session
13952 "txn" : the variable is shared with the transaction (request and response)
13953 "req" : the variable is shared only during request processing
13954 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013955 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013956 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013957
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013958djb2([<avalanche>])
13959 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
13960 hash function. Optionally, it is possible to apply a full avalanche hash
13961 function to the output if the optional <avalanche> argument equals 1. This
13962 converter uses the same functions as used by the various hash-based load
13963 balancing algorithms, so it will provide exactly the same results. It is
13964 mostly intended for debugging, but can be used as a stick-table entry to
13965 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013966 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
13967 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013968
Willy Tarreau97707872015-01-27 15:12:13 +010013969even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013970 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010013971 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
13972
Marcin Deranek9631a282018-04-16 14:30:46 +020013973field(<index>,<delimiters>[,<count>])
13974 Extracts the substring at the given index counting from the beginning
13975 (positive index) or from the end (negative index) considering given delimiters
13976 from an input string. Indexes start at 1 or -1 and delimiters are a string
13977 formatted list of chars. Optionally you can specify <count> of fields to
13978 extract (default: 1). Value of 0 indicates extraction of all remaining
13979 fields.
13980
13981 Example :
13982 str(f1_f2_f3__f5),field(5,_) # f5
13983 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
13984 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
13985 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
13986 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010013987
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013988hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010013989 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013990 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010013991 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 +020013992 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010013993
Dragan Dosen3f957b22017-10-24 09:27:34 +020013994hex2i
13995 Converts a hex string containing two hex digits per input byte to an
John Roesler27c754c2019-07-10 15:45:51 -050013996 integer. If the input value cannot be converted, then zero is returned.
Dragan Dosen3f957b22017-10-24 09:27:34 +020013997
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013998http_date([<offset>])
13999 Converts an integer supposed to contain a date since epoch to a string
14000 representing this date in a format suitable for use in HTTP header fields. If
14001 an offset value is specified, then it is a number of seconds that is added to
14002 the date before the conversion is operated. This is particularly useful to
14003 emit Date header fields, Expires values in responses when combined with a
14004 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014005
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014006in_table(<table>)
14007 Uses the string representation of the input sample to perform a look up in
14008 the specified table. If the key is not found in the table, a boolean false
14009 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010014010 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014011 or not a source IP address or an Authorization header was already seen).
14012
Tim Duesterhus1478aa72018-01-25 16:24:51 +010014013ipmask(<mask4>, [<mask6>])
14014 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014015 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010014016 table entries and as such use the same server. The mask4 can be passed in
14017 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
14018 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
14019 If no mask6 is given IPv6 addresses will fail to convert for backwards
14020 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014021
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014022json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010014023 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014024 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020014025 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014026 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
14027 of errors:
14028 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
14029 bytes, ...)
14030 - invalid range (the decoded value is within a UTF-8 prohibited range),
14031 - code overlong (the value is encoded with more bytes than necessary).
14032
14033 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
14034 character is greater than 0xffff because the JSON string escape specification
14035 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
14036 in 4 variants designated by a combination of two suffix letters : "p" for
14037 "permissive" and "s" for "silently ignore". The behaviors of the decoders
14038 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010014039 - "ascii" : never fails;
14040 - "utf8" : fails on any detected errors;
14041 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014042 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010014043 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014044 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
14045 characters corresponding to the other errors.
14046
14047 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010014048 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014049
14050 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014051 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020014052 capture request header user-agent len 150
14053 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020014054
14055 Input request from client 127.0.0.1:
14056 GET / HTTP/1.0
14057 User-Agent: Very "Ugly" UA 1/2
14058
14059 Output log:
14060 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
14061
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014062language(<value>[,<default>])
14063 Returns the value with the highest q-factor from a list as extracted from the
14064 "accept-language" header using "req.fhdr". Values with no q-factor have a
14065 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
14066 belong to the list of semi-colon delimited <values> will be considered. The
14067 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
14068 given list and a default value is provided, it is returned. Note that language
14069 names may have a variant after a dash ('-'). If this variant is present in the
14070 list, it will be matched, but if it is not, only the base language is checked.
14071 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010014072 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014073 ordering of the values in the request counts, as the first value among
14074 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020014075
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014076 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020014077
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014078 # this configuration switches to the backend matching a
14079 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020014080
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014081 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
14082 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
14083 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
14084 use_backend spanish if es
14085 use_backend french if fr
14086 use_backend english if en
14087 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020014088
Willy Tarreau60a2ee72017-12-15 07:13:48 +010014089length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010014090 Get the length of the string. This can only be placed after a string
14091 sample fetch function or after a transformation keyword returning a string
14092 type. The result is of type integer.
14093
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014094lower
14095 Convert a string sample to lower case. This can only be placed after a string
14096 sample fetch function or after a transformation keyword returning a string
14097 type. The result is of type string.
14098
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014099ltime(<format>[,<offset>])
14100 Converts an integer supposed to contain a date since epoch to a string
14101 representing this date in local time using a format defined by the <format>
14102 string using strftime(3). The purpose is to allow any date format to be used
14103 in logs. An optional <offset> in seconds may be applied to the input date
14104 (positive or negative). See the strftime() man page for the format supported
14105 by your operating system. See also the utime converter.
14106
14107 Example :
14108
14109 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014110 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014111 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
14112
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014113map(<map_file>[,<default_value>])
14114map_<match_type>(<map_file>[,<default_value>])
14115map_<match_type>_<output_type>(<map_file>[,<default_value>])
14116 Search the input value from <map_file> using the <match_type> matching method,
14117 and return the associated value converted to the type <output_type>. If the
14118 input value cannot be found in the <map_file>, the converter returns the
14119 <default_value>. If the <default_value> is not set, the converter fails and
14120 acts as if no input value could be fetched. If the <match_type> is not set, it
14121 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
14122 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
14123 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014124
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014125 It is important to avoid overlapping between the keys : IP addresses and
14126 strings are stored in trees, so the first of the finest match will be used.
14127 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014128
Tim Düsterhus4896c442016-11-29 02:15:19 +010014129 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014130 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014131
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014132 input type | match method | output type str | output type int | output type ip
14133 -----------+--------------+-----------------+-----------------+---------------
14134 str | str | map_str | map_str_int | map_str_ip
14135 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020014136 str | beg | map_beg | map_beg_int | map_end_ip
14137 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014138 str | sub | map_sub | map_sub_int | map_sub_ip
14139 -----------+--------------+-----------------+-----------------+---------------
14140 str | dir | map_dir | map_dir_int | map_dir_ip
14141 -----------+--------------+-----------------+-----------------+---------------
14142 str | dom | map_dom | map_dom_int | map_dom_ip
14143 -----------+--------------+-----------------+-----------------+---------------
14144 str | end | map_end | map_end_int | map_end_ip
14145 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080014146 str | reg | map_reg | map_reg_int | map_reg_ip
14147 -----------+--------------+-----------------+-----------------+---------------
14148 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014149 -----------+--------------+-----------------+-----------------+---------------
14150 int | int | map_int | map_int_int | map_int_ip
14151 -----------+--------------+-----------------+-----------------+---------------
14152 ip | ip | map_ip | map_ip_int | map_ip_ip
14153 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014154
Thierry Fournier8feaa662016-02-10 22:55:20 +010014155 The special map called "map_regm" expect matching zone in the regular
14156 expression and modify the output replacing back reference (like "\1") by
14157 the corresponding match text.
14158
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014159 The file contains one key + value per line. Lines which start with '#' are
14160 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
14161 is then the first "word" (series of non-space/tabs characters), and the value
14162 is what follows this series of space/tab till the end of the line excluding
14163 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014164
Thierry FOURNIER060762e2014-04-23 13:29:15 +020014165 Example :
14166
14167 # this is a comment and is ignored
14168 2.22.246.0/23 United Kingdom \n
14169 <-><-----------><--><------------><---->
14170 | | | | `- trailing spaces ignored
14171 | | | `---------- value
14172 | | `-------------------- middle spaces ignored
14173 | `---------------------------- key
14174 `------------------------------------ leading spaces ignored
14175
Willy Tarreau97707872015-01-27 15:12:13 +010014176mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014177 Divides the input value of type signed integer by <value>, and returns the
14178 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014179 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014180 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014181 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014182 "sess" : the variable is shared with the whole session
14183 "txn" : the variable is shared with the transaction (request and response)
14184 "req" : the variable is shared only during request processing
14185 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014186 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014187 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014188
14189mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014190 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020014191 the product as an signed integer. In case of overflow, the largest possible
14192 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014193 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014194 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014195 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014196 "sess" : the variable is shared with the whole session
14197 "txn" : the variable is shared with the transaction (request and response)
14198 "req" : the variable is shared only during request processing
14199 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014200 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014201 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014202
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010014203nbsrv
14204 Takes an input value of type string, interprets it as a backend name and
14205 returns the number of usable servers in that backend. Can be used in places
14206 where we want to look up a backend from a dynamic name, like a result of a
14207 map lookup.
14208
Willy Tarreau97707872015-01-27 15:12:13 +010014209neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014210 Takes the input value of type signed integer, computes the opposite value,
14211 and returns the remainder as an signed integer. 0 is identity. This operator
14212 is provided for reversed subtracts : in order to subtract the input from a
14213 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010014214
14215not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014216 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010014217 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010014218 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010014219 absence of a flag).
14220
14221odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014222 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010014223 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
14224
14225or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014226 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014227 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010014228 numeric value or a variable name. The name of the variable starts with an
14229 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014230 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014231 "sess" : the variable is shared with the whole session
14232 "txn" : the variable is shared with the transaction (request and response)
14233 "req" : the variable is shared only during request processing
14234 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014235 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014236 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014237
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014238protobuf(<field_number>,[<field_type>])
14239 This extracts the protocol buffers message field in raw mode of an input binary
14240 sample representation of a protocol buffer message with <field_number> as field
14241 number (dotted notation) if <field_type> is not present, or as an integer sample
14242 if this field is present (see also "ungrpc" below).
14243 The list of the authorized types is the following one: "int32", "int64", "uint32",
14244 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14245 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14246 "float" for the wire type 5. Note that "string" is considered as a length-delimited
14247 type, so it does not require any <field_type> argument to be extracted.
14248 More information may be found here about the protocol buffers message field types:
14249 https://developers.google.com/protocol-buffers/docs/encoding
14250
Willy Tarreauc4dc3502015-01-23 20:39:28 +010014251regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010014252 Applies a regex-based substitution to the input string. It does the same
14253 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
14254 default it will replace in the input string the first occurrence of the
14255 largest part matching the regular expression <regex> with the substitution
14256 string <subst>. It is possible to replace all occurrences instead by adding
14257 the flag "g" in the third argument <flags>. It is also possible to make the
14258 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
14259 string, it is made up from the concatenation of all desired flags. Thus if
14260 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
14261 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010014262 configuration parser, some characters such as closing parenthesis, closing
14263 square brackets or comma are not possible to use in the arguments. The first
14264 use of this converter is to replace certain characters or sequence of
14265 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010014266
14267 Example :
14268
14269 # de-duplicate "/" in header "x-path".
14270 # input: x-path: /////a///b/c/xzxyz/
14271 # output: x-path: /a/b/c/xzxyz/
14272 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
14273
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014274capture-req(<id>)
14275 Capture the string entry in the request slot <id> and returns the entry as
14276 is. If the slot doesn't exist, the capture fails silently.
14277
14278 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014279 "http-response capture", "capture.req.hdr" and
14280 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014281
14282capture-res(<id>)
14283 Capture the string entry in the response slot <id> and returns the entry as
14284 is. If the slot doesn't exist, the capture fails silently.
14285
14286 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014287 "http-response capture", "capture.req.hdr" and
14288 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014289
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014290sdbm([<avalanche>])
14291 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
14292 hash function. Optionally, it is possible to apply a full avalanche hash
14293 function to the output if the optional <avalanche> argument equals 1. This
14294 converter uses the same functions as used by the various hash-based load
14295 balancing algorithms, so it will provide exactly the same results. It is
14296 mostly intended for debugging, but can be used as a stick-table entry to
14297 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014298 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
14299 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014300
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014301set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010014302 Sets a variable with the input content and returns the content on the output
14303 as-is. The variable keeps the value and the associated input type. The name of
14304 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014305 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014306 "sess" : the variable is shared with the whole session
14307 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014308 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014309 "req" : the variable is shared only during request processing,
14310 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014311 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014312 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014313
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020014314sha1
14315 Converts a binary input sample to a SHA1 digest. The result is a binary
14316 sample with length of 20 bytes.
14317
Tim Duesterhusca097c12018-04-27 21:18:45 +020014318strcmp(<var>)
14319 Compares the contents of <var> with the input value of type string. Returns
14320 the result as a signed integer compatible with strcmp(3): 0 if both strings
14321 are identical. A value less than 0 if the left string is lexicographically
14322 smaller than the right string or if the left string is shorter. A value greater
14323 than 0 otherwise (right string greater than left string or the right string is
14324 shorter).
14325
14326 Example :
14327
14328 http-request set-var(txn.host) hdr(host)
14329 # Check whether the client is attempting domain fronting.
14330 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
14331
14332
Willy Tarreau97707872015-01-27 15:12:13 +010014333sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014334 Subtracts <value> from the input value of type signed integer, and returns
14335 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014336 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010014337 or a variable name. The name of the variable starts with an indication about
14338 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014339 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014340 "sess" : the variable is shared with the whole session
14341 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014342 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014343 "req" : the variable is shared only during request processing,
14344 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014345 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014346 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014347
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014348table_bytes_in_rate(<table>)
14349 Uses the string representation of the input sample to perform a look up in
14350 the specified table. If the key is not found in the table, integer value zero
14351 is returned. Otherwise the converter returns the average client-to-server
14352 bytes rate associated with the input sample in the designated table, measured
14353 in amount of bytes over the period configured in the table. See also the
14354 sc_bytes_in_rate sample fetch keyword.
14355
14356
14357table_bytes_out_rate(<table>)
14358 Uses the string representation of the input sample to perform a look up in
14359 the specified table. If the key is not found in the table, integer value zero
14360 is returned. Otherwise the converter returns the average server-to-client
14361 bytes rate associated with the input sample in the designated table, measured
14362 in amount of bytes over the period configured in the table. See also the
14363 sc_bytes_out_rate sample fetch keyword.
14364
14365table_conn_cnt(<table>)
14366 Uses the string representation of the input sample to perform a look up in
14367 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014368 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014369 connections associated with the input sample in the designated table. See
14370 also the sc_conn_cnt sample fetch keyword.
14371
14372table_conn_cur(<table>)
14373 Uses the string representation of the input sample to perform a look up in
14374 the specified table. If the key is not found in the table, integer value zero
14375 is returned. Otherwise the converter returns the current amount of concurrent
14376 tracked connections associated with the input sample in the designated table.
14377 See also the sc_conn_cur sample fetch keyword.
14378
14379table_conn_rate(<table>)
14380 Uses the string representation of the input sample to perform a look up in
14381 the specified table. If the key is not found in the table, integer value zero
14382 is returned. Otherwise the converter returns the average incoming connection
14383 rate associated with the input sample in the designated table. See also the
14384 sc_conn_rate sample fetch keyword.
14385
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014386table_gpt0(<table>)
14387 Uses the string representation of the input sample to perform a look up in
14388 the specified table. If the key is not found in the table, boolean value zero
14389 is returned. Otherwise the converter returns the current value of the first
14390 general purpose tag associated with the input sample in the designated table.
14391 See also the sc_get_gpt0 sample fetch keyword.
14392
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014393table_gpc0(<table>)
14394 Uses the string representation of the input sample to perform a look up in
14395 the specified table. If the key is not found in the table, integer value zero
14396 is returned. Otherwise the converter returns the current value of the first
14397 general purpose counter associated with the input sample in the designated
14398 table. See also the sc_get_gpc0 sample fetch keyword.
14399
14400table_gpc0_rate(<table>)
14401 Uses the string representation of the input sample to perform a look up in
14402 the specified table. If the key is not found in the table, integer value zero
14403 is returned. Otherwise the converter returns the frequency which the gpc0
14404 counter was incremented over the configured period in the table, associated
14405 with the input sample in the designated table. See also the sc_get_gpc0_rate
14406 sample fetch keyword.
14407
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014408table_gpc1(<table>)
14409 Uses the string representation of the input sample to perform a look up in
14410 the specified table. If the key is not found in the table, integer value zero
14411 is returned. Otherwise the converter returns the current value of the second
14412 general purpose counter associated with the input sample in the designated
14413 table. See also the sc_get_gpc1 sample fetch keyword.
14414
14415table_gpc1_rate(<table>)
14416 Uses the string representation of the input sample to perform a look up in
14417 the specified table. If the key is not found in the table, integer value zero
14418 is returned. Otherwise the converter returns the frequency which the gpc1
14419 counter was incremented over the configured period in the table, associated
14420 with the input sample in the designated table. See also the sc_get_gpc1_rate
14421 sample fetch keyword.
14422
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014423table_http_err_cnt(<table>)
14424 Uses the string representation of the input sample to perform a look up in
14425 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014426 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014427 errors associated with the input sample in the designated table. See also the
14428 sc_http_err_cnt sample fetch keyword.
14429
14430table_http_err_rate(<table>)
14431 Uses the string representation of the input sample to perform a look up in
14432 the specified table. If the key is not found in the table, integer value zero
14433 is returned. Otherwise the average rate of HTTP errors associated with the
14434 input sample in the designated table, measured in amount of errors over the
14435 period configured in the table. See also the sc_http_err_rate sample fetch
14436 keyword.
14437
14438table_http_req_cnt(<table>)
14439 Uses the string representation of the input sample to perform a look up in
14440 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014441 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014442 requests associated with the input sample in the designated table. See also
14443 the sc_http_req_cnt sample fetch keyword.
14444
14445table_http_req_rate(<table>)
14446 Uses the string representation of the input sample to perform a look up in
14447 the specified table. If the key is not found in the table, integer value zero
14448 is returned. Otherwise the average rate of HTTP requests associated with the
14449 input sample in the designated table, measured in amount of requests over the
14450 period configured in the table. See also the sc_http_req_rate sample fetch
14451 keyword.
14452
14453table_kbytes_in(<table>)
14454 Uses the string representation of the input sample to perform a look up in
14455 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014456 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014457 to-server data associated with the input sample in the designated table,
14458 measured in kilobytes. The test is currently performed on 32-bit integers,
14459 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
14460 keyword.
14461
14462table_kbytes_out(<table>)
14463 Uses the string representation of the input sample to perform a look up in
14464 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014465 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014466 to-client data associated with the input sample in the designated table,
14467 measured in kilobytes. The test is currently performed on 32-bit integers,
14468 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
14469 keyword.
14470
14471table_server_id(<table>)
14472 Uses the string representation of the input sample to perform a look up in
14473 the specified table. If the key is not found in the table, integer value zero
14474 is returned. Otherwise the converter returns the server ID associated with
14475 the input sample in the designated table. A server ID is associated to a
14476 sample by a "stick" rule when a connection to a server succeeds. A server ID
14477 zero means that no server is associated with this key.
14478
14479table_sess_cnt(<table>)
14480 Uses the string representation of the input sample to perform a look up in
14481 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014482 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014483 sessions associated with the input sample in the designated table. Note that
14484 a session here refers to an incoming connection being accepted by the
14485 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
14486 keyword.
14487
14488table_sess_rate(<table>)
14489 Uses the string representation of the input sample to perform a look up in
14490 the specified table. If the key is not found in the table, integer value zero
14491 is returned. Otherwise the converter returns the average incoming session
14492 rate associated with the input sample in the designated table. Note that a
14493 session here refers to an incoming connection being accepted by the
14494 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
14495 keyword.
14496
14497table_trackers(<table>)
14498 Uses the string representation of the input sample to perform a look up in
14499 the specified table. If the key is not found in the table, integer value zero
14500 is returned. Otherwise the converter returns the current amount of concurrent
14501 connections tracking the same key as the input sample in the designated
14502 table. It differs from table_conn_cur in that it does not rely on any stored
14503 information but on the table's reference count (the "use" value which is
14504 returned by "show table" on the CLI). This may sometimes be more suited for
14505 layer7 tracking. It can be used to tell a server how many concurrent
14506 connections there are from a given address for example. See also the
14507 sc_trackers sample fetch keyword.
14508
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014509upper
14510 Convert a string sample to upper case. This can only be placed after a string
14511 sample fetch function or after a transformation keyword returning a string
14512 type. The result is of type string.
14513
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020014514url_dec
14515 Takes an url-encoded string provided as input and returns the decoded
14516 version as output. The input and the output are of type string.
14517
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014518ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014519 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014520 sample representation of a gRPC message with <field_number> as field number
14521 (dotted notation) if <field_type> is not present, or as an integer sample if this
14522 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014523 The list of the authorized types is the following one: "int32", "int64", "uint32",
14524 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14525 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14526 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014527 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014528 More information may be found here about the protocol buffers message field types:
14529 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014530
14531 Example:
14532 // with such a protocol buffer .proto file content adapted from
14533 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
14534
14535 message Point {
14536 int32 latitude = 1;
14537 int32 longitude = 2;
14538 }
14539
14540 message PPoint {
14541 Point point = 59;
14542 }
14543
14544 message Rectangle {
14545 // One corner of the rectangle.
14546 PPoint lo = 48;
14547 // The other corner of the rectangle.
14548 PPoint hi = 49;
14549 }
14550
14551 let's say a body request is made of a "Rectangle" object value (two PPoint
14552 protocol buffers messages), the four protocol buffers fields could be
14553 extracted with these "ungrpc" directives:
14554
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014555 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
14556 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
John Roesler27c754c2019-07-10 15:45:51 -050014557 req.body,ungrpc(49.59.1,int32) # "latitude" of "hi" second PPoint
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014558 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
14559
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014560 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014561
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014562 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014563
John Roesler27c754c2019-07-10 15:45:51 -050014564 As a gRPC message is always made of a gRPC header followed by protocol buffers
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014565 messages, in the previous example the "latitude" of "lo" first PPoint
14566 could be extracted with these equivalent directives:
14567
14568 req.body,ungrpc(48.59),protobuf(1,int32)
14569 req.body,ungrpc(48),protobuf(59.1,int32)
14570 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
14571
14572 Note that the first convert must be "ungrpc", the remaining ones must be
14573 "protobuf" and only the last one may have or not a second argument to
14574 interpret the previous binary sample.
14575
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014576
Christopher Faulet85d79c92016-11-09 16:54:56 +010014577unset-var(<var name>)
14578 Unsets a variable if the input content is defined. The name of the variable
14579 starts with an indication about its scope. The scopes allowed are:
14580 "proc" : the variable is shared with the whole process
14581 "sess" : the variable is shared with the whole session
14582 "txn" : the variable is shared with the transaction (request and
14583 response),
14584 "req" : the variable is shared only during request processing,
14585 "res" : the variable is shared only during response processing.
14586 This prefix is followed by a name. The separator is a '.'. The name may only
14587 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
14588
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014589utime(<format>[,<offset>])
14590 Converts an integer supposed to contain a date since epoch to a string
14591 representing this date in UTC time using a format defined by the <format>
14592 string using strftime(3). The purpose is to allow any date format to be used
14593 in logs. An optional <offset> in seconds may be applied to the input date
14594 (positive or negative). See the strftime() man page for the format supported
14595 by your operating system. See also the ltime converter.
14596
14597 Example :
14598
14599 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014600 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014601 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
14602
Marcin Deranek9631a282018-04-16 14:30:46 +020014603word(<index>,<delimiters>[,<count>])
14604 Extracts the nth word counting from the beginning (positive index) or from
14605 the end (negative index) considering given delimiters from an input string.
14606 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
Jerome Magnind1fa5fa2020-01-28 13:33:44 +010014607 Delimiters at the beginning or end of the input string are ignored.
Marcin Deranek9631a282018-04-16 14:30:46 +020014608 Optionally you can specify <count> of words to extract (default: 1).
14609 Value of 0 indicates extraction of all remaining words.
14610
14611 Example :
14612 str(f1_f2_f3__f5),word(4,_) # f5
14613 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
14614 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
14615 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
14616 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Jerome Magnind1fa5fa2020-01-28 13:33:44 +010014617 str(/f1/f2/f3/f4),word(1,/) # f1
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010014618
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014619wt6([<avalanche>])
14620 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
14621 hash function. Optionally, it is possible to apply a full avalanche hash
14622 function to the output if the optional <avalanche> argument equals 1. This
14623 converter uses the same functions as used by the various hash-based load
14624 balancing algorithms, so it will provide exactly the same results. It is
14625 mostly intended for debugging, but can be used as a stick-table entry to
14626 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014627 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
14628 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014629
Willy Tarreau97707872015-01-27 15:12:13 +010014630xor(<value>)
14631 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014632 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014633 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014634 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014635 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014636 "sess" : the variable is shared with the whole session
14637 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014638 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014639 "req" : the variable is shared only during request processing,
14640 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014641 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014642 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014643
Thierry FOURNIER01e09742016-12-26 11:46:11 +010014644xxh32([<seed>])
14645 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
14646 variant of the XXHash hash function. This hash supports a seed which defaults
14647 to zero but a different value maybe passed as the <seed> argument. This hash
14648 is known to be very good and very fast so it can be used to hash URLs and/or
14649 URL parameters for use as stick-table keys to collect statistics with a low
14650 collision rate, though care must be taken as the algorithm is not considered
14651 as cryptographically secure.
14652
14653xxh64([<seed>])
14654 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
14655 variant of the XXHash hash function. This hash supports a seed which defaults
14656 to zero but a different value maybe passed as the <seed> argument. This hash
14657 is known to be very good and very fast so it can be used to hash URLs and/or
14658 URL parameters for use as stick-table keys to collect statistics with a low
14659 collision rate, though care must be taken as the algorithm is not considered
14660 as cryptographically secure.
14661
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014662
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200146637.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020014664--------------------------------------------
14665
14666A first set of sample fetch methods applies to internal information which does
14667not even relate to any client information. These ones are sometimes used with
14668"monitor-fail" directives to report an internal status to external watchers.
14669The sample fetch methods described in this section are usable anywhere.
14670
14671always_false : boolean
14672 Always returns the boolean "false" value. It may be used with ACLs as a
14673 temporary replacement for another one when adjusting configurations.
14674
14675always_true : boolean
14676 Always returns the boolean "true" value. It may be used with ACLs as a
14677 temporary replacement for another one when adjusting configurations.
14678
14679avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014680 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020014681 divided by the number of active servers. The current backend is used if no
14682 backend is specified. This is very similar to "queue" except that the size of
14683 the farm is considered, in order to give a more accurate measurement of the
14684 time it may take for a new connection to be processed. The main usage is with
14685 ACL to return a sorry page to new users when it becomes certain they will get
14686 a degraded service, or to pass to the backend servers in a header so that
14687 they decide to work in degraded mode or to disable some functions to speed up
14688 the processing a bit. Note that in the event there would not be any active
14689 server anymore, twice the number of queued connections would be considered as
14690 the measured value. This is a fair estimate, as we expect one server to get
14691 back soon anyway, but we still prefer to send new traffic to another backend
14692 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
14693 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010014694
Willy Tarreau74ca5042013-06-11 23:12:07 +020014695be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020014696 Applies to the number of currently established connections on the backend,
14697 possibly including the connection being evaluated. If no backend name is
14698 specified, the current one is used. But it is also possible to check another
14699 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014700 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
14701
14702be_conn_free([<backend>]) : integer
14703 Returns an integer value corresponding to the number of available connections
14704 across available servers in the backend. Queue slots are not included. Backup
14705 servers are also not included, unless all other servers are down. If no
14706 backend name is specified, the current one is used. But it is also possible
14707 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014708 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
14709 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014710
14711 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
14712 (meaning unlimited), then this fetch clearly does not make sense, in which
14713 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014714
Willy Tarreau74ca5042013-06-11 23:12:07 +020014715be_sess_rate([<backend>]) : integer
14716 Returns an integer value corresponding to the sessions creation rate on the
14717 backend, in number of new sessions per second. This is used with ACLs to
14718 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010014719 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014720 online dictionary). It can also be useful to add this element to logs using a
14721 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014722
14723 Example :
14724 # Redirect to an error page if the dictionary is requested too often
14725 backend dynamic
14726 mode http
14727 acl being_scanned be_sess_rate gt 100
14728 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010014729
Davor Ocelice9ed2812017-12-25 17:49:28 +010014730bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014731 Returns a binary chain. The input is the hexadecimal representation
14732 of the string.
14733
14734bool(<bool>) : bool
14735 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
14736 'false' and '0' are the same. 'true' and '1' are the same.
14737
Willy Tarreau74ca5042013-06-11 23:12:07 +020014738connslots([<backend>]) : integer
14739 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014740 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020014741 connections on all servers and the maximum queue size. This is probably only
14742 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050014743
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014744 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020014745 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014746 usage; see "use_backend" keyword) can be redirected to a different backend.
14747
Willy Tarreau55165fe2009-05-10 12:02:55 +020014748 'connslots' = number of available server connection slots, + number of
14749 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014750
Willy Tarreaua36af912009-10-10 12:02:45 +020014751 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020014752 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020014753 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020014754 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010014755 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020014756 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020014757 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014758
Willy Tarreau55165fe2009-05-10 12:02:55 +020014759 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
14760 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020014761 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020014762 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014763
Willy Tarreau70fe9442018-11-22 16:07:39 +010014764cpu_calls : integer
14765 Returns the number of calls to the task processing the stream or current
14766 request since it was allocated. This number is reset for each new request on
14767 the same connections in case of HTTP keep-alive. This value should usually be
14768 low and stable (around 2 calls for a typically simple request) but may become
14769 high if some processing (compression, caching or analysis) is performed. This
14770 is purely for performance monitoring purposes.
14771
14772cpu_ns_avg : integer
14773 Returns the average number of nanoseconds spent in each call to the task
14774 processing the stream or current request. This number is reset for each new
14775 request on the same connections in case of HTTP keep-alive. This value
14776 indicates the overall cost of processing the request or the connection for
14777 each call. There is no good nor bad value but the time spent in a call
14778 automatically causes latency for other processing (see lat_ns_avg below),
14779 and may affect other connection's apparent response time. Certain operations
14780 like compression, complex regex matching or heavy Lua operations may directly
14781 affect this value, and having it in the logs will make it easier to spot the
14782 faulty processing that needs to be fixed to recover decent performance.
14783 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
14784
14785cpu_ns_tot : integer
14786 Returns the total number of nanoseconds spent in each call to the task
14787 processing the stream or current request. This number is reset for each new
14788 request on the same connections in case of HTTP keep-alive. This value
14789 indicates the overall cost of processing the request or the connection for
14790 each call. There is no good nor bad value but the time spent in a call
14791 automatically causes latency for other processing (see lat_ns_avg below),
14792 induces CPU costs on the machine, and may affect other connection's apparent
14793 response time. Certain operations like compression, complex regex matching or
14794 heavy Lua operations may directly affect this value, and having it in the
14795 logs will make it easier to spot the faulty processing that needs to be fixed
14796 to recover decent performance. The value may be artificially high due to a
14797 high cpu_calls count, for example when processing many HTTP chunks, and for
14798 this reason it is often preferred to log cpu_ns_avg instead.
14799
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014800date([<offset>]) : integer
14801 Returns the current date as the epoch (number of seconds since 01/01/1970).
14802 If an offset value is specified, then it is a number of seconds that is added
14803 to the current date before returning the value. This is particularly useful
14804 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020014805 It is useful combined with the http_date converter.
14806
14807 Example :
14808
14809 # set an expires header to now+1 hour in every response
14810 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014811
Etienne Carrierea792a0a2018-01-17 13:43:24 +010014812date_us : integer
14813 Return the microseconds part of the date (the "second" part is returned by
14814 date sample). This sample is coherent with the date sample as it is comes
14815 from the same timeval structure.
14816
Willy Tarreaud716f9b2017-10-13 11:03:15 +020014817distcc_body(<token>[,<occ>]) : binary
14818 Parses a distcc message and returns the body associated to occurrence #<occ>
14819 of the token <token>. Occurrences start at 1, and when unspecified, any may
14820 match though in practice only the first one is checked for now. This can be
14821 used to extract file names or arguments in files built using distcc through
14822 haproxy. Please refer to distcc's protocol documentation for the complete
14823 list of supported tokens.
14824
14825distcc_param(<token>[,<occ>]) : integer
14826 Parses a distcc message and returns the parameter associated to occurrence
14827 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
14828 any may match though in practice only the first one is checked for now. This
14829 can be used to extract certain information such as the protocol version, the
14830 file size or the argument in files built using distcc through haproxy.
14831 Another use case consists in waiting for the start of the preprocessed file
14832 contents before connecting to the server to avoid keeping idle connections.
14833 Please refer to distcc's protocol documentation for the complete list of
14834 supported tokens.
14835
14836 Example :
14837 # wait up to 20s for the pre-processed file to be uploaded
14838 tcp-request inspect-delay 20s
14839 tcp-request content accept if { distcc_param(DOTI) -m found }
14840 # send large files to the big farm
14841 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
14842
Willy Tarreau595ec542013-06-12 21:34:28 +020014843env(<name>) : string
14844 Returns a string containing the value of environment variable <name>. As a
14845 reminder, environment variables are per-process and are sampled when the
14846 process starts. This can be useful to pass some information to a next hop
14847 server, or with ACLs to take specific action when the process is started a
14848 certain way.
14849
14850 Examples :
14851 # Pass the Via header to next hop with the local hostname in it
14852 http-request add-header Via 1.1\ %[env(HOSTNAME)]
14853
14854 # reject cookie-less requests when the STOP environment variable is set
14855 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
14856
Willy Tarreau74ca5042013-06-11 23:12:07 +020014857fe_conn([<frontend>]) : integer
14858 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010014859 possibly including the connection being evaluated. If no frontend name is
14860 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014861 frontend. It can be used to return a sorry page before hard-blocking, or to
14862 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010014863 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020014864 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
14865 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020014866
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020014867fe_req_rate([<frontend>]) : integer
14868 Returns an integer value corresponding to the number of HTTP requests per
14869 second sent to a frontend. This number can differ from "fe_sess_rate" in
14870 situations where client-side keep-alive is enabled.
14871
Willy Tarreau74ca5042013-06-11 23:12:07 +020014872fe_sess_rate([<frontend>]) : integer
14873 Returns an integer value corresponding to the sessions creation rate on the
14874 frontend, in number of new sessions per second. This is used with ACLs to
14875 limit the incoming session rate to an acceptable range in order to prevent
14876 abuse of service at the earliest moment, for example when combined with other
14877 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
14878 down below the limit. It can also be useful to add this element to logs using
14879 a log-format directive. See also the "rate-limit sessions" directive for use
14880 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014881
14882 Example :
14883 # This frontend limits incoming mails to 10/s with a max of 100
14884 # concurrent connections. We accept any connection below 10/s, and
14885 # force excess clients to wait for 100 ms. Since clients are limited to
14886 # 100 max, there cannot be more than 10 incoming mails per second.
14887 frontend mail
14888 bind :25
14889 mode tcp
14890 maxconn 100
14891 acl too_fast fe_sess_rate ge 10
14892 tcp-request inspect-delay 100ms
14893 tcp-request content accept if ! too_fast
14894 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010014895
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010014896hostname : string
14897 Returns the system hostname.
14898
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014899int(<integer>) : signed integer
14900 Returns a signed integer.
14901
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014902ipv4(<ipv4>) : ipv4
14903 Returns an ipv4.
14904
14905ipv6(<ipv6>) : ipv6
14906 Returns an ipv6.
14907
Willy Tarreau70fe9442018-11-22 16:07:39 +010014908lat_ns_avg : integer
14909 Returns the average number of nanoseconds spent between the moment the task
14910 handling the stream is woken up and the moment it is effectively called. This
14911 number is reset for each new request on the same connections in case of HTTP
14912 keep-alive. This value indicates the overall latency inflicted to the current
14913 request by all other requests being processed in parallel, and is a direct
14914 indicator of perceived performance due to noisy neighbours. In order to keep
14915 the value low, it is possible to reduce the scheduler's run queue depth using
14916 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14917 once using "tune.maxpollevents", to decrease the stream's nice value using
14918 the "nice" option on the "bind" lines or in the frontend, or to look for
14919 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14920 whose processing needs to be adjusted or fixed. Compression of large buffers
14921 could be a culprit, like heavy regex or long lists of regex.
14922 Note: this value is exactly lat_ns_tot divided by cpu_calls.
14923
14924lat_ns_tot : integer
14925 Returns the total number of nanoseconds spent between the moment the task
14926 handling the stream is woken up and the moment it is effectively called. This
14927 number is reset for each new request on the same connections in case of HTTP
14928 keep-alive. This value indicates the overall latency inflicted to the current
14929 request by all other requests being processed in parallel, and is a direct
14930 indicator of perceived performance due to noisy neighbours. In order to keep
14931 the value low, it is possible to reduce the scheduler's run queue depth using
14932 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14933 once using "tune.maxpollevents", to decrease the stream's nice value using
14934 the "nice" option on the "bind" lines or in the frontend, or to look for
14935 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14936 whose processing needs to be adjusted or fixed. Compression of large buffers
14937 could be a culprit, like heavy regex or long lists of regex. Note: while it
14938 may intuitively seem that the total latency adds to a transfer time, it is
14939 almost never true because while a task waits for the CPU, network buffers
14940 continue to fill up and the next call will process more at once. The value
14941 may be artificially high due to a high cpu_calls count, for example when
14942 processing many HTTP chunks, and for this reason it is often preferred to log
14943 lat_ns_avg instead, which is a more relevant performance indicator.
14944
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014945meth(<method>) : method
14946 Returns a method.
14947
Willy Tarreau0f30d262014-11-24 16:02:05 +010014948nbproc : integer
14949 Returns an integer value corresponding to the number of processes that were
14950 started (it equals the global "nbproc" setting). This is useful for logging
14951 and debugging purposes.
14952
Willy Tarreau74ca5042013-06-11 23:12:07 +020014953nbsrv([<backend>]) : integer
14954 Returns an integer value corresponding to the number of usable servers of
14955 either the current backend or the named backend. This is mostly used with
14956 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010014957 switch to an alternate backend when the number of servers is too low to
14958 to handle some load. It is useful to report a failure when combined with
14959 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014960
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040014961prio_class : integer
14962 Returns the priority class of the current session for http mode or connection
14963 for tcp mode. The value will be that set by the last call to "http-request
14964 set-priority-class" or "tcp-request content set-priority-class".
14965
14966prio_offset : integer
14967 Returns the priority offset of the current session for http mode or
14968 connection for tcp mode. The value will be that set by the last call to
14969 "http-request set-priority-offset" or "tcp-request content
14970 set-priority-offset".
14971
Willy Tarreau0f30d262014-11-24 16:02:05 +010014972proc : integer
14973 Returns an integer value corresponding to the position of the process calling
14974 the function, between 1 and global.nbproc. This is useful for logging and
14975 debugging purposes.
14976
Willy Tarreau74ca5042013-06-11 23:12:07 +020014977queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014978 Returns the total number of queued connections of the designated backend,
14979 including all the connections in server queues. If no backend name is
14980 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014981 one. This is useful with ACLs or to pass statistics to backend servers. This
14982 can be used to take actions when queuing goes above a known level, generally
14983 indicating a surge of traffic or a massive slowdown on the servers. One
14984 possible action could be to reject new users but still accept old ones. See
14985 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
14986
Willy Tarreau84310e22014-02-14 11:59:04 +010014987rand([<range>]) : integer
14988 Returns a random integer value within a range of <range> possible values,
14989 starting at zero. If the range is not specified, it defaults to 2^32, which
14990 gives numbers between 0 and 4294967295. It can be useful to pass some values
14991 needed to take some routing decisions for example, or just for debugging
14992 purposes. This random must not be used for security purposes.
14993
Luca Schimweg77306662019-09-10 15:42:52 +020014994uuid([<version>]) : string
14995 Returns a UUID following the RFC4122 standard. If the version is not
14996 specified, a UUID version 4 (fully random) is returned.
14997 Currently, only version 4 is supported.
14998
Willy Tarreau74ca5042013-06-11 23:12:07 +020014999srv_conn([<backend>/]<server>) : integer
15000 Returns an integer value corresponding to the number of currently established
15001 connections on the designated server, possibly including the connection being
15002 evaluated. If <backend> is omitted, then the server is looked up in the
15003 current backend. It can be used to use a specific farm when one server is
15004 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040015005 connections with it. See also the "fe_conn", "be_conn", "queue", and
15006 "srv_conn_free" fetch methods.
15007
15008srv_conn_free([<backend>/]<server>) : integer
15009 Returns an integer value corresponding to the number of available connections
15010 on the designated server, possibly including the connection being evaluated.
15011 The value does not include queue slots. If <backend> is omitted, then the
15012 server is looked up in the current backend. It can be used to use a specific
15013 farm when one server is full, or to inform the server about our view of the
15014 number of active connections with it. See also the "be_conn_free" and
15015 "srv_conn" fetch methods.
15016
15017 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
15018 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015019
15020srv_is_up([<backend>/]<server>) : boolean
15021 Returns true when the designated server is UP, and false when it is either
15022 DOWN or in maintenance mode. If <backend> is omitted, then the server is
15023 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010015024 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015025 availability). Another possible use which is more of a hack consists in
15026 using dummy servers as boolean variables that can be enabled or disabled from
15027 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
15028
Willy Tarreauff2b7af2017-10-13 11:46:26 +020015029srv_queue([<backend>/]<server>) : integer
15030 Returns an integer value corresponding to the number of connections currently
15031 pending in the designated server's queue. If <backend> is omitted, then the
15032 server is looked up in the current backend. It can sometimes be used together
15033 with the "use-server" directive to force to use a known faster server when it
15034 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
15035 fetch methods.
15036
Willy Tarreau74ca5042013-06-11 23:12:07 +020015037srv_sess_rate([<backend>/]<server>) : integer
15038 Returns an integer corresponding to the sessions creation rate on the
15039 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015040 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020015041 used with ACLs but can make sense with logs too. This is used to switch to an
15042 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010015043 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020015044 overloading servers).
15045
15046 Example :
15047 # Redirect to a separate back
15048 acl srv1_full srv_sess_rate(be1/srv1) gt 50
15049 acl srv2_full srv_sess_rate(be1/srv2) gt 50
15050 use_backend be2 if srv1_full or srv2_full
15051
Willy Tarreau0f30d262014-11-24 16:02:05 +010015052stopping : boolean
15053 Returns TRUE if the process calling the function is currently stopping. This
15054 can be useful for logging, or for relaxing certain checks or helping close
15055 certain connections upon graceful shutdown.
15056
Thierry FOURNIERcc103292015-06-06 19:30:17 +020015057str(<string>) : string
15058 Returns a string.
15059
Willy Tarreau74ca5042013-06-11 23:12:07 +020015060table_avl([<table>]) : integer
15061 Returns the total number of available entries in the current proxy's
15062 stick-table or in the designated stick-table. See also table_cnt.
15063
15064table_cnt([<table>]) : integer
15065 Returns the total number of entries currently in use in the current proxy's
15066 stick-table or in the designated stick-table. See also src_conn_cnt and
15067 table_avl for other entry counting methods.
15068
Christopher Faulet34adb2a2017-11-21 21:45:38 +010015069thread : integer
15070 Returns an integer value corresponding to the position of the thread calling
15071 the function, between 0 and (global.nbthread-1). This is useful for logging
15072 and debugging purposes.
15073
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015074var(<var-name>) : undefined
15075 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010015076 sample fetch fails. The name of the variable starts with an indication
15077 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010015078 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010015079 "sess" : the variable is shared with the whole session
15080 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015081 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010015082 "req" : the variable is shared only during request processing,
15083 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015084 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010015085 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020015086
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200150877.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020015088----------------------------------
15089
15090The layer 4 usually describes just the transport layer which in haproxy is
15091closest to the connection, where no content is yet made available. The fetch
15092methods described here are usable as low as the "tcp-request connection" rule
15093sets unless they require some future information. Those generally include
15094TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015095the incoming connection. For retrieving a value from a sticky counters, the
15096counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020015097"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
15098used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
15099can be specified as the first integer argument when using the "sc_" prefix.
15100Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
15101table may be specified with the "sc*" form, in which case the currently
15102tracked key will be looked up into this alternate table instead of the table
15103currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015104
Jérôme Magnin35e53a62019-01-16 14:38:37 +010015105bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010015106 Returns the backend connection's HTTP major version encoding, which may be 1
15107 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
15108 encoding and not the version present in the request header.
15109
Willy Tarreau74ca5042013-06-11 23:12:07 +020015110be_id : integer
15111 Returns an integer containing the current backend's id. It can be used in
15112 frontends with responses to check which backend processed the request.
15113
Marcin Deranekd2471c22016-12-12 14:08:05 +010015114be_name : string
15115 Returns a string containing the current backend's name. It can be used in
15116 frontends with responses to check which backend processed the request.
15117
Willy Tarreau74ca5042013-06-11 23:12:07 +020015118dst : ip
15119 This is the destination IPv4 address of the connection on the client side,
15120 which is the address the client connected to. It can be useful when running
15121 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
15122 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015123 RFC 4291. When the incoming connection passed through address translation or
15124 redirection involving connection tracking, the original destination address
15125 before the redirection will be reported. On Linux systems, the source and
15126 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
15127 is set, because a late response may reopen a timed out connection and switch
15128 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015129
15130dst_conn : integer
15131 Returns an integer value corresponding to the number of currently established
15132 connections on the same socket including the one being evaluated. It is
15133 normally used with ACLs but can as well be used to pass the information to
15134 servers in an HTTP header or in logs. It can be used to either return a sorry
15135 page before hard-blocking, or to use a specific backend to drain new requests
15136 when the socket is considered saturated. This offers the ability to assign
15137 different limits to different listening ports or addresses. See also the
15138 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015139
Willy Tarreau16e01562016-08-09 16:46:18 +020015140dst_is_local : boolean
15141 Returns true if the destination address of the incoming connection is local
15142 to the system, or false if the address doesn't exist on the system, meaning
15143 that it was intercepted in transparent mode. It can be useful to apply
15144 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010015145 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020015146 be delivered only on this address, or SSH access could be locally redirected.
15147 Please note that the check involves a few system calls, so it's better to do
15148 it only once per connection.
15149
Willy Tarreau74ca5042013-06-11 23:12:07 +020015150dst_port : integer
15151 Returns an integer value corresponding to the destination TCP port of the
15152 connection on the client side, which is the port the client connected to.
15153 This might be used when running in transparent mode, when assigning dynamic
15154 ports to some clients for a whole application session, to stick all users to
15155 a same server, or to pass the destination port information to a server using
15156 an HTTP header.
15157
Willy Tarreau60ca10a2017-08-18 15:26:54 +020015158fc_http_major : integer
15159 Reports the front connection's HTTP major version encoding, which may be 1
15160 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
15161 encoding and not on the version present in the request header.
15162
Emeric Brun4f603012017-01-05 15:11:44 +010015163fc_rcvd_proxy : boolean
15164 Returns true if the client initiated the connection with a PROXY protocol
15165 header.
15166
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020015167fc_rtt(<unit>) : integer
15168 Returns the Round Trip Time (RTT) measured by the kernel for the client
15169 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
15170 can be set to "ms" for milliseconds or "us" for microseconds. If the server
15171 connection is not established, if the connection is not TCP or if the
15172 operating system does not support TCP_INFO, for example Linux kernels before
15173 2.4, the sample fetch fails.
15174
15175fc_rttvar(<unit>) : integer
15176 Returns the Round Trip Time (RTT) variance measured by the kernel for the
15177 client connection. <unit> is facultative, by default the unit is milliseconds.
15178 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
15179 server connection is not established, if the connection is not TCP or if the
15180 operating system does not support TCP_INFO, for example Linux kernels before
15181 2.4, the sample fetch fails.
15182
Christopher Faulet297df182019-10-17 14:40:48 +020015183fc_unacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015184 Returns the unacked counter measured by the kernel for the client connection.
15185 If the server connection is not established, if the connection is not TCP or
15186 if the operating system does not support TCP_INFO, for example Linux kernels
15187 before 2.4, the sample fetch fails.
15188
Christopher Faulet297df182019-10-17 14:40:48 +020015189fc_sacked : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015190 Returns the sacked counter measured by the kernel for the client connection.
15191 If the server connection is not established, if the connection is not TCP or
15192 if the operating system does not support TCP_INFO, for example Linux kernels
15193 before 2.4, the sample fetch fails.
15194
Christopher Faulet297df182019-10-17 14:40:48 +020015195fc_retrans : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015196 Returns the retransmits counter measured by the kernel for the client
15197 connection. If the server connection is not established, if the connection is
15198 not TCP or if the operating system does not support TCP_INFO, for example
15199 Linux kernels before 2.4, the sample fetch fails.
15200
Christopher Faulet297df182019-10-17 14:40:48 +020015201fc_fackets : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015202 Returns the fack counter measured by the kernel for the client
15203 connection. If the server connection is not established, if the connection is
15204 not TCP or if the operating system does not support TCP_INFO, for example
15205 Linux kernels before 2.4, the sample fetch fails.
15206
Christopher Faulet297df182019-10-17 14:40:48 +020015207fc_lost : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015208 Returns the lost counter measured by the kernel for the client
15209 connection. If the server connection is not established, if the connection is
15210 not TCP or if the operating system does not support TCP_INFO, for example
15211 Linux kernels before 2.4, the sample fetch fails.
15212
Christopher Faulet297df182019-10-17 14:40:48 +020015213fc_reordering : integer
Joe Williams30fcd392016-08-10 07:06:44 -070015214 Returns the reordering counter measured by the kernel for the client
15215 connection. If the server connection is not established, if the connection is
15216 not TCP or if the operating system does not support TCP_INFO, for example
15217 Linux kernels before 2.4, the sample fetch fails.
15218
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020015219fe_defbe : string
15220 Returns a string containing the frontend's default backend name. It can be
15221 used in frontends to check which backend will handle requests by default.
15222
Willy Tarreau74ca5042013-06-11 23:12:07 +020015223fe_id : integer
15224 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010015225 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020015226 coming via a same frontend to the same server.
15227
Marcin Deranekd2471c22016-12-12 14:08:05 +010015228fe_name : string
15229 Returns a string containing the current frontend's name. It can be used in
15230 backends to check from which frontend it was called, or to stick all users
15231 coming via a same frontend to the same server.
15232
Cyril Bonté62ba8702014-04-22 23:52:25 +020015233sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015234sc0_bytes_in_rate([<table>]) : integer
15235sc1_bytes_in_rate([<table>]) : integer
15236sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015237 Returns the average client-to-server bytes rate from the currently tracked
15238 counters, measured in amount of bytes over the period configured in the
15239 table. See also src_bytes_in_rate.
15240
Cyril Bonté62ba8702014-04-22 23:52:25 +020015241sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015242sc0_bytes_out_rate([<table>]) : integer
15243sc1_bytes_out_rate([<table>]) : integer
15244sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015245 Returns the average server-to-client bytes rate from the currently tracked
15246 counters, measured in amount of bytes over the period configured in the
15247 table. See also src_bytes_out_rate.
15248
Cyril Bonté62ba8702014-04-22 23:52:25 +020015249sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015250sc0_clr_gpc0([<table>]) : integer
15251sc1_clr_gpc0([<table>]) : integer
15252sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020015253 Clears the first General Purpose Counter associated to the currently tracked
15254 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010015255 stored value is zero, so first invocation will always return zero. This is
15256 typically used as a second ACL in an expression in order to mark a connection
15257 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015258
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015259 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015260 # block if 5 consecutive requests continue to come faster than 10 sess
15261 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015262 acl abuse sc0_http_req_rate gt 10
15263 acl kill sc0_inc_gpc0 gt 5
15264 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015265 tcp-request connection accept if !abuse save
15266 tcp-request connection reject if abuse kill
15267
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015268sc_clr_gpc1(<ctr>[,<table>]) : integer
15269sc0_clr_gpc1([<table>]) : integer
15270sc1_clr_gpc1([<table>]) : integer
15271sc2_clr_gpc1([<table>]) : integer
15272 Clears the second General Purpose Counter associated to the currently tracked
15273 counters, and returns its previous value. Before the first invocation, the
15274 stored value is zero, so first invocation will always return zero. This is
15275 typically used as a second ACL in an expression in order to mark a connection
15276 when a first ACL was verified.
15277
Cyril Bonté62ba8702014-04-22 23:52:25 +020015278sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015279sc0_conn_cnt([<table>]) : integer
15280sc1_conn_cnt([<table>]) : integer
15281sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015282 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015283 counters. See also src_conn_cnt.
15284
Cyril Bonté62ba8702014-04-22 23:52:25 +020015285sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015286sc0_conn_cur([<table>]) : integer
15287sc1_conn_cur([<table>]) : integer
15288sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015289 Returns the current amount of concurrent connections tracking the same
15290 tracked counters. This number is automatically incremented when tracking
15291 begins and decremented when tracking stops. See also src_conn_cur.
15292
Cyril Bonté62ba8702014-04-22 23:52:25 +020015293sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015294sc0_conn_rate([<table>]) : integer
15295sc1_conn_rate([<table>]) : integer
15296sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015297 Returns the average connection rate from the currently tracked counters,
15298 measured in amount of connections over the period configured in the table.
15299 See also src_conn_rate.
15300
Cyril Bonté62ba8702014-04-22 23:52:25 +020015301sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015302sc0_get_gpc0([<table>]) : integer
15303sc1_get_gpc0([<table>]) : integer
15304sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015305 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015306 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015307
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015308sc_get_gpc1(<ctr>[,<table>]) : integer
15309sc0_get_gpc1([<table>]) : integer
15310sc1_get_gpc1([<table>]) : integer
15311sc2_get_gpc1([<table>]) : integer
15312 Returns the value of the second General Purpose Counter associated to the
15313 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
15314
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015315sc_get_gpt0(<ctr>[,<table>]) : integer
15316sc0_get_gpt0([<table>]) : integer
15317sc1_get_gpt0([<table>]) : integer
15318sc2_get_gpt0([<table>]) : integer
15319 Returns the value of the first General Purpose Tag associated to the
15320 currently tracked counters. See also src_get_gpt0.
15321
Cyril Bonté62ba8702014-04-22 23:52:25 +020015322sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015323sc0_gpc0_rate([<table>]) : integer
15324sc1_gpc0_rate([<table>]) : integer
15325sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015326 Returns the average increment rate of the first General Purpose Counter
15327 associated to the currently tracked counters. It reports the frequency
15328 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015329 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15330 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15331 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020015332
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015333sc_gpc1_rate(<ctr>[,<table>]) : integer
15334sc0_gpc1_rate([<table>]) : integer
15335sc1_gpc1_rate([<table>]) : integer
15336sc2_gpc1_rate([<table>]) : integer
15337 Returns the average increment rate of the second General Purpose Counter
15338 associated to the currently tracked counters. It reports the frequency
15339 which the gpc1 counter was incremented over the configured period. See also
15340 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15341 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15342 be returned, as "gpc1" only holds the event count.
15343
Cyril Bonté62ba8702014-04-22 23:52:25 +020015344sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015345sc0_http_err_cnt([<table>]) : integer
15346sc1_http_err_cnt([<table>]) : integer
15347sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015348 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015349 counters. This includes the both request errors and 4xx error responses.
15350 See also src_http_err_cnt.
15351
Cyril Bonté62ba8702014-04-22 23:52:25 +020015352sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015353sc0_http_err_rate([<table>]) : integer
15354sc1_http_err_rate([<table>]) : integer
15355sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015356 Returns the average rate of HTTP errors from the currently tracked counters,
15357 measured in amount of errors over the period configured in the table. This
15358 includes the both request errors and 4xx error responses. See also
15359 src_http_err_rate.
15360
Cyril Bonté62ba8702014-04-22 23:52:25 +020015361sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015362sc0_http_req_cnt([<table>]) : integer
15363sc1_http_req_cnt([<table>]) : integer
15364sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015365 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015366 counters. This includes every started request, valid or not. See also
15367 src_http_req_cnt.
15368
Cyril Bonté62ba8702014-04-22 23:52:25 +020015369sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015370sc0_http_req_rate([<table>]) : integer
15371sc1_http_req_rate([<table>]) : integer
15372sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015373 Returns the average rate of HTTP requests from the currently tracked
15374 counters, measured in amount of requests over the period configured in
15375 the table. This includes every started request, valid or not. See also
15376 src_http_req_rate.
15377
Cyril Bonté62ba8702014-04-22 23:52:25 +020015378sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015379sc0_inc_gpc0([<table>]) : integer
15380sc1_inc_gpc0([<table>]) : integer
15381sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015382 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010015383 tracked counters, and returns its new value. Before the first invocation,
15384 the stored value is zero, so first invocation will increase it to 1 and will
15385 return 1. This is typically used as a second ACL in an expression in order
15386 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020015387
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015388 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015389 acl abuse sc0_http_req_rate gt 10
15390 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015391 tcp-request connection reject if abuse kill
15392
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015393sc_inc_gpc1(<ctr>[,<table>]) : integer
15394sc0_inc_gpc1([<table>]) : integer
15395sc1_inc_gpc1([<table>]) : integer
15396sc2_inc_gpc1([<table>]) : integer
15397 Increments the second General Purpose Counter associated to the currently
15398 tracked counters, and returns its new value. Before the first invocation,
15399 the stored value is zero, so first invocation will increase it to 1 and will
15400 return 1. This is typically used as a second ACL in an expression in order
15401 to mark a connection when a first ACL was verified.
15402
Cyril Bonté62ba8702014-04-22 23:52:25 +020015403sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015404sc0_kbytes_in([<table>]) : integer
15405sc1_kbytes_in([<table>]) : integer
15406sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015407 Returns the total amount of client-to-server data from the currently tracked
15408 counters, measured in kilobytes. The test is currently performed on 32-bit
15409 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020015410
Cyril Bonté62ba8702014-04-22 23:52:25 +020015411sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015412sc0_kbytes_out([<table>]) : integer
15413sc1_kbytes_out([<table>]) : integer
15414sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015415 Returns the total amount of server-to-client data from the currently tracked
15416 counters, measured in kilobytes. The test is currently performed on 32-bit
15417 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020015418
Cyril Bonté62ba8702014-04-22 23:52:25 +020015419sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015420sc0_sess_cnt([<table>]) : integer
15421sc1_sess_cnt([<table>]) : integer
15422sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015423 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020015424 into sessions, which means that they were accepted by a "tcp-request
15425 connection" rule, from the currently tracked counters. A backend may count
15426 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015427 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020015428 with the client. See also src_sess_cnt.
15429
Cyril Bonté62ba8702014-04-22 23:52:25 +020015430sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015431sc0_sess_rate([<table>]) : integer
15432sc1_sess_rate([<table>]) : integer
15433sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015434 Returns the average session rate from the currently tracked counters,
15435 measured in amount of sessions over the period configured in the table. A
15436 session is a connection that got past the early "tcp-request connection"
15437 rules. A backend may count more sessions than connections because each
15438 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015439 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020015440
Cyril Bonté62ba8702014-04-22 23:52:25 +020015441sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020015442sc0_tracked([<table>]) : boolean
15443sc1_tracked([<table>]) : boolean
15444sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020015445 Returns true if the designated session counter is currently being tracked by
15446 the current session. This can be useful when deciding whether or not we want
15447 to set some values in a header passed to the server.
15448
Cyril Bonté62ba8702014-04-22 23:52:25 +020015449sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015450sc0_trackers([<table>]) : integer
15451sc1_trackers([<table>]) : integer
15452sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010015453 Returns the current amount of concurrent connections tracking the same
15454 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015455 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010015456 that it does not rely on any stored information but on the table's reference
15457 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015458 may sometimes be more suited for layer7 tracking. It can be used to tell a
15459 server how many concurrent connections there are from a given address for
15460 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010015461
Willy Tarreau74ca5042013-06-11 23:12:07 +020015462so_id : integer
15463 Returns an integer containing the current listening socket's id. It is useful
15464 in frontends involving many "bind" lines, or to stick all users coming via a
15465 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015466
Willy Tarreau74ca5042013-06-11 23:12:07 +020015467src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010015468 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020015469 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
15470 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
15471 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015472 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
15473 directive is used, it can be the address of a client behind another
15474 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015475 "tcp-request connection" which sees the real address. When the incoming
15476 connection passed through address translation or redirection involving
15477 connection tracking, the original destination address before the redirection
15478 will be reported. On Linux systems, the source and destination may seldom
15479 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
15480 response may reopen a timed out connection and switch what is believed to be
15481 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015482
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010015483 Example:
15484 # add an HTTP header in requests with the originating address' country
15485 http-request set-header X-Country %[src,map_ip(geoip.lst)]
15486
Willy Tarreau74ca5042013-06-11 23:12:07 +020015487src_bytes_in_rate([<table>]) : integer
15488 Returns the average bytes rate from the incoming connection's source address
15489 in the current proxy's stick-table or in the designated stick-table, measured
15490 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015491 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015492
Willy Tarreau74ca5042013-06-11 23:12:07 +020015493src_bytes_out_rate([<table>]) : integer
15494 Returns the average bytes rate to the incoming connection's source address in
15495 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015496 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015497 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015498
Willy Tarreau74ca5042013-06-11 23:12:07 +020015499src_clr_gpc0([<table>]) : integer
15500 Clears the first General Purpose Counter associated to the incoming
15501 connection's source address in the current proxy's stick-table or in the
15502 designated stick-table, and returns its previous value. If the address is not
15503 found, an entry is created and 0 is returned. This is typically used as a
15504 second ACL in an expression in order to mark a connection when a first ACL
15505 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015506
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015507 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015508 # block if 5 consecutive requests continue to come faster than 10 sess
15509 # per second, and reset the counter as soon as the traffic slows down.
15510 acl abuse src_http_req_rate gt 10
15511 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010015512 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015513 tcp-request connection accept if !abuse save
15514 tcp-request connection reject if abuse kill
15515
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015516src_clr_gpc1([<table>]) : integer
15517 Clears the second General Purpose Counter associated to the incoming
15518 connection's source address in the current proxy's stick-table or in the
15519 designated stick-table, and returns its previous value. If the address is not
15520 found, an entry is created and 0 is returned. This is typically used as a
15521 second ACL in an expression in order to mark a connection when a first ACL
15522 was verified.
15523
Willy Tarreau74ca5042013-06-11 23:12:07 +020015524src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015525 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020015526 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015527 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015528 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015529
Willy Tarreau74ca5042013-06-11 23:12:07 +020015530src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015531 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015532 current incoming connection's source address in the current proxy's
15533 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015534 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015535
Willy Tarreau74ca5042013-06-11 23:12:07 +020015536src_conn_rate([<table>]) : integer
15537 Returns the average connection rate from the incoming connection's source
15538 address in the current proxy's stick-table or in the designated stick-table,
15539 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015540 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015541
Willy Tarreau74ca5042013-06-11 23:12:07 +020015542src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015543 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015544 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015545 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015546 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015547
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015548src_get_gpc1([<table>]) : integer
15549 Returns the value of the second General Purpose Counter associated to the
15550 incoming connection's source address in the current proxy's stick-table or in
15551 the designated stick-table. If the address is not found, zero is returned.
15552 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
15553
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015554src_get_gpt0([<table>]) : integer
15555 Returns the value of the first General Purpose Tag associated to the
15556 incoming connection's source address in the current proxy's stick-table or in
15557 the designated stick-table. If the address is not found, zero is returned.
15558 See also sc/sc0/sc1/sc2_get_gpt0.
15559
Willy Tarreau74ca5042013-06-11 23:12:07 +020015560src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015561 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020015562 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015563 stick-table or in the designated stick-table. It reports the frequency
15564 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015565 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15566 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15567 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015568
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015569src_gpc1_rate([<table>]) : integer
15570 Returns the average increment rate of the second General Purpose Counter
15571 associated to the incoming connection's source address in the current proxy's
15572 stick-table or in the designated stick-table. It reports the frequency
15573 which the gpc1 counter was incremented over the configured period. See also
15574 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15575 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15576 be returned, as "gpc1" only holds the event count.
15577
Willy Tarreau74ca5042013-06-11 23:12:07 +020015578src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015579 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015580 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020015581 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015582 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020015583 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015584
Willy Tarreau74ca5042013-06-11 23:12:07 +020015585src_http_err_rate([<table>]) : integer
15586 Returns the average rate of HTTP errors from the incoming connection's source
15587 address in the current proxy's stick-table or in the designated stick-table,
15588 measured in amount of errors over the period configured in the table. This
15589 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015590 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015591
Willy Tarreau74ca5042013-06-11 23:12:07 +020015592src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015593 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015594 source address in the current proxy's stick-table or in the designated stick-
15595 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015596 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015597
Willy Tarreau74ca5042013-06-11 23:12:07 +020015598src_http_req_rate([<table>]) : integer
15599 Returns the average rate of HTTP requests from the incoming connection's
15600 source address in the current proxy's stick-table or in the designated stick-
15601 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020015602 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015603 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015604
Willy Tarreau74ca5042013-06-11 23:12:07 +020015605src_inc_gpc0([<table>]) : integer
15606 Increments the first General Purpose Counter associated to the incoming
15607 connection's source address in the current proxy's stick-table or in the
15608 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015609 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015610 This is typically used as a second ACL in an expression in order to mark a
15611 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020015612
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015613 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020015614 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010015615 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015616 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020015617
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015618src_inc_gpc1([<table>]) : integer
15619 Increments the second General Purpose Counter associated to the incoming
15620 connection's source address in the current proxy's stick-table or in the
15621 designated stick-table, and returns its new value. If the address is not
15622 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
15623 This is typically used as a second ACL in an expression in order to mark a
15624 connection when a first ACL was verified.
15625
Willy Tarreau16e01562016-08-09 16:46:18 +020015626src_is_local : boolean
15627 Returns true if the source address of the incoming connection is local to the
15628 system, or false if the address doesn't exist on the system, meaning that it
15629 comes from a remote machine. Note that UNIX addresses are considered local.
15630 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015631 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020015632 note that the check involves a few system calls, so it's better to do it only
15633 once per connection.
15634
Willy Tarreau74ca5042013-06-11 23:12:07 +020015635src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015636 Returns the total amount of data received from the incoming connection's
15637 source address in the current proxy's stick-table or in the designated
15638 stick-table, measured in kilobytes. If the address is not found, zero is
15639 returned. The test is currently performed on 32-bit integers, which limits
15640 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015641
Willy Tarreau74ca5042013-06-11 23:12:07 +020015642src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015643 Returns the total amount of data sent to the incoming connection's source
15644 address in the current proxy's stick-table or in the designated stick-table,
15645 measured in kilobytes. If the address is not found, zero is returned. The
15646 test is currently performed on 32-bit integers, which limits values to 4
15647 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015648
Willy Tarreau74ca5042013-06-11 23:12:07 +020015649src_port : integer
15650 Returns an integer value corresponding to the TCP source port of the
15651 connection on the client side, which is the port the client connected from.
15652 Usage of this function is very limited as modern protocols do not care much
15653 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010015654
Willy Tarreau74ca5042013-06-11 23:12:07 +020015655src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015656 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020015657 connection's source IPv4 address in the current proxy's stick-table or in the
15658 designated stick-table, that were transformed into sessions, which means that
15659 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015660 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015661
Willy Tarreau74ca5042013-06-11 23:12:07 +020015662src_sess_rate([<table>]) : integer
15663 Returns the average session rate from the incoming connection's source
15664 address in the current proxy's stick-table or in the designated stick-table,
15665 measured in amount of sessions over the period configured in the table. A
15666 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015667 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015668
Willy Tarreau74ca5042013-06-11 23:12:07 +020015669src_updt_conn_cnt([<table>]) : integer
15670 Creates or updates the entry associated to the incoming connection's source
15671 address in the current proxy's stick-table or in the designated stick-table.
15672 This table must be configured to store the "conn_cnt" data type, otherwise
15673 the match will be ignored. The current count is incremented by one, and the
15674 expiration timer refreshed. The updated count is returned, so this match
15675 can't return zero. This was used to reject service abusers based on their
15676 source address. Note: it is recommended to use the more complete "track-sc*"
15677 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015678
15679 Example :
15680 # This frontend limits incoming SSH connections to 3 per 10 second for
15681 # each source address, and rejects excess connections until a 10 second
15682 # silence is observed. At most 20 addresses are tracked.
15683 listen ssh
15684 bind :22
15685 mode tcp
15686 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020015687 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020015688 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015689 server local 127.0.0.1:22
15690
Willy Tarreau74ca5042013-06-11 23:12:07 +020015691srv_id : integer
15692 Returns an integer containing the server's id when processing the response.
15693 While it's almost only used with ACLs, it may be used for logging or
15694 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020015695
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200156967.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020015697----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020015698
Willy Tarreau74ca5042013-06-11 23:12:07 +020015699The layer 5 usually describes just the session layer which in haproxy is
15700closest to the session once all the connection handshakes are finished, but
15701when no content is yet made available. The fetch methods described here are
15702usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015703future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020015704
Ben Shillitof25e8e52016-12-02 14:25:37 +00001570551d.all(<prop>[,<prop>*]) : string
15706 Returns values for the properties requested as a string, where values are
15707 separated by the delimiter specified with "51degrees-property-separator".
15708 The device is identified using all the important HTTP headers from the
15709 request. The function can be passed up to five property names, and if a
15710 property name can't be found, the value "NoData" is returned.
15711
15712 Example :
15713 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
15714 # containing the three properties requested using all relevant headers from
15715 # the request.
15716 frontend http-in
15717 bind *:8081
15718 default_backend servers
15719 http-request set-header X-51D-DeviceTypeMobileTablet \
15720 %[51d.all(DeviceType,IsMobile,IsTablet)]
15721
Emeric Brun645ae792014-04-30 14:21:06 +020015722ssl_bc : boolean
15723 Returns true when the back connection was made via an SSL/TLS transport
15724 layer and is locally deciphered. This means the outgoing connection was made
15725 other a server with the "ssl" option.
15726
15727ssl_bc_alg_keysize : integer
15728 Returns the symmetric cipher key size supported in bits when the outgoing
15729 connection was made over an SSL/TLS transport layer.
15730
Olivier Houchard6b77f492018-11-22 18:18:29 +010015731ssl_bc_alpn : string
15732 This extracts the Application Layer Protocol Negotiation field from an
15733 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020015734 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010015735 server. The SSL library must have been built with support for TLS
15736 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15737 not advertised unless the "alpn" keyword on the "server" line specifies a
15738 protocol list. Also, nothing forces the server to pick a protocol from this
15739 list, any other one may be requested. The TLS ALPN extension is meant to
15740 replace the TLS NPN extension. See also "ssl_bc_npn".
15741
Emeric Brun645ae792014-04-30 14:21:06 +020015742ssl_bc_cipher : string
15743 Returns the name of the used cipher when the outgoing connection was made
15744 over an SSL/TLS transport layer.
15745
Patrick Hemmer65674662019-06-04 08:13:03 -040015746ssl_bc_client_random : binary
15747 Returns the client random of the back connection when the incoming connection
15748 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15749 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15750
Emeric Brun74f7ffa2018-02-19 16:14:12 +010015751ssl_bc_is_resumed : boolean
15752 Returns true when the back connection was made over an SSL/TLS transport
15753 layer and the newly created SSL session was resumed using a cached
15754 session or a TLS ticket.
15755
Olivier Houchard6b77f492018-11-22 18:18:29 +010015756ssl_bc_npn : string
15757 This extracts the Next Protocol Negotiation field from an outgoing connection
15758 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020015759 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010015760 built with support for TLS extensions enabled (check haproxy -vv). Note that
15761 the TLS NPN extension is not advertised unless the "npn" keyword on the
15762 "server" line specifies a protocol list. Also, nothing forces the server to
15763 pick a protocol from this list, any other one may be used. Please note that
15764 the TLS NPN extension was replaced with ALPN.
15765
Emeric Brun645ae792014-04-30 14:21:06 +020015766ssl_bc_protocol : string
15767 Returns the name of the used protocol when the outgoing connection was made
15768 over an SSL/TLS transport layer.
15769
Emeric Brunb73a9b02014-04-30 18:49:19 +020015770ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020015771 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015772 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15773 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020015774
Patrick Hemmer65674662019-06-04 08:13:03 -040015775ssl_bc_server_random : binary
15776 Returns the server random of the back connection when the incoming connection
15777 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15778 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15779
Emeric Brun645ae792014-04-30 14:21:06 +020015780ssl_bc_session_id : binary
15781 Returns the SSL ID of the back connection when the outgoing connection was
15782 made over an SSL/TLS transport layer. It is useful to log if we want to know
15783 if session was reused or not.
15784
Patrick Hemmere0275472018-04-28 19:15:51 -040015785ssl_bc_session_key : binary
15786 Returns the SSL session master key of the back connection when the outgoing
15787 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15788 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15789 BoringSSL.
15790
Emeric Brun645ae792014-04-30 14:21:06 +020015791ssl_bc_use_keysize : integer
15792 Returns the symmetric cipher key size used in bits when the outgoing
15793 connection was made over an SSL/TLS transport layer.
15794
Willy Tarreau74ca5042013-06-11 23:12:07 +020015795ssl_c_ca_err : integer
15796 When the incoming connection was made over an SSL/TLS transport layer,
15797 returns the ID of the first error detected during verification of the client
15798 certificate at depth > 0, or 0 if no error was encountered during this
15799 verification process. Please refer to your SSL library's documentation to
15800 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020015801
Willy Tarreau74ca5042013-06-11 23:12:07 +020015802ssl_c_ca_err_depth : integer
15803 When the incoming connection was made over an SSL/TLS transport layer,
15804 returns the depth in the CA chain of the first error detected during the
15805 verification of the client certificate. If no error is encountered, 0 is
15806 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015807
Emeric Brun43e79582014-10-29 19:03:26 +010015808ssl_c_der : binary
15809 Returns the DER formatted certificate presented by the client when the
15810 incoming connection was made over an SSL/TLS transport layer. When used for
15811 an ACL, the value(s) to match against can be passed in hexadecimal form.
15812
Willy Tarreau74ca5042013-06-11 23:12:07 +020015813ssl_c_err : integer
15814 When the incoming connection was made over an SSL/TLS transport layer,
15815 returns the ID of the first error detected during verification at depth 0, or
15816 0 if no error was encountered during this verification process. Please refer
15817 to your SSL library's documentation to find the exhaustive list of error
15818 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020015819
Willy Tarreau74ca5042013-06-11 23:12:07 +020015820ssl_c_i_dn([<entry>[,<occ>]]) : string
15821 When the incoming connection was made over an SSL/TLS transport layer,
15822 returns the full distinguished name of the issuer of the certificate
15823 presented by the client when no <entry> is specified, or the value of the
15824 first given entry found from the beginning of the DN. If a positive/negative
15825 occurrence number is specified as the optional second argument, it returns
15826 the value of the nth given entry value from the beginning/end of the DN.
15827 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
15828 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020015829
Willy Tarreau74ca5042013-06-11 23:12:07 +020015830ssl_c_key_alg : string
15831 Returns the name of the algorithm used to generate the key of the certificate
15832 presented by the client when the incoming connection was made over an SSL/TLS
15833 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020015834
Willy Tarreau74ca5042013-06-11 23:12:07 +020015835ssl_c_notafter : string
15836 Returns the end date presented by the client as a formatted string
15837 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15838 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020015839
Willy Tarreau74ca5042013-06-11 23:12:07 +020015840ssl_c_notbefore : string
15841 Returns the start date presented by the client as a formatted string
15842 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15843 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015844
Willy Tarreau74ca5042013-06-11 23:12:07 +020015845ssl_c_s_dn([<entry>[,<occ>]]) : string
15846 When the incoming connection was made over an SSL/TLS transport layer,
15847 returns the full distinguished name of the subject of the certificate
15848 presented by the client when no <entry> is specified, or the value of the
15849 first given entry found from the beginning of the DN. If a positive/negative
15850 occurrence number is specified as the optional second argument, it returns
15851 the value of the nth given entry value from the beginning/end of the DN.
15852 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
15853 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015854
Willy Tarreau74ca5042013-06-11 23:12:07 +020015855ssl_c_serial : binary
15856 Returns the serial of the certificate presented by the client when the
15857 incoming connection was made over an SSL/TLS transport layer. When used for
15858 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015859
Willy Tarreau74ca5042013-06-11 23:12:07 +020015860ssl_c_sha1 : binary
15861 Returns the SHA-1 fingerprint of the certificate presented by the client when
15862 the incoming connection was made over an SSL/TLS transport layer. This can be
15863 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015864 Note that the output is binary, so if you want to pass that signature to the
15865 server, you need to encode it in hex or base64, such as in the example below:
15866
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015867 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015868 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020015869
Willy Tarreau74ca5042013-06-11 23:12:07 +020015870ssl_c_sig_alg : string
15871 Returns the name of the algorithm used to sign the certificate presented by
15872 the client when the incoming connection was made over an SSL/TLS transport
15873 layer.
Emeric Brun87855892012-10-17 17:39:35 +020015874
Willy Tarreau74ca5042013-06-11 23:12:07 +020015875ssl_c_used : boolean
15876 Returns true if current SSL session uses a client certificate even if current
15877 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020015878
Willy Tarreau74ca5042013-06-11 23:12:07 +020015879ssl_c_verify : integer
15880 Returns the verify result error ID when the incoming connection was made over
15881 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
15882 refer to your SSL library's documentation for an exhaustive list of error
15883 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020015884
Willy Tarreau74ca5042013-06-11 23:12:07 +020015885ssl_c_version : integer
15886 Returns the version of the certificate presented by the client when the
15887 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020015888
Emeric Brun43e79582014-10-29 19:03:26 +010015889ssl_f_der : binary
15890 Returns the DER formatted certificate presented by the frontend when the
15891 incoming connection was made over an SSL/TLS transport layer. When used for
15892 an ACL, the value(s) to match against can be passed in hexadecimal form.
15893
Willy Tarreau74ca5042013-06-11 23:12:07 +020015894ssl_f_i_dn([<entry>[,<occ>]]) : string
15895 When the incoming connection was made over an SSL/TLS transport layer,
15896 returns the full distinguished name of the issuer of the certificate
15897 presented by the frontend when no <entry> is specified, or the value of the
15898 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020015899 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020015900 the value of the nth given entry value from the beginning/end of the DN.
15901 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
15902 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020015903
Willy Tarreau74ca5042013-06-11 23:12:07 +020015904ssl_f_key_alg : string
15905 Returns the name of the algorithm used to generate the key of the certificate
15906 presented by the frontend when the incoming connection was made over an
15907 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015908
Willy Tarreau74ca5042013-06-11 23:12:07 +020015909ssl_f_notafter : string
15910 Returns the end date presented by the frontend as a formatted string
15911 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15912 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015913
Willy Tarreau74ca5042013-06-11 23:12:07 +020015914ssl_f_notbefore : string
15915 Returns the start date presented by the frontend as a formatted string
15916 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15917 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020015918
Willy Tarreau74ca5042013-06-11 23:12:07 +020015919ssl_f_s_dn([<entry>[,<occ>]]) : string
15920 When the incoming connection was made over an SSL/TLS transport layer,
15921 returns the full distinguished name of the subject of the certificate
15922 presented by the frontend when no <entry> is specified, or the value of the
15923 first given entry found from the beginning of the DN. If a positive/negative
15924 occurrence number is specified as the optional second argument, it returns
15925 the value of the nth given entry value from the beginning/end of the DN.
15926 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
15927 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020015928
Willy Tarreau74ca5042013-06-11 23:12:07 +020015929ssl_f_serial : binary
15930 Returns the serial of the certificate presented by the frontend when the
15931 incoming connection was made over an SSL/TLS transport layer. When used for
15932 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020015933
Emeric Brun55f4fa82014-04-30 17:11:25 +020015934ssl_f_sha1 : binary
15935 Returns the SHA-1 fingerprint of the certificate presented by the frontend
15936 when the incoming connection was made over an SSL/TLS transport layer. This
15937 can be used to know which certificate was chosen using SNI.
15938
Willy Tarreau74ca5042013-06-11 23:12:07 +020015939ssl_f_sig_alg : string
15940 Returns the name of the algorithm used to sign the certificate presented by
15941 the frontend when the incoming connection was made over an SSL/TLS transport
15942 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015943
Willy Tarreau74ca5042013-06-11 23:12:07 +020015944ssl_f_version : integer
15945 Returns the version of the certificate presented by the frontend when the
15946 incoming connection was made over an SSL/TLS transport layer.
15947
15948ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015949 Returns true when the front connection was made via an SSL/TLS transport
15950 layer and is locally deciphered. This means it has matched a socket declared
15951 with a "bind" line having the "ssl" option.
15952
Willy Tarreau74ca5042013-06-11 23:12:07 +020015953 Example :
15954 # This passes "X-Proto: https" to servers when client connects over SSL
15955 listen http-https
15956 bind :80
15957 bind :443 ssl crt /etc/haproxy.pem
15958 http-request add-header X-Proto https if { ssl_fc }
15959
15960ssl_fc_alg_keysize : integer
15961 Returns the symmetric cipher key size supported in bits when the incoming
15962 connection was made over an SSL/TLS transport layer.
15963
15964ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015965 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020015966 incoming connection made via a TLS transport layer and locally deciphered by
15967 haproxy. The result is a string containing the protocol name advertised by
15968 the client. The SSL library must have been built with support for TLS
15969 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15970 not advertised unless the "alpn" keyword on the "bind" line specifies a
15971 protocol list. Also, nothing forces the client to pick a protocol from this
15972 list, any other one may be requested. The TLS ALPN extension is meant to
15973 replace the TLS NPN extension. See also "ssl_fc_npn".
15974
Willy Tarreau74ca5042013-06-11 23:12:07 +020015975ssl_fc_cipher : string
15976 Returns the name of the used cipher when the incoming connection was made
15977 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020015978
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015979ssl_fc_cipherlist_bin : binary
15980 Returns the binary form of the client hello cipher list. The maximum returned
15981 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015982 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015983
15984ssl_fc_cipherlist_hex : string
15985 Returns the binary form of the client hello cipher list encoded as
15986 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015987 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015988
15989ssl_fc_cipherlist_str : string
15990 Returns the decoded text form of the client hello cipher list. The maximum
15991 number of ciphers returned is according with the value of
15992 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010015993 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020015994 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015995
15996ssl_fc_cipherlist_xxh : integer
15997 Returns a xxh64 of the cipher list. This hash can be return only is the value
15998 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015999 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010016000
Patrick Hemmer65674662019-06-04 08:13:03 -040016001ssl_fc_client_random : binary
16002 Returns the client random of the front connection when the incoming connection
16003 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
16004 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
16005
Willy Tarreau74ca5042013-06-11 23:12:07 +020016006ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020016007 Returns true if a client certificate is present in an incoming connection over
16008 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010016009 Note: on SSL session resumption with Session ID or TLS ticket, client
16010 certificate is not present in the current connection but may be retrieved
16011 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
16012 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020016013
Olivier Houchardccaa7de2017-10-02 11:51:03 +020016014ssl_fc_has_early : boolean
16015 Returns true if early data were sent, and the handshake didn't happen yet. As
16016 it has security implications, it is useful to be able to refuse those, or
16017 wait until the handshake happened.
16018
Willy Tarreau74ca5042013-06-11 23:12:07 +020016019ssl_fc_has_sni : boolean
16020 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020016021 in an incoming connection was made over an SSL/TLS transport layer. Returns
16022 true when the incoming connection presents a TLS SNI field. This requires
John Roesler27c754c2019-07-10 15:45:51 -050016023 that the SSL library is built with support for TLS extensions enabled (check
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020016024 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020016025
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020016026ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020016027 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010016028 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
16029 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020016030
Willy Tarreau74ca5042013-06-11 23:12:07 +020016031ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016032 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020016033 made via a TLS transport layer and locally deciphered by haproxy. The result
16034 is a string containing the protocol name advertised by the client. The SSL
16035 library must have been built with support for TLS extensions enabled (check
16036 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
16037 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
16038 forces the client to pick a protocol from this list, any other one may be
16039 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020016040
Willy Tarreau74ca5042013-06-11 23:12:07 +020016041ssl_fc_protocol : string
16042 Returns the name of the used protocol when the incoming connection was made
16043 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020016044
Emeric Brunb73a9b02014-04-30 18:49:19 +020016045ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040016046 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020016047 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
16048 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040016049
Patrick Hemmer65674662019-06-04 08:13:03 -040016050ssl_fc_server_random : binary
16051 Returns the server random of the front connection when the incoming connection
16052 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
16053 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
16054
Willy Tarreau74ca5042013-06-11 23:12:07 +020016055ssl_fc_session_id : binary
16056 Returns the SSL ID of the front connection when the incoming connection was
16057 made over an SSL/TLS transport layer. It is useful to stick a given client to
16058 a server. It is important to note that some browsers refresh their session ID
16059 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020016060
Patrick Hemmere0275472018-04-28 19:15:51 -040016061ssl_fc_session_key : binary
16062 Returns the SSL session master key of the front connection when the incoming
16063 connection was made over an SSL/TLS transport layer. It is useful to decrypt
16064 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
16065 BoringSSL.
16066
16067
Willy Tarreau74ca5042013-06-11 23:12:07 +020016068ssl_fc_sni : string
16069 This extracts the Server Name Indication TLS extension (SNI) field from an
16070 incoming connection made via an SSL/TLS transport layer and locally
16071 deciphered by haproxy. The result (when present) typically is a string
16072 matching the HTTPS host name (253 chars or less). The SSL library must have
16073 been built with support for TLS extensions enabled (check haproxy -vv).
16074
16075 This fetch is different from "req_ssl_sni" above in that it applies to the
16076 connection being deciphered by haproxy and not to SSL contents being blindly
16077 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
John Roesler27c754c2019-07-10 15:45:51 -050016078 requires that the SSL library is built with support for TLS extensions
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020016079 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020016080
Willy Tarreau74ca5042013-06-11 23:12:07 +020016081 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020016082 ssl_fc_sni_end : suffix match
16083 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020016084
Willy Tarreau74ca5042013-06-11 23:12:07 +020016085ssl_fc_use_keysize : integer
16086 Returns the symmetric cipher key size used in bits when the incoming
16087 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016088
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016089
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200160907.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016091------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020016092
Willy Tarreau74ca5042013-06-11 23:12:07 +020016093Fetching samples from buffer contents is a bit different from the previous
16094sample fetches above because the sampled data are ephemeral. These data can
16095only be used when they're available and will be lost when they're forwarded.
16096For this reason, samples fetched from buffer contents during a request cannot
16097be used in a response for example. Even while the data are being fetched, they
16098can change. Sometimes it is necessary to set some delays or combine multiple
16099sample fetch methods to ensure that the expected data are complete and usable,
16100for example through TCP request content inspection. Please see the "tcp-request
16101content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020016102
Willy Tarreau74ca5042013-06-11 23:12:07 +020016103payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010016104 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016105 "stick on", "stick match"), and for "res.payload" when used in the context of
16106 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010016107
Willy Tarreau74ca5042013-06-11 23:12:07 +020016108payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
16109 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010016110 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020016111 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010016112
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020016113req.hdrs : string
16114 Returns the current request headers as string including the last empty line
16115 separating headers from the request body. The last empty line can be used to
16116 detect a truncated header block. This sample fetch is useful for some SPOE
16117 headers analyzers and for advanced logging.
16118
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020016119req.hdrs_bin : binary
16120 Returns the current request headers contained in preparsed binary form. This
16121 is useful for offloading some processing with SPOE. Each string is described
16122 by a length followed by the number of bytes indicated in the length. The
16123 length is represented using the variable integer encoding detailed in the
16124 SPOE documentation. The end of the list is marked by a couple of empty header
16125 names and values (length of 0 for both).
16126
16127 *(<str:header-name><str:header-value>)<empty string><empty string>
16128
16129 int: refer to the SPOE documentation for the encoding
16130 str: <int:length><bytes>
16131
Willy Tarreau74ca5042013-06-11 23:12:07 +020016132req.len : integer
16133req_len : integer (deprecated)
16134 Returns an integer value corresponding to the number of bytes present in the
16135 request buffer. This is mostly used in ACL. It is important to understand
16136 that this test does not return false as long as the buffer is changing. This
16137 means that a check with equality to zero will almost always immediately match
16138 at the beginning of the session, while a test for more data will wait for
16139 that data to come in and return false only when haproxy is certain that no
16140 more data will come in. This test was designed to be used with TCP request
16141 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016142
Willy Tarreau74ca5042013-06-11 23:12:07 +020016143req.payload(<offset>,<length>) : binary
16144 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020016145 in the request buffer. As a special case, if the <length> argument is zero,
16146 the the whole buffer from <offset> to the end is extracted. This can be used
16147 with ACLs in order to check for the presence of some content in a buffer at
16148 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016149
Willy Tarreau74ca5042013-06-11 23:12:07 +020016150 ACL alternatives :
16151 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016152
Willy Tarreau74ca5042013-06-11 23:12:07 +020016153req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16154 This extracts a binary block whose size is specified at <offset1> for <length>
16155 bytes, and which starts at <offset2> if specified or just after the length in
16156 the request buffer. The <offset2> parameter also supports relative offsets if
16157 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016158
Willy Tarreau74ca5042013-06-11 23:12:07 +020016159 ACL alternatives :
16160 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016161
Willy Tarreau74ca5042013-06-11 23:12:07 +020016162 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016163
Willy Tarreau74ca5042013-06-11 23:12:07 +020016164req.proto_http : boolean
16165req_proto_http : boolean (deprecated)
16166 Returns true when data in the request buffer look like HTTP and correctly
16167 parses as such. It is the same parser as the common HTTP request parser which
16168 is used so there should be no surprises. The test does not match until the
16169 request is complete, failed or timed out. This test may be used to report the
16170 protocol in TCP logs, but the biggest use is to block TCP request analysis
16171 until a complete HTTP request is present in the buffer, for example to track
16172 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016173
Willy Tarreau74ca5042013-06-11 23:12:07 +020016174 Example:
16175 # track request counts per "base" (concatenation of Host+URL)
16176 tcp-request inspect-delay 10s
16177 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020016178 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020016179
Willy Tarreau74ca5042013-06-11 23:12:07 +020016180req.rdp_cookie([<name>]) : string
16181rdp_cookie([<name>]) : string (deprecated)
16182 When the request buffer looks like the RDP protocol, extracts the RDP cookie
16183 <name>, or any cookie if unspecified. The parser only checks for the first
16184 cookie, as illustrated in the RDP protocol specification. The cookie name is
16185 case insensitive. Generally the "MSTS" cookie name will be used, as it can
16186 contain the user name of the client connecting to the server if properly
16187 configured on the client. The "MSTSHASH" cookie is often used as well for
16188 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016189
Willy Tarreau74ca5042013-06-11 23:12:07 +020016190 This differs from "balance rdp-cookie" in that any balancing algorithm may be
16191 used and thus the distribution of clients to backend servers is not linked to
16192 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
16193 such as "balance roundrobin" or "balance leastconn" will lead to a more even
16194 distribution of clients to backend servers than the hash used by "balance
16195 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016196
Willy Tarreau74ca5042013-06-11 23:12:07 +020016197 ACL derivatives :
16198 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016199
Willy Tarreau74ca5042013-06-11 23:12:07 +020016200 Example :
16201 listen tse-farm
16202 bind 0.0.0.0:3389
16203 # wait up to 5s for an RDP cookie in the request
16204 tcp-request inspect-delay 5s
16205 tcp-request content accept if RDP_COOKIE
16206 # apply RDP cookie persistence
16207 persist rdp-cookie
16208 # Persist based on the mstshash cookie
16209 # This is only useful makes sense if
16210 # balance rdp-cookie is not used
16211 stick-table type string size 204800
16212 stick on req.rdp_cookie(mstshash)
16213 server srv1 1.1.1.1:3389
16214 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016215
Willy Tarreau74ca5042013-06-11 23:12:07 +020016216 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
16217 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016218
Willy Tarreau74ca5042013-06-11 23:12:07 +020016219req.rdp_cookie_cnt([name]) : integer
16220rdp_cookie_cnt([name]) : integer (deprecated)
16221 Tries to parse the request buffer as RDP protocol, then returns an integer
16222 corresponding to the number of RDP cookies found. If an optional cookie name
16223 is passed, only cookies matching this name are considered. This is mostly
16224 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016225
Willy Tarreau74ca5042013-06-11 23:12:07 +020016226 ACL derivatives :
16227 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016228
Alex Zorin4afdd132018-12-30 13:56:28 +110016229req.ssl_alpn : string
16230 Returns a string containing the values of the Application-Layer Protocol
16231 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
16232 ClientHello message. Note that this only applies to raw contents found in the
16233 request buffer and not to the contents deciphered via an SSL data layer, so
16234 this will not work with "bind" lines having the "ssl" option. This is useful
16235 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020016236 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110016237
16238 Examples :
16239 # Wait for a client hello for at most 5 seconds
16240 tcp-request inspect-delay 5s
16241 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020016242 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110016243 default_backend bk_default
16244
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016245req.ssl_ec_ext : boolean
16246 Returns a boolean identifying if client sent the Supported Elliptic Curves
16247 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020016248 message. This can be used to present ECC compatible clients with EC
16249 certificate and to use RSA for all others, on the same IP address. Note that
16250 this only applies to raw contents found in the request buffer and not to
16251 contents deciphered via an SSL data layer, so this will not work with "bind"
16252 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016253
Willy Tarreau74ca5042013-06-11 23:12:07 +020016254req.ssl_hello_type : integer
16255req_ssl_hello_type : integer (deprecated)
16256 Returns an integer value containing the type of the SSL hello message found
16257 in the request buffer if the buffer contains data that parse as a complete
16258 SSL (v3 or superior) client hello message. Note that this only applies to raw
16259 contents found in the request buffer and not to contents deciphered via an
16260 SSL data layer, so this will not work with "bind" lines having the "ssl"
16261 option. This is mostly used in ACL to detect presence of an SSL hello message
16262 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016263
Willy Tarreau74ca5042013-06-11 23:12:07 +020016264req.ssl_sni : string
16265req_ssl_sni : string (deprecated)
16266 Returns a string containing the value of the Server Name TLS extension sent
16267 by a client in a TLS stream passing through the request buffer if the buffer
16268 contains data that parse as a complete SSL (v3 or superior) client hello
16269 message. Note that this only applies to raw contents found in the request
16270 buffer and not to contents deciphered via an SSL data layer, so this will not
16271 work with "bind" lines having the "ssl" option. SNI normally contains the
16272 name of the host the client tries to connect to (for recent browsers). SNI is
16273 useful for allowing or denying access to certain hosts when SSL/TLS is used
16274 by the client. This test was designed to be used with TCP request content
16275 inspection. If content switching is needed, it is recommended to first wait
16276 for a complete client hello (type 1), like in the example below. See also
16277 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016278
Willy Tarreau74ca5042013-06-11 23:12:07 +020016279 ACL derivatives :
16280 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016281
Willy Tarreau74ca5042013-06-11 23:12:07 +020016282 Examples :
16283 # Wait for a client hello for at most 5 seconds
16284 tcp-request inspect-delay 5s
16285 tcp-request content accept if { req_ssl_hello_type 1 }
16286 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
16287 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016288
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053016289req.ssl_st_ext : integer
16290 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
16291 Returns 1 if the client sent SessionTicket TLS Extension
16292 Returns 2 if the client also sent non-zero length TLS SessionTicket
16293 Note that this only applies to raw contents found in the request buffer and
16294 not to contents deciphered via an SSL data layer, so this will not work with
16295 "bind" lines having the "ssl" option. This can for example be used to detect
16296 whether the client sent a SessionTicket or not and stick it accordingly, if
16297 no SessionTicket then stick on SessionID or don't stick as there's no server
16298 side state is there when SessionTickets are in use.
16299
Willy Tarreau74ca5042013-06-11 23:12:07 +020016300req.ssl_ver : integer
16301req_ssl_ver : integer (deprecated)
16302 Returns an integer value containing the version of the SSL/TLS protocol of a
16303 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
16304 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
16305 composed of the major version multiplied by 65536, added to the minor
16306 version. Note that this only applies to raw contents found in the request
16307 buffer and not to contents deciphered via an SSL data layer, so this will not
16308 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010016309 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020016310 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016311
Willy Tarreau74ca5042013-06-11 23:12:07 +020016312 ACL derivatives :
16313 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016314
Willy Tarreau47e8eba2013-09-11 23:28:46 +020016315res.len : integer
16316 Returns an integer value corresponding to the number of bytes present in the
16317 response buffer. This is mostly used in ACL. It is important to understand
16318 that this test does not return false as long as the buffer is changing. This
16319 means that a check with equality to zero will almost always immediately match
16320 at the beginning of the session, while a test for more data will wait for
16321 that data to come in and return false only when haproxy is certain that no
16322 more data will come in. This test was designed to be used with TCP response
16323 content inspection.
16324
Willy Tarreau74ca5042013-06-11 23:12:07 +020016325res.payload(<offset>,<length>) : binary
16326 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020016327 in the response buffer. As a special case, if the <length> argument is zero,
16328 the the whole buffer from <offset> to the end is extracted. This can be used
16329 with ACLs in order to check for the presence of some content in a buffer at
16330 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016331
Willy Tarreau74ca5042013-06-11 23:12:07 +020016332res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16333 This extracts a binary block whose size is specified at <offset1> for <length>
16334 bytes, and which starts at <offset2> if specified or just after the length in
16335 the response buffer. The <offset2> parameter also supports relative offsets
16336 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016337
Willy Tarreau74ca5042013-06-11 23:12:07 +020016338 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016339
Willy Tarreau971f7b62015-09-29 14:06:59 +020016340res.ssl_hello_type : integer
16341rep_ssl_hello_type : integer (deprecated)
16342 Returns an integer value containing the type of the SSL hello message found
16343 in the response buffer if the buffer contains data that parses as a complete
16344 SSL (v3 or superior) hello message. Note that this only applies to raw
16345 contents found in the response buffer and not to contents deciphered via an
16346 SSL data layer, so this will not work with "server" lines having the "ssl"
16347 option. This is mostly used in ACL to detect presence of an SSL hello message
16348 that is supposed to contain an SSL session ID usable for stickiness.
16349
Willy Tarreau74ca5042013-06-11 23:12:07 +020016350wait_end : boolean
16351 This fetch either returns true when the inspection period is over, or does
16352 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016353 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020016354 actions, such as a delayed reject for some special addresses. Since it either
16355 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016356 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020016357 "WAIT_END" is always usable without prior declaration. This test was designed
16358 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016359
Willy Tarreau74ca5042013-06-11 23:12:07 +020016360 Examples :
16361 # delay every incoming request by 2 seconds
16362 tcp-request inspect-delay 2s
16363 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010016364
Willy Tarreau74ca5042013-06-11 23:12:07 +020016365 # don't immediately tell bad guys they are rejected
16366 tcp-request inspect-delay 10s
16367 acl goodguys src 10.0.0.0/24
16368 acl badguys src 10.0.1.0/24
16369 tcp-request content accept if goodguys
16370 tcp-request content reject if badguys WAIT_END
16371 tcp-request content reject
16372
16373
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200163747.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016375--------------------------------------
16376
16377It is possible to fetch samples from HTTP contents, requests and responses.
16378This application layer is also called layer 7. It is only possible to fetch the
16379data in this section when a full HTTP request or response has been parsed from
16380its respective request or response buffer. This is always the case with all
16381HTTP specific rules and for sections running with "mode http". When using TCP
16382content inspection, it may be necessary to support an inspection delay in order
16383to let the request or response come in first. These fetches may require a bit
16384more CPU resources than the layer 4 ones, but not much since the request and
16385response are indexed.
16386
16387base : string
16388 This returns the concatenation of the first Host header and the path part of
16389 the request, which starts at the first slash and ends before the question
16390 mark. It can be useful in virtual hosted environments to detect URL abuses as
16391 well as to improve shared caches efficiency. Using this with a limited size
16392 stick table also allows one to collect statistics about most commonly
16393 requested objects by host/path. With ACLs it can allow simple content
16394 switching rules involving the host and the path at the same time, such as
16395 "www.example.com/favicon.ico". See also "path" and "uri".
16396
16397 ACL derivatives :
16398 base : exact string match
16399 base_beg : prefix match
16400 base_dir : subdir match
16401 base_dom : domain match
16402 base_end : suffix match
16403 base_len : length match
16404 base_reg : regex match
16405 base_sub : substring match
16406
16407base32 : integer
16408 This returns a 32-bit hash of the value returned by the "base" fetch method
16409 above. This is useful to track per-URL activity on high traffic sites without
16410 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016411 memory. The output type is an unsigned integer. The hash function used is
16412 SDBM with full avalanche on the output. Technically, base32 is exactly equal
16413 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020016414
16415base32+src : binary
16416 This returns the concatenation of the base32 fetch above and the src fetch
16417 below. The resulting type is of type binary, with a size of 8 or 20 bytes
16418 depending on the source address family. This can be used to track per-IP,
16419 per-URL counters.
16420
William Lallemand65ad6e12014-01-31 15:08:02 +010016421capture.req.hdr(<idx>) : string
16422 This extracts the content of the header captured by the "capture request
16423 header", idx is the position of the capture keyword in the configuration.
16424 The first entry is an index of 0. See also: "capture request header".
16425
16426capture.req.method : string
16427 This extracts the METHOD of an HTTP request. It can be used in both request
16428 and response. Unlike "method", it can be used in both request and response
16429 because it's allocated.
16430
16431capture.req.uri : string
16432 This extracts the request's URI, which starts at the first slash and ends
16433 before the first space in the request (without the host part). Unlike "path"
16434 and "url", it can be used in both request and response because it's
16435 allocated.
16436
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016437capture.req.ver : string
16438 This extracts the request's HTTP version and returns either "HTTP/1.0" or
16439 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
16440 logs because it relies on a persistent flag.
16441
William Lallemand65ad6e12014-01-31 15:08:02 +010016442capture.res.hdr(<idx>) : string
16443 This extracts the content of the header captured by the "capture response
16444 header", idx is the position of the capture keyword in the configuration.
16445 The first entry is an index of 0.
16446 See also: "capture response header"
16447
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016448capture.res.ver : string
16449 This extracts the response's HTTP version and returns either "HTTP/1.0" or
16450 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
16451 persistent flag.
16452
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016453req.body : binary
16454 This returns the HTTP request's available body as a block of data. It
16455 requires that the request body has been buffered made available using
16456 "option http-buffer-request". In case of chunked-encoded body, currently only
16457 the first chunk is analyzed.
16458
Thierry FOURNIER9826c772015-05-20 15:50:54 +020016459req.body_param([<name>) : string
16460 This fetch assumes that the body of the POST request is url-encoded. The user
16461 can check if the "content-type" contains the value
16462 "application/x-www-form-urlencoded". This extracts the first occurrence of the
16463 parameter <name> in the body, which ends before '&'. The parameter name is
16464 case-sensitive. If no name is given, any parameter will match, and the first
16465 one will be returned. The result is a string corresponding to the value of the
16466 parameter <name> as presented in the request body (no URL decoding is
16467 performed). Note that the ACL version of this fetch iterates over multiple
16468 parameters and will iteratively report all parameters values if no name is
16469 given.
16470
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016471req.body_len : integer
16472 This returns the length of the HTTP request's available body in bytes. It may
16473 be lower than the advertised length if the body is larger than the buffer. It
16474 requires that the request body has been buffered made available using
16475 "option http-buffer-request".
16476
16477req.body_size : integer
16478 This returns the advertised length of the HTTP request's body in bytes. It
16479 will represent the advertised Content-Length header, or the size of the first
16480 chunk in case of chunked encoding. In order to parse the chunks, it requires
16481 that the request body has been buffered made available using
16482 "option http-buffer-request".
16483
Willy Tarreau74ca5042013-06-11 23:12:07 +020016484req.cook([<name>]) : string
16485cook([<name>]) : string (deprecated)
16486 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16487 header line from the request, and returns its value as string. If no name is
16488 specified, the first cookie value is returned. When used with ACLs, all
16489 matching cookies are evaluated. Spaces around the name and the value are
16490 ignored as requested by the Cookie header specification (RFC6265). The cookie
16491 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
16492 well return an empty value if it is present. Use the "found" match to detect
16493 presence. Use the res.cook() variant for response cookies sent by the server.
16494
16495 ACL derivatives :
16496 cook([<name>]) : exact string match
16497 cook_beg([<name>]) : prefix match
16498 cook_dir([<name>]) : subdir match
16499 cook_dom([<name>]) : domain match
16500 cook_end([<name>]) : suffix match
16501 cook_len([<name>]) : length match
16502 cook_reg([<name>]) : regex match
16503 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016504
Willy Tarreau74ca5042013-06-11 23:12:07 +020016505req.cook_cnt([<name>]) : integer
16506cook_cnt([<name>]) : integer (deprecated)
16507 Returns an integer value representing the number of occurrences of the cookie
16508 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016509
Willy Tarreau74ca5042013-06-11 23:12:07 +020016510req.cook_val([<name>]) : integer
16511cook_val([<name>]) : integer (deprecated)
16512 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16513 header line from the request, and converts its value to an integer which is
16514 returned. If no name is specified, the first cookie value is returned. When
16515 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020016516
Willy Tarreau74ca5042013-06-11 23:12:07 +020016517cookie([<name>]) : string (deprecated)
16518 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16519 header line from the request, or a "Set-Cookie" header from the response, and
16520 returns its value as a string. A typical use is to get multiple clients
16521 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020016522 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020016523 multi-peer synchronization and state keeping across restarts. If no name is
16524 specified, the first cookie value is returned. This fetch should not be used
16525 anymore and should be replaced by req.cook() or res.cook() instead as it
16526 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016527
Willy Tarreau74ca5042013-06-11 23:12:07 +020016528hdr([<name>[,<occ>]]) : string
16529 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
16530 used on responses. Please refer to these respective fetches for more details.
16531 In case of doubt about the fetch direction, please use the explicit ones.
16532 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016533 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016534
Willy Tarreau74ca5042013-06-11 23:12:07 +020016535req.fhdr(<name>[,<occ>]) : string
16536 This extracts the last occurrence of header <name> in an HTTP request. When
16537 used from an ACL, all occurrences are iterated over until a match is found.
16538 Optionally, a specific occurrence might be specified as a position number.
16539 Positive values indicate a position from the first occurrence, with 1 being
16540 the first one. Negative values indicate positions relative to the last one,
16541 with -1 being the last one. It differs from req.hdr() in that any commas
16542 present in the value are returned and are not used as delimiters. This is
16543 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016544
Willy Tarreau74ca5042013-06-11 23:12:07 +020016545req.fhdr_cnt([<name>]) : integer
16546 Returns an integer value representing the number of occurrences of request
16547 header field name <name>, or the total number of header fields if <name> is
16548 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
16549 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016550
Willy Tarreau74ca5042013-06-11 23:12:07 +020016551req.hdr([<name>[,<occ>]]) : string
16552 This extracts the last occurrence of header <name> in an HTTP request. When
16553 used from an ACL, all occurrences are iterated over until a match is found.
16554 Optionally, a specific occurrence might be specified as a position number.
16555 Positive values indicate a position from the first occurrence, with 1 being
16556 the first one. Negative values indicate positions relative to the last one,
16557 with -1 being the last one. A typical use is with the X-Forwarded-For header
16558 once converted to IP, associated with an IP stick-table. The function
16559 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000016560 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020016561 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010016562 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010016563
Willy Tarreau74ca5042013-06-11 23:12:07 +020016564 ACL derivatives :
16565 hdr([<name>[,<occ>]]) : exact string match
16566 hdr_beg([<name>[,<occ>]]) : prefix match
16567 hdr_dir([<name>[,<occ>]]) : subdir match
16568 hdr_dom([<name>[,<occ>]]) : domain match
16569 hdr_end([<name>[,<occ>]]) : suffix match
16570 hdr_len([<name>[,<occ>]]) : length match
16571 hdr_reg([<name>[,<occ>]]) : regex match
16572 hdr_sub([<name>[,<occ>]]) : substring match
16573
16574req.hdr_cnt([<name>]) : integer
16575hdr_cnt([<header>]) : integer (deprecated)
16576 Returns an integer value representing the number of occurrences of request
16577 header field name <name>, or the total number of header field values if
16578 <name> is not specified. It is important to remember that one header line may
16579 count as several headers if it has several values. The function considers any
16580 comma as a delimiter for distinct values. If full-line headers are desired
16581 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
16582 detect presence, absence or abuse of a specific header, as well as to block
16583 request smuggling attacks by rejecting requests which contain more than one
16584 of certain headers. See "req.hdr" for more information on header matching.
16585
16586req.hdr_ip([<name>[,<occ>]]) : ip
16587hdr_ip([<name>[,<occ>]]) : ip (deprecated)
16588 This extracts the last occurrence of header <name> in an HTTP request,
16589 converts it to an IPv4 or IPv6 address and returns this address. When used
16590 with ACLs, all occurrences are checked, and if <name> is omitted, every value
16591 of every header is checked. Optionally, a specific occurrence might be
16592 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010016593 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020016594 positions relative to the last one, with -1 being the last one. A typical use
16595 is with the X-Forwarded-For and X-Client-IP headers.
16596
16597req.hdr_val([<name>[,<occ>]]) : integer
16598hdr_val([<name>[,<occ>]]) : integer (deprecated)
16599 This extracts the last occurrence of header <name> in an HTTP request, and
16600 converts it to an integer value. When used with ACLs, all occurrences are
16601 checked, and if <name> is omitted, every value of every header is checked.
16602 Optionally, a specific occurrence might be specified as a position number.
16603 Positive values indicate a position from the first occurrence, with 1 being
16604 the first one. Negative values indicate positions relative to the last one,
16605 with -1 being the last one. A typical use is with the X-Forwarded-For header.
16606
Frédéric Lécailleec891192019-02-26 15:02:35 +010016607
16608
Willy Tarreau74ca5042013-06-11 23:12:07 +020016609http_auth(<userlist>) : boolean
16610 Returns a boolean indicating whether the authentication data received from
16611 the client match a username & password stored in the specified userlist. This
16612 fetch function is not really useful outside of ACLs. Currently only http
16613 basic auth is supported.
16614
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016615http_auth_group(<userlist>) : string
16616 Returns a string corresponding to the user name found in the authentication
16617 data received from the client if both the user name and password are valid
16618 according to the specified userlist. The main purpose is to use it in ACLs
16619 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016620 This fetch function is not really useful outside of ACLs. Currently only http
16621 basic auth is supported.
16622
16623 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016624 http_auth_group(<userlist>) : group ...
16625 Returns true when the user extracted from the request and whose password is
16626 valid according to the specified userlist belongs to at least one of the
16627 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016628
16629http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020016630 Returns true when the request being processed is the first one of the
16631 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020016632 from some requests when a request is not the first one, or to help grouping
16633 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020016634
Willy Tarreau74ca5042013-06-11 23:12:07 +020016635method : integer + string
16636 Returns an integer value corresponding to the method in the HTTP request. For
16637 example, "GET" equals 1 (check sources to establish the matching). Value 9
16638 means "other method" and may be converted to a string extracted from the
16639 stream. This should not be used directly as a sample, this is only meant to
16640 be used from ACLs, which transparently convert methods from patterns to these
16641 integer + string values. Some predefined ACL already check for most common
16642 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016643
Willy Tarreau74ca5042013-06-11 23:12:07 +020016644 ACL derivatives :
16645 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016646
Willy Tarreau74ca5042013-06-11 23:12:07 +020016647 Example :
16648 # only accept GET and HEAD requests
16649 acl valid_method method GET HEAD
16650 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020016651
Willy Tarreau74ca5042013-06-11 23:12:07 +020016652path : string
16653 This extracts the request's URL path, which starts at the first slash and
16654 ends before the question mark (without the host part). A typical use is with
16655 prefetch-capable caches, and with portals which need to aggregate multiple
16656 information from databases and keep them in caches. Note that with outgoing
16657 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010016658 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020016659 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016660
Willy Tarreau74ca5042013-06-11 23:12:07 +020016661 ACL derivatives :
16662 path : exact string match
16663 path_beg : prefix match
16664 path_dir : subdir match
16665 path_dom : domain match
16666 path_end : suffix match
16667 path_len : length match
16668 path_reg : regex match
16669 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016670
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016671query : string
16672 This extracts the request's query string, which starts after the first
16673 question mark. If no question mark is present, this fetch returns nothing. If
16674 a question mark is present but nothing follows, it returns an empty string.
16675 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010016676 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016677 which stops before the question mark.
16678
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016679req.hdr_names([<delim>]) : string
16680 This builds a string made from the concatenation of all header names as they
16681 appear in the request when the rule is evaluated. The default delimiter is
16682 the comma (',') but it may be overridden as an optional argument <delim>. In
16683 this case, only the first character of <delim> is considered.
16684
Willy Tarreau74ca5042013-06-11 23:12:07 +020016685req.ver : string
16686req_ver : string (deprecated)
16687 Returns the version string from the HTTP request, for example "1.1". This can
16688 be useful for logs, but is mostly there for ACL. Some predefined ACL already
16689 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016690
Willy Tarreau74ca5042013-06-11 23:12:07 +020016691 ACL derivatives :
16692 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020016693
Willy Tarreau74ca5042013-06-11 23:12:07 +020016694res.comp : boolean
16695 Returns the boolean "true" value if the response has been compressed by
16696 HAProxy, otherwise returns boolean "false". This may be used to add
16697 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016698
Willy Tarreau74ca5042013-06-11 23:12:07 +020016699res.comp_algo : string
16700 Returns a string containing the name of the algorithm used if the response
16701 was compressed by HAProxy, for example : "deflate". This may be used to add
16702 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016703
Willy Tarreau74ca5042013-06-11 23:12:07 +020016704res.cook([<name>]) : string
16705scook([<name>]) : string (deprecated)
16706 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16707 header line from the response, and returns its value as string. If no name is
16708 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016709
Willy Tarreau74ca5042013-06-11 23:12:07 +020016710 ACL derivatives :
16711 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016712
Willy Tarreau74ca5042013-06-11 23:12:07 +020016713res.cook_cnt([<name>]) : integer
16714scook_cnt([<name>]) : integer (deprecated)
16715 Returns an integer value representing the number of occurrences of the cookie
16716 <name> in the response, or all cookies if <name> is not specified. This is
16717 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016718
Willy Tarreau74ca5042013-06-11 23:12:07 +020016719res.cook_val([<name>]) : integer
16720scook_val([<name>]) : integer (deprecated)
16721 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16722 header line from the response, and converts its value to an integer which is
16723 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016724
Willy Tarreau74ca5042013-06-11 23:12:07 +020016725res.fhdr([<name>[,<occ>]]) : string
16726 This extracts the last occurrence of header <name> in an HTTP response, or of
16727 the last header if no <name> is specified. Optionally, a specific occurrence
16728 might be specified as a position number. Positive values indicate a position
16729 from the first occurrence, with 1 being the first one. Negative values
16730 indicate positions relative to the last one, with -1 being the last one. It
16731 differs from res.hdr() in that any commas present in the value are returned
16732 and are not used as delimiters. If this is not desired, the res.hdr() fetch
16733 should be used instead. This is sometimes useful with headers such as Date or
16734 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016735
Willy Tarreau74ca5042013-06-11 23:12:07 +020016736res.fhdr_cnt([<name>]) : integer
16737 Returns an integer value representing the number of occurrences of response
16738 header field name <name>, or the total number of header fields if <name> is
16739 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
16740 the number of full line headers and does not stop on commas. If this is not
16741 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016742
Willy Tarreau74ca5042013-06-11 23:12:07 +020016743res.hdr([<name>[,<occ>]]) : string
16744shdr([<name>[,<occ>]]) : string (deprecated)
16745 This extracts the last occurrence of header <name> in an HTTP response, or of
16746 the last header if no <name> is specified. Optionally, a specific occurrence
16747 might be specified as a position number. Positive values indicate a position
16748 from the first occurrence, with 1 being the first one. Negative values
16749 indicate positions relative to the last one, with -1 being the last one. This
16750 can be useful to learn some data into a stick-table. The function considers
16751 any comma as a delimiter for distinct values. If this is not desired, the
16752 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016753
Willy Tarreau74ca5042013-06-11 23:12:07 +020016754 ACL derivatives :
16755 shdr([<name>[,<occ>]]) : exact string match
16756 shdr_beg([<name>[,<occ>]]) : prefix match
16757 shdr_dir([<name>[,<occ>]]) : subdir match
16758 shdr_dom([<name>[,<occ>]]) : domain match
16759 shdr_end([<name>[,<occ>]]) : suffix match
16760 shdr_len([<name>[,<occ>]]) : length match
16761 shdr_reg([<name>[,<occ>]]) : regex match
16762 shdr_sub([<name>[,<occ>]]) : substring match
16763
16764res.hdr_cnt([<name>]) : integer
16765shdr_cnt([<name>]) : integer (deprecated)
16766 Returns an integer value representing the number of occurrences of response
16767 header field name <name>, or the total number of header fields if <name> is
16768 not specified. The function considers any comma as a delimiter for distinct
16769 values. If this is not desired, the res.fhdr_cnt() fetch should be used
16770 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016771
Willy Tarreau74ca5042013-06-11 23:12:07 +020016772res.hdr_ip([<name>[,<occ>]]) : ip
16773shdr_ip([<name>[,<occ>]]) : ip (deprecated)
16774 This extracts the last occurrence of header <name> in an HTTP response,
16775 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
16776 specific occurrence might be specified as a position number. Positive values
16777 indicate a position from the first occurrence, with 1 being the first one.
16778 Negative values indicate positions relative to the last one, with -1 being
16779 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016780
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016781res.hdr_names([<delim>]) : string
16782 This builds a string made from the concatenation of all header names as they
16783 appear in the response when the rule is evaluated. The default delimiter is
16784 the comma (',') but it may be overridden as an optional argument <delim>. In
16785 this case, only the first character of <delim> is considered.
16786
Willy Tarreau74ca5042013-06-11 23:12:07 +020016787res.hdr_val([<name>[,<occ>]]) : integer
16788shdr_val([<name>[,<occ>]]) : integer (deprecated)
16789 This extracts the last occurrence of header <name> in an HTTP response, and
16790 converts it to an integer value. Optionally, a specific occurrence might be
16791 specified as a position number. Positive values indicate a position from the
16792 first occurrence, with 1 being the first one. Negative values indicate
16793 positions relative to the last one, with -1 being the last one. This can be
16794 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016795
Willy Tarreau74ca5042013-06-11 23:12:07 +020016796res.ver : string
16797resp_ver : string (deprecated)
16798 Returns the version string from the HTTP response, for example "1.1". This
16799 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020016800
Willy Tarreau74ca5042013-06-11 23:12:07 +020016801 ACL derivatives :
16802 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016803
Willy Tarreau74ca5042013-06-11 23:12:07 +020016804set-cookie([<name>]) : string (deprecated)
16805 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16806 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020016807 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020016808 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016809
Willy Tarreau74ca5042013-06-11 23:12:07 +020016810 This fetch function is deprecated and has been superseded by the "res.cook"
16811 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016812
Willy Tarreau74ca5042013-06-11 23:12:07 +020016813status : integer
16814 Returns an integer containing the HTTP status code in the HTTP response, for
16815 example, 302. It is mostly used within ACLs and integer ranges, for example,
16816 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016817
Thierry Fournier0e00dca2016-04-07 15:47:40 +020016818unique-id : string
16819 Returns the unique-id attached to the request. The directive
16820 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
16821 fails. Note that the unique-id is usually used with HTTP requests, however this
16822 sample fetch can be used with other protocols. Obviously, if it is used with
16823 other protocols than HTTP, the unique-id-format directive must not contain
16824 HTTP parts. See: unique-id-format and unique-id-header
16825
Willy Tarreau74ca5042013-06-11 23:12:07 +020016826url : string
16827 This extracts the request's URL as presented in the request. A typical use is
16828 with prefetch-capable caches, and with portals which need to aggregate
16829 multiple information from databases and keep them in caches. With ACLs, using
16830 "path" is preferred over using "url", because clients may send a full URL as
16831 is normally done with proxies. The only real use is to match "*" which does
16832 not match in "path", and for which there is already a predefined ACL. See
16833 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016834
Willy Tarreau74ca5042013-06-11 23:12:07 +020016835 ACL derivatives :
16836 url : exact string match
16837 url_beg : prefix match
16838 url_dir : subdir match
16839 url_dom : domain match
16840 url_end : suffix match
16841 url_len : length match
16842 url_reg : regex match
16843 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016844
Willy Tarreau74ca5042013-06-11 23:12:07 +020016845url_ip : ip
16846 This extracts the IP address from the request's URL when the host part is
16847 presented as an IP address. Its use is very limited. For instance, a
16848 monitoring system might use this field as an alternative for the source IP in
16849 order to test what path a given source address would follow, or to force an
16850 entry in a table for a given source address. With ACLs it can be used to
16851 restrict access to certain systems through a proxy, for example when combined
16852 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016853
Willy Tarreau74ca5042013-06-11 23:12:07 +020016854url_port : integer
16855 This extracts the port part from the request's URL. Note that if the port is
16856 not specified in the request, port 80 is assumed. With ACLs it can be used to
16857 restrict access to certain systems through a proxy, for example when combined
16858 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016859
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016860urlp([<name>[,<delim>]]) : string
16861url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020016862 This extracts the first occurrence of the parameter <name> in the query
16863 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016864 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
16865 any parameter will match, and the first one will be returned. The result is
16866 a string corresponding to the value of the parameter <name> as presented in
16867 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020016868 stickiness based on a client ID, to extract an application cookie passed as a
16869 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016870 this fetch iterates over multiple parameters and will iteratively report all
16871 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016872
Willy Tarreau74ca5042013-06-11 23:12:07 +020016873 ACL derivatives :
16874 urlp(<name>[,<delim>]) : exact string match
16875 urlp_beg(<name>[,<delim>]) : prefix match
16876 urlp_dir(<name>[,<delim>]) : subdir match
16877 urlp_dom(<name>[,<delim>]) : domain match
16878 urlp_end(<name>[,<delim>]) : suffix match
16879 urlp_len(<name>[,<delim>]) : length match
16880 urlp_reg(<name>[,<delim>]) : regex match
16881 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016882
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016883
Willy Tarreau74ca5042013-06-11 23:12:07 +020016884 Example :
16885 # match http://example.com/foo?PHPSESSIONID=some_id
16886 stick on urlp(PHPSESSIONID)
16887 # match http://example.com/foo;JSESSIONID=some_id
16888 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016889
Jarno Huuskonen676f6222017-03-30 09:19:45 +030016890urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020016891 See "urlp" above. This one extracts the URL parameter <name> in the request
16892 and converts it to an integer value. This can be used for session stickiness
16893 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020016894
Dragan Dosen0070cd52016-06-16 12:19:49 +020016895url32 : integer
16896 This returns a 32-bit hash of the value obtained by concatenating the first
16897 Host header and the whole URL including parameters (not only the path part of
16898 the request, as in the "base32" fetch above). This is useful to track per-URL
16899 activity. A shorter hash is stored, saving a lot of memory. The output type
16900 is an unsigned integer.
16901
16902url32+src : binary
16903 This returns the concatenation of the "url32" fetch and the "src" fetch. The
16904 resulting type is of type binary, with a size of 8 or 20 bytes depending on
16905 the source address family. This can be used to track per-IP, per-URL counters.
16906
Willy Tarreau198a7442008-01-17 12:05:32 +010016907
Willy Tarreau74ca5042013-06-11 23:12:07 +0200169087.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016909---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016910
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016911Some predefined ACLs are hard-coded so that they do not have to be declared in
16912every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020016913order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010016914
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016915ACL name Equivalent to Usage
16916---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016917FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020016918HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016919HTTP_1.0 req_ver 1.0 match HTTP version 1.0
16920HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010016921HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
16922HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
16923HTTP_URL_SLASH url_beg / match URL beginning with "/"
16924HTTP_URL_STAR url * match URL equal to "*"
16925LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016926METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016927METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016928METH_GET method GET HEAD match HTTP GET or HEAD method
16929METH_HEAD method HEAD match HTTP HEAD method
16930METH_OPTIONS method OPTIONS match HTTP OPTIONS method
16931METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016932METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016933METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020016934RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016935REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010016936TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016937WAIT_END wait_end wait for end of content analysis
16938---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016939
Willy Tarreaub937b7e2010-01-12 15:27:54 +010016940
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169418. Logging
16942----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010016943
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016944One of HAProxy's strong points certainly lies is its precise logs. It probably
16945provides the finest level of information available for such a product, which is
16946very important for troubleshooting complex environments. Standard information
16947provided in logs include client ports, TCP/HTTP state timers, precise session
16948state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016949to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016950headers.
16951
16952In order to improve administrators reactivity, it offers a great transparency
16953about encountered problems, both internal and external, and it is possible to
16954send logs to different sources at the same time with different level filters :
16955
16956 - global process-level logs (system errors, start/stop, etc..)
16957 - per-instance system and internal errors (lack of resource, bugs, ...)
16958 - per-instance external troubles (servers up/down, max connections)
16959 - per-instance activity (client connections), either at the establishment or
16960 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010016961 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060016962 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016963
16964The ability to distribute different levels of logs to different log servers
16965allow several production teams to interact and to fix their problems as soon
16966as possible. For example, the system team might monitor system-wide errors,
16967while the application team might be monitoring the up/down for their servers in
16968real time, and the security team might analyze the activity logs with one hour
16969delay.
16970
16971
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169728.1. Log levels
16973---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016974
Simon Hormandf791f52011-05-29 15:01:10 +090016975TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016976source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090016977HTTP request, HTTP return code, number of bytes transmitted, conditions
16978in which the session ended, and even exchanged cookies values. For example
16979track a particular user's problems. All messages may be sent to up to two
16980syslog servers. Check the "log" keyword in section 4.2 for more information
16981about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016982
16983
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169848.2. Log formats
16985----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016986
William Lallemand48940402012-01-30 16:47:22 +010016987HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090016988and will be detailed in the following sections. A few of them may vary
16989slightly with the configuration, due to indicators specific to certain
16990options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016991
16992 - the default format, which is very basic and very rarely used. It only
16993 provides very basic information about the incoming connection at the moment
16994 it is accepted : source IP:port, destination IP:port, and frontend-name.
16995 This mode will eventually disappear so it will not be described to great
16996 extents.
16997
16998 - the TCP format, which is more advanced. This format is enabled when "option
16999 tcplog" is set on the frontend. HAProxy will then usually wait for the
17000 connection to terminate before logging. This format provides much richer
17001 information, such as timers, connection counts, queue size, etc... This
17002 format is recommended for pure TCP proxies.
17003
17004 - the HTTP format, which is the most advanced for HTTP proxying. This format
17005 is enabled when "option httplog" is set on the frontend. It provides the
17006 same information as the TCP format with some HTTP-specific fields such as
17007 the request, the status code, and captures of headers and cookies. This
17008 format is recommended for HTTP proxies.
17009
Emeric Brun3a058f32009-06-30 18:26:00 +020017010 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
17011 fields arranged in the same order as the CLF format. In this mode, all
17012 timers, captures, flags, etc... appear one per field after the end of the
17013 common fields, in the same order they appear in the standard HTTP format.
17014
William Lallemand48940402012-01-30 16:47:22 +010017015 - the custom log format, allows you to make your own log line.
17016
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017017Next sections will go deeper into details for each of these formats. Format
17018specification will be performed on a "field" basis. Unless stated otherwise, a
17019field is a portion of text delimited by any number of spaces. Since syslog
17020servers are susceptible of inserting fields at the beginning of a line, it is
17021always assumed that the first field is the one containing the process name and
17022identifier.
17023
17024Note : Since log lines may be quite long, the log examples in sections below
17025 might be broken into multiple lines. The example log lines will be
17026 prefixed with 3 closing angle brackets ('>>>') and each time a log is
17027 broken into multiple lines, each non-final line will end with a
17028 backslash ('\') and the next line will start indented by two characters.
17029
17030
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170318.2.1. Default log format
17032-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017033
17034This format is used when no specific option is set. The log is emitted as soon
17035as the connection is accepted. One should note that this currently is the only
17036format which logs the request's destination IP and ports.
17037
17038 Example :
17039 listen www
17040 mode http
17041 log global
17042 server srv1 127.0.0.1:8000
17043
17044 >>> Feb 6 12:12:09 localhost \
17045 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
17046 (www/HTTP)
17047
17048 Field Format Extract from the example above
17049 1 process_name '[' pid ']:' haproxy[14385]:
17050 2 'Connect from' Connect from
17051 3 source_ip ':' source_port 10.0.1.2:33312
17052 4 'to' to
17053 5 destination_ip ':' destination_port 10.0.3.31:8012
17054 6 '(' frontend_name '/' mode ')' (www/HTTP)
17055
17056Detailed fields description :
17057 - "source_ip" is the IP address of the client which initiated the connection.
17058 - "source_port" is the TCP port of the client which initiated the connection.
17059 - "destination_ip" is the IP address the client connected to.
17060 - "destination_port" is the TCP port the client connected to.
17061 - "frontend_name" is the name of the frontend (or listener) which received
17062 and processed the connection.
17063 - "mode is the mode the frontend is operating (TCP or HTTP).
17064
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017065In case of a UNIX socket, the source and destination addresses are marked as
17066"unix:" and the ports reflect the internal ID of the socket which accepted the
17067connection (the same ID as reported in the stats).
17068
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017069It is advised not to use this deprecated format for newer installations as it
17070will eventually disappear.
17071
17072
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170738.2.2. TCP log format
17074---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017075
17076The TCP format is used when "option tcplog" is specified in the frontend, and
17077is the recommended format for pure TCP proxies. It provides a lot of precious
17078information for troubleshooting. Since this format includes timers and byte
17079counts, the log is normally emitted at the end of the session. It can be
17080emitted earlier if "option logasap" is specified, which makes sense in most
17081environments with long sessions such as remote terminals. Sessions which match
17082the "monitor" rules are never logged. It is also possible not to emit logs for
17083sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017084specifying "option dontlognull" in the frontend. Successful connections will
17085not be logged if "option dontlog-normal" is specified in the frontend. A few
17086fields may slightly vary depending on some configuration options, those are
17087marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017088
17089 Example :
17090 frontend fnt
17091 mode tcp
17092 option tcplog
17093 log global
17094 default_backend bck
17095
17096 backend bck
17097 server srv1 127.0.0.1:8000
17098
17099 >>> Feb 6 12:12:56 localhost \
17100 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
17101 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
17102
17103 Field Format Extract from the example above
17104 1 process_name '[' pid ']:' haproxy[14387]:
17105 2 client_ip ':' client_port 10.0.1.2:33313
17106 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
17107 4 frontend_name fnt
17108 5 backend_name '/' server_name bck/srv1
17109 6 Tw '/' Tc '/' Tt* 0/0/5007
17110 7 bytes_read* 212
17111 8 termination_state --
17112 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
17113 10 srv_queue '/' backend_queue 0/0
17114
17115Detailed fields description :
17116 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017117 connection to haproxy. If the connection was accepted on a UNIX socket
17118 instead, the IP address would be replaced with the word "unix". Note that
17119 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017120 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010017121 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017122 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017123
17124 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017125 If the connection was accepted on a UNIX socket instead, the port would be
17126 replaced with the ID of the accepting socket, which is also reported in the
17127 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017128
17129 - "accept_date" is the exact date when the connection was received by haproxy
17130 (which might be very slightly different from the date observed on the
17131 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020017132 the same date which may appear in any upstream firewall's log. When used in
17133 HTTP mode, the accept_date field will be reset to the first moment the
17134 connection is ready to receive a new request (end of previous response for
17135 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017136
17137 - "frontend_name" is the name of the frontend (or listener) which received
17138 and processed the connection.
17139
17140 - "backend_name" is the name of the backend (or listener) which was selected
17141 to manage the connection to the server. This will be the same as the
17142 frontend if no switching rule has been applied, which is common for TCP
17143 applications.
17144
17145 - "server_name" is the name of the last server to which the connection was
17146 sent, which might differ from the first one if there were connection errors
17147 and a redispatch occurred. Note that this server belongs to the backend
17148 which processed the request. If the connection was aborted before reaching
17149 a server, "<NOSRV>" is indicated instead of a server name.
17150
17151 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17152 It can be "-1" if the connection was aborted before reaching the queue.
17153 See "Timers" below for more details.
17154
17155 - "Tc" is the total time in milliseconds spent waiting for the connection to
17156 establish to the final server, including retries. It can be "-1" if the
17157 connection was aborted before a connection could be established. See
17158 "Timers" below for more details.
17159
17160 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017161 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017162 "option logasap" was specified, then the time counting stops at the moment
17163 the log is emitted. In this case, a '+' sign is prepended before the value,
17164 indicating that the final one will be larger. See "Timers" below for more
17165 details.
17166
17167 - "bytes_read" is the total number of bytes transmitted from the server to
17168 the client when the log is emitted. If "option logasap" is specified, the
17169 this value will be prefixed with a '+' sign indicating that the final one
17170 may be larger. Please note that this value is a 64-bit counter, so log
17171 analysis tools must be able to handle it without overflowing.
17172
17173 - "termination_state" is the condition the session was in when the session
17174 ended. This indicates the session state, which side caused the end of
17175 session to happen, and for what reason (timeout, error, ...). The normal
17176 flags should be "--", indicating the session was closed by either end with
17177 no data remaining in buffers. See below "Session state at disconnection"
17178 for more details.
17179
17180 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017181 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017182 limits have been reached. For instance, if actconn is close to 512 when
17183 multiple connection errors occur, chances are high that the system limits
17184 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017185 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017186
17187 - "feconn" is the total number of concurrent connections on the frontend when
17188 the session was logged. It is useful to estimate the amount of resource
17189 required to sustain high loads, and to detect when the frontend's "maxconn"
17190 has been reached. Most often when this value increases by huge jumps, it is
17191 because there is congestion on the backend servers, but sometimes it can be
17192 caused by a denial of service attack.
17193
17194 - "beconn" is the total number of concurrent connections handled by the
17195 backend when the session was logged. It includes the total number of
17196 concurrent connections active on servers as well as the number of
17197 connections pending in queues. It is useful to estimate the amount of
17198 additional servers needed to support high loads for a given application.
17199 Most often when this value increases by huge jumps, it is because there is
17200 congestion on the backend servers, but sometimes it can be caused by a
17201 denial of service attack.
17202
17203 - "srv_conn" is the total number of concurrent connections still active on
17204 the server when the session was logged. It can never exceed the server's
17205 configured "maxconn" parameter. If this value is very often close or equal
17206 to the server's "maxconn", it means that traffic regulation is involved a
17207 lot, meaning that either the server's maxconn value is too low, or that
17208 there aren't enough servers to process the load with an optimal response
17209 time. When only one of the server's "srv_conn" is high, it usually means
17210 that this server has some trouble causing the connections to take longer to
17211 be processed than on other servers.
17212
17213 - "retries" is the number of connection retries experienced by this session
17214 when trying to connect to the server. It must normally be zero, unless a
17215 server is being stopped at the same moment the connection was attempted.
17216 Frequent retries generally indicate either a network problem between
17217 haproxy and the server, or a misconfigured system backlog on the server
17218 preventing new connections from being queued. This field may optionally be
17219 prefixed with a '+' sign, indicating that the session has experienced a
17220 redispatch after the maximal retry count has been reached on the initial
17221 server. In this case, the server name appearing in the log is the one the
17222 connection was redispatched to, and not the first one, though both may
17223 sometimes be the same in case of hashing for instance. So as a general rule
17224 of thumb, when a '+' is present in front of the retry count, this count
17225 should not be attributed to the logged server.
17226
17227 - "srv_queue" is the total number of requests which were processed before
17228 this one in the server queue. It is zero when the request has not gone
17229 through the server queue. It makes it possible to estimate the approximate
17230 server's response time by dividing the time spent in queue by the number of
17231 requests in the queue. It is worth noting that if a session experiences a
17232 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017233 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017234 backend queue unless a redispatch occurs.
17235
17236 - "backend_queue" is the total number of requests which were processed before
17237 this one in the backend's global queue. It is zero when the request has not
17238 gone through the global queue. It makes it possible to estimate the average
17239 queue length, which easily translates into a number of missing servers when
17240 divided by a server's "maxconn" parameter. It is worth noting that if a
17241 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017242 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017243 through both the server queue and the backend queue unless a redispatch
17244 occurs.
17245
17246
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200172478.2.3. HTTP log format
17248----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017249
17250The HTTP format is the most complete and the best suited for HTTP proxies. It
17251is enabled by when "option httplog" is specified in the frontend. It provides
17252the same level of information as the TCP format with additional features which
17253are specific to the HTTP protocol. Just like the TCP format, the log is usually
17254emitted at the end of the session, unless "option logasap" is specified, which
17255generally only makes sense for download sites. A session which matches the
17256"monitor" rules will never logged. It is also possible not to log sessions for
17257which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017258frontend. Successful connections will not be logged if "option dontlog-normal"
17259is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017260
17261Most fields are shared with the TCP log, some being different. A few fields may
17262slightly vary depending on some configuration options. Those ones are marked
17263with a star ('*') after the field name below.
17264
17265 Example :
17266 frontend http-in
17267 mode http
17268 option httplog
17269 log global
17270 default_backend bck
17271
17272 backend static
17273 server srv1 127.0.0.1:8000
17274
17275 >>> Feb 6 12:14:14 localhost \
17276 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
17277 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 +010017278 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017279
17280 Field Format Extract from the example above
17281 1 process_name '[' pid ']:' haproxy[14389]:
17282 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017283 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017284 4 frontend_name http-in
17285 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017286 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017287 7 status_code 200
17288 8 bytes_read* 2750
17289 9 captured_request_cookie -
17290 10 captured_response_cookie -
17291 11 termination_state ----
17292 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
17293 13 srv_queue '/' backend_queue 0/0
17294 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
17295 15 '{' captured_response_headers* '}' {}
17296 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010017297
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017298Detailed fields description :
17299 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017300 connection to haproxy. If the connection was accepted on a UNIX socket
17301 instead, the IP address would be replaced with the word "unix". Note that
17302 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017303 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010017304 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017305 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017306
17307 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017308 If the connection was accepted on a UNIX socket instead, the port would be
17309 replaced with the ID of the accepting socket, which is also reported in the
17310 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017311
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017312 - "request_date" is the exact date when the first byte of the HTTP request
17313 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017314
17315 - "frontend_name" is the name of the frontend (or listener) which received
17316 and processed the connection.
17317
17318 - "backend_name" is the name of the backend (or listener) which was selected
17319 to manage the connection to the server. This will be the same as the
17320 frontend if no switching rule has been applied.
17321
17322 - "server_name" is the name of the last server to which the connection was
17323 sent, which might differ from the first one if there were connection errors
17324 and a redispatch occurred. Note that this server belongs to the backend
17325 which processed the request. If the request was aborted before reaching a
17326 server, "<NOSRV>" is indicated instead of a server name. If the request was
17327 intercepted by the stats subsystem, "<STATS>" is indicated instead.
17328
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017329 - "TR" is the total time in milliseconds spent waiting for a full HTTP
17330 request from the client (not counting body) after the first byte was
17331 received. It can be "-1" if the connection was aborted before a complete
John Roesler27c754c2019-07-10 15:45:51 -050017332 request could be received or a bad request was received. It should
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017333 always be very small because a request generally fits in one single packet.
17334 Large times here generally indicate network issues between the client and
Willy Tarreau590a0512018-09-05 11:56:48 +020017335 haproxy or requests being typed by hand. See section 8.4 "Timing Events"
17336 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017337
17338 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17339 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020017340 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017341
17342 - "Tc" is the total time in milliseconds spent waiting for the connection to
17343 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020017344 request was aborted before a connection could be established. See section
17345 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017346
17347 - "Tr" is the total time in milliseconds spent waiting for the server to send
17348 a full HTTP response, not counting data. It can be "-1" if the request was
17349 aborted before a complete response could be received. It generally matches
17350 the server's processing time for the request, though it may be altered by
17351 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020017352 "GET" requests generally indicate an overloaded server. See section 8.4
17353 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017354
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017355 - "Ta" is the time the request remained active in haproxy, which is the total
17356 time in milliseconds elapsed between the first byte of the request was
17357 received and the last byte of response was sent. It covers all possible
17358 processing except the handshake (see Th) and idle time (see Ti). There is
17359 one exception, if "option logasap" was specified, then the time counting
17360 stops at the moment the log is emitted. In this case, a '+' sign is
17361 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020017362 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017363
17364 - "status_code" is the HTTP status code returned to the client. This status
17365 is generally set by the server, but it might also be set by haproxy when
17366 the server cannot be reached or when its response is blocked by haproxy.
17367
17368 - "bytes_read" is the total number of bytes transmitted to the client when
17369 the log is emitted. This does include HTTP headers. If "option logasap" is
John Roesler27c754c2019-07-10 15:45:51 -050017370 specified, this value will be prefixed with a '+' sign indicating that
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017371 the final one may be larger. Please note that this value is a 64-bit
17372 counter, so log analysis tools must be able to handle it without
17373 overflowing.
17374
17375 - "captured_request_cookie" is an optional "name=value" entry indicating that
17376 the client had this cookie in the request. The cookie name and its maximum
17377 length are defined by the "capture cookie" statement in the frontend
17378 configuration. The field is a single dash ('-') when the option is not
17379 set. Only one cookie may be captured, it is generally used to track session
17380 ID exchanges between a client and a server to detect session crossing
17381 between clients due to application bugs. For more details, please consult
17382 the section "Capturing HTTP headers and cookies" below.
17383
17384 - "captured_response_cookie" is an optional "name=value" entry indicating
17385 that the server has returned a cookie with its response. The cookie name
17386 and its maximum length are defined by the "capture cookie" statement in the
17387 frontend configuration. The field is a single dash ('-') when the option is
17388 not set. Only one cookie may be captured, it is generally used to track
17389 session ID exchanges between a client and a server to detect session
17390 crossing between clients due to application bugs. For more details, please
17391 consult the section "Capturing HTTP headers and cookies" below.
17392
17393 - "termination_state" is the condition the session was in when the session
17394 ended. This indicates the session state, which side caused the end of
17395 session to happen, for what reason (timeout, error, ...), just like in TCP
17396 logs, and information about persistence operations on cookies in the last
17397 two characters. The normal flags should begin with "--", indicating the
17398 session was closed by either end with no data remaining in buffers. See
17399 below "Session state at disconnection" for more details.
17400
17401 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017402 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017403 limits have been reached. For instance, if actconn is close to 512 or 1024
17404 when multiple connection errors occur, chances are high that the system
17405 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017406 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017407 system.
17408
17409 - "feconn" is the total number of concurrent connections on the frontend when
17410 the session was logged. It is useful to estimate the amount of resource
17411 required to sustain high loads, and to detect when the frontend's "maxconn"
17412 has been reached. Most often when this value increases by huge jumps, it is
17413 because there is congestion on the backend servers, but sometimes it can be
17414 caused by a denial of service attack.
17415
17416 - "beconn" is the total number of concurrent connections handled by the
17417 backend when the session was logged. It includes the total number of
17418 concurrent connections active on servers as well as the number of
17419 connections pending in queues. It is useful to estimate the amount of
17420 additional servers needed to support high loads for a given application.
17421 Most often when this value increases by huge jumps, it is because there is
17422 congestion on the backend servers, but sometimes it can be caused by a
17423 denial of service attack.
17424
17425 - "srv_conn" is the total number of concurrent connections still active on
17426 the server when the session was logged. It can never exceed the server's
17427 configured "maxconn" parameter. If this value is very often close or equal
17428 to the server's "maxconn", it means that traffic regulation is involved a
17429 lot, meaning that either the server's maxconn value is too low, or that
17430 there aren't enough servers to process the load with an optimal response
17431 time. When only one of the server's "srv_conn" is high, it usually means
17432 that this server has some trouble causing the requests to take longer to be
17433 processed than on other servers.
17434
17435 - "retries" is the number of connection retries experienced by this session
17436 when trying to connect to the server. It must normally be zero, unless a
17437 server is being stopped at the same moment the connection was attempted.
17438 Frequent retries generally indicate either a network problem between
17439 haproxy and the server, or a misconfigured system backlog on the server
17440 preventing new connections from being queued. This field may optionally be
17441 prefixed with a '+' sign, indicating that the session has experienced a
17442 redispatch after the maximal retry count has been reached on the initial
17443 server. In this case, the server name appearing in the log is the one the
17444 connection was redispatched to, and not the first one, though both may
17445 sometimes be the same in case of hashing for instance. So as a general rule
17446 of thumb, when a '+' is present in front of the retry count, this count
17447 should not be attributed to the logged server.
17448
17449 - "srv_queue" is the total number of requests which were processed before
17450 this one in the server queue. It is zero when the request has not gone
17451 through the server queue. It makes it possible to estimate the approximate
17452 server's response time by dividing the time spent in queue by the number of
17453 requests in the queue. It is worth noting that if a session experiences a
17454 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017455 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017456 backend queue unless a redispatch occurs.
17457
17458 - "backend_queue" is the total number of requests which were processed before
17459 this one in the backend's global queue. It is zero when the request has not
17460 gone through the global queue. It makes it possible to estimate the average
17461 queue length, which easily translates into a number of missing servers when
17462 divided by a server's "maxconn" parameter. It is worth noting that if a
17463 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017464 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017465 through both the server queue and the backend queue unless a redispatch
17466 occurs.
17467
17468 - "captured_request_headers" is a list of headers captured in the request due
17469 to the presence of the "capture request header" statement in the frontend.
17470 Multiple headers can be captured, they will be delimited by a vertical bar
17471 ('|'). When no capture is enabled, the braces do not appear, causing a
17472 shift of remaining fields. It is important to note that this field may
17473 contain spaces, and that using it requires a smarter log parser than when
17474 it's not used. Please consult the section "Capturing HTTP headers and
17475 cookies" below for more details.
17476
17477 - "captured_response_headers" is a list of headers captured in the response
17478 due to the presence of the "capture response header" statement in the
17479 frontend. Multiple headers can be captured, they will be delimited by a
17480 vertical bar ('|'). When no capture is enabled, the braces do not appear,
17481 causing a shift of remaining fields. It is important to note that this
17482 field may contain spaces, and that using it requires a smarter log parser
17483 than when it's not used. Please consult the section "Capturing HTTP headers
17484 and cookies" below for more details.
17485
17486 - "http_request" is the complete HTTP request line, including the method,
17487 request and HTTP version string. Non-printable characters are encoded (see
17488 below the section "Non-printable characters"). This is always the last
17489 field, and it is always delimited by quotes and is the only one which can
17490 contain quotes. If new fields are added to the log format, they will be
17491 added before this field. This field might be truncated if the request is
17492 huge and does not fit in the standard syslog buffer (1024 characters). This
17493 is the reason why this field must always remain the last one.
17494
17495
Cyril Bontédc4d9032012-04-08 21:57:39 +0200174968.2.4. Custom log format
17497------------------------
William Lallemand48940402012-01-30 16:47:22 +010017498
Willy Tarreau2beef582012-12-20 17:22:52 +010017499The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010017500mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010017501
Davor Ocelice9ed2812017-12-25 17:49:28 +010017502HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010017503Variables can take arguments using braces ('{}'), and multiple arguments are
17504separated by commas within the braces. Flags may be added or removed by
17505prefixing them with a '+' or '-' sign.
17506
17507Special variable "%o" may be used to propagate its flags to all other
17508variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010017509("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010017510
Willy Tarreauc8368452012-12-21 00:09:23 +010017511If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020017512as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010017513less common information such as the client's SSL certificate's DN, or to log
17514the key that would be used to store an entry into a stick table.
17515
William Lallemand48940402012-01-30 16:47:22 +010017516Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017517In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010017518in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010017519
Dragan Dosen835b9212016-02-12 13:23:03 +010017520Note: when using the RFC5424 syslog message format, the characters '"',
17521'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
17522https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
17523such cases, the use of the flag "E" should be considered.
17524
William Lallemand48940402012-01-30 16:47:22 +010017525Flags are :
17526 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017527 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010017528 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
17529 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010017530
17531 Example:
17532
17533 log-format %T\ %t\ Some\ Text
17534 log-format %{+Q}o\ %t\ %s\ %{-Q}r
17535
Dragan Dosen835b9212016-02-12 13:23:03 +010017536 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
17537
William Lallemand48940402012-01-30 16:47:22 +010017538At the moment, the default HTTP format is defined this way :
17539
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017540 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
17541 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010017542
William Lallemandbddd4fd2012-02-27 11:23:10 +010017543the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010017544
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017545 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
17546 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
17547 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010017548
William Lallemandbddd4fd2012-02-27 11:23:10 +010017549and the default TCP format is defined this way :
17550
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017551 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
17552 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010017553
William Lallemand48940402012-01-30 16:47:22 +010017554Please refer to the table below for currently defined variables :
17555
William Lallemandbddd4fd2012-02-27 11:23:10 +010017556 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017557 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017558 +---+------+-----------------------------------------------+-------------+
17559 | | %o | special variable, apply flags on all next var | |
17560 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010017561 | | %B | bytes_read (from server to client) | numeric |
17562 | H | %CC | captured_request_cookie | string |
17563 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020017564 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017565 | H | %HM | HTTP method (ex: POST) | string |
17566 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000017567 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017568 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
17569 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010017570 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020017571 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017572 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017573 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017574 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020017575 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080017576 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017577 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
17578 | H | %Ti | idle time before the HTTP request | numeric |
17579 | H | %Tq | Th + Ti + TR | numeric |
17580 | H | %TR | time to receive the full request from 1st byte| numeric |
17581 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017582 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017583 | | %Tt | Tt | numeric |
17584 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017585 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017586 | | %ac | actconn | numeric |
17587 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017588 | | %bc | beconn (backend concurrent connections) | numeric |
17589 | | %bi | backend_source_ip (connecting address) | IP |
17590 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017591 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017592 | | %ci | client_ip (accepted address) | IP |
17593 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017594 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017595 | | %fc | feconn (frontend concurrent connections) | numeric |
17596 | | %fi | frontend_ip (accepting address) | IP |
17597 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020017598 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020017599 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017600 | | %hr | captured_request_headers default style | string |
17601 | | %hrl | captured_request_headers CLF style | string list |
17602 | | %hs | captured_response_headers default style | string |
17603 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020017604 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017605 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017606 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017607 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010017608 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017609 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017610 | | %sc | srv_conn (server concurrent connections) | numeric |
17611 | | %si | server_IP (target address) | IP |
17612 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017613 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017614 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
17615 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017616 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017617 | H | %tr | date_time of HTTP request | date |
17618 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020017619 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017620 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017621 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017622 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010017623
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017624 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010017625
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017626
176278.2.5. Error log format
17628-----------------------
17629
17630When an incoming connection fails due to an SSL handshake or an invalid PROXY
17631protocol header, haproxy will log the event using a shorter, fixed line format.
17632By default, logs are emitted at the LOG_INFO level, unless the option
17633"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017634will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017635logged if the "dontlognull" option is set.
17636
17637The format looks like this :
17638
17639 >>> Dec 3 18:27:14 localhost \
17640 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
17641 Connection error during SSL handshake
17642
17643 Field Format Extract from the example above
17644 1 process_name '[' pid ']:' haproxy[6103]:
17645 2 client_ip ':' client_port 127.0.0.1:56059
17646 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
17647 4 frontend_name "/" bind_name ":" frt/f1:
17648 5 message Connection error during SSL handshake
17649
17650These fields just provide minimal information to help debugging connection
17651failures.
17652
17653
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200176548.3. Advanced logging options
17655-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017656
17657Some advanced logging options are often looked for but are not easy to find out
17658just by looking at the various options. Here is an entry point for the few
17659options which can enable better logging. Please refer to the keywords reference
17660for more information about their usage.
17661
17662
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200176638.3.1. Disabling logging of external tests
17664------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017665
17666It is quite common to have some monitoring tools perform health checks on
17667haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
17668commercial load-balancer, and sometimes it will simply be a more complete
17669monitoring system such as Nagios. When the tests are very frequent, users often
17670ask how to disable logging for those checks. There are three possibilities :
17671
17672 - if connections come from everywhere and are just TCP probes, it is often
17673 desired to simply disable logging of connections without data exchange, by
17674 setting "option dontlognull" in the frontend. It also disables logging of
17675 port scans, which may or may not be desired.
17676
17677 - if the connection come from a known source network, use "monitor-net" to
17678 declare this network as monitoring only. Any host in this network will then
17679 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017680 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017681 such as other load-balancers.
17682
17683 - if the tests are performed on a known URI, use "monitor-uri" to declare
17684 this URI as dedicated to monitoring. Any host sending this request will
17685 only get the result of a health-check, and the request will not be logged.
17686
17687
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200176888.3.2. Logging before waiting for the session to terminate
17689----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017690
17691The problem with logging at end of connection is that you have no clue about
17692what is happening during very long sessions, such as remote terminal sessions
17693or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010017694"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017695just before data transfer begins. This means that in case of TCP, it will still
17696log the connection status to the server, and in case of HTTP, it will log just
17697after processing the server headers. In this case, the number of bytes reported
17698is the number of header bytes sent to the client. In order to avoid confusion
17699with normal logs, the total time field and the number of bytes are prefixed
17700with a '+' sign which means that real numbers are certainly larger.
17701
17702
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177038.3.3. Raising log level upon errors
17704------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017705
17706Sometimes it is more convenient to separate normal traffic from errors logs,
17707for instance in order to ease error monitoring from log files. When the option
17708"log-separate-errors" is used, connections which experience errors, timeouts,
17709retries, redispatches or HTTP status codes 5xx will see their syslog level
17710raised from "info" to "err". This will help a syslog daemon store the log in
17711a separate file. It is very important to keep the errors in the normal traffic
17712file too, so that log ordering is not altered. You should also be careful if
17713you already have configured your syslog daemon to store all logs higher than
17714"notice" in an "admin" file, because the "err" level is higher than "notice".
17715
17716
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177178.3.4. Disabling logging of successful connections
17718--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017719
17720Although this may sound strange at first, some large sites have to deal with
17721multiple thousands of logs per second and are experiencing difficulties keeping
17722them intact for a long time or detecting errors within them. If the option
17723"dontlog-normal" is set on the frontend, all normal connections will not be
17724logged. In this regard, a normal connection is defined as one without any
17725error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
17726and a response with a status 5xx is not considered normal and will be logged
17727too. Of course, doing is is really discouraged as it will remove most of the
17728useful information from the logs. Do this only if you have no other
17729alternative.
17730
17731
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200177328.4. Timing events
17733------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017734
17735Timers provide a great help in troubleshooting network problems. All values are
17736reported in milliseconds (ms). These timers should be used in conjunction with
17737the session termination flags. In TCP mode with "option tcplog" set on the
17738frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017739mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
17740addition, three other measures are provided, "Th", "Ti", and "Tq".
17741
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010017742Timings events in HTTP mode:
17743
17744 first request 2nd request
17745 |<-------------------------------->|<-------------- ...
17746 t tr t tr ...
17747 ---|----|----|----|----|----|----|----|----|--
17748 : Th Ti TR Tw Tc Tr Td : Ti ...
17749 :<---- Tq ---->: :
17750 :<-------------- Tt -------------->:
17751 :<--------- Ta --------->:
17752
17753Timings events in TCP mode:
17754
17755 TCP session
17756 |<----------------->|
17757 t t
17758 ---|----|----|----|----|---
17759 | Th Tw Tc Td |
17760 |<------ Tt ------->|
17761
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017762 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017763 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017764 only happen once during the whole connection's lifetime. A large time here
17765 may indicate that the client only pre-established the connection without
17766 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010017767 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020017768 SSL handshake was very expensive to compute. Please note that this time is
17769 reported only before the first request, so it is safe to average it over
17770 all request to calculate the amortized value. The second and subsequent
17771 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017772
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017773 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
17774 counts between the end of the handshakes and the first byte of the HTTP
17775 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020017776 to count after the end of the transmission the previous response. When a
17777 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
17778 after the previous request. Some browsers pre-establish connections to a
17779 server in order to reduce the latency of a future request, and keep them
17780 pending until they need it. This delay will be reported as the idle time. A
17781 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017782
17783 - TR: total time to get the client request (HTTP mode only). It's the time
17784 elapsed between the first bytes received and the moment the proxy received
17785 the empty line marking the end of the HTTP headers. The value "-1"
17786 indicates that the end of headers has never been seen. This happens when
17787 the client closes prematurely or times out. This time is usually very short
17788 since most requests fit in a single packet. A large time may indicate a
17789 request typed by hand during a test.
17790
17791 - Tq: total time to get the client request from the accept date or since the
17792 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010017793 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 +020017794 returns -1 as well. This timer used to be very useful before the arrival of
17795 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
17796 it in favor of TR nowadays, as the idle time adds a lot of noise to the
17797 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017798
17799 - Tw: total time spent in the queues waiting for a connection slot. It
17800 accounts for backend queue as well as the server queues, and depends on the
17801 queue size, and the time needed for the server to complete previous
17802 requests. The value "-1" means that the request was killed before reaching
17803 the queue, which is generally what happens with invalid or denied requests.
17804
17805 - Tc: total time to establish the TCP connection to the server. It's the time
17806 elapsed between the moment the proxy sent the connection request, and the
17807 moment it was acknowledged by the server, or between the TCP SYN packet and
17808 the matching SYN/ACK packet in return. The value "-1" means that the
17809 connection never established.
17810
17811 - Tr: server response time (HTTP mode only). It's the time elapsed between
17812 the moment the TCP connection was established to the server and the moment
17813 the server sent its complete response headers. It purely shows its request
17814 processing time, without the network overhead due to the data transmission.
17815 It is worth noting that when the client has data to send to the server, for
17816 instance during a POST request, the time already runs, and this can distort
17817 apparent response time. For this reason, it's generally wise not to trust
17818 too much this field for POST requests initiated from clients behind an
17819 untrusted network. A value of "-1" here means that the last the response
17820 header (empty line) was never seen, most likely because the server timeout
17821 stroke before the server managed to process the request.
17822
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017823 - Ta: total active time for the HTTP request, between the moment the proxy
17824 received the first byte of the request header and the emission of the last
17825 byte of the response body. The exception is when the "logasap" option is
17826 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
17827 a '+' sign. From this field, we can deduce "Td", the data transmission time,
17828 by subtracting other timers when valid :
17829
17830 Td = Ta - (TR + Tw + Tc + Tr)
17831
17832 Timers with "-1" values have to be excluded from this equation. Note that
17833 "Ta" can never be negative.
17834
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017835 - Tt: total session duration time, between the moment the proxy accepted it
17836 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017837 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
17838 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017839 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017840
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017841 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017842
17843 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017844 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
17845 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017846
17847These timers provide precious indications on trouble causes. Since the TCP
17848protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
17849that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017850due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
17851"Tt" is close to a timeout value specified in the configuration, it often means
17852that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017853
17854Most common cases :
17855
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017856 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
17857 the client and the proxy. This is very rare on local networks but might
17858 happen when clients are on far remote networks and send large requests. It
17859 may happen that values larger than usual appear here without any network
17860 cause. Sometimes, during an attack or just after a resource starvation has
17861 ended, haproxy may accept thousands of connections in a few milliseconds.
17862 The time spent accepting these connections will inevitably slightly delay
17863 processing of other connections, and it can happen that request times in the
17864 order of a few tens of milliseconds are measured after a few thousands of
17865 new connections have been accepted at once. Using one of the keep-alive
17866 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020017867 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017868
17869 - If "Tc" is close to 3000, a packet has probably been lost between the
17870 server and the proxy during the server connection phase. This value should
17871 always be very low, such as 1 ms on local networks and less than a few tens
17872 of ms on remote networks.
17873
Willy Tarreau55165fe2009-05-10 12:02:55 +020017874 - If "Tr" is nearly always lower than 3000 except some rare values which seem
17875 to be the average majored by 3000, there are probably some packets lost
17876 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017877
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017878 - If "Ta" is large even for small byte counts, it generally is because
17879 neither the client nor the server decides to close the connection while
17880 haproxy is running in tunnel mode and both have agreed on a keep-alive
17881 connection mode. In order to solve this issue, it will be needed to specify
17882 one of the HTTP options to manipulate keep-alive or close options on either
17883 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
17884 important when connection regulation is used with the "maxconn" option on
17885 the servers, since no new connection will be sent to the server until
17886 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017887
17888Other noticeable HTTP log cases ('xx' means any value to be ignored) :
17889
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017890 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017891 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017892 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017893
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017894 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017895 or it aborted too early. Check the session termination flags
17896 then "timeout http-request" and "timeout client" settings.
17897
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017898 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017899 servers were out of order, because the request was invalid
17900 or forbidden by ACL rules. Check the session termination
17901 flags.
17902
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017903 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
17904 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017905 Check the session termination flags, then check the
17906 "timeout connect" setting. Note that the tarpit action might
17907 return similar-looking patterns, with "Tw" equal to the time
17908 the client connection was maintained open.
17909
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017910 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017911 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017912 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017913 termination flags, then check the "timeout server" setting.
17914
17915
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179168.5. Session state at disconnection
17917-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017918
17919TCP and HTTP logs provide a session termination indicator in the
17920"termination_state" field, just before the number of active connections. It is
179212-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
17922each of which has a special meaning :
17923
17924 - On the first character, a code reporting the first event which caused the
17925 session to terminate :
17926
17927 C : the TCP session was unexpectedly aborted by the client.
17928
17929 S : the TCP session was unexpectedly aborted by the server, or the
17930 server explicitly refused it.
17931
17932 P : the session was prematurely aborted by the proxy, because of a
17933 connection limit enforcement, because a DENY filter was matched,
17934 because of a security check which detected and blocked a dangerous
17935 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010017936 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020017937
17938 L : the session was locally processed by haproxy and was not passed to
17939 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017940
17941 R : a resource on the proxy has been exhausted (memory, sockets, source
17942 ports, ...). Usually, this appears during the connection phase, and
17943 system logs should contain a copy of the precise error. If this
17944 happens, it must be considered as a very serious anomaly which
17945 should be fixed as soon as possible by any means.
17946
17947 I : an internal error was identified by the proxy during a self-check.
17948 This should NEVER happen, and you are encouraged to report any log
17949 containing this, because this would almost certainly be a bug. It
17950 would be wise to preventively restart the process after such an
17951 event too, in case it would be caused by memory corruption.
17952
Simon Horman752dc4a2011-06-21 14:34:59 +090017953 D : the session was killed by haproxy because the server was detected
17954 as down and was configured to kill all connections when going down.
17955
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070017956 U : the session was killed by haproxy on this backup server because an
17957 active server was detected as up and was configured to kill all
17958 backup connections when going up.
17959
Willy Tarreaua2a64e92011-09-07 23:01:56 +020017960 K : the session was actively killed by an admin operating on haproxy.
17961
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017962 c : the client-side timeout expired while waiting for the client to
17963 send or receive data.
17964
17965 s : the server-side timeout expired while waiting for the server to
17966 send or receive data.
17967
17968 - : normal session completion, both the client and the server closed
17969 with nothing left in the buffers.
17970
17971 - on the second character, the TCP or HTTP session state when it was closed :
17972
Willy Tarreauf7b30a92010-12-06 22:59:17 +010017973 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017974 (HTTP mode only). Nothing was sent to any server.
17975
17976 Q : the proxy was waiting in the QUEUE for a connection slot. This can
17977 only happen when servers have a 'maxconn' parameter set. It can
17978 also happen in the global queue after a redispatch consecutive to
17979 a failed attempt to connect to a dying server. If no redispatch is
17980 reported, then no connection attempt was made to any server.
17981
17982 C : the proxy was waiting for the CONNECTION to establish on the
17983 server. The server might at most have noticed a connection attempt.
17984
17985 H : the proxy was waiting for complete, valid response HEADERS from the
17986 server (HTTP only).
17987
17988 D : the session was in the DATA phase.
17989
17990 L : the proxy was still transmitting LAST data to the client while the
17991 server had already finished. This one is very rare as it can only
17992 happen when the client dies while receiving the last packets.
17993
17994 T : the request was tarpitted. It has been held open with the client
17995 during the whole "timeout tarpit" duration or until the client
17996 closed, both of which will be reported in the "Tw" timer.
17997
17998 - : normal session completion after end of data transfer.
17999
18000 - the third character tells whether the persistence cookie was provided by
18001 the client (only in HTTP mode) :
18002
18003 N : the client provided NO cookie. This is usually the case for new
18004 visitors, so counting the number of occurrences of this flag in the
18005 logs generally indicate a valid trend for the site frequentation.
18006
18007 I : the client provided an INVALID cookie matching no known server.
18008 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020018009 cookies between HTTP/HTTPS sites, persistence conditionally
18010 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018011
18012 D : the client provided a cookie designating a server which was DOWN,
18013 so either "option persist" was used and the client was sent to
18014 this server, or it was not set and the client was redispatched to
18015 another server.
18016
Willy Tarreau996a92c2010-10-13 19:30:47 +020018017 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018018 server.
18019
Willy Tarreau996a92c2010-10-13 19:30:47 +020018020 E : the client provided a valid cookie, but with a last date which was
18021 older than what is allowed by the "maxidle" cookie parameter, so
18022 the cookie is consider EXPIRED and is ignored. The request will be
18023 redispatched just as if there was no cookie.
18024
18025 O : the client provided a valid cookie, but with a first date which was
18026 older than what is allowed by the "maxlife" cookie parameter, so
18027 the cookie is consider too OLD and is ignored. The request will be
18028 redispatched just as if there was no cookie.
18029
Willy Tarreauc89ccb62012-04-05 21:18:22 +020018030 U : a cookie was present but was not used to select the server because
18031 some other server selection mechanism was used instead (typically a
18032 "use-server" rule).
18033
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018034 - : does not apply (no cookie set in configuration).
18035
18036 - the last character reports what operations were performed on the persistence
18037 cookie returned by the server (only in HTTP mode) :
18038
18039 N : NO cookie was provided by the server, and none was inserted either.
18040
18041 I : no cookie was provided by the server, and the proxy INSERTED one.
18042 Note that in "cookie insert" mode, if the server provides a cookie,
18043 it will still be overwritten and reported as "I" here.
18044
Willy Tarreau996a92c2010-10-13 19:30:47 +020018045 U : the proxy UPDATED the last date in the cookie that was presented by
18046 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030018047 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020018048 date indicated in the cookie. If any other change happens, such as
18049 a redispatch, then the cookie will be marked as inserted instead.
18050
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018051 P : a cookie was PROVIDED by the server and transmitted as-is.
18052
18053 R : the cookie provided by the server was REWRITTEN by the proxy, which
18054 happens in "cookie rewrite" or "cookie prefix" modes.
18055
18056 D : the cookie provided by the server was DELETED by the proxy.
18057
18058 - : does not apply (no cookie set in configuration).
18059
Willy Tarreau996a92c2010-10-13 19:30:47 +020018060The combination of the two first flags gives a lot of information about what
18061was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018062helpful to detect server saturation, network troubles, local system resource
18063starvation, attacks, etc...
18064
18065The most common termination flags combinations are indicated below. They are
18066alphabetically sorted, with the lowercase set just after the upper case for
18067easier finding and understanding.
18068
18069 Flags Reason
18070
18071 -- Normal termination.
18072
18073 CC The client aborted before the connection could be established to the
18074 server. This can happen when haproxy tries to connect to a recently
18075 dead (or unchecked) server, and the client aborts while haproxy is
18076 waiting for the server to respond or for "timeout connect" to expire.
18077
18078 CD The client unexpectedly aborted during data transfer. This can be
18079 caused by a browser crash, by an intermediate equipment between the
18080 client and haproxy which decided to actively break the connection,
18081 by network routing issues between the client and haproxy, or by a
18082 keep-alive session between the server and the client terminated first
18083 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010018084
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018085 cD The client did not send nor acknowledge any data for as long as the
18086 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020018087 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018088
18089 CH The client aborted while waiting for the server to start responding.
18090 It might be the server taking too long to respond or the client
18091 clicking the 'Stop' button too fast.
18092
18093 cH The "timeout client" stroke while waiting for client data during a
18094 POST request. This is sometimes caused by too large TCP MSS values
18095 for PPPoE networks which cannot transport full-sized packets. It can
18096 also happen when client timeout is smaller than server timeout and
18097 the server takes too long to respond.
18098
18099 CQ The client aborted while its session was queued, waiting for a server
18100 with enough empty slots to accept it. It might be that either all the
18101 servers were saturated or that the assigned server was taking too
18102 long a time to respond.
18103
18104 CR The client aborted before sending a full HTTP request. Most likely
18105 the request was typed by hand using a telnet client, and aborted
18106 too early. The HTTP status code is likely a 400 here. Sometimes this
18107 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020018108 and the client. "option http-ignore-probes" can be used to ignore
18109 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018110
18111 cR The "timeout http-request" stroke before the client sent a full HTTP
18112 request. This is sometimes caused by too large TCP MSS values on the
18113 client side for PPPoE networks which cannot transport full-sized
18114 packets, or by clients sending requests by hand and not typing fast
18115 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020018116 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020018117 some browsers started to implement a "pre-connect" feature consisting
18118 in speculatively connecting to some recently visited web sites just
18119 in case the user would like to visit them. This results in many
18120 connections being established to web sites, which end up in 408
18121 Request Timeout if the timeout strikes first, or 400 Bad Request when
18122 the browser decides to close them first. These ones pollute the log
18123 and feed the error counters. Some versions of some browsers have even
18124 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010018125 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020018126 http-ignore-probes" in the frontend, resulting in connections with
18127 zero data transfer to be totally ignored. This will definitely hide
18128 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018129
18130 CT The client aborted while its session was tarpitted. It is important to
18131 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020018132 wrong tarpit rules have been written. If a lot of them happen, it
18133 might make sense to lower the "timeout tarpit" value to something
18134 closer to the average reported "Tw" timer, in order not to consume
18135 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018136
Willy Tarreau570f2212013-06-10 16:42:09 +020018137 LR The request was intercepted and locally handled by haproxy. Generally
18138 it means that this was a redirect or a stats request.
18139
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018140 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018141 the TCP connection (the proxy received a TCP RST or an ICMP message
18142 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010018143 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018144 or no ARP response on local network). When this happens in HTTP mode,
18145 the status code is likely a 502 or 503 here.
18146
18147 sC The "timeout connect" stroke before a connection to the server could
18148 complete. When this happens in HTTP mode, the status code is likely a
18149 503 or 504 here.
18150
18151 SD The connection to the server died with an error during the data
18152 transfer. This usually means that haproxy has received an RST from
18153 the server or an ICMP message from an intermediate equipment while
18154 exchanging data with the server. This can be caused by a server crash
18155 or by a network issue on an intermediate equipment.
18156
18157 sD The server did not send nor acknowledge any data for as long as the
18158 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010018159 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018160 load-balancers, ...), as well as keep-alive sessions maintained
18161 between the client and the server expiring first on haproxy.
18162
18163 SH The server aborted before sending its full HTTP response headers, or
18164 it crashed while processing the request. Since a server aborting at
18165 this moment is very rare, it would be wise to inspect its logs to
18166 control whether it crashed and why. The logged request may indicate a
18167 small set of faulty requests, demonstrating bugs in the application.
18168 Sometimes this might also be caused by an IDS killing the connection
18169 between haproxy and the server.
18170
18171 sH The "timeout server" stroke before the server could return its
18172 response headers. This is the most common anomaly, indicating too
18173 long transactions, probably caused by server or database saturation.
18174 The immediate workaround consists in increasing the "timeout server"
18175 setting, but it is important to keep in mind that the user experience
18176 will suffer from these long response times. The only long term
18177 solution is to fix the application.
18178
18179 sQ The session spent too much time in queue and has been expired. See
18180 the "timeout queue" and "timeout connect" settings to find out how to
18181 fix this if it happens too often. If it often happens massively in
18182 short periods, it may indicate general problems on the affected
18183 servers due to I/O or database congestion, or saturation caused by
18184 external attacks.
18185
18186 PC The proxy refused to establish a connection to the server because the
18187 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020018188 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018189 so that it does not happen anymore. This status is very rare and
18190 might happen when the global "ulimit-n" parameter is forced by hand.
18191
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018192 PD The proxy blocked an incorrectly formatted chunked encoded message in
18193 a request or a response, after the server has emitted its headers. In
18194 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010018195 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020018196 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018197
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018198 PH The proxy blocked the server's response, because it was invalid,
18199 incomplete, dangerous (cache control), or matched a security filter.
18200 In any case, an HTTP 502 error is sent to the client. One possible
18201 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010018202 containing unauthorized characters. It is also possible but quite
18203 rare, that the proxy blocked a chunked-encoding request from the
18204 client due to an invalid syntax, before the server responded. In this
18205 case, an HTTP 400 error is sent to the client and reported in the
18206 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018207
18208 PR The proxy blocked the client's HTTP request, either because of an
18209 invalid HTTP syntax, in which case it returned an HTTP 400 error to
18210 the client, or because a deny filter matched, in which case it
18211 returned an HTTP 403 error.
18212
18213 PT The proxy blocked the client's request and has tarpitted its
18214 connection before returning it a 500 server error. Nothing was sent
18215 to the server. The connection was maintained open for as long as
18216 reported by the "Tw" timer field.
18217
18218 RC A local resource has been exhausted (memory, sockets, source ports)
18219 preventing the connection to the server from establishing. The error
18220 logs will tell precisely what was missing. This is very rare and can
18221 only be solved by proper system tuning.
18222
Willy Tarreau996a92c2010-10-13 19:30:47 +020018223The combination of the two last flags gives a lot of information about how
18224persistence was handled by the client, the server and by haproxy. This is very
18225important to troubleshoot disconnections, when users complain they have to
18226re-authenticate. The commonly encountered flags are :
18227
18228 -- Persistence cookie is not enabled.
18229
18230 NN No cookie was provided by the client, none was inserted in the
18231 response. For instance, this can be in insert mode with "postonly"
18232 set on a GET request.
18233
18234 II A cookie designating an invalid server was provided by the client,
18235 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040018236 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020018237 value can be presented by a client when no other server knows it.
18238
18239 NI No cookie was provided by the client, one was inserted in the
18240 response. This typically happens for first requests from every user
18241 in "insert" mode, which makes it an easy way to count real users.
18242
18243 VN A cookie was provided by the client, none was inserted in the
18244 response. This happens for most responses for which the client has
18245 already got a cookie.
18246
18247 VU A cookie was provided by the client, with a last visit date which is
18248 not completely up-to-date, so an updated cookie was provided in
18249 response. This can also happen if there was no date at all, or if
18250 there was a date but the "maxidle" parameter was not set, so that the
18251 cookie can be switched to unlimited time.
18252
18253 EI A cookie was provided by the client, with a last visit date which is
18254 too old for the "maxidle" parameter, so the cookie was ignored and a
18255 new cookie was inserted in the response.
18256
18257 OI A cookie was provided by the client, with a first visit date which is
18258 too old for the "maxlife" parameter, so the cookie was ignored and a
18259 new cookie was inserted in the response.
18260
18261 DI The server designated by the cookie was down, a new server was
18262 selected and a new cookie was emitted in the response.
18263
18264 VI The server designated by the cookie was not marked dead but could not
18265 be reached. A redispatch happened and selected another one, which was
18266 then advertised in the response.
18267
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018268
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200182698.6. Non-printable characters
18270-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018271
18272In order not to cause trouble to log analysis tools or terminals during log
18273consulting, non-printable characters are not sent as-is into log files, but are
18274converted to the two-digits hexadecimal representation of their ASCII code,
18275prefixed by the character '#'. The only characters that can be logged without
18276being escaped are comprised between 32 and 126 (inclusive). Obviously, the
18277escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
18278is the same for the character '"' which becomes "#22", as well as '{', '|' and
18279'}' when logging headers.
18280
18281Note that the space character (' ') is not encoded in headers, which can cause
18282issues for tools relying on space count to locate fields. A typical header
18283containing spaces is "User-Agent".
18284
18285Last, it has been observed that some syslog daemons such as syslog-ng escape
18286the quote ('"') with a backslash ('\'). The reverse operation can safely be
18287performed since no quote may appear anywhere else in the logs.
18288
18289
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200182908.7. Capturing HTTP cookies
18291---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018292
18293Cookie capture simplifies the tracking a complete user session. This can be
18294achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018295section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018296cookie will simultaneously be checked in the request ("Cookie:" header) and in
18297the response ("Set-Cookie:" header). The respective values will be reported in
18298the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018299locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018300not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
18301user switches to a new session for example, because the server will reassign it
18302a new cookie. It is also possible to detect if a server unexpectedly sets a
18303wrong cookie to a client, leading to session crossing.
18304
18305 Examples :
18306 # capture the first cookie whose name starts with "ASPSESSION"
18307 capture cookie ASPSESSION len 32
18308
18309 # capture the first cookie whose name is exactly "vgnvisitor"
18310 capture cookie vgnvisitor= len 32
18311
18312
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200183138.8. Capturing HTTP headers
18314---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018315
18316Header captures are useful to track unique request identifiers set by an upper
18317proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
18318the response, one can search for information about the response length, how the
18319server asked the cache to behave, or an object location during a redirection.
18320
18321Header captures are performed using the "capture request header" and "capture
18322response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018323section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018324
18325It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018326time. Non-existent headers are logged as empty strings, and if one header
18327appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018328are grouped within braces '{' and '}' in the same order as they were declared,
18329and delimited with a vertical bar '|' without any space. Response headers
18330follow the same representation, but are displayed after a space following the
18331request headers block. These blocks are displayed just before the HTTP request
18332in the logs.
18333
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020018334As a special case, it is possible to specify an HTTP header capture in a TCP
18335frontend. The purpose is to enable logging of headers which will be parsed in
18336an HTTP backend if the request is then switched to this HTTP backend.
18337
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018338 Example :
18339 # This instance chains to the outgoing proxy
18340 listen proxy-out
18341 mode http
18342 option httplog
18343 option logasap
18344 log global
18345 server cache1 192.168.1.1:3128
18346
18347 # log the name of the virtual server
18348 capture request header Host len 20
18349
18350 # log the amount of data uploaded during a POST
18351 capture request header Content-Length len 10
18352
18353 # log the beginning of the referrer
18354 capture request header Referer len 20
18355
18356 # server name (useful for outgoing proxies only)
18357 capture response header Server len 20
18358
18359 # logging the content-length is useful with "option logasap"
18360 capture response header Content-Length len 10
18361
Davor Ocelice9ed2812017-12-25 17:49:28 +010018362 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018363 capture response header Cache-Control len 8
18364
18365 # the Via header will report the next proxy's name
18366 capture response header Via len 20
18367
18368 # log the URL location during a redirection
18369 capture response header Location len 20
18370
18371 >>> Aug 9 20:26:09 localhost \
18372 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
18373 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
18374 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
18375 "GET http://fr.adserver.yahoo.com/"
18376
18377 >>> Aug 9 20:30:46 localhost \
18378 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
18379 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
18380 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018381 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018382
18383 >>> Aug 9 20:30:46 localhost \
18384 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
18385 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
18386 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
18387 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018388 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018389
18390
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200183918.9. Examples of logs
18392---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018393
18394These are real-world examples of logs accompanied with an explanation. Some of
18395them have been made up by hand. The syslog part has been removed for better
18396reading. Their sole purpose is to explain how to decipher them.
18397
18398 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
18399 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
18400 "HEAD / HTTP/1.0"
18401
18402 => long request (6.5s) entered by hand through 'telnet'. The server replied
18403 in 147 ms, and the session ended normally ('----')
18404
18405 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
18406 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
18407 0/9 "HEAD / HTTP/1.0"
18408
18409 => Idem, but the request was queued in the global queue behind 9 other
18410 requests, and waited there for 1230 ms.
18411
18412 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
18413 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
18414 "GET /image.iso HTTP/1.0"
18415
18416 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018417 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018418 14 ms, 243 bytes of headers were sent to the client, and total time from
18419 accept to first data byte is 30 ms.
18420
18421 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
18422 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
18423 "GET /cgi-bin/bug.cgi? HTTP/1.0"
18424
18425 => the proxy blocked a server response either because of an "rspdeny" or
18426 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020018427 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018428 risked being cached. In this case, the response is replaced with a "502
18429 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
18430 to return the 502 and not the server.
18431
18432 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018433 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 +010018434
18435 => the client never completed its request and aborted itself ("C---") after
18436 8.5s, while the proxy was waiting for the request headers ("-R--").
18437 Nothing was sent to any server.
18438
18439 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
18440 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
18441
18442 => The client never completed its request, which was aborted by the
18443 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010018444 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018445 send a 408 return code to the client.
18446
18447 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
18448 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
18449
18450 => This log was produced with "option tcplog". The client timed out after
18451 5 seconds ("c----").
18452
18453 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
18454 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018455 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018456
18457 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018458 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018459 (config says 'retries 3'), and no redispatch (otherwise we would have
18460 seen "/+3"). Status code 503 was returned to the client. There were 115
18461 connections on this server, 202 connections on this proxy, and 205 on
18462 the global process. It is possible that the server refused the
18463 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010018464
Willy Tarreau52b2d222011-09-07 23:48:48 +020018465
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200184669. Supported filters
18467--------------------
18468
18469Here are listed officially supported filters with the list of parameters they
18470accept. Depending on compile options, some of these filters might be
18471unavailable. The list of available filters is reported in haproxy -vv.
18472
18473See also : "filter"
18474
184759.1. Trace
18476----------
18477
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018478filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018479
18480 Arguments:
18481 <name> is an arbitrary name that will be reported in
18482 messages. If no name is provided, "TRACE" is used.
18483
18484 <random-parsing> enables the random parsing of data exchanged between
18485 the client and the server. By default, this filter
18486 parses all available data. With this parameter, it
18487 only parses a random amount of the available data.
18488
Davor Ocelice9ed2812017-12-25 17:49:28 +010018489 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018490 default, this filter forwards all previously parsed
18491 data. With this parameter, it only forwards a random
18492 amount of the parsed data.
18493
Davor Ocelice9ed2812017-12-25 17:49:28 +010018494 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018495
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018496This filter can be used as a base to develop new filters. It defines all
18497callbacks and print a message on the standard error stream (stderr) with useful
18498information for all of them. It may be useful to debug the activity of other
18499filters or, quite simply, HAProxy's activity.
18500
18501Using <random-parsing> and/or <random-forwarding> parameters is a good way to
18502tests the behavior of a filter that parses data exchanged between a client and
18503a server by adding some latencies in the processing.
18504
18505
185069.2. HTTP compression
18507---------------------
18508
18509filter compression
18510
18511The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
18512keyword must still be used to enable and configure the HTTP compression. And
Christopher Faulet27d93c32018-12-15 22:32:02 +010018513when no other filter is used, it is enough. When used with the cache enabled,
18514it is also enough. In this case, the compression is always done after the
18515response is stored in the cache. But it is mandatory to explicitly use a filter
18516line to enable the HTTP compression when at least one filter other than the
18517cache is used for the same listener/frontend/backend. This is important to know
18518the filters evaluation order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018519
Christopher Faulet27d93c32018-12-15 22:32:02 +010018520See also : "compression" and section 9.4 about the cache filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018521
18522
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200185239.3. Stream Processing Offload Engine (SPOE)
18524--------------------------------------------
18525
18526filter spoe [engine <name>] config <file>
18527
18528 Arguments :
18529
18530 <name> is the engine name that will be used to find the right scope in
18531 the configuration file. If not provided, all the file will be
18532 parsed.
18533
18534 <file> is the path of the engine configuration file. This file can
18535 contain configuration of several engines. In this case, each
18536 part must be placed in its own scope.
18537
18538The Stream Processing Offload Engine (SPOE) is a filter communicating with
18539external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010018540streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018541exchanged with them are configured in dedicated files, for the main part. It
18542also requires dedicated backends, defined in HAProxy configuration.
18543
18544SPOE communicates with external components using an in-house binary protocol,
18545the Stream Processing Offload Protocol (SPOP).
18546
Tim Düsterhus4896c442016-11-29 02:15:19 +010018547For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018548"doc/SPOE.txt".
18549
18550Important note:
18551 The SPOE filter is highly experimental for now and was not heavily
18552 tested. It is really not production ready. So use it carefully.
18553
Christopher Faulet99a17a22018-12-11 09:18:27 +0100185549.4. Cache
18555----------
18556
18557filter cache <name>
18558
18559 Arguments :
18560
18561 <name> is name of the cache section this filter will use.
18562
18563The cache uses a filter to store cacheable responses. The HTTP rules
18564"cache-store" and "cache-use" must be used to define how and when to use a
John Roesler27c754c2019-07-10 15:45:51 -050018565cache. By default the corresponding filter is implicitly defined. And when no
Christopher Faulet27d93c32018-12-15 22:32:02 +010018566other filters than cache or compression are used, it is enough. In such case,
18567the compression filter is always evaluated after the cache filter. But it is
18568mandatory to explicitly use a filter line to use a cache when at least one
18569filter other than the compression is used for the same
18570listener/frontend/backend. This is important to know the filters evaluation
18571order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018572
Christopher Faulet27d93c32018-12-15 22:32:02 +010018573See also : section 9.2 about the compression filter and section 10 about cache.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018574
William Lallemand86d0df02017-11-24 21:36:45 +01001857510. Cache
18576---------
18577
18578HAProxy provides a cache, which was designed to perform cache on small objects
18579(favicon, css...). This is a minimalist low-maintenance cache which runs in
18580RAM.
18581
18582The cache is based on a memory which is shared between processes and threads,
Cyril Bonté7b888f12017-11-26 22:24:31 +010018583this memory is split in blocks of 1k.
William Lallemand86d0df02017-11-24 21:36:45 +010018584
18585If an object is not used anymore, it can be deleted to store a new object
18586independently of its expiration date. The oldest objects are deleted first
18587when we try to allocate a new one.
18588
Cyril Bonté7b888f12017-11-26 22:24:31 +010018589The cache uses a hash of the host header and the URI as the key.
William Lallemand86d0df02017-11-24 21:36:45 +010018590
18591It's possible to view the status of a cache using the Unix socket command
18592"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
18593for more details.
18594
18595When an object is delivered from the cache, the server name in the log is
18596replaced by "<CACHE>".
18597
Cyril Bonté7b888f12017-11-26 22:24:31 +01001859810.1. Limitation
18599----------------
William Lallemand86d0df02017-11-24 21:36:45 +010018600
18601The cache won't store and won't deliver objects in these cases:
18602
18603- If the response is not a 200
18604- If the response contains a Vary header
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018605- If the Content-Length + the headers size is greater than "max-object-size"
William Lallemand86d0df02017-11-24 21:36:45 +010018606- If the response is not cacheable
18607
18608- If the request is not a GET
18609- If the HTTP version of the request is smaller than 1.1
William Lallemand8a16fe02018-05-22 11:04:33 +020018610- If the request contains an Authorization header
William Lallemand86d0df02017-11-24 21:36:45 +010018611
Christopher Faulet99a17a22018-12-11 09:18:27 +010018612Caution!: For HAProxy version prior to 1.9, due to the limitation of the
18613filters, it is not recommended to use the cache with other filters. Using them
18614can cause undefined behavior if they modify the response (compression for
18615example). For HAProxy 1.9 and greater, it is safe, for HTX proxies only (see
18616"option http-use-htx" for details).
William Lallemand86d0df02017-11-24 21:36:45 +010018617
Cyril Bonté7b888f12017-11-26 22:24:31 +01001861810.2. Setup
18619-----------
William Lallemand86d0df02017-11-24 21:36:45 +010018620
18621To setup a cache, you must define a cache section and use it in a proxy with
18622the corresponding http-request and response actions.
18623
Cyril Bonté7b888f12017-11-26 22:24:31 +01001862410.2.1. Cache section
18625---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018626
18627cache <name>
18628 Declare a cache section, allocate a shared cache memory named <name>, the
18629 size of cache is mandatory.
18630
18631total-max-size <megabytes>
Davor Ocelice9ed2812017-12-25 17:49:28 +010018632 Define the size in RAM of the cache in megabytes. This size is split in
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018633 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
William Lallemand86d0df02017-11-24 21:36:45 +010018634
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018635max-object-size <bytes>
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018636 Define the maximum size of the objects to be cached. Must not be greater than
18637 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
18638 All objects with sizes larger than "max-object-size" will not be cached.
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018639
William Lallemand86d0df02017-11-24 21:36:45 +010018640max-age <seconds>
18641 Define the maximum expiration duration. The expiration is set has the lowest
18642 value between the s-maxage or max-age (in this order) directive in the
18643 Cache-Control response header and this value. The default value is 60
18644 seconds, which means that you can't cache an object more than 60 seconds by
18645 default.
18646
Cyril Bonté7b888f12017-11-26 22:24:31 +01001864710.2.2. Proxy section
18648---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018649
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018650http-request cache-use <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018651 Try to deliver a cached object from the cache <name>. This directive is also
18652 mandatory to store the cache as it calculates the cache hash. If you want to
18653 use a condition for both storage and delivering that's a good idea to put it
18654 after this one.
18655
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018656http-response cache-store <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018657 Store an http-response within the cache. The storage of the response headers
18658 is done at this step, which means you can use others http-response actions
18659 to modify headers before or after the storage of the response. This action
18660 is responsible for the setup of the cache storage filter.
18661
18662
18663Example:
18664
18665 backend bck1
18666 mode http
18667
18668 http-request cache-use foobar
18669 http-response cache-store foobar
18670 server srv1 127.0.0.1:80
18671
18672 cache foobar
18673 total-max-size 4
18674 max-age 240
18675
Willy Tarreau0ba27502007-12-24 16:55:16 +010018676/*
18677 * Local variables:
18678 * fill-column: 79
18679 * End:
18680 */