blob: ebb19b1366ae51b08ad993b5541c4a4660951ffd [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
Christopher Faulet59a54be2019-06-26 15:59:03 +02007 2019/06/26
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
492* HAPROXY_CLI: configured listeners adresses of the stats socket for every
493 processes, separated by semicolons.
494
495* HAPROXY_MASTER_CLI: In master-worker mode, listeners adresses of the master
496 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
Willy Tarreau6a06a402007-07-15 20:15:28 +0200587 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200588 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100589 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200590 - lua-load
William Lallemand27edc4b2019-05-07 17:49:33 +0200591 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200592 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200593 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200594 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200595 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100596 - presetenv
597 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200598 - uid
599 - ulimit-n
600 - user
Willy Tarreau636848a2019-04-15 19:38:50 +0200601 - set-dumpable
Willy Tarreau1d549722016-02-16 12:41:57 +0100602 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200603 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200604 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200605 - ssl-default-bind-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200606 - ssl-default-bind-options
607 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200608 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200609 - ssl-default-server-options
610 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100611 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100612 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100613 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100614 - 51degrees-data-file
615 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200616 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200617 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200618 - wurfl-data-file
619 - wurfl-information-list
620 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200621 - wurfl-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100622
Willy Tarreau6a06a402007-07-15 20:15:28 +0200623 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200624 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200625 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200626 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100627 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100628 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100629 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200630 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200631 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200632 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200633 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200634 - noepoll
635 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +0000636 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +0200637 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100638 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300639 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000640 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +0100641 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +0200642 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200643 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200644 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000645 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000646 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200647 - tune.buffers.limit
648 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200649 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200650 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100651 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200652 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200653 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200654 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100655 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200656 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200657 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100658 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100659 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100660 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100661 - tune.lua.session-timeout
662 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200663 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100664 - tune.maxaccept
665 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200666 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200667 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200668 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100669 - tune.rcvbuf.client
670 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100671 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200672 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100673 - tune.sndbuf.client
674 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100675 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100676 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200677 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100678 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200679 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200680 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100681 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200682 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100683 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200684 - tune.vars.reqres-max-size
685 - tune.vars.sess-max-size
686 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100687 - tune.zlib.memlevel
688 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100689
Willy Tarreau6a06a402007-07-15 20:15:28 +0200690 * Debugging
691 - debug
692 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200693
694
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006953.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200696------------------------------------
697
Emeric Brunc8e8d122012-10-02 18:42:10 +0200698ca-base <dir>
699 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200700 relative path is used with "ca-file" or "crl-file" directives. Absolute
701 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200702
Willy Tarreau6a06a402007-07-15 20:15:28 +0200703chroot <jail dir>
704 Changes current directory to <jail dir> and performs a chroot() there before
705 dropping privileges. This increases the security level in case an unknown
706 vulnerability would be exploited, since it would make it very hard for the
707 attacker to exploit the system. This only works when the process is started
708 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100709 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100710
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100711cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
712 On Linux 2.6 and above, it is possible to bind a process or a thread to a
713 specific CPU set. This means that the process or the thread will never run on
714 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
715 sets. The first argument is a process set, eventually followed by a thread
716 set. These sets have the format
717
718 all | odd | even | number[-[number]]
719
720 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100721 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100722 ignored. It is possible to specify a range with two such number delimited by
723 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100724 "all", only odd numbers using "odd" or even numbers using "even", just like
725 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100726 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 +0100727 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100728 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100729 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100730 specified. Each "cpu-map" directive will replace the previous ones when they
731 overlap. A thread will be bound on the intersection of its mapping and the
732 one of the process on which it is attached. If the intersection is null, no
733 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100734
Christopher Fauletff4121f2017-11-22 16:38:49 +0100735 Ranges can be partially defined. The higher bound can be omitted. In such
736 case, it is replaced by the corresponding maximum value, 32 or 64 depending
737 on the machine's word size.
738
Christopher Faulet26028f62017-11-22 15:01:51 +0100739 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100740 automatically bind a process or a thread to a CPU by incrementing
741 process/thread and CPU sets. To be valid, both sets must have the same
742 size. No matter the declaration order of the CPU sets, it will be bound from
743 the lowest to the highest bound. Having a process and a thread range with the
744 "auto:" prefix is not supported. Only one range is supported, the other one
745 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100746
747 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100748 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
749
750 cpu-map 1/all 0-3 # bind all threads of the first process on the
751 # first 4 CPUs
752
753 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
754 # or "cpu-map 1-32 0-31" depending on the machine's
755 # word size.
756
Christopher Faulet26028f62017-11-22 15:01:51 +0100757 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100758 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100759 cpu-map auto:1-4 0-3
760 cpu-map auto:1-4 0-1 2-3
761 cpu-map auto:1-4 3 2 1 0
762
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100763 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
764 # and so on.
765 cpu-map auto:1/1-4 0-3
766 cpu-map auto:1/1-4 0-1 2-3
767 cpu-map auto:1/1-4 3 2 1 0
768
Davor Ocelice9ed2812017-12-25 17:49:28 +0100769 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100770 cpu-map auto:all 0-63
771 cpu-map auto:even 0-31
772 cpu-map auto:odd 32-63
773
774 # invalid cpu-map because process and CPU sets have different sizes.
775 cpu-map auto:1-4 0 # invalid
776 cpu-map auto:1 0-3 # invalid
777
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100778 # invalid cpu-map because automatic binding is used with a process range
779 # and a thread range.
780 cpu-map auto:all/all 0 # invalid
781 cpu-map auto:all/1-4 0 # invalid
782 cpu-map auto:1-4/all 0 # invalid
783
Emeric Brunc8e8d122012-10-02 18:42:10 +0200784crt-base <dir>
785 Assigns a default directory to fetch SSL certificates from when a relative
786 path is used with "crtfile" directives. Absolute locations specified after
787 "crtfile" prevail and ignore "crt-base".
788
Willy Tarreau6a06a402007-07-15 20:15:28 +0200789daemon
790 Makes the process fork into background. This is the recommended mode of
791 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100792 disabled by the command line "-db" argument. This option is ignored in
793 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200794
David Carlier8167f302015-06-01 13:50:06 +0200795deviceatlas-json-file <path>
796 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100797 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200798
799deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100800 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200801 optional and set to 0 by default if not set.
802
803deviceatlas-separator <char>
804 Sets the character separator for the API properties results. This directive
805 is optional and set to | by default if not set.
806
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100807deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200808 Sets the client cookie's name used for the detection if the DeviceAtlas
809 Client-side component was used during the request. This directive is optional
810 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100811
Simon Horman98637e52014-06-20 12:30:16 +0900812external-check
813 Allows the use of an external agent to perform health checks.
814 This is disabled by default as a security precaution.
815 See "option external-check".
816
Willy Tarreau6a06a402007-07-15 20:15:28 +0200817gid <number>
818 Changes the process' group ID to <number>. It is recommended that the group
819 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
820 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100821 Note that if haproxy is started from a user having supplementary groups, it
822 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200823 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100824
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100825hard-stop-after <time>
826 Defines the maximum time allowed to perform a clean soft-stop.
827
828 Arguments :
829 <time> is the maximum time (by default in milliseconds) for which the
830 instance will remain alive when a soft-stop is received via the
831 SIGUSR1 signal.
832
833 This may be used to ensure that the instance will quit even if connections
834 remain opened during a soft-stop (for example with long timeouts for a proxy
835 in tcp mode). It applies both in TCP and HTTP mode.
836
837 Example:
838 global
839 hard-stop-after 30s
840
Willy Tarreau6a06a402007-07-15 20:15:28 +0200841group <group name>
842 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
843 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100844
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200845log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
846 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100847 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100848 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100849 configured with "log global".
850
851 <address> can be one of:
852
Willy Tarreau2769aa02007-12-27 18:26:09 +0100853 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100854 no port is specified, 514 is used by default (the standard syslog
855 port).
856
David du Colombier24bb5f52011-03-17 10:40:23 +0100857 - An IPv6 address followed by a colon and optionally a UDP port. If
858 no port is specified, 514 is used by default (the standard syslog
859 port).
860
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100861 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +0100862 considerations for chroot (be sure the path is accessible inside
863 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100864 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100865
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100866 - A file descriptor number in the form "fd@<number>", which may point
867 to a pipe, terminal, or socket. In this case unbuffered logs are used
868 and one writev() call per log is performed. This is a bit expensive
869 but acceptable for most workloads. Messages sent this way will not be
870 truncated but may be dropped, in which case the DroppedLogs counter
871 will be incremented. The writev() call is atomic even on pipes for
872 messages up to PIPE_BUF size, which POSIX recommends to be at least
873 512 and which is 4096 bytes on most modern operating systems. Any
874 larger message may be interleaved with messages from other processes.
875 Exceptionally for debugging purposes the file descriptor may also be
876 directed to a file, but doing so will significantly slow haproxy down
877 as non-blocking calls will be ignored. Also there will be no way to
878 purge nor rotate this file without restarting the process. Note that
879 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +0100880 for use with a TCP syslog server. See also the "short" and "raw"
881 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100882
883 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
884 "fd@2", see above.
885
William Lallemandb2f07452015-05-12 14:27:13 +0200886 You may want to reference some environment variables in the address
887 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100888
Willy Tarreau18324f52014-06-27 18:10:07 +0200889 <length> is an optional maximum line length. Log lines larger than this value
890 will be truncated before being sent. The reason is that syslog
891 servers act differently on log line length. All servers support the
892 default value of 1024, but some servers simply drop larger lines
893 while others do log them. If a server supports long lines, it may
894 make sense to set this value here in order to avoid truncating long
895 lines. Similarly, if a server drops long lines, it is preferable to
896 truncate them before sending them. Accepted values are 80 to 65535
897 inclusive. The default value of 1024 is generally fine for all
898 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +0100899 JSON-formatted logs may require larger values. You may also need to
900 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +0200901
Dragan Dosen7ad31542015-09-28 17:16:47 +0200902 <format> is the log format used when generating syslog messages. It may be
903 one of the following :
904
905 rfc3164 The RFC3164 syslog message format. This is the default.
906 (https://tools.ietf.org/html/rfc3164)
907
908 rfc5424 The RFC5424 syslog message format.
909 (https://tools.ietf.org/html/rfc5424)
910
Willy Tarreaue8746a02018-11-12 08:45:00 +0100911 short A message containing only a level between angle brackets such as
912 '<3>', followed by the text. The PID, date, time, process name
913 and system name are omitted. This is designed to be used with a
914 local log server. This format is compatible with what the systemd
915 logger consumes.
916
Willy Tarreauc1b06452018-11-12 11:57:56 +0100917 raw A message containing only the text. The level, PID, date, time,
918 process name and system name are omitted. This is designed to be
919 used in containers or during development, where the severity only
920 depends on the file descriptor used (stdout/stderr).
921
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200922 <ranges> A list of comma-separated ranges to identify the logs to sample.
923 This is used to balance the load of the logs to send to the log
924 server. The limits of the ranges cannot be null. They are numbered
925 from 1. The size or period (in number of logs) of the sample must be
926 set with <sample_size> parameter.
927
928 <sample_size>
929 The size of the sample in number of logs to consider when balancing
930 their logging loads. It is used to balance the load of the logs to
931 send to the syslog server. This size must be greater or equal to the
932 maximum of the high limits of the ranges.
933 (see also <ranges> parameter).
934
Robert Tsai81ae1952007-12-05 10:47:29 +0100935 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200936
Willy Tarreaue8746a02018-11-12 08:45:00 +0100937 kern user mail daemon auth syslog lpr news
938 uucp cron auth2 ftp ntp audit alert cron2
939 local0 local1 local2 local3 local4 local5 local6 local7
940
Willy Tarreauc1b06452018-11-12 11:57:56 +0100941 Note that the facility is ignored for the "short" and "raw"
942 formats, but still required as a positional field. It is
943 recommended to use "daemon" in this case to make it clear that
944 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200945
946 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200947 all messages are sent. If a maximum level is specified, only messages with a
948 severity at least as important as this level will be sent. An optional minimum
949 level can be specified. If it is set, logs emitted with a more severe level
950 than this one will be capped to this level. This is used to avoid sending
951 "emerg" messages on all terminals on some default syslog configurations.
952 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200953
Cyril Bontédc4d9032012-04-08 21:57:39 +0200954 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200955
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100956log-send-hostname [<string>]
957 Sets the hostname field in the syslog header. If optional "string" parameter
958 is set the header is set to the string contents, otherwise uses the hostname
959 of the system. Generally used if one is not relaying logs through an
960 intermediate syslog server or for simply customizing the hostname printed in
961 the logs.
962
Kevinm48936af2010-12-22 16:08:21 +0000963log-tag <string>
964 Sets the tag field in the syslog header to this string. It defaults to the
965 program name as launched from the command line, which usually is "haproxy".
966 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100967 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000968
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100969lua-load <file>
970 This global directive loads and executes a Lua file. This directive can be
971 used multiple times.
972
William Lallemand4cfede82017-11-24 22:02:34 +0100973master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +0200974 Master-worker mode. It is equivalent to the command line "-W" argument.
975 This mode will launch a "master" which will monitor the "workers". Using
976 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100977 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +0200978 or daemon mode. It is recommended to use this mode with multiprocess and
979 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +0100980 By default, if a worker exits with a bad return code, in the case of a
981 segfault for example, all workers will be killed, and the master will leave.
982 It is convenient to combine this behavior with Restart=on-failure in a
983 systemd unit file in order to relaunch the whole process. If you don't want
984 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +0200985
William Lallemand4cfede82017-11-24 22:02:34 +0100986 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +0200987
William Lallemand27edc4b2019-05-07 17:49:33 +0200988mworker-max-reloads <number>
989 In master-worker mode, this option limits the number of time a worker can
990 survive to a reload. If the worker did not left after a reload, once its
991 number of reloads is greater than this number, the worker will receive a
992 SIGTERM. This option helps to keep under control the number of workers.
993 See also "show proc" in the Management Guide.
994
Willy Tarreau6a06a402007-07-15 20:15:28 +0200995nbproc <number>
996 Creates <number> processes when going daemon. This requires the "daemon"
997 mode. By default, only one process is created, which is the recommended mode
998 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +0100999 process, it may be needed to fork multiple daemons. When set to a value
1000 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreau1f672a82019-01-26 14:20:55 +01001001 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon" and
1002 "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001003
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001004nbthread <number>
1005 This setting is only available when support for threads was built in. It
Willy Tarreau26f6ae12019-02-02 12:56:15 +01001006 makes haproxy run on <number> threads. This is exclusive with "nbproc". While
1007 "nbproc" historically used to be the only way to use multiple processors, it
1008 also involved a number of shortcomings related to the lack of synchronization
1009 between processes (health-checks, peers, stick-tables, stats, ...) which do
1010 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +01001011 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
1012 HAProxy is started in foreground. On some platforms supporting CPU affinity,
1013 when nbproc is not used, the default "nbthread" value is automatically set to
1014 the number of CPUs the process is bound to upon startup. This means that the
1015 thread count can easily be adjusted from the calling process using commands
1016 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1017 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001018
Willy Tarreau6a06a402007-07-15 20:15:28 +02001019pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001020 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +02001021 the "-p" command line argument. The file must be accessible to the user
1022 starting the process. See also "daemon".
1023
Willy Tarreau1d549722016-02-16 12:41:57 +01001024presetenv <name> <value>
1025 Sets environment variable <name> to value <value>. If the variable exists, it
1026 is NOT overwritten. The changes immediately take effect so that the next line
1027 in the configuration file sees the new value. See also "setenv", "resetenv",
1028 and "unsetenv".
1029
1030resetenv [<name> ...]
1031 Removes all environment variables except the ones specified in argument. It
1032 allows to use a clean controlled environment before setting new values with
1033 setenv or unsetenv. Please note that some internal functions may make use of
1034 some environment variables, such as time manipulation functions, but also
1035 OpenSSL or even external checks. This must be used with extreme care and only
1036 after complete validation. The changes immediately take effect so that the
1037 next line in the configuration file sees the new environment. See also
1038 "setenv", "presetenv", and "unsetenv".
1039
Christopher Fauletff4121f2017-11-22 16:38:49 +01001040stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001041 Limits the stats socket to a certain set of processes numbers. By default the
1042 stats socket is bound to all processes, causing a warning to be emitted when
1043 nbproc is greater than 1 because there is no way to select the target process
1044 when connecting. However, by using this setting, it becomes possible to pin
1045 the stats socket to a specific set of processes, typically the first one. The
1046 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001047 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001048 word size (32 or 64). Ranges can be partially defined. The higher bound can
1049 be omitted. In such case, it is replaced by the corresponding maximum
1050 value. A better option consists in using the "process" setting of the "stats
1051 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001052
Baptiste Assmann5626f482015-08-23 10:00:10 +02001053server-state-base <directory>
1054 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001055 file names which do not start with a '/'. See also "server-state-file",
1056 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001057
1058server-state-file <file>
1059 Specifies the path to the file containing state of servers. If the path starts
1060 with a slash ('/'), it is considered absolute, otherwise it is considered
1061 relative to the directory specified using "server-state-base" (if set) or to
1062 the current directory. Before reloading HAProxy, it is possible to save the
1063 servers' current state using the stats command "show servers state". The
1064 output of this command must be written in the file pointed by <file>. When
1065 starting up, before handling traffic, HAProxy will read, load and apply state
1066 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001067 configuration. See also "server-state-base" and "show servers state",
1068 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001069
Willy Tarreau1d549722016-02-16 12:41:57 +01001070setenv <name> <value>
1071 Sets environment variable <name> to value <value>. If the variable exists, it
1072 is overwritten. The changes immediately take effect so that the next line in
1073 the configuration file sees the new value. See also "presetenv", "resetenv",
1074 and "unsetenv".
1075
Willy Tarreau636848a2019-04-15 19:38:50 +02001076set-dumpable
1077 This option is better left disabled by default and enabled only upon a
1078 developer's request. It has no impact on performance nor stability but will
1079 try hard to re-enable core dumps that were possibly disabled by file size
1080 limitations (ulimit -f), core size limitations (ulimit -c), or "dumpability"
1081 of a process after changing its UID/GID (such as /proc/sys/fs/suid_dumpable
1082 on Linux). Core dumps might still be limited by the current directory's
1083 permissions (check what directory the file is started from), the chroot
1084 directory's permission (it may be needed to temporarily disable the chroot
1085 directive or to move it to a dedicated writable location), or any other
1086 system-specific constraint. For example, some Linux flavours are notorious
1087 for replacing the default core file with a path to an executable not even
1088 installed on the system (check /proc/sys/kernel/core_pattern). Often, simply
1089 writing "core", "core.%p" or "/var/log/core/core.%p" addresses the issue.
1090 When trying to enable this option waiting for a rare issue to re-appear, it's
1091 often a good idea to first try to obtain such a dump by issuing, for example,
1092 "kill -11" to the haproxy process and verify that it leaves a core where
1093 expected when dying.
1094
Willy Tarreau610f04b2014-02-13 11:36:41 +01001095ssl-default-bind-ciphers <ciphers>
1096 This setting is only available when support for OpenSSL was built in. It sets
1097 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001098 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001099 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001100 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1101 information and recommendations see e.g.
1102 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1103 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1104 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1105 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001106
1107ssl-default-bind-ciphersuites <ciphersuites>
1108 This setting is only available when support for OpenSSL was built in and
1109 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1110 describing the list of cipher algorithms ("cipher suite") that are negotiated
1111 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1112 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001113 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1114 cipher configuration for TLSv1.2 and earlier, please check the
1115 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001116 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001117
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001118ssl-default-bind-options [<option>]...
1119 This setting is only available when support for OpenSSL was built in. It sets
1120 default ssl-options to force on all "bind" lines. Please check the "bind"
1121 keyword to see available options.
1122
1123 Example:
1124 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001125 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001126
Willy Tarreau610f04b2014-02-13 11:36:41 +01001127ssl-default-server-ciphers <ciphers>
1128 This setting is only available when support for OpenSSL was built in. It
1129 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001130 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001131 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001132 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1133 information and recommendations see e.g.
1134 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1135 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1136 For TLSv1.3 cipher configuration, please check the
1137 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1138 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001139
1140ssl-default-server-ciphersuites <ciphersuites>
1141 This setting is only available when support for OpenSSL was built in and
1142 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1143 string describing the list of cipher algorithms that are negotiated during
1144 the TLSv1.3 handshake with the server, for all "server" lines which do not
1145 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001146 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1147 cipher configuration for TLSv1.2 and earlier, please check the
1148 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1149 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001150
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001151ssl-default-server-options [<option>]...
1152 This setting is only available when support for OpenSSL was built in. It sets
1153 default ssl-options to force on all "server" lines. Please check the "server"
1154 keyword to see available options.
1155
Remi Gacogne47783ef2015-05-29 15:53:22 +02001156ssl-dh-param-file <file>
1157 This setting is only available when support for OpenSSL was built in. It sets
1158 the default DH parameters that are used during the SSL/TLS handshake when
1159 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001160 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001161 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001162 are not specified either by using ssl-dh-param-file or by setting them
1163 directly in the certificate file, pre-generated DH parameters of the size
1164 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1165 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001166 Custom DH parameters may be generated by using the OpenSSL command
1167 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1168 parameters should not be considered secure anymore.
1169
Emeric Brun850efd52014-01-29 12:24:34 +01001170ssl-server-verify [none|required]
1171 The default behavior for SSL verify on servers side. If specified to 'none',
1172 servers certificates are not verified. The default is 'required' except if
1173 forced using cmdline option '-dV'.
1174
Willy Tarreauabb175f2012-09-24 12:43:26 +02001175stats socket [<address:port>|<path>] [param*]
1176 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1177 Connections to this socket will return various statistics outputs and even
1178 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001179 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001180 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001181
Willy Tarreauabb175f2012-09-24 12:43:26 +02001182 All parameters supported by "bind" lines are supported, for instance to
1183 restrict access to some users or their access rights. Please consult
1184 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001185
1186stats timeout <timeout, in milliseconds>
1187 The default timeout on the stats socket is set to 10 seconds. It is possible
1188 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001189 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001190
1191stats maxconn <connections>
1192 By default, the stats socket is limited to 10 concurrent connections. It is
1193 possible to change this value with "stats maxconn".
1194
Willy Tarreau6a06a402007-07-15 20:15:28 +02001195uid <number>
1196 Changes the process' user ID to <number>. It is recommended that the user ID
1197 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1198 be started with superuser privileges in order to be able to switch to another
1199 one. See also "gid" and "user".
1200
1201ulimit-n <number>
1202 Sets the maximum number of per-process file-descriptors to <number>. By
1203 default, it is automatically computed, so it is recommended not to use this
1204 option.
1205
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001206unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1207 [ group <group> ] [ gid <gid> ]
1208
1209 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1210 This is mainly used to simplify declaration of those UNIX sockets and reduce
1211 the risk of errors, since those settings are most commonly required but are
1212 also process-specific. The <prefix> setting can be used to force all socket
1213 path to be relative to that directory. This might be needed to access another
1214 component's chroot. Note that those paths are resolved before haproxy chroots
1215 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1216 all have the same meaning as their homonyms used by the "bind" statement. If
1217 both are specified, the "bind" statement has priority, meaning that the
1218 "unix-bind" settings may be seen as process-wide default settings.
1219
Willy Tarreau1d549722016-02-16 12:41:57 +01001220unsetenv [<name> ...]
1221 Removes environment variables specified in arguments. This can be useful to
1222 hide some sensitive information that are occasionally inherited from the
1223 user's environment during some operations. Variables which did not exist are
1224 silently ignored so that after the operation, it is certain that none of
1225 these variables remain. The changes immediately take effect so that the next
1226 line in the configuration file will not see these variables. See also
1227 "setenv", "presetenv", and "resetenv".
1228
Willy Tarreau6a06a402007-07-15 20:15:28 +02001229user <user name>
1230 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1231 See also "uid" and "group".
1232
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001233node <name>
1234 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1235
1236 This statement is useful in HA configurations where two or more processes or
1237 servers share the same IP address. By setting a different node-name on all
1238 nodes, it becomes easy to immediately spot what server is handling the
1239 traffic.
1240
1241description <text>
1242 Add a text that describes the instance.
1243
1244 Please note that it is required to escape certain characters (# for example)
1245 and this text is inserted into a html page so you should avoid using
1246 "<" and ">" characters.
1247
Thomas Holmesdb04f192015-05-18 13:21:39 +0100124851degrees-data-file <file path>
1249 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001250 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001251
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001252 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001253 compiled with USE_51DEGREES.
1254
Ben Shillitof25e8e52016-12-02 14:25:37 +0000125551degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001256 A list of 51Degrees property names to be load from the dataset. A full list
1257 of names is available on the 51Degrees website:
1258 https://51degrees.com/resources/property-dictionary
1259
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001260 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001261 compiled with USE_51DEGREES.
1262
Dragan Dosen93b38d92015-06-29 16:43:25 +0200126351degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001264 A char that will be appended to every property value in a response header
1265 containing 51Degrees results. If not set that will be set as ','.
1266
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001267 Please note that this option is only available when haproxy has been
1268 compiled with USE_51DEGREES.
1269
127051degrees-cache-size <number>
1271 Sets the size of the 51Degrees converter cache to <number> entries. This
1272 is an LRU cache which reminds previous device detections and their results.
1273 By default, this cache is disabled.
1274
1275 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001276 compiled with USE_51DEGREES.
1277
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001278wurfl-data-file <file path>
1279 The path of the WURFL data file to provide device detection services. The
1280 file should be accessible by HAProxy with relevant permissions.
1281
1282 Please note that this option is only available when haproxy has been compiled
1283 with USE_WURFL=1.
1284
1285wurfl-information-list [<capability>]*
1286 A space-delimited list of WURFL capabilities, virtual capabilities, property
1287 names we plan to use in injected headers. A full list of capability and
1288 virtual capability names is available on the Scientiamobile website :
1289
1290 https://www.scientiamobile.com/wurflCapability
1291
1292 Valid WURFL properties are:
1293 - wurfl_id Contains the device ID of the matched device.
1294
1295 - wurfl_root_id Contains the device root ID of the matched
1296 device.
1297
1298 - wurfl_isdevroot Tells if the matched device is a root device.
1299 Possible values are "TRUE" or "FALSE".
1300
1301 - wurfl_useragent The original useragent coming with this
1302 particular web request.
1303
1304 - wurfl_api_version Contains a string representing the currently
1305 used Libwurfl API version.
1306
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001307 - wurfl_info A string containing information on the parsed
1308 wurfl.xml and its full path.
1309
1310 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1311 WURFL has been loaded successfully.
1312
1313 - wurfl_normalized_useragent The normalized useragent.
1314
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001315 Please note that this option is only available when haproxy has been compiled
1316 with USE_WURFL=1.
1317
1318wurfl-information-list-separator <char>
1319 A char that will be used to separate values in a response header containing
1320 WURFL results. If not set that a comma (',') will be used by default.
1321
1322 Please note that this option is only available when haproxy has been compiled
1323 with USE_WURFL=1.
1324
1325wurfl-patch-file [<file path>]
1326 A list of WURFL patch file paths. Note that patches are loaded during startup
1327 thus before the chroot.
1328
1329 Please note that this option is only available when haproxy has been compiled
1330 with USE_WURFL=1.
1331
paulborilebad132c2019-04-18 11:57:04 +02001332wurfl-cache-size <size>
1333 Sets the WURFL Useragent cache size. For faster lookups, already processed user
1334 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001335 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02001336 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001337
1338 Please note that this option is only available when haproxy has been compiled
1339 with USE_WURFL=1.
1340
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013413.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001342-----------------------
1343
Willy Tarreaubeb859a2018-11-22 18:07:59 +01001344busy-polling
1345 In some situations, especially when dealing with low latency on processors
1346 supporting a variable frequency or when running inside virtual machines, each
1347 time the process waits for an I/O using the poller, the processor goes back
1348 to sleep or is offered to another VM for a long time, and it causes
1349 excessively high latencies. This option provides a solution preventing the
1350 processor from sleeping by always using a null timeout on the pollers. This
1351 results in a significant latency reduction (30 to 100 microseconds observed)
1352 at the expense of a risk to overheat the processor. It may even be used with
1353 threads, in which case improperly bound threads may heavily conflict,
1354 resulting in a worse performance and high values for the CPU stolen fields
1355 in "show info" output, indicating which threads are misconfigured. It is
1356 important not to let the process run on the same processor as the network
1357 interrupts when this option is used. It is also better to avoid using it on
1358 multiple CPU threads sharing the same core. This option is disabled by
1359 default. If it has been enabled, it may still be forcibly disabled by
1360 prefixing it with the "no" keyword. It is ignored by the "select" and
1361 "poll" pollers.
1362
Willy Tarreau1746eec2014-04-25 10:46:47 +02001363max-spread-checks <delay in milliseconds>
1364 By default, haproxy tries to spread the start of health checks across the
1365 smallest health check interval of all the servers in a farm. The principle is
1366 to avoid hammering services running on the same server. But when using large
1367 check intervals (10 seconds or more), the last servers in the farm take some
1368 time before starting to be tested, which can be a problem. This parameter is
1369 used to enforce an upper bound on delay between the first and the last check,
1370 even if the servers' check intervals are larger. When servers run with
1371 shorter intervals, their intervals will be respected though.
1372
Willy Tarreau6a06a402007-07-15 20:15:28 +02001373maxconn <number>
1374 Sets the maximum per-process number of concurrent connections to <number>. It
1375 is equivalent to the command-line argument "-n". Proxies will stop accepting
1376 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001377 automatically adjusted according to this value. See also "ulimit-n". Note:
1378 the "select" poller cannot reliably use more than 1024 file descriptors on
1379 some platforms. If your platform only supports select and reports "select
1380 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01001381 below 500 in general). If this value is not set, it will automatically be
1382 calculated based on the current file descriptors limit reported by the
1383 "ulimit -n" command, possibly reduced to a lower value if a memory limit
1384 is enforced, based on the buffer size, memory allocated to compression, SSL
1385 cache size, and use or not of SSL and the associated maxsslconn (which can
1386 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001387
Willy Tarreau81c25d02011-09-07 15:17:21 +02001388maxconnrate <number>
1389 Sets the maximum per-process number of connections per second to <number>.
1390 Proxies will stop accepting connections when this limit is reached. It can be
1391 used to limit the global capacity regardless of each frontend capacity. It is
1392 important to note that this can only be used as a service protection measure,
1393 as there will not necessarily be a fair share between frontends when the
1394 limit is reached, so it's a good idea to also limit each frontend to some
1395 value close to its expected share. Also, lowering tune.maxaccept can improve
1396 fairness.
1397
William Lallemandd85f9172012-11-09 17:05:39 +01001398maxcomprate <number>
1399 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001400 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001401 level will be decreased during the session. If the maximum is reached at the
1402 beginning of a session, the session will not compress at all. If the maximum
1403 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001404 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001405 default value.
1406
William Lallemand072a2bf2012-11-20 17:01:01 +01001407maxcompcpuusage <number>
1408 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1409 for new requests or decreasing the compression level of current requests.
1410 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1411 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1412 case of multiple processes (nbproc > 1), each process manages its individual
1413 usage. A value of 100 disable the limit. The default value is 100. Setting
1414 a lower value will prevent the compression work from slowing the whole
1415 process down and from introducing high latencies.
1416
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001417maxpipes <number>
1418 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1419 are only used by kernel-based tcp splicing. Since a pipe contains two file
1420 descriptors, the "ulimit-n" value will be increased accordingly. The default
1421 value is maxconn/4, which seems to be more than enough for most heavy usages.
1422 The splice code dynamically allocates and releases pipes, and can fall back
1423 to standard copy, so setting this value too low may only impact performance.
1424
Willy Tarreau93e7c002013-10-07 18:51:07 +02001425maxsessrate <number>
1426 Sets the maximum per-process number of sessions per second to <number>.
1427 Proxies will stop accepting connections when this limit is reached. It can be
1428 used to limit the global capacity regardless of each frontend capacity. It is
1429 important to note that this can only be used as a service protection measure,
1430 as there will not necessarily be a fair share between frontends when the
1431 limit is reached, so it's a good idea to also limit each frontend to some
1432 value close to its expected share. Also, lowering tune.maxaccept can improve
1433 fairness.
1434
Willy Tarreau403edff2012-09-06 11:58:37 +02001435maxsslconn <number>
1436 Sets the maximum per-process number of concurrent SSL connections to
1437 <number>. By default there is no SSL-specific limit, which means that the
1438 global maxconn setting will apply to all connections. Setting this limit
1439 avoids having openssl use too much memory and crash when malloc returns NULL
1440 (since it unfortunately does not reliably check for such conditions). Note
1441 that the limit applies both to incoming and outgoing connections, so one
1442 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001443 If this value is not set, but a memory limit is enforced, this value will be
1444 automatically computed based on the memory limit, maxconn, the buffer size,
1445 memory allocated to compression, SSL cache size, and use of SSL in either
1446 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1447 when there is a memory limit, haproxy will automatically adjust these values
1448 so that 100% of the connections can be made over SSL with no risk, and will
1449 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001450
Willy Tarreaue43d5322013-10-07 20:01:52 +02001451maxsslrate <number>
1452 Sets the maximum per-process number of SSL sessions per second to <number>.
1453 SSL listeners will stop accepting connections when this limit is reached. It
1454 can be used to limit the global SSL CPU usage regardless of each frontend
1455 capacity. It is important to note that this can only be used as a service
1456 protection measure, as there will not necessarily be a fair share between
1457 frontends when the limit is reached, so it's a good idea to also limit each
1458 frontend to some value close to its expected share. It is also important to
1459 note that the sessions are accounted before they enter the SSL stack and not
1460 after, which also protects the stack against bad handshakes. Also, lowering
1461 tune.maxaccept can improve fairness.
1462
William Lallemand9d5f5482012-11-07 16:12:57 +01001463maxzlibmem <number>
1464 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1465 When the maximum amount is reached, future sessions will not compress as long
1466 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001467 The default value is 0. The value is available in bytes on the UNIX socket
1468 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1469 "ZlibMemUsage" in bytes.
1470
Willy Tarreau6a06a402007-07-15 20:15:28 +02001471noepoll
1472 Disables the use of the "epoll" event polling system on Linux. It is
1473 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001474 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001475
1476nokqueue
1477 Disables the use of the "kqueue" event polling system on BSD. It is
1478 equivalent to the command-line argument "-dk". The next polling system
1479 used will generally be "poll". See also "nopoll".
1480
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001481noevports
1482 Disables the use of the event ports event polling system on SunOS systems
1483 derived from Solaris 10 and later. It is equivalent to the command-line
1484 argument "-dv". The next polling system used will generally be "poll". See
1485 also "nopoll".
1486
Willy Tarreau6a06a402007-07-15 20:15:28 +02001487nopoll
1488 Disables the use of the "poll" event polling system. It is equivalent to the
1489 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001490 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001491 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
1492 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001493
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001494nosplice
1495 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001496 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001497 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001498 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001499 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1500 be used. This option makes it easier to globally disable kernel splicing in
1501 case of doubt. See also "option splice-auto", "option splice-request" and
1502 "option splice-response".
1503
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001504nogetaddrinfo
1505 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1506 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1507
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001508noreuseport
1509 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1510 command line argument "-dR".
1511
Willy Tarreaud2d33482019-04-25 17:09:07 +02001512profiling.tasks { auto | on | off }
1513 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
1514 the profiling automatically turns on a thread when it starts to suffer from
1515 an average latency of 1000 microseconds or higher as reported in the
1516 "avg_loop_us" activity field, and automatically turns off when the latency
1517 reutnrs below 990 microseconds (this value is an average over the last 1024
1518 loops so it does not vary quickly and tends to significantly smooth short
1519 spikes). It may also spontaneously trigger from time to time on overloaded
1520 systems, containers, or virtual machines, or when the system swaps (which
1521 must absolutely never happen on a load balancer).
1522
1523 CPU profiling per task can be very convenient to report where the time is
1524 spent and which requests have what effect on which other request. Enabling
1525 it will typically affect the overall's performance by less than 1%, thus it
1526 is recommended to leave it to the default 'auto' value so that it only
1527 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01001528 supporting the clock_gettime(2) syscall with clock identifiers
1529 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
1530 be zero. This option may be changed at run time using "set profiling" on the
1531 CLI.
1532
Willy Tarreaufe255b72007-10-14 23:09:26 +02001533spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001534 Sometimes it is desirable to avoid sending agent and health checks to
1535 servers at exact intervals, for instance when many logical servers are
1536 located on the same physical server. With the help of this parameter, it
1537 becomes possible to add some randomness in the check interval between 0
1538 and +/- 50%. A value between 2 and 5 seems to show good results. The
1539 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001540
Davor Ocelice9ed2812017-12-25 17:49:28 +01001541ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001542 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001543 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001544 multiple times, it will simply enable multiple crypto engines. Referencing an
1545 unsupported engine will prevent haproxy from starting. Note that many engines
1546 will lead to lower HTTPS performance than pure software with recent
1547 processors. The optional command "algo" sets the default algorithms an ENGINE
1548 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001549 of "ALL" uses the engine for all cryptographic operations. If no list of
1550 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001551 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1552 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1553 openssl configuration file uses:
1554 https://www.openssl.org/docs/man1.0.2/apps/config.html
1555
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001556ssl-mode-async
1557 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001558 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001559 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1560 doesn't support moving read/write buffers and is not compliant with
1561 haproxy's buffer management. So the asynchronous mode is disabled on
1562 read/write operations (it is only enabled during initial and reneg
1563 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001564
Willy Tarreau33cb0652014-12-23 22:52:37 +01001565tune.buffers.limit <number>
1566 Sets a hard limit on the number of buffers which may be allocated per process.
1567 The default value is zero which means unlimited. The minimum non-zero value
1568 will always be greater than "tune.buffers.reserve" and should ideally always
1569 be about twice as large. Forcing this value can be particularly useful to
1570 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001571 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001572 another one to be released by another session. Since buffers are dynamically
1573 allocated and released, the waiting time is very short and not perceptible
1574 provided that limits remain reasonable. In fact sometimes reducing the limit
1575 may even increase performance by increasing the CPU cache's efficiency. Tests
1576 have shown good results on average HTTP traffic with a limit to 1/10 of the
1577 expected global maxconn setting, which also significantly reduces memory
1578 usage. The memory savings come from the fact that a number of connections
1579 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1580 advised to do so by an haproxy core developer.
1581
Willy Tarreau1058ae72014-12-23 22:40:40 +01001582tune.buffers.reserve <number>
1583 Sets the number of buffers which are pre-allocated and reserved for use only
1584 during memory shortage conditions resulting in failed memory allocations. The
1585 minimum value is 2 and is also the default. There is no reason a user would
1586 want to change this value, it's mostly aimed at haproxy core developers.
1587
Willy Tarreau27a674e2009-08-17 07:23:33 +02001588tune.bufsize <number>
1589 Sets the buffer size to this size (in bytes). Lower values allow more
1590 sessions to coexist in the same amount of RAM, and higher values allow some
1591 applications with very large cookies to work. The default value is 16384 and
1592 can be changed at build time. It is strongly recommended not to change this
1593 from the default value, as very low values will break some services such as
1594 statistics, and values larger than default size will increase memory usage,
1595 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001596 parameter should be decreased by the same factor as this one is increased. In
1597 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1598 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001599 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Willy Tarreauc77d3642018-12-12 06:19:42 +01001600 than this size, haproxy will return HTTP 502 (Bad Gateway). Note that the
1601 value set using this parameter will automatically be rounded up to the next
1602 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02001603
Willy Tarreau43961d52010-10-04 20:39:20 +02001604tune.chksize <number>
1605 Sets the check buffer size to this size (in bytes). Higher values may help
1606 find string or regex patterns in very large pages, though doing so may imply
1607 more memory and CPU usage. The default value is 16384 and can be changed at
1608 build time. It is not recommended to change this value, but to use better
1609 checks whenever possible.
1610
William Lallemandf3747832012-11-09 12:33:10 +01001611tune.comp.maxlevel <number>
1612 Sets the maximum compression level. The compression level affects CPU
1613 usage during compression. This value affects CPU usage during compression.
1614 Each session using compression initializes the compression algorithm with
1615 this value. The default value is 1.
1616
Willy Tarreauc299e1e2019-02-27 11:35:12 +01001617tune.fail-alloc
1618 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
1619 allocation attempt fails. Must be between 0 (no failure) and 100 (no
1620 success). This is useful to debug and make sure memory failures are handled
1621 gracefully.
1622
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001623tune.h2.header-table-size <number>
1624 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1625 cannot be larger than 65536 bytes. A larger value may help certain clients
1626 send more compact requests, depending on their capabilities. This amount of
1627 memory is consumed for each HTTP/2 connection. It is recommended not to
1628 change it.
1629
Willy Tarreaue6baec02017-07-27 11:45:11 +02001630tune.h2.initial-window-size <number>
1631 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001632 can upload before waiting for an acknowledgment from haproxy. This setting
1633 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001634 The default value is 65535, which roughly allows up to 5 Mbps of upload
1635 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1636 over a 1-ms local network. It can make sense to increase this value to allow
1637 faster uploads, or to reduce it to increase fairness when dealing with many
1638 clients. It doesn't affect resource usage.
1639
Willy Tarreau5242ef82017-07-27 11:47:28 +02001640tune.h2.max-concurrent-streams <number>
1641 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1642 number of outstanding requests on a single connection). The default value is
1643 100. A larger one may slightly improve page load time for complex sites when
1644 visited over high latency networks, but increases the amount of resources a
1645 single client may allocate. A value of zero disables the limit so a single
1646 client may create as many streams as allocatable by haproxy. It is highly
1647 recommended not to change this value.
1648
Willy Tarreaua24b35c2019-02-21 13:24:36 +01001649tune.h2.max-frame-size <number>
1650 Sets the HTTP/2 maximum frame size that haproxy announces it is willing to
1651 receive to its peers. The default value is the largest between 16384 and the
1652 buffer size (tune.bufsize). In any case, haproxy will not announce support
1653 for frame sizes larger than buffers. The main purpose of this setting is to
1654 allow to limit the maximum frame size setting when using large buffers. Too
1655 large frame sizes might have performance impact or cause some peers to
1656 misbehave. It is highly recommended not to change this value.
1657
Willy Tarreau193b8c62012-11-22 00:17:38 +01001658tune.http.cookielen <number>
1659 Sets the maximum length of captured cookies. This is the maximum value that
1660 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1661 will automatically be truncated to this one. It is important not to set too
1662 high a value because all cookie captures still allocate this size whatever
1663 their configured value (they share a same pool). This value is per request
1664 per response, so the memory allocated is twice this value per connection.
1665 When not specified, the limit is set to 63 characters. It is recommended not
1666 to change this value.
1667
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001668tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001669 Sets the maximum length of request URI in logs. This prevents truncating long
1670 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001671 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001672 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001673 configuration directives too.
1674 The default value is 1024.
1675
Willy Tarreauac1932d2011-10-24 19:14:41 +02001676tune.http.maxhdr <number>
1677 Sets the maximum number of headers in a request. When a request comes with a
1678 number of headers greater than this value (including the first line), it is
1679 rejected with a "400 Bad Request" status code. Similarly, too large responses
1680 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1681 for all usages, considering that the widely deployed Apache server uses the
1682 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001683 a buggy application to work by the time it gets fixed. The accepted range is
1684 1..32767. Keep in mind that each new header consumes 32bits of memory for
1685 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001686
Willy Tarreau7e312732014-02-12 16:35:14 +01001687tune.idletimer <timeout>
1688 Sets the duration after which haproxy will consider that an empty buffer is
1689 probably associated with an idle stream. This is used to optimally adjust
1690 some packet sizes while forwarding large and small data alternatively. The
1691 decision to use splice() or to send large buffers in SSL is modulated by this
1692 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1693 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001694 which seems to correctly detect end user pauses (e.g. read a page before
Willy Tarreau7e312732014-02-12 16:35:14 +01001695 clicking). There should be not reason for changing this value. Please check
1696 tune.ssl.maxrecord below.
1697
Willy Tarreau7ac908b2019-02-27 12:02:18 +01001698tune.listener.multi-queue { on | off }
1699 Enables ('on') or disables ('off') the listener's multi-queue accept which
1700 spreads the incoming traffic to all threads a "bind" line is allowed to run
1701 on instead of taking them for itself. This provides a smoother traffic
1702 distribution and scales much better, especially in environments where threads
1703 may be unevenly loaded due to external activity (network interrupts colliding
1704 with one thread for example). This option is enabled by default, but it may
1705 be forcefully disabled for troubleshooting or for situations where it is
1706 estimated that the operating system already provides a good enough
1707 distribution and connections are extremely short-lived.
1708
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001709tune.lua.forced-yield <number>
1710 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001711 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001712 HAProxy scheduler to process other tasks like accepting connections or
1713 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001714 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001715 lowered. If the Lua code is quite long and its result is absolutely required
1716 to process the data, the <number> can be increased.
1717
Willy Tarreau32f61e22015-03-18 17:54:59 +01001718tune.lua.maxmem
1719 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1720 default it is zero which means unlimited. It is important to set a limit to
1721 ensure that a bug in a script will not result in the system running out of
1722 memory.
1723
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001724tune.lua.session-timeout <timeout>
1725 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001726 preventing infinite loops or spending too much time in Lua. This timeout
1727 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001728 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001729
1730tune.lua.task-timeout <timeout>
1731 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1732 dedicated to the tasks. By default, this timeout isn't set because a task may
1733 remain alive during of the lifetime of HAProxy. For example, a task used to
1734 check servers.
1735
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001736tune.lua.service-timeout <timeout>
1737 This is the execution timeout for the Lua services. This is useful for
1738 preventing infinite loops or spending too much time in Lua. This timeout
1739 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001740 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001741
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001742tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001743 Sets the maximum number of consecutive connections a process may accept in a
1744 row before switching to other work. In single process mode, higher numbers
1745 give better performance at high connection rates. However in multi-process
1746 modes, keeping a bit of fairness between processes generally is better to
1747 increase performance. This value applies individually to each listener, so
1748 that the number of processes a listener is bound to is taken into account.
1749 This value defaults to 64. In multi-process mode, it is divided by twice
1750 the number of processes the listener is bound to. Setting this value to -1
1751 completely disables the limitation. It should normally not be needed to tweak
1752 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001753
1754tune.maxpollevents <number>
1755 Sets the maximum amount of events that can be processed at once in a call to
1756 the polling system. The default value is adapted to the operating system. It
1757 has been noticed that reducing it below 200 tends to slightly decrease
1758 latency at the expense of network bandwidth, and increasing it above 200
1759 tends to trade latency for slightly increased bandwidth.
1760
Willy Tarreau27a674e2009-08-17 07:23:33 +02001761tune.maxrewrite <number>
1762 Sets the reserved buffer space to this size in bytes. The reserved space is
1763 used for header rewriting or appending. The first reads on sockets will never
1764 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1765 bufsize, though that does not make much sense since there are rarely large
1766 numbers of headers to add. Setting it too high prevents processing of large
1767 requests or responses. Setting it too low prevents addition of new headers
1768 to already large requests or to POST requests. It is generally wise to set it
1769 to about 1024. It is automatically readjusted to half of bufsize if it is
1770 larger than that. This means you don't have to worry about it when changing
1771 bufsize.
1772
Willy Tarreauf3045d22015-04-29 16:24:50 +02001773tune.pattern.cache-size <number>
1774 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1775 cache which reminds previous lookups and their results. It is used by ACLs
1776 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1777 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1778 strings. It applies to pattern expressions which means that it will be able
1779 to memorize the result of a lookup among all the patterns specified on a
1780 configuration line (including all those loaded from files). It automatically
1781 invalidates entries which are updated using HTTP actions or on the CLI. The
1782 default cache size is set to 10000 entries, which limits its footprint to
1783 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1784 risk of collision in this cache, which is in the order of the size of the
1785 cache divided by 2^64. Typically, at 10000 requests per second with the
1786 default cache size of 10000 entries, there's 1% chance that a brute force
1787 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1788 This is considered much lower than the risk of a memory corruption caused by
1789 aging components. If this is not acceptable, the cache can be disabled by
1790 setting this parameter to 0.
1791
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001792tune.pipesize <number>
1793 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1794 are the default size for the system. But sometimes when using TCP splicing,
1795 it can improve performance to increase pipe sizes, especially if it is
1796 suspected that pipes are not filled and that many calls to splice() are
1797 performed. This has an impact on the kernel's memory footprint, so this must
1798 not be changed if impacts are not understood.
1799
Olivier Houchard88698d92019-04-16 19:07:22 +02001800tune.pool-low-fd-ratio <number>
1801 This setting sets the max number of file descriptors (in percentage) used by
1802 haproxy globally against the maximum number of file descriptors haproxy can
1803 use before we stop putting connection into the idle pool for reuse. The
1804 default is 20.
1805
1806tune.pool-high-fd-ratio <number>
1807 This setting sets the max number of file descriptors (in percentage) used by
1808 haproxy globally against the maximum number of file descriptors haproxy can
1809 use before we start killing idle connections when we can't reuse a connection
1810 and we have to create a new one. The default is 25 (one quarter of the file
1811 descriptor will mean that roughly half of the maximum front connections can
1812 keep an idle connection behind, anything beyond this probably doesn't make
1813 much sense in the general case when targetting connection reuse).
1814
Willy Tarreaue803de22010-01-21 17:43:04 +01001815tune.rcvbuf.client <number>
1816tune.rcvbuf.server <number>
1817 Forces the kernel socket receive buffer size on the client or the server side
1818 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1819 and backends. It should normally never be set, and the default size (0) lets
1820 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001821 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001822 order to save kernel memory by preventing it from buffering too large amounts
1823 of received data. Lower values will significantly increase CPU usage though.
1824
Willy Tarreaub22fc302015-12-14 12:04:35 +01001825tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001826 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01001827 socket buffers. One of them is that a read returns more than <recv_enough>
1828 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1829 may be changed by this setting to better deal with workloads involving lots
1830 of short messages such as telnet or SSH sessions.
1831
Olivier Houchard1599b802018-05-24 18:59:04 +02001832tune.runqueue-depth <number>
1833 Sets the maxinum amount of task that can be processed at once when running
1834 tasks. The default value is 200. Increasing it may incur latency when
1835 dealing with I/Os, making it too small can incur extra overhead.
1836
Willy Tarreaue803de22010-01-21 17:43:04 +01001837tune.sndbuf.client <number>
1838tune.sndbuf.server <number>
1839 Forces the kernel socket send buffer size on the client or the server side to
1840 the specified value in bytes. This value applies to all TCP/HTTP frontends
1841 and backends. It should normally never be set, and the default size (0) lets
1842 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001843 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001844 order to save kernel memory by preventing it from buffering too large amounts
1845 of received data. Lower values will significantly increase CPU usage though.
1846 Another use case is to prevent write timeouts with extremely slow clients due
1847 to the kernel waiting for a large part of the buffer to be read before
1848 notifying haproxy again.
1849
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001850tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001851 Sets the size of the global SSL session cache, in a number of blocks. A block
1852 is large enough to contain an encoded session without peer certificate.
1853 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001854 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001855 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01001856 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01001857 and reassigned. Higher values reduce the occurrence of such a purge, hence
1858 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1859 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001860 and are shared between all processes if "nbproc" is greater than 1. Setting
1861 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001862
Emeric Brun8dc60392014-05-09 13:52:00 +02001863tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02001864 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02001865 should normally not be used since it will force many renegotiations due to
1866 clients hitting a random process. But it may be required on some operating
1867 systems where none of the SSL cache synchronization method may be used. In
1868 this case, adding a first layer of hash-based load balancing before the SSL
1869 layer might limit the impact of the lack of session sharing.
1870
Emeric Brun4f65bff2012-11-16 15:11:00 +01001871tune.ssl.lifetime <timeout>
1872 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001873 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001874 does not guarantee that sessions will last that long, because if the cache is
1875 full, the longest idle sessions will be purged despite their configured
1876 lifetime. The real usefulness of this setting is to prevent sessions from
1877 being used for too long.
1878
Willy Tarreaubfd59462013-02-21 07:46:09 +01001879tune.ssl.maxrecord <number>
1880 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1881 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1882 data only once it has received a full record. With large records, it means
1883 that clients might have to download up to 16kB of data before starting to
1884 process them. Limiting the value can improve page load times on browsers
1885 located over high latency or low bandwidth networks. It is suggested to find
1886 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1887 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1888 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1889 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001890 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01001891 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001892
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001893tune.ssl.default-dh-param <number>
1894 Sets the maximum size of the Diffie-Hellman parameters used for generating
1895 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1896 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1897 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1898 this maximum value. Default value if 1024. Only 1024 or higher values are
1899 allowed. Higher values will increase the CPU load, and values greater than
1900 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001901 used if static Diffie-Hellman parameters are supplied either directly
1902 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001903
Christopher Faulet31af49d2015-06-09 17:29:50 +02001904tune.ssl.ssl-ctx-cache-size <number>
1905 Sets the size of the cache used to store generated certificates to <number>
1906 entries. This is a LRU cache. Because generating a SSL certificate
1907 dynamically is expensive, they are cached. The default cache size is set to
1908 1000 entries.
1909
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01001910tune.ssl.capture-cipherlist-size <number>
1911 Sets the maximum size of the buffer used for capturing client-hello cipher
1912 list. If the value is 0 (default value) the capture is disabled, otherwise
1913 a buffer is allocated for each SSL/TLS connection.
1914
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001915tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001916tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001917tune.vars.reqres-max-size <size>
1918tune.vars.sess-max-size <size>
1919tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001920 These five tunes help to manage the maximum amount of memory used by the
1921 variables system. "global" limits the overall amount of memory available for
1922 all scopes. "proc" limits the memory for the process scope, "sess" limits the
1923 memory for the session scope, "txn" for the transaction scope, and "reqres"
1924 limits the memory for each request or response processing.
1925 Memory accounting is hierarchical, meaning more coarse grained limits include
1926 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
1927 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001928
Daniel Schneller0b547052016-03-21 20:46:57 +01001929 For example, when "tune.vars.sess-max-size" is limited to 100,
1930 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1931 100 either. If we create a variable "txn.var" that contains 100 bytes,
1932 all available space is consumed.
1933 Notice that exceeding the limits at runtime will not result in an error
1934 message, but values might be cut off or corrupted. So make sure to accurately
1935 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001936
William Lallemanda509e4c2012-11-07 16:54:34 +01001937tune.zlib.memlevel <number>
1938 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001939 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001940 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01001941 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01001942 between 1 and 9. The default value is 8.
1943
1944tune.zlib.windowsize <number>
1945 Sets the window size (the size of the history buffer) as a parameter of the
1946 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01001947 in better compression at the expense of memory usage. Can be a value between
1948 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001949
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019503.3. Debugging
1951--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001952
1953debug
1954 Enables debug mode which dumps to stdout all exchanges, and disables forking
1955 into background. It is the equivalent of the command-line argument "-d". It
1956 should never be used in a production configuration since it may prevent full
1957 system startup.
1958
1959quiet
1960 Do not display any message during startup. It is equivalent to the command-
1961 line argument "-q".
1962
Emeric Brunf099e792010-09-27 12:05:28 +02001963
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010019643.4. Userlists
1965--------------
1966It is possible to control access to frontend/backend/listen sections or to
1967http stats by allowing only authenticated and authorized users. To do this,
1968it is required to create at least one userlist and to define users.
1969
1970userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001971 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001972 used to store authentication & authorization data for independent customers.
1973
1974group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001975 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001976 attach users to this group by using a comma separated list of names
1977 proceeded by "users" keyword.
1978
Cyril Bontéf0c60612010-02-06 14:44:47 +01001979user <username> [password|insecure-password <password>]
1980 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001981 Adds user <username> to the current userlist. Both secure (encrypted) and
1982 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001983 evaluated using the crypt(3) function, so depending on the system's
1984 capabilities, different algorithms are supported. For example, modern Glibc
1985 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
1986 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001987
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001988 Attention: Be aware that using encrypted passwords might cause significantly
1989 increased CPU usage, depending on the number of requests, and the algorithm
1990 used. For any of the hashed variants, the password for each request must
1991 be processed through the chosen algorithm, before it can be compared to the
1992 value specified in the config file. Most current algorithms are deliberately
1993 designed to be expensive to compute to achieve resistance against brute
1994 force attacks. They do not simply salt/hash the clear text password once,
1995 but thousands of times. This can quickly become a major factor in haproxy's
1996 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001997
1998 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001999 userlist L1
2000 group G1 users tiger,scott
2001 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002002
Cyril Bontéf0c60612010-02-06 14:44:47 +01002003 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
2004 user scott insecure-password elgato
2005 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002006
Cyril Bontéf0c60612010-02-06 14:44:47 +01002007 userlist L2
2008 group G1
2009 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002010
Cyril Bontéf0c60612010-02-06 14:44:47 +01002011 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
2012 user scott insecure-password elgato groups G1,G2
2013 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002014
2015 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002016
Emeric Brunf099e792010-09-27 12:05:28 +02002017
20183.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002019----------
Emeric Brun94900952015-06-11 18:25:54 +02002020It is possible to propagate entries of any data-types in stick-tables between
2021several haproxy instances over TCP connections in a multi-master fashion. Each
2022instance pushes its local updates and insertions to remote peers. The pushed
2023values overwrite remote ones without aggregation. Interrupted exchanges are
2024automatically detected and recovered from the last known point.
2025In addition, during a soft restart, the old process connects to the new one
2026using such a TCP connection to push all its entries before the new process
2027tries to connect to other peers. That ensures very fast replication during a
2028reload, it typically takes a fraction of a second even for large tables.
2029Note that Server IDs are used to identify servers remotely, so it is important
2030that configurations look similar or at least that the same IDs are forced on
2031each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002032
2033peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002034 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002035 which is referenced by one or more stick-tables.
2036
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002037bind [<address>]:<port_range> [, ...] [param*]
2038 Defines the binding parameters of the local peer of this "peers" section.
2039 Such lines are not supported with "peer" line in the same "peers" section.
2040
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002041disabled
2042 Disables a peers section. It disables both listening and any synchronization
2043 related to this section. This is provided to disable synchronization of stick
2044 tables without having to comment out all "peers" references.
2045
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002046default-bind [param*]
2047 Defines the binding parameters for the local peer, excepted its address.
2048
2049default-server [param*]
2050 Change default options for a server in a "peers" section.
2051
2052 Arguments:
2053 <param*> is a list of parameters for this server. The "default-server"
2054 keyword accepts an important number of options and has a complete
2055 section dedicated to it. Please refer to section 5 for more
2056 details.
2057
2058
2059 See also: "server" and section 5 about server options
2060
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002061enable
2062 This re-enables a disabled peers section which was previously disabled.
2063
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002064peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002065 Defines a peer inside a peers section.
2066 If <peername> is set to the local peer name (by default hostname, or forced
2067 using "-L" command line option), haproxy will listen for incoming remote peer
2068 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
2069 to join the remote peer, and <peername> is used at the protocol level to
2070 identify and validate the remote peer on the server side.
2071
2072 During a soft restart, local peer <ip>:<port> is used by the old instance to
2073 connect the new one and initiate a complete replication (teaching process).
2074
2075 It is strongly recommended to have the exact same peers declaration on all
2076 peers and to only rely on the "-L" command line argument to change the local
2077 peer name. This makes it easier to maintain coherent configuration files
2078 across all peers.
2079
William Lallemandb2f07452015-05-12 14:27:13 +02002080 You may want to reference some environment variables in the address
2081 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01002082
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002083 Note: "peer" keyword may transparently be replaced by "server" keyword (see
2084 "server" keyword explanation below).
2085
2086server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02002087 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002088 with a support for all "server" parameters found in 5.2 paragraph.
2089 If the underlying peer is local, <ip>:<port> parameters must not be present.
2090 These parameters must be provided on a "bind" line (see "bind" keyword
2091 of this "peers" section).
2092 Some of these parameters are irrelevant for "peers" sections.
2093
2094
Cyril Bontédc4d9032012-04-08 21:57:39 +02002095 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002096 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02002097 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002098 peer haproxy1 192.168.0.1:1024
2099 peer haproxy2 192.168.0.2:1024
2100 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02002101
2102 backend mybackend
2103 mode tcp
2104 balance roundrobin
2105 stick-table type ip size 20k peers mypeers
2106 stick on src
2107
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002108 server srv1 192.168.0.30:80
2109 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02002110
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002111 Example:
2112 peers mypeers
2113 bind 127.0.0.11:10001 ssl crt mycerts/pem
2114 default-server ssl verify none
2115 server hostA 127.0.0.10:10000
2116 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02002117
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002118
2119table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
2120 size <size> [expire <expire>] [nopurge] [store <data_type>]*
2121
2122 Configure a stickiness table for the current section. This line is parsed
2123 exactly the same way as the "stick-table" keyword in others section, except
2124 for the "peers" argument which is not required here and with an aditionnal
2125 mandatory first parameter to designate the stick-table. Contrary to others
2126 sections, there may be several "table" lines in "peers" sections (see also
2127 "stick-table" keyword).
2128
2129 Also be aware of the fact that "peers" sections have their own stick-table
2130 namespaces to avoid collisions between stick-table names identical in
2131 different "peers" section. This is internally handled prepending the "peers"
2132 sections names to the name of the stick-tables followed by a '/' character.
2133 If somewhere else in the configuration file you have to refer to such
2134 stick-tables declared in "peers" sections you must use the prefixed version
2135 of the stick-table name as follows:
2136
2137 peers mypeers
2138 peer A ...
2139 peer B ...
2140 table t1 ...
2141
2142 frontend fe1
2143 tcp-request content track-sc0 src table mypeers/t1
2144
2145 This is also this prefixed version of the stick-table names which must be
2146 used to refer to stick-tables through the CLI.
2147
2148 About "peers" protocol, as only "peers" belonging to the same section may
2149 communicate with each others, there is no need to do such a distinction.
2150 Several "peers" sections may declare stick-tables with the same name.
2151 This is shorter version of the stick-table name which is sent over the network.
2152 There is only a '/' character as prefix to avoid stick-table name collisions between
2153 stick-tables declared as backends and stick-table declared in "peers" sections
2154 as follows in this weird but supported configuration:
2155
2156 peers mypeers
2157 peer A ...
2158 peer B ...
2159 table t1 type string size 10m store gpc0
2160
2161 backend t1
2162 stick-table type string size 10m store gpc0 peers mypeers
2163
2164 Here "t1" table declared in "mypeeers" section has "mypeers/t1" as global name.
2165 "t1" table declared as a backend as "t1" as global name. But at peer protocol
2166 level the former table is named "/t1", the latter is again named "t1".
2167
Simon Horman51a1cf62015-02-03 13:00:44 +090021683.6. Mailers
2169------------
2170It is possible to send email alerts when the state of servers changes.
2171If configured email alerts are sent to each mailer that is configured
2172in a mailers section. Email is sent to mailers using SMTP.
2173
Pieter Baauw386a1272015-08-16 15:26:24 +02002174mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09002175 Creates a new mailer list with the name <mailersect>. It is an
2176 independent section which is referenced by one or more proxies.
2177
2178mailer <mailername> <ip>:<port>
2179 Defines a mailer inside a mailers section.
2180
2181 Example:
2182 mailers mymailers
2183 mailer smtp1 192.168.0.1:587
2184 mailer smtp2 192.168.0.2:587
2185
2186 backend mybackend
2187 mode tcp
2188 balance roundrobin
2189
2190 email-alert mailers mymailers
2191 email-alert from test1@horms.org
2192 email-alert to test2@horms.org
2193
2194 server srv1 192.168.0.30:80
2195 server srv2 192.168.0.31:80
2196
Pieter Baauw235fcfc2016-02-13 15:33:40 +01002197timeout mail <time>
2198 Defines the time available for a mail/connection to be made and send to
2199 the mail-server. If not defined the default value is 10 seconds. To allow
2200 for at least two SYN-ACK packets to be send during initial TCP handshake it
2201 is advised to keep this value above 4 seconds.
2202
2203 Example:
2204 mailers mymailers
2205 timeout mail 20s
2206 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09002207
William Lallemandc9515522019-06-12 16:32:11 +020022083.7. Programs
2209-------------
2210In master-worker mode, it is possible to launch external binaries with the
2211master, these processes are called programs. These programs are launched and
2212managed the same way as the workers.
2213
2214During a reload of HAProxy, those processes are dealing with the same
2215sequence as a worker:
2216
2217 - the master is re-executed
2218 - the master sends a SIGUSR1 signal to the program
2219 - if "option start-on-reload" is not disabled, the master launches a new
2220 instance of the program
2221
2222During a stop, or restart, a SIGTERM is sent to the programs.
2223
2224program <name>
2225 This is a new program section, this section will create an instance <name>
2226 which is visible in "show proc" on the master CLI. (See "9.4. Master CLI" in
2227 the management guide).
2228
2229command <command> [arguments*]
2230 Define the command to start with optional arguments. The command is looked
2231 up in the current PATH if it does not include an absolute path. This is a
2232 mandatory option of the program section. Arguments containing spaces must
2233 be enclosed in quotes or double quotes or be prefixed by a backslash.
2234
2235option start-on-reload
2236no option start-on-reload
2237 Start (or not) a new instance of the program upon a reload of the master.
2238 The default is to start a new instance. This option may only be used in a
2239 program section.
2240
2241
Willy Tarreauc57f0e22009-05-10 13:12:33 +020022424. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02002243----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002244
Willy Tarreau6a06a402007-07-15 20:15:28 +02002245Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02002246 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02002247 - frontend <name>
2248 - backend <name>
2249 - listen <name>
2250
2251A "defaults" section sets default parameters for all other sections following
2252its declaration. Those default parameters are reset by the next "defaults"
2253section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01002254section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002255
2256A "frontend" section describes a set of listening sockets accepting client
2257connections.
2258
2259A "backend" section describes a set of servers to which the proxy will connect
2260to forward incoming connections.
2261
2262A "listen" section defines a complete proxy with its frontend and backend
2263parts combined in one section. It is generally useful for TCP-only traffic.
2264
Willy Tarreau0ba27502007-12-24 16:55:16 +01002265All proxy names must be formed from upper and lower case letters, digits,
2266'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
2267case-sensitive, which means that "www" and "WWW" are two different proxies.
2268
2269Historically, all proxy names could overlap, it just caused troubles in the
2270logs. Since the introduction of content switching, it is mandatory that two
2271proxies with overlapping capabilities (frontend/backend) have different names.
2272However, it is still permitted that a frontend and a backend share the same
2273name, as this configuration seems to be commonly encountered.
2274
2275Right now, two major proxy modes are supported : "tcp", also known as layer 4,
2276and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002277bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002278protocol, and can interact with it by allowing, blocking, switching, adding,
2279modifying, or removing arbitrary contents in requests or responses, based on
2280arbitrary criteria.
2281
Willy Tarreau70dffda2014-01-30 03:07:23 +01002282In HTTP mode, the processing applied to requests and responses flowing over
2283a connection depends in the combination of the frontend's HTTP options and
Christopher Faulet315b39c2018-09-21 16:26:19 +02002284the backend's. HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01002285
2286 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
2287 requests and responses are processed, and connections remain open but idle
2288 between responses and new requests.
2289
2290 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
2291 1.0 to 1.5-dev21 : only the first request and response are processed, and
2292 everything else is forwarded with no analysis at all. This mode should not
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002293 be used as it creates lots of trouble with logging and HTTP processing.
2294 And because it cannot work in HTTP/2, this option is deprecated and it is
2295 only supported on legacy HTTP frontends. In HTX, it is ignored and a
2296 warning is emitted during HAProxy startup.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002297
Willy Tarreau70dffda2014-01-30 03:07:23 +01002298 - SCL: server close ("option http-server-close") : the server-facing
2299 connection is closed after the end of the response is received, but the
2300 client-facing connection remains open.
2301
Christopher Faulet315b39c2018-09-21 16:26:19 +02002302 - CLO: close ("option httpclose"): the connection is closed after the end of
2303 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002304
2305The effective mode that will be applied to a connection passing through a
2306frontend and a backend can be determined by both proxy modes according to the
2307following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02002308weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002309
Christopher Faulet315b39c2018-09-21 16:26:19 +02002310 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01002311
Christopher Faulet315b39c2018-09-21 16:26:19 +02002312 | KAL | SCL | CLO
2313 ----+-----+-----+----
2314 KAL | KAL | SCL | CLO
2315 ----+-----+-----+----
2316 TUN | TUN | SCL | CLO
2317 Frontend ----+-----+-----+----
2318 mode SCL | SCL | SCL | CLO
2319 ----+-----+-----+----
2320 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01002321
Willy Tarreau0ba27502007-12-24 16:55:16 +01002322
Willy Tarreau70dffda2014-01-30 03:07:23 +01002323
Willy Tarreauc57f0e22009-05-10 13:12:33 +020023244.1. Proxy keywords matrix
2325--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002326
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002327The following list of keywords is supported. Most of them may only be used in a
2328limited set of section types. Some of them are marked as "deprecated" because
2329they are inherited from an old syntax which may be confusing or functionally
2330limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002331marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002332option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002333and must be disabled for a specific instance. Such options may also be prefixed
2334with "default" in order to restore default settings regardless of what has been
2335specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002336
Willy Tarreau6a06a402007-07-15 20:15:28 +02002337
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002338 keyword defaults frontend listen backend
2339------------------------------------+----------+----------+---------+---------
2340acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002341backlog X X X -
2342balance X - X X
2343bind - X X -
2344bind-process X X X X
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002345block (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002346capture cookie - X X -
2347capture request header - X X -
2348capture response header - X X -
2349clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002350compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002351contimeout (deprecated) X - X X
2352cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002353declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002354default-server X - X X
2355default_backend X X X -
2356description - X X X
2357disabled X X X X
2358dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002359email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002360email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002361email-alert mailers X X X X
2362email-alert myhostname X X X X
2363email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002364enabled X X X X
2365errorfile X X X X
2366errorloc X X X X
2367errorloc302 X X X X
2368-- keyword -------------------------- defaults - frontend - listen -- backend -
2369errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002370force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002371filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002372fullconn X - X X
2373grace X X X X
2374hash-type X - X X
2375http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002376http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002377http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002378http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002379http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002380http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002381http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002382id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002383ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002384load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002385log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002386log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002387log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002388log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002389max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002390maxconn X X X -
2391mode X X X X
2392monitor fail - X X -
2393monitor-net X X X -
2394monitor-uri X X X -
2395option abortonclose (*) X - X X
2396option accept-invalid-http-request (*) X X X -
2397option accept-invalid-http-response (*) X - X X
2398option allbackups (*) X - X X
2399option checkcache (*) X - X X
2400option clitcpka (*) X X X -
2401option contstats (*) X X X -
2402option dontlog-normal (*) X X X -
2403option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002404-- keyword -------------------------- defaults - frontend - listen -- backend -
2405option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002406option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002407option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002408option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002409option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02002410option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002411option http-server-close (*) X X X X
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002412option http-tunnel (deprecated) (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002413option http-use-proxy-header (*) X X X -
Willy Tarreau68ad3a42018-10-22 11:49:15 +02002414option http-use-htx (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002415option httpchk X - X X
2416option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01002417option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002418option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002419option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002420option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002421option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002422option log-health-checks (*) X - X X
2423option log-separate-errors (*) X X X -
2424option logasap (*) X X X -
2425option mysql-check X - X X
2426option nolinger (*) X X X X
2427option originalto X X X X
2428option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002429option pgsql-check X - X X
2430option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002431option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002432option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002433option smtpchk X - X X
2434option socket-stats (*) X X X -
2435option splice-auto (*) X X X X
2436option splice-request (*) X X X X
2437option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002438option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002439option srvtcpka (*) X - X X
2440option ssl-hello-chk X - X X
2441-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002442option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002443option tcp-smart-accept (*) X X X -
2444option tcp-smart-connect (*) X - X X
2445option tcpka X X X X
2446option tcplog X X X X
2447option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002448external-check command X - X X
2449external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002450persist rdp-cookie X - X X
2451rate-limit sessions X X X -
2452redirect - X X X
2453redisp (deprecated) X - X X
2454redispatch (deprecated) X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002455reqadd (deprecated) - X X X
2456reqallow (deprecated) - X X X
2457reqdel (deprecated) - X X X
2458reqdeny (deprecated) - X X X
2459reqiallow (deprecated) - X X X
2460reqidel (deprecated) - X X X
2461reqideny (deprecated) - X X X
2462reqipass (deprecated) - X X X
2463reqirep (deprecated) - X X X
2464reqitarpit (deprecated) - X X X
2465reqpass (deprecated) - X X X
2466reqrep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002467-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau96d51952019-05-22 20:34:35 +02002468reqtarpit (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002469retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02002470retry-on X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002471rspadd (deprecated) - X X X
2472rspdel (deprecated) - X X X
2473rspdeny (deprecated) - X X X
2474rspidel (deprecated) - X X X
2475rspideny (deprecated) - X X X
2476rspirep (deprecated) - X X X
2477rsprep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002478server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002479server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002480server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002481source X - X X
2482srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002483stats admin - X X X
2484stats auth X X X X
2485stats enable X X X X
2486stats hide-version X X X X
2487stats http-request - X X X
2488stats realm X X X X
2489stats refresh X X X X
2490stats scope X X X X
2491stats show-desc X X X X
2492stats show-legends X X X X
2493stats show-node X X X X
2494stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002495-- keyword -------------------------- defaults - frontend - listen -- backend -
2496stick match - - X X
2497stick on - - X X
2498stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002499stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002500stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002501tcp-check connect - - X X
2502tcp-check expect - - X X
2503tcp-check send - - X X
2504tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002505tcp-request connection - X X -
2506tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002507tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002508tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002509tcp-response content - - X X
2510tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002511timeout check X - X X
2512timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002513timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002514timeout clitimeout (deprecated) X X X -
2515timeout connect X - X X
2516timeout contimeout (deprecated) X - X X
2517timeout http-keep-alive X X X X
2518timeout http-request X X X X
2519timeout queue X - X X
2520timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002521timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002522timeout srvtimeout (deprecated) X - X X
2523timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002524timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002525transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002526unique-id-format X X X -
2527unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002528use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002529use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002530------------------------------------+----------+----------+---------+---------
2531 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002532
Willy Tarreau0ba27502007-12-24 16:55:16 +01002533
Willy Tarreauc57f0e22009-05-10 13:12:33 +020025344.2. Alphabetically sorted keywords reference
2535---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002536
2537This section provides a description of each keyword and its usage.
2538
2539
2540acl <aclname> <criterion> [flags] [operator] <value> ...
2541 Declare or complete an access list.
2542 May be used in sections : defaults | frontend | listen | backend
2543 no | yes | yes | yes
2544 Example:
2545 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2546 acl invalid_src src_port 0:1023
2547 acl local_dst hdr(host) -i localhost
2548
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002549 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002550
2551
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002552backlog <conns>
2553 Give hints to the system about the approximate listen backlog desired size
2554 May be used in sections : defaults | frontend | listen | backend
2555 yes | yes | yes | no
2556 Arguments :
2557 <conns> is the number of pending connections. Depending on the operating
2558 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002559 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002560
2561 In order to protect against SYN flood attacks, one solution is to increase
2562 the system's SYN backlog size. Depending on the system, sometimes it is just
2563 tunable via a system parameter, sometimes it is not adjustable at all, and
2564 sometimes the system relies on hints given by the application at the time of
2565 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2566 to the listen() syscall. On systems which can make use of this value, it can
2567 sometimes be useful to be able to specify a different value, hence this
2568 backlog parameter.
2569
2570 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2571 used as a hint and the system accepts up to the smallest greater power of
2572 two, and never more than some limits (usually 32768).
2573
2574 See also : "maxconn" and the target operating system's tuning guide.
2575
2576
Willy Tarreau0ba27502007-12-24 16:55:16 +01002577balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002578balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002579 Define the load balancing algorithm to be used in a backend.
2580 May be used in sections : defaults | frontend | listen | backend
2581 yes | no | yes | yes
2582 Arguments :
2583 <algorithm> is the algorithm used to select a server when doing load
2584 balancing. This only applies when no persistence information
2585 is available, or when a connection is redispatched to another
2586 server. <algorithm> may be one of the following :
2587
2588 roundrobin Each server is used in turns, according to their weights.
2589 This is the smoothest and fairest algorithm when the server's
2590 processing time remains equally distributed. This algorithm
2591 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002592 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002593 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002594 large farms, when a server becomes up after having been down
2595 for a very short time, it may sometimes take a few hundreds
2596 requests for it to be re-integrated into the farm and start
2597 receiving traffic. This is normal, though very rare. It is
2598 indicated here in case you would have the chance to observe
2599 it, so that you don't worry.
2600
2601 static-rr Each server is used in turns, according to their weights.
2602 This algorithm is as similar to roundrobin except that it is
2603 static, which means that changing a server's weight on the
2604 fly will have no effect. On the other hand, it has no design
2605 limitation on the number of servers, and when a server goes
2606 up, it is always immediately reintroduced into the farm, once
2607 the full map is recomputed. It also uses slightly less CPU to
2608 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002609
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002610 leastconn The server with the lowest number of connections receives the
2611 connection. Round-robin is performed within groups of servers
2612 of the same load to ensure that all servers will be used. Use
2613 of this algorithm is recommended where very long sessions are
2614 expected, such as LDAP, SQL, TSE, etc... but is not very well
2615 suited for protocols using short sessions such as HTTP. This
2616 algorithm is dynamic, which means that server weights may be
2617 adjusted on the fly for slow starts for instance.
2618
Willy Tarreauf09c6602012-02-13 17:12:08 +01002619 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002620 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002621 identifier to the highest (see server parameter "id"), which
2622 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002623 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002624 not make sense to use this algorithm without setting maxconn.
2625 The purpose of this algorithm is to always use the smallest
2626 number of servers so that extra servers can be powered off
2627 during non-intensive hours. This algorithm ignores the server
2628 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002629 or IMAP than HTTP, though it can be useful there too. In
2630 order to use this algorithm efficiently, it is recommended
2631 that a cloud controller regularly checks server usage to turn
2632 them off when unused, and regularly checks backend queue to
2633 turn new servers on when the queue inflates. Alternatively,
2634 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002635
Willy Tarreau0ba27502007-12-24 16:55:16 +01002636 source The source IP address is hashed and divided by the total
2637 weight of the running servers to designate which server will
2638 receive the request. This ensures that the same client IP
2639 address will always reach the same server as long as no
2640 server goes down or up. If the hash result changes due to the
2641 number of running servers changing, many clients will be
2642 directed to a different server. This algorithm is generally
2643 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002644 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002645 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002646 static by default, which means that changing a server's
2647 weight on the fly will have no effect, but this can be
2648 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002649
Oskar Stolc8dc41842012-05-19 10:19:54 +01002650 uri This algorithm hashes either the left part of the URI (before
2651 the question mark) or the whole URI (if the "whole" parameter
2652 is present) and divides the hash value by the total weight of
2653 the running servers. The result designates which server will
2654 receive the request. This ensures that the same URI will
2655 always be directed to the same server as long as no server
2656 goes up or down. This is used with proxy caches and
2657 anti-virus proxies in order to maximize the cache hit rate.
2658 Note that this algorithm may only be used in an HTTP backend.
2659 This algorithm is static by default, which means that
2660 changing a server's weight on the fly will have no effect,
2661 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002662
Oskar Stolc8dc41842012-05-19 10:19:54 +01002663 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002664 "depth", both followed by a positive integer number. These
2665 options may be helpful when it is needed to balance servers
2666 based on the beginning of the URI only. The "len" parameter
2667 indicates that the algorithm should only consider that many
2668 characters at the beginning of the URI to compute the hash.
2669 Note that having "len" set to 1 rarely makes sense since most
2670 URIs start with a leading "/".
2671
2672 The "depth" parameter indicates the maximum directory depth
2673 to be used to compute the hash. One level is counted for each
2674 slash in the request. If both parameters are specified, the
2675 evaluation stops when either is reached.
2676
Willy Tarreau0ba27502007-12-24 16:55:16 +01002677 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002678 the query string of each HTTP GET request.
2679
2680 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002681 request entity will be searched for the parameter argument,
2682 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002683 ('?') in the URL. The message body will only start to be
2684 analyzed once either the advertised amount of data has been
2685 received or the request buffer is full. In the unlikely event
2686 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002687 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002688 be randomly balanced if at all. This keyword used to support
2689 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002690
2691 If the parameter is found followed by an equal sign ('=') and
2692 a value, then the value is hashed and divided by the total
2693 weight of the running servers. The result designates which
2694 server will receive the request.
2695
2696 This is used to track user identifiers in requests and ensure
2697 that a same user ID will always be sent to the same server as
2698 long as no server goes up or down. If no value is found or if
2699 the parameter is not found, then a round robin algorithm is
2700 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002701 backend. This algorithm is static by default, which means
2702 that changing a server's weight on the fly will have no
2703 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002704
Cyril Bontédc4d9032012-04-08 21:57:39 +02002705 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2706 request. Just as with the equivalent ACL 'hdr()' function,
2707 the header name in parenthesis is not case sensitive. If the
2708 header is absent or if it does not contain any value, the
2709 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002710
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002711 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002712 reducing the hash algorithm to the main domain part with some
2713 specific headers such as 'Host'. For instance, in the Host
2714 value "haproxy.1wt.eu", only "1wt" will be considered.
2715
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002716 This algorithm is static by default, which means that
2717 changing a server's weight on the fly will have no effect,
2718 but this can be changed using "hash-type".
2719
Willy Tarreau21c741a2019-01-14 18:14:27 +01002720 random
2721 random(<draws>)
2722 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02002723 hashing function. This means that the servers' weights are
2724 respected, dynamic weight changes immediately take effect, as
2725 well as new server additions. Random load balancing can be
2726 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01002727 or removed as it may avoid the hammering effect that could
2728 result from roundrobin or leastconn in this situation. The
2729 hash-balance-factor directive can be used to further improve
2730 fairness of the load balancing, especially in situations
2731 where servers show highly variable response times. When an
2732 argument <draws> is present, it must be an integer value one
2733 or greater, indicating the number of draws before selecting
2734 the least loaded of these servers. It was indeed demonstrated
2735 that picking the least loaded of two servers is enough to
2736 significantly improve the fairness of the algorithm, by
2737 always avoiding to pick the most loaded server within a farm
2738 and getting rid of any bias that could be induced by the
2739 unfair distribution of the consistent list. Higher values N
2740 will take away N-1 of the highest loaded servers at the
2741 expense of performance. With very high values, the algorithm
2742 will converge towards the leastconn's result but much slower.
2743 The default value is 2, which generally shows very good
2744 distribution and performance. This algorithm is also known as
2745 the Power of Two Random Choices and is described here :
2746 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02002747
Emeric Brun736aa232009-06-30 17:56:00 +02002748 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002749 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002750 The RDP cookie <name> (or "mstshash" if omitted) will be
2751 looked up and hashed for each incoming TCP request. Just as
2752 with the equivalent ACL 'req_rdp_cookie()' function, the name
2753 is not case-sensitive. This mechanism is useful as a degraded
2754 persistence mode, as it makes it possible to always send the
2755 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002756 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002757 used instead.
2758
2759 Note that for this to work, the frontend must ensure that an
2760 RDP cookie is already present in the request buffer. For this
2761 you must use 'tcp-request content accept' rule combined with
2762 a 'req_rdp_cookie_cnt' ACL.
2763
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002764 This algorithm is static by default, which means that
2765 changing a server's weight on the fly will have no effect,
2766 but this can be changed using "hash-type".
2767
Cyril Bontédc4d9032012-04-08 21:57:39 +02002768 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002769
Willy Tarreau0ba27502007-12-24 16:55:16 +01002770 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002771 algorithms. Right now, only "url_param" and "uri" support an
2772 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002773
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002774 The load balancing algorithm of a backend is set to roundrobin when no other
2775 algorithm, mode nor option have been set. The algorithm may only be set once
2776 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002777
Lukas Tribus80512b12018-10-27 20:07:40 +02002778 With authentication schemes that require the same connection like NTLM, URI
2779 based alghoritms must not be used, as they would cause subsequent requests
2780 to be routed to different backend servers, breaking the invalid assumptions
2781 NTLM relies on.
2782
Willy Tarreau0ba27502007-12-24 16:55:16 +01002783 Examples :
2784 balance roundrobin
2785 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002786 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002787 balance hdr(User-Agent)
2788 balance hdr(host)
2789 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002790
2791 Note: the following caveats and limitations on using the "check_post"
2792 extension with "url_param" must be considered :
2793
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002794 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002795 to determine if the parameters will be found in the body or entity which
2796 may contain binary data. Therefore another method may be required to
2797 restrict consideration of POST requests that have no URL parameters in
2798 the body. (see acl reqideny http_end)
2799
2800 - using a <max_wait> value larger than the request buffer size does not
2801 make sense and is useless. The buffer size is set at build time, and
2802 defaults to 16 kB.
2803
2804 - Content-Encoding is not supported, the parameter search will probably
2805 fail; and load balancing will fall back to Round Robin.
2806
2807 - Expect: 100-continue is not supported, load balancing will fall back to
2808 Round Robin.
2809
Lukas Tribus23953682017-04-28 13:24:30 +00002810 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002811 If the entire parameter value is not present in the first chunk, the
2812 selection of server is undefined (actually, defined by how little
2813 actually appeared in the first chunk).
2814
2815 - This feature does not support generation of a 100, 411 or 501 response.
2816
2817 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002818 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002819 white space or control characters are found, indicating the end of what
2820 might be a URL parameter list. This is probably not a concern with SGML
2821 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002822
Willy Tarreau294d0f02015-08-10 19:40:12 +02002823 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002824
2825
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002826bind [<address>]:<port_range> [, ...] [param*]
2827bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002828 Define one or several listening addresses and/or ports in a frontend.
2829 May be used in sections : defaults | frontend | listen | backend
2830 no | yes | yes | no
2831 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002832 <address> is optional and can be a host name, an IPv4 address, an IPv6
2833 address, or '*'. It designates the address the frontend will
2834 listen on. If unset, all IPv4 addresses of the system will be
2835 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002836 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002837 Optionally, an address family prefix may be used before the
2838 address to force the family regardless of the address format,
2839 which can be useful to specify a path to a unix socket with
2840 no slash ('/'). Currently supported prefixes are :
2841 - 'ipv4@' -> address is always IPv4
2842 - 'ipv6@' -> address is always IPv6
2843 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002844 - 'abns@' -> address is in abstract namespace (Linux only).
2845 Note: since abstract sockets are not "rebindable", they
2846 do not cope well with multi-process mode during
2847 soft-restart, so it is better to avoid them if
2848 nbproc is greater than 1. The effect is that if the
2849 new process fails to start, only one of the old ones
2850 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002851 - 'fd@<n>' -> use file descriptor <n> inherited from the
2852 parent. The fd must be bound and may or may not already
2853 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02002854 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
2855 connected unix socket or of a socketpair. The bind waits
2856 to receive a FD over the unix socket and uses it as if it
2857 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02002858 You may want to reference some environment variables in the
2859 address parameter, see section 2.3 about environment
2860 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002861
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002862 <port_range> is either a unique TCP port, or a port range for which the
2863 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002864 above. The port is mandatory for TCP listeners. Note that in
2865 the case of an IPv6 address, the port is always the number
2866 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002867 - a numerical port (ex: '80')
2868 - a dash-delimited ports range explicitly stating the lower
2869 and upper bounds (ex: '2000-2100') which are included in
2870 the range.
2871
2872 Particular care must be taken against port ranges, because
2873 every <address:port> couple consumes one socket (= a file
2874 descriptor), so it's easy to consume lots of descriptors
2875 with a simple range, and to run out of sockets. Also, each
2876 <address:port> couple must be used only once among all
2877 instances running on a same system. Please note that binding
2878 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002879 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002880 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002881
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002882 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002883 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002884 receive UNIX connections on the socket located at this place.
2885 The path must begin with a slash and by default is absolute.
2886 It can be relative to the prefix defined by "unix-bind" in
2887 the global section. Note that the total length of the prefix
2888 followed by the socket path cannot exceed some system limits
2889 for UNIX sockets, which commonly are set to 107 characters.
2890
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002891 <param*> is a list of parameters common to all sockets declared on the
2892 same line. These numerous parameters depend on OS and build
2893 options and have a complete section dedicated to them. Please
2894 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002895
Willy Tarreau0ba27502007-12-24 16:55:16 +01002896 It is possible to specify a list of address:port combinations delimited by
2897 commas. The frontend will then listen on all of these addresses. There is no
2898 fixed limit to the number of addresses and ports which can be listened on in
2899 a frontend, as well as there is no limit to the number of "bind" statements
2900 in a frontend.
2901
2902 Example :
2903 listen http_proxy
2904 bind :80,:443
2905 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002906 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002907
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002908 listen http_https_proxy
2909 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002910 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002911
Willy Tarreau24709282013-03-10 21:32:12 +01002912 listen http_https_proxy_explicit
2913 bind ipv6@:80
2914 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2915 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2916
Willy Tarreaudad36a32013-03-11 01:20:04 +01002917 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002918 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002919
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002920 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2921 sun_path length is used for the address length. Some other programs
2922 such as socat use the string length only by default. Pass the option
2923 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2924 make it compatible with HAProxy's.
2925
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002926 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002927 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002928
2929
Christopher Fauletff4121f2017-11-22 16:38:49 +01002930bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002931 Limit visibility of an instance to a certain set of processes numbers.
2932 May be used in sections : defaults | frontend | listen | backend
2933 yes | yes | yes | yes
2934 Arguments :
2935 all All process will see this instance. This is the default. It
2936 may be used to override a default value.
2937
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002938 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002939 option may be combined with other numbers.
2940
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002941 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002942 option may be combined with other numbers. Do not use it
2943 with less than 2 processes otherwise some instances might be
2944 missing from all processes.
2945
Christopher Fauletff4121f2017-11-22 16:38:49 +01002946 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002947 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01002948 the machine's word size. Ranges can be partially defined. The
2949 higher bound can be omitted. In such case, it is replaced by
2950 the corresponding maximum value. If a proxy is bound to
2951 process numbers greater than the configured global.nbproc, it
2952 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02002953 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002954
2955 This keyword limits binding of certain instances to certain processes. This
2956 is useful in order not to have too many processes listening to the same
2957 ports. For instance, on a dual-core machine, it might make sense to set
2958 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2959 and 'even' instances.
2960
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002961 At the moment, it is not possible to reference more than 32 or 64 processes
2962 using this keyword, but this should be more than enough for most setups.
2963 Please note that 'all' really means all processes regardless of the machine's
2964 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002965
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002966 Each "bind" line may further be limited to a subset of the proxy's processes,
2967 please consult the "process" bind keyword in section 5.1.
2968
Willy Tarreaub369a042014-09-16 13:21:03 +02002969 When a frontend has no explicit "bind-process" line, it tries to bind to all
2970 the processes referenced by its "bind" lines. That means that frontends can
2971 easily adapt to their listeners' processes.
2972
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002973 If some backends are referenced by frontends bound to other processes, the
2974 backend automatically inherits the frontend's processes.
2975
2976 Example :
2977 listen app_ip1
2978 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002979 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002980
2981 listen app_ip2
2982 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002983 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002984
2985 listen management
2986 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002987 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002988
Willy Tarreau110ecc12012-11-15 17:50:01 +01002989 listen management
2990 bind 10.0.0.4:80
2991 bind-process 1-4
2992
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002993 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002994
2995
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002996block { if | unless } <condition> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002997 Block a layer 7 request if/unless a condition is matched
2998 May be used in sections : defaults | frontend | listen | backend
2999 no | yes | yes | yes
3000
3001 The HTTP request will be blocked very early in the layer 7 processing
3002 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003003 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02003004 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01003005 conditions are met or not met. There is no fixed limit to the number of
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03003006 "block" statements per instance. To block connections at layer 4 (without
3007 sending a 403 error) see "tcp-request connection reject" and
3008 "tcp-request content reject" rules.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003009
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02003010 This form is deprecated, do not use it in any new configuration, use the new
3011 "http-request deny" instead.
3012
Willy Tarreau0ba27502007-12-24 16:55:16 +01003013 Example:
3014 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
3015 acl invalid_src src_port 0:1023
3016 acl local_dst hdr(host) -i localhost
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +03003017 # block is deprecated. Use http-request deny instead:
3018 #block if invalid_src || local_dst
3019 http-request deny if invalid_src || local_dst
Willy Tarreau0ba27502007-12-24 16:55:16 +01003020
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03003021 See also : section 7 about ACL usage, "http-request deny",
3022 "http-response deny", "tcp-request connection reject" and
3023 "tcp-request content reject".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003024
3025capture cookie <name> len <length>
3026 Capture and log a cookie in the request and in the response.
3027 May be used in sections : defaults | frontend | listen | backend
3028 no | yes | yes | no
3029 Arguments :
3030 <name> is the beginning of the name of the cookie to capture. In order
3031 to match the exact name, simply suffix the name with an equal
3032 sign ('='). The full name will appear in the logs, which is
3033 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01003034 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003035
3036 <length> is the maximum number of characters to report in the logs, which
3037 include the cookie name, the equal sign and the value, all in the
3038 standard "name=value" form. The string will be truncated on the
3039 right if it exceeds <length>.
3040
3041 Only the first cookie is captured. Both the "cookie" request headers and the
3042 "set-cookie" response headers are monitored. This is particularly useful to
3043 check for application bugs causing session crossing or stealing between
3044 users, because generally the user's cookies can only change on a login page.
3045
3046 When the cookie was not presented by the client, the associated log column
3047 will report "-". When a request does not cause a cookie to be assigned by the
3048 server, a "-" is reported in the response column.
3049
3050 The capture is performed in the frontend only because it is necessary that
3051 the log format does not change for a given frontend depending on the
3052 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01003053 "capture cookie" statement in a frontend. The maximum capture length is set
3054 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
3055 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003056
3057 Example:
3058 capture cookie ASPSESSION len 32
3059
3060 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003061 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003062
3063
3064capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003065 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003066 May be used in sections : defaults | frontend | listen | backend
3067 no | yes | yes | no
3068 Arguments :
3069 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003070 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003071 appear in the requests, with the first letter of each word in
3072 upper case. The header name will not appear in the logs, only the
3073 value is reported, but the position in the logs is respected.
3074
3075 <length> is the maximum number of characters to extract from the value and
3076 report in the logs. The string will be truncated on the right if
3077 it exceeds <length>.
3078
Willy Tarreau4460d032012-11-21 23:37:37 +01003079 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003080 value will be added to the logs between braces ('{}'). If multiple headers
3081 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003082 in the same order they were declared in the configuration. Non-existent
3083 headers will be logged just as an empty string. Common uses for request
3084 header captures include the "Host" field in virtual hosting environments, the
3085 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003086 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003087 environments to find where the request came from.
3088
3089 Note that when capturing headers such as "User-agent", some spaces may be
3090 logged, making the log analysis more difficult. Thus be careful about what
3091 you log if you know your log parser is not smart enough to rely on the
3092 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003093
Willy Tarreau0900abb2012-11-22 00:21:46 +01003094 There is no limit to the number of captured request headers nor to their
3095 length, though it is wise to keep them low to limit memory usage per session.
3096 In order to keep log format consistent for a same frontend, header captures
3097 can only be declared in a frontend. It is not possible to specify a capture
3098 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003099
3100 Example:
3101 capture request header Host len 15
3102 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01003103 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01003104
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003105 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003106 about logging.
3107
3108
3109capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003110 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003111 May be used in sections : defaults | frontend | listen | backend
3112 no | yes | yes | no
3113 Arguments :
3114 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003115 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003116 appear in the response, with the first letter of each word in
3117 upper case. The header name will not appear in the logs, only the
3118 value is reported, but the position in the logs is respected.
3119
3120 <length> is the maximum number of characters to extract from the value and
3121 report in the logs. The string will be truncated on the right if
3122 it exceeds <length>.
3123
Willy Tarreau4460d032012-11-21 23:37:37 +01003124 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003125 result will be added to the logs between braces ('{}') after the captured
3126 request headers. If multiple headers are captured, they will be delimited by
3127 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003128 the configuration. Non-existent headers will be logged just as an empty
3129 string. Common uses for response header captures include the "Content-length"
3130 header which indicates how many bytes are expected to be returned, the
3131 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003132
Willy Tarreau0900abb2012-11-22 00:21:46 +01003133 There is no limit to the number of captured response headers nor to their
3134 length, though it is wise to keep them low to limit memory usage per session.
3135 In order to keep log format consistent for a same frontend, header captures
3136 can only be declared in a frontend. It is not possible to specify a capture
3137 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003138
3139 Example:
3140 capture response header Content-length len 9
3141 capture response header Location len 15
3142
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003143 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003144 about logging.
3145
3146
Cyril Bontéf0c60612010-02-06 14:44:47 +01003147clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003148 Set the maximum inactivity time on the client side.
3149 May be used in sections : defaults | frontend | listen | backend
3150 yes | yes | yes | no
3151 Arguments :
3152 <timeout> is the timeout value is specified in milliseconds by default, but
3153 can be in any other unit if the number is suffixed by the unit,
3154 as explained at the top of this document.
3155
3156 The inactivity timeout applies when the client is expected to acknowledge or
3157 send data. In HTTP mode, this timeout is particularly important to consider
3158 during the first phase, when the client sends the request, and during the
3159 response while it is reading data sent by the server. The value is specified
3160 in milliseconds by default, but can be in any other unit if the number is
3161 suffixed by the unit, as specified at the top of this document. In TCP mode
3162 (and to a lesser extent, in HTTP mode), it is highly recommended that the
3163 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003164 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01003165 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +01003166 (e.g. 4 or 5 seconds).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003167
3168 This parameter is specific to frontends, but can be specified once for all in
3169 "defaults" sections. This is in fact one of the easiest solutions not to
3170 forget about it. An unspecified timeout results in an infinite timeout, which
3171 is not recommended. Such a usage is accepted and works but reports a warning
3172 during startup because it may results in accumulation of expired sessions in
3173 the system if the system's timeouts are not configured either.
3174
3175 This parameter is provided for compatibility but is currently deprecated.
3176 Please use "timeout client" instead.
3177
Willy Tarreau036fae02008-01-06 13:24:40 +01003178 See also : "timeout client", "timeout http-request", "timeout server", and
3179 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003180
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003181compression algo <algorithm> ...
3182compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02003183compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02003184 Enable HTTP compression.
3185 May be used in sections : defaults | frontend | listen | backend
3186 yes | yes | yes | yes
3187 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003188 algo is followed by the list of supported compression algorithms.
3189 type is followed by the list of MIME types that will be compressed.
3190 offload makes haproxy work as a compression offloader only (see notes).
3191
3192 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01003193 identity this is mostly for debugging, and it was useful for developing
3194 the compression feature. Identity does not apply any change on
3195 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003196
Willy Tarreauc91840a2015-03-28 17:00:39 +01003197 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01003198 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01003199
3200 deflate same as "gzip", but with deflate algorithm and zlib format.
3201 Note that this algorithm has ambiguous support on many
3202 browsers and no support at all from recent ones. It is
3203 strongly recommended not to use it for anything else than
3204 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01003205 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003206
Willy Tarreauc91840a2015-03-28 17:00:39 +01003207 raw-deflate same as "deflate" without the zlib wrapper, and used as an
3208 alternative when the browser wants "deflate". All major
3209 browsers understand it and despite violating the standards,
3210 it is known to work better than "deflate", at least on MSIE
3211 and some versions of Safari. Do not use it in conjunction
3212 with "deflate", use either one or the other since both react
3213 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01003214 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003215
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04003216 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003217 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003218 If backend servers support HTTP compression, these directives
3219 will be no-op: haproxy will see the compressed response and will not
3220 compress again. If backend servers do not support HTTP compression and
3221 there is Accept-Encoding header in request, haproxy will compress the
3222 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02003223
3224 The "offload" setting makes haproxy remove the Accept-Encoding header to
3225 prevent backend servers from compressing responses. It is strongly
3226 recommended not to do this because this means that all the compression work
3227 will be done on the single point where haproxy is located. However in some
3228 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003229 with broken HTTP compression implementation which can't be turned off.
3230 In that case haproxy can be used to prevent that gateway from emitting
3231 invalid payloads. In this case, simply removing the header in the
3232 configuration does not work because it applies before the header is parsed,
3233 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02003234 then be used for such scenarios. Note: for now, the "offload" setting is
3235 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02003236
William Lallemand05097442012-11-20 12:14:28 +01003237 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003238 * the request does not advertise a supported compression algorithm in the
3239 "Accept-Encoding" header
3240 * the response message is not HTTP/1.1
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003241 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003242 * response contain neither a "Content-Length" header nor a
3243 "Transfer-Encoding" whose last value is "chunked"
3244 * response contains a "Content-Type" header whose first value starts with
3245 "multipart"
3246 * the response contains the "no-transform" value in the "Cache-control"
3247 header
3248 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
3249 and later
3250 * The response contains a "Content-Encoding" header, indicating that the
3251 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003252 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01003253
Tim Duesterhusb229f012019-01-29 16:38:56 +01003254 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01003255
William Lallemand82fe75c2012-10-23 10:25:10 +02003256 Examples :
3257 compression algo gzip
3258 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01003259
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003260
Cyril Bontéf0c60612010-02-06 14:44:47 +01003261contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003262 Set the maximum time to wait for a connection attempt to a server to succeed.
3263 May be used in sections : defaults | frontend | listen | backend
3264 yes | no | yes | yes
3265 Arguments :
3266 <timeout> is the timeout value is specified in milliseconds by default, but
3267 can be in any other unit if the number is suffixed by the unit,
3268 as explained at the top of this document.
3269
3270 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003271 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01003272 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +01003273 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003274 connect timeout also presets the queue timeout to the same value if this one
3275 has not been specified. Historically, the contimeout was also used to set the
3276 tarpit timeout in a listen section, which is not possible in a pure frontend.
3277
3278 This parameter is specific to backends, but can be specified once for all in
3279 "defaults" sections. This is in fact one of the easiest solutions not to
3280 forget about it. An unspecified timeout results in an infinite timeout, which
3281 is not recommended. Such a usage is accepted and works but reports a warning
3282 during startup because it may results in accumulation of failed sessions in
3283 the system if the system's timeouts are not configured either.
3284
3285 This parameter is provided for backwards compatibility but is currently
3286 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
3287 instead.
3288
3289 See also : "timeout connect", "timeout queue", "timeout tarpit",
3290 "timeout server", "contimeout".
3291
3292
Willy Tarreau55165fe2009-05-10 12:02:55 +02003293cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003294 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3295 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Olivier Houchard4e694042017-03-14 20:01:29 +01003296 [ dynamic ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003297 Enable cookie-based persistence in a backend.
3298 May be used in sections : defaults | frontend | listen | backend
3299 yes | no | yes | yes
3300 Arguments :
3301 <name> is the name of the cookie which will be monitored, modified or
3302 inserted in order to bring persistence. This cookie is sent to
3303 the client via a "Set-Cookie" header in the response, and is
3304 brought back by the client in a "Cookie" header in all requests.
3305 Special care should be taken to choose a name which does not
3306 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003307 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003308 HTTP/HTTPS), care should be taken to use different cookie names
3309 between all backends if persistence between them is not desired.
3310
3311 rewrite This keyword indicates that the cookie will be provided by the
3312 server and that haproxy will have to modify its value to set the
3313 server's identifier in it. This mode is handy when the management
3314 of complex combinations of "Set-cookie" and "Cache-control"
3315 headers is left to the application. The application can then
3316 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003317 cookie. Since all responses should be monitored, this mode
3318 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003319 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003320 start with this mode for new deployments. This keyword is
3321 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003322
3323 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003324 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003325
Willy Tarreaua79094d2010-08-31 22:54:15 +02003326 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003327 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02003328 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003329 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003330 existing configurations running in the "rewrite" mode. The cookie
3331 will only be a session cookie and will not be stored on the
3332 client's disk. By default, unless the "indirect" option is added,
3333 the server will see the cookies emitted by the client. Due to
3334 caching effects, it is generally wise to add the "nocache" or
3335 "postonly" keywords (see below). The "insert" keyword is not
3336 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003337
3338 prefix This keyword indicates that instead of relying on a dedicated
3339 cookie for the persistence, an existing one will be completed.
3340 This may be needed in some specific environments where the client
3341 does not support more than one single cookie and the application
3342 already needs it. In this case, whenever the server sets a cookie
3343 named <name>, it will be prefixed with the server's identifier
3344 and a delimiter. The prefix will be removed from all client
3345 requests so that the server still finds the cookie it emitted.
3346 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003347 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003348 not compatible with "rewrite" and "insert". Note: it is highly
3349 recommended not to use "indirect" with "prefix", otherwise server
3350 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003351
Willy Tarreaua79094d2010-08-31 22:54:15 +02003352 indirect When this option is specified, no cookie will be emitted to a
3353 client which already has a valid one for the server which has
3354 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003355 it will be removed, unless the "preserve" option is also set. In
3356 "insert" mode, this will additionally remove cookies from the
3357 requests transmitted to the server, making the persistence
3358 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003359 Note: it is highly recommended not to use "indirect" with
3360 "prefix", otherwise server cookie updates would not be sent to
3361 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003362
3363 nocache This option is recommended in conjunction with the insert mode
3364 when there is a cache between the client and HAProxy, as it
3365 ensures that a cacheable response will be tagged non-cacheable if
3366 a cookie needs to be inserted. This is important because if all
3367 persistence cookies are added on a cacheable home page for
3368 instance, then all customers will then fetch the page from an
3369 outer cache and will all share the same persistence cookie,
3370 leading to one server receiving much more traffic than others.
3371 See also the "insert" and "postonly" options.
3372
3373 postonly This option ensures that cookie insertion will only be performed
3374 on responses to POST requests. It is an alternative to the
3375 "nocache" option, because POST responses are not cacheable, so
3376 this ensures that the persistence cookie will never get cached.
3377 Since most sites do not need any sort of persistence before the
3378 first POST which generally is a login request, this is a very
3379 efficient method to optimize caching without risking to find a
3380 persistence cookie in the cache.
3381 See also the "insert" and "nocache" options.
3382
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003383 preserve This option may only be used with "insert" and/or "indirect". It
3384 allows the server to emit the persistence cookie itself. In this
3385 case, if a cookie is found in the response, haproxy will leave it
3386 untouched. This is useful in order to end persistence after a
3387 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003388 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003389 the past. By combining this mechanism with the "disable-on-404"
3390 check option, it is possible to perform a completely graceful
3391 shutdown because users will definitely leave the server after
3392 they logout.
3393
Willy Tarreau4992dd22012-05-31 21:02:17 +02003394 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3395 when a cookie is inserted. This attribute is used so that a
3396 user agent doesn't share the cookie with non-HTTP components.
3397 Please check RFC6265 for more information on this attribute.
3398
3399 secure This option tells haproxy to add a "Secure" cookie attribute when
3400 a cookie is inserted. This attribute is used so that a user agent
3401 never emits this cookie over non-secure channels, which means
3402 that a cookie learned with this flag will be presented only over
3403 SSL/TLS connections. Please check RFC6265 for more information on
3404 this attribute.
3405
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003406 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003407 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003408 name. If the domain begins with a dot, the browser is allowed to
3409 use it for any host ending with that name. It is also possible to
3410 specify several domain names by invoking this option multiple
3411 times. Some browsers might have small limits on the number of
3412 domains, so be careful when doing that. For the record, sending
3413 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003414
Willy Tarreau996a92c2010-10-13 19:30:47 +02003415 maxidle This option allows inserted cookies to be ignored after some idle
3416 time. It only works with insert-mode cookies. When a cookie is
3417 sent to the client, the date this cookie was emitted is sent too.
3418 Upon further presentations of this cookie, if the date is older
3419 than the delay indicated by the parameter (in seconds), it will
3420 be ignored. Otherwise, it will be refreshed if needed when the
3421 response is sent to the client. This is particularly useful to
3422 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003423 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003424 this option is set and a cookie has no date, it is always
3425 accepted, but gets refreshed in the response. This maintains the
3426 ability for admins to access their sites. Cookies that have a
3427 date in the future further than 24 hours are ignored. Doing so
3428 lets admins fix timezone issues without risking kicking users off
3429 the site.
3430
3431 maxlife This option allows inserted cookies to be ignored after some life
3432 time, whether they're in use or not. It only works with insert
3433 mode cookies. When a cookie is first sent to the client, the date
3434 this cookie was emitted is sent too. Upon further presentations
3435 of this cookie, if the date is older than the delay indicated by
3436 the parameter (in seconds), it will be ignored. If the cookie in
3437 the request has no date, it is accepted and a date will be set.
3438 Cookies that have a date in the future further than 24 hours are
3439 ignored. Doing so lets admins fix timezone issues without risking
3440 kicking users off the site. Contrary to maxidle, this value is
3441 not refreshed, only the first visit date counts. Both maxidle and
3442 maxlife may be used at the time. This is particularly useful to
3443 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003444 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003445 is stronger than the maxidle method in that it forces a
3446 redispatch after some absolute delay.
3447
Olivier Houchard4e694042017-03-14 20:01:29 +01003448 dynamic Activate dynamic cookies. When used, a session cookie is
3449 dynamically created for each server, based on the IP and port
3450 of the server, and a secret key, specified in the
3451 "dynamic-cookie-key" backend directive.
3452 The cookie will be regenerated each time the IP address change,
3453 and is only generated for IPv4/IPv6.
3454
Willy Tarreau0ba27502007-12-24 16:55:16 +01003455 There can be only one persistence cookie per HTTP backend, and it can be
3456 declared in a defaults section. The value of the cookie will be the value
3457 indicated after the "cookie" keyword in a "server" statement. If no cookie
3458 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003459
Willy Tarreau0ba27502007-12-24 16:55:16 +01003460 Examples :
3461 cookie JSESSIONID prefix
3462 cookie SRV insert indirect nocache
3463 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003464 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003465
Willy Tarreau294d0f02015-08-10 19:40:12 +02003466 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003467
Willy Tarreau983e01e2010-01-11 18:42:06 +01003468
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003469declare capture [ request | response ] len <length>
3470 Declares a capture slot.
3471 May be used in sections : defaults | frontend | listen | backend
3472 no | yes | yes | no
3473 Arguments:
3474 <length> is the length allowed for the capture.
3475
3476 This declaration is only available in the frontend or listen section, but the
3477 reserved slot can be used in the backends. The "request" keyword allocates a
3478 capture slot for use in the request, and "response" allocates a capture slot
3479 for use in the response.
3480
3481 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003482 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003483 "http-request capture" and "http-response capture".
3484
3485
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003486default-server [param*]
3487 Change default options for a server in a backend
3488 May be used in sections : defaults | frontend | listen | backend
3489 yes | no | yes | yes
3490 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003491 <param*> is a list of parameters for this server. The "default-server"
3492 keyword accepts an important number of options and has a complete
3493 section dedicated to it. Please refer to section 5 for more
3494 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003495
Willy Tarreau983e01e2010-01-11 18:42:06 +01003496 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003497 default-server inter 1000 weight 13
3498
3499 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003500
Willy Tarreau983e01e2010-01-11 18:42:06 +01003501
Willy Tarreau0ba27502007-12-24 16:55:16 +01003502default_backend <backend>
3503 Specify the backend to use when no "use_backend" rule has been matched.
3504 May be used in sections : defaults | frontend | listen | backend
3505 yes | yes | yes | no
3506 Arguments :
3507 <backend> is the name of the backend to use.
3508
3509 When doing content-switching between frontend and backends using the
3510 "use_backend" keyword, it is often useful to indicate which backend will be
3511 used when no rule has matched. It generally is the dynamic backend which
3512 will catch all undetermined requests.
3513
Willy Tarreau0ba27502007-12-24 16:55:16 +01003514 Example :
3515
3516 use_backend dynamic if url_dyn
3517 use_backend static if url_css url_img extension_img
3518 default_backend dynamic
3519
Willy Tarreau98d04852015-05-26 12:18:29 +02003520 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003521
Willy Tarreau0ba27502007-12-24 16:55:16 +01003522
Baptiste Assmann27f51342013-10-09 06:51:49 +02003523description <string>
3524 Describe a listen, frontend or backend.
3525 May be used in sections : defaults | frontend | listen | backend
3526 no | yes | yes | yes
3527 Arguments : string
3528
3529 Allows to add a sentence to describe the related object in the HAProxy HTML
3530 stats page. The description will be printed on the right of the object name
3531 it describes.
3532 No need to backslash spaces in the <string> arguments.
3533
3534
Willy Tarreau0ba27502007-12-24 16:55:16 +01003535disabled
3536 Disable a proxy, frontend or backend.
3537 May be used in sections : defaults | frontend | listen | backend
3538 yes | yes | yes | yes
3539 Arguments : none
3540
3541 The "disabled" keyword is used to disable an instance, mainly in order to
3542 liberate a listening port or to temporarily disable a service. The instance
3543 will still be created and its configuration will be checked, but it will be
3544 created in the "stopped" state and will appear as such in the statistics. It
3545 will not receive any traffic nor will it send any health-checks or logs. It
3546 is possible to disable many instances at once by adding the "disabled"
3547 keyword in a "defaults" section.
3548
3549 See also : "enabled"
3550
3551
Willy Tarreau5ce94572010-06-07 14:35:41 +02003552dispatch <address>:<port>
3553 Set a default server address
3554 May be used in sections : defaults | frontend | listen | backend
3555 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003556 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003557
3558 <address> is the IPv4 address of the default server. Alternatively, a
3559 resolvable hostname is supported, but this name will be resolved
3560 during start-up.
3561
3562 <ports> is a mandatory port specification. All connections will be sent
3563 to this port, and it is not permitted to use port offsets as is
3564 possible with normal servers.
3565
Willy Tarreau787aed52011-04-15 06:45:37 +02003566 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003567 server can take the connection. In the past it was used to forward non
3568 persistent connections to an auxiliary load balancer. Due to its simple
3569 syntax, it has also been used for simple TCP relays. It is recommended not to
3570 use it for more clarity, and to use the "server" directive instead.
3571
3572 See also : "server"
3573
Olivier Houchard4e694042017-03-14 20:01:29 +01003574
3575dynamic-cookie-key <string>
3576 Set the dynamic cookie secret key for a backend.
3577 May be used in sections : defaults | frontend | listen | backend
3578 yes | no | yes | yes
3579 Arguments : The secret key to be used.
3580
3581 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003582 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003583 specified on the "server" line), using a hash of the IP address of the
3584 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003585 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003586 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003587
Willy Tarreau0ba27502007-12-24 16:55:16 +01003588enabled
3589 Enable a proxy, frontend or backend.
3590 May be used in sections : defaults | frontend | listen | backend
3591 yes | yes | yes | yes
3592 Arguments : none
3593
3594 The "enabled" keyword is used to explicitly enable an instance, when the
3595 defaults has been set to "disabled". This is very rarely used.
3596
3597 See also : "disabled"
3598
3599
3600errorfile <code> <file>
3601 Return a file contents instead of errors generated by HAProxy
3602 May be used in sections : defaults | frontend | listen | backend
3603 yes | yes | yes | yes
3604 Arguments :
3605 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003606 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3607 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003608
3609 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003610 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003611 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003612 error pages, and to use absolute paths, since files are read
3613 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003614
3615 It is important to understand that this keyword is not meant to rewrite
3616 errors returned by the server, but errors detected and returned by HAProxy.
3617 This is why the list of supported errors is limited to a small set.
3618
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003619 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3620
Willy Tarreau0ba27502007-12-24 16:55:16 +01003621 The files are returned verbatim on the TCP socket. This allows any trick such
3622 as redirections to another URL or site, as well as tricks to clean cookies,
3623 force enable or disable caching, etc... The package provides default error
3624 files returning the same contents as default errors.
3625
Willy Tarreau59140a22009-02-22 12:02:30 +01003626 The files should not exceed the configured buffer size (BUFSIZE), which
3627 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003628 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003629 loops between the client and HAProxy when all servers are down, causing an
3630 error to be returned instead of an image. For better HTTP compliance, it is
3631 recommended that all header lines end with CR-LF and not LF alone.
3632
Willy Tarreau0ba27502007-12-24 16:55:16 +01003633 The files are read at the same time as the configuration and kept in memory.
3634 For this reason, the errors continue to be returned even when the process is
3635 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003636 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003637 403 status code and interrogating a blocked URL.
3638
3639 See also : "errorloc", "errorloc302", "errorloc303"
3640
Willy Tarreau59140a22009-02-22 12:02:30 +01003641 Example :
3642 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003643 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003644 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3645 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3646
Willy Tarreau2769aa02007-12-27 18:26:09 +01003647
3648errorloc <code> <url>
3649errorloc302 <code> <url>
3650 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3651 May be used in sections : defaults | frontend | listen | backend
3652 yes | yes | yes | yes
3653 Arguments :
3654 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003655 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3656 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003657
3658 <url> it is the exact contents of the "Location" header. It may contain
3659 either a relative URI to an error page hosted on the same site,
3660 or an absolute URI designating an error page on another site.
3661 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003662 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003663
3664 It is important to understand that this keyword is not meant to rewrite
3665 errors returned by the server, but errors detected and returned by HAProxy.
3666 This is why the list of supported errors is limited to a small set.
3667
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003668 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3669
Willy Tarreau2769aa02007-12-27 18:26:09 +01003670 Note that both keyword return the HTTP 302 status code, which tells the
3671 client to fetch the designated URL using the same HTTP method. This can be
3672 quite problematic in case of non-GET methods such as POST, because the URL
3673 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003674 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003675 status code, indicating to the client that the URL must be fetched with a GET
3676 request.
3677
3678 See also : "errorfile", "errorloc303"
3679
3680
3681errorloc303 <code> <url>
3682 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3683 May be used in sections : defaults | frontend | listen | backend
3684 yes | yes | yes | yes
3685 Arguments :
3686 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003687 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3688 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003689
3690 <url> it is the exact contents of the "Location" header. It may contain
3691 either a relative URI to an error page hosted on the same site,
3692 or an absolute URI designating an error page on another site.
3693 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003694 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003695
3696 It is important to understand that this keyword is not meant to rewrite
3697 errors returned by the server, but errors detected and returned by HAProxy.
3698 This is why the list of supported errors is limited to a small set.
3699
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003700 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3701
Willy Tarreau2769aa02007-12-27 18:26:09 +01003702 Note that both keyword return the HTTP 303 status code, which tells the
3703 client to fetch the designated URL using the same HTTP GET method. This
3704 solves the usual problems associated with "errorloc" and the 302 code. It is
3705 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003706 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003707
3708 See also : "errorfile", "errorloc", "errorloc302"
3709
3710
Simon Horman51a1cf62015-02-03 13:00:44 +09003711email-alert from <emailaddr>
3712 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003713 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003714 May be used in sections: defaults | frontend | listen | backend
3715 yes | yes | yes | yes
3716
3717 Arguments :
3718
3719 <emailaddr> is the from email address to use when sending email alerts
3720
3721 Also requires "email-alert mailers" and "email-alert to" to be set
3722 and if so sending email alerts is enabled for the proxy.
3723
Simon Horman64e34162015-02-06 11:11:57 +09003724 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003725 "email-alert myhostname", "email-alert to", section 3.6 about
3726 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003727
3728
3729email-alert level <level>
3730 Declare the maximum log level of messages for which email alerts will be
3731 sent. This acts as a filter on the sending of email alerts.
3732 May be used in sections: defaults | frontend | listen | backend
3733 yes | yes | yes | yes
3734
3735 Arguments :
3736
3737 <level> One of the 8 syslog levels:
3738 emerg alert crit err warning notice info debug
3739 The above syslog levels are ordered from lowest to highest.
3740
3741 By default level is alert
3742
3743 Also requires "email-alert from", "email-alert mailers" and
3744 "email-alert to" to be set and if so sending email alerts is enabled
3745 for the proxy.
3746
Simon Horman1421e212015-04-30 13:10:35 +09003747 Alerts are sent when :
3748
3749 * An un-paused server is marked as down and <level> is alert or lower
3750 * A paused server is marked as down and <level> is notice or lower
3751 * A server is marked as up or enters the drain state and <level>
3752 is notice or lower
3753 * "option log-health-checks" is enabled, <level> is info or lower,
3754 and a health check status update occurs
3755
Simon Horman64e34162015-02-06 11:11:57 +09003756 See also : "email-alert from", "email-alert mailers",
3757 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003758 section 3.6 about mailers.
3759
3760
3761email-alert mailers <mailersect>
3762 Declare the mailers to be used when sending email alerts
3763 May be used in sections: defaults | frontend | listen | backend
3764 yes | yes | yes | yes
3765
3766 Arguments :
3767
3768 <mailersect> is the name of the mailers section to send email alerts.
3769
3770 Also requires "email-alert from" and "email-alert to" to be set
3771 and if so sending email alerts is enabled for the proxy.
3772
Simon Horman64e34162015-02-06 11:11:57 +09003773 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3774 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003775
3776
3777email-alert myhostname <hostname>
3778 Declare the to hostname address to be used when communicating with
3779 mailers.
3780 May be used in sections: defaults | frontend | listen | backend
3781 yes | yes | yes | yes
3782
3783 Arguments :
3784
Baptiste Assmann738bad92015-12-21 15:27:53 +01003785 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003786
3787 By default the systems hostname is used.
3788
3789 Also requires "email-alert from", "email-alert mailers" and
3790 "email-alert to" to be set and if so sending email alerts is enabled
3791 for the proxy.
3792
Simon Horman64e34162015-02-06 11:11:57 +09003793 See also : "email-alert from", "email-alert level", "email-alert mailers",
3794 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003795
3796
3797email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003798 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003799 header of email alerts. This is the address that email alerts are sent to.
3800 May be used in sections: defaults | frontend | listen | backend
3801 yes | yes | yes | yes
3802
3803 Arguments :
3804
3805 <emailaddr> is the to email address to use when sending email alerts
3806
3807 Also requires "email-alert mailers" and "email-alert to" to be set
3808 and if so sending email alerts is enabled for the proxy.
3809
Simon Horman64e34162015-02-06 11:11:57 +09003810 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003811 "email-alert myhostname", section 3.6 about mailers.
3812
3813
Willy Tarreau4de91492010-01-22 19:10:05 +01003814force-persist { if | unless } <condition>
3815 Declare a condition to force persistence on down servers
3816 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003817 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003818
3819 By default, requests are not dispatched to down servers. It is possible to
3820 force this using "option persist", but it is unconditional and redispatches
3821 to a valid server if "option redispatch" is set. That leaves with very little
3822 possibilities to force some requests to reach a server which is artificially
3823 marked down for maintenance operations.
3824
3825 The "force-persist" statement allows one to declare various ACL-based
3826 conditions which, when met, will cause a request to ignore the down status of
3827 a server and still try to connect to it. That makes it possible to start a
3828 server, still replying an error to the health checks, and run a specially
3829 configured browser to test the service. Among the handy methods, one could
3830 use a specific source IP address, or a specific cookie. The cookie also has
3831 the advantage that it can easily be added/removed on the browser from a test
3832 page. Once the service is validated, it is then possible to open the service
3833 to the world by returning a valid response to health checks.
3834
3835 The forced persistence is enabled when an "if" condition is met, or unless an
3836 "unless" condition is met. The final redispatch is always disabled when this
3837 is used.
3838
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003839 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003840 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003841
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003842
3843filter <name> [param*]
3844 Add the filter <name> in the filter list attached to the proxy.
3845 May be used in sections : defaults | frontend | listen | backend
3846 no | yes | yes | yes
3847 Arguments :
3848 <name> is the name of the filter. Officially supported filters are
3849 referenced in section 9.
3850
Tim Düsterhus4896c442016-11-29 02:15:19 +01003851 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003852 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01003853 filter. Please refer to the documentation of the corresponding
3854 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003855
3856 Multiple occurrences of the filter line can be used for the same proxy. The
3857 same filter can be referenced many times if needed.
3858
3859 Example:
3860 listen
3861 bind *:80
3862
3863 filter trace name BEFORE-HTTP-COMP
3864 filter compression
3865 filter trace name AFTER-HTTP-COMP
3866
3867 compression algo gzip
3868 compression offload
3869
3870 server srv1 192.168.0.1:80
3871
3872 See also : section 9.
3873
Willy Tarreau4de91492010-01-22 19:10:05 +01003874
Willy Tarreau2769aa02007-12-27 18:26:09 +01003875fullconn <conns>
3876 Specify at what backend load the servers will reach their maxconn
3877 May be used in sections : defaults | frontend | listen | backend
3878 yes | no | yes | yes
3879 Arguments :
3880 <conns> is the number of connections on the backend which will make the
3881 servers use the maximal number of connections.
3882
Willy Tarreau198a7442008-01-17 12:05:32 +01003883 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003884 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003885 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003886 load. The server will then always accept at least <minconn> connections,
3887 never more than <maxconn>, and the limit will be on the ramp between both
3888 values when the backend has less than <conns> concurrent connections. This
3889 makes it possible to limit the load on the servers during normal loads, but
3890 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003891 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003892
Willy Tarreaufbb78422011-06-05 15:38:35 +02003893 Since it's hard to get this value right, haproxy automatically sets it to
3894 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003895 backend (based on "use_backend" and "default_backend" rules). That way it's
3896 safe to leave it unset. However, "use_backend" involving dynamic names are
3897 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003898
Willy Tarreau2769aa02007-12-27 18:26:09 +01003899 Example :
3900 # The servers will accept between 100 and 1000 concurrent connections each
3901 # and the maximum of 1000 will be reached when the backend reaches 10000
3902 # connections.
3903 backend dynamic
3904 fullconn 10000
3905 server srv1 dyn1:80 minconn 100 maxconn 1000
3906 server srv2 dyn2:80 minconn 100 maxconn 1000
3907
3908 See also : "maxconn", "server"
3909
3910
3911grace <time>
3912 Maintain a proxy operational for some time after a soft stop
3913 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003914 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003915 Arguments :
3916 <time> is the time (by default in milliseconds) for which the instance
3917 will remain operational with the frontend sockets still listening
3918 when a soft-stop is received via the SIGUSR1 signal.
3919
3920 This may be used to ensure that the services disappear in a certain order.
3921 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003922 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003923 needed by the equipment to detect the failure.
3924
3925 Note that currently, there is very little benefit in using this parameter,
3926 and it may in fact complicate the soft-reconfiguration process more than
3927 simplify it.
3928
Willy Tarreau0ba27502007-12-24 16:55:16 +01003929
Andrew Rodland17be45e2016-10-25 17:04:12 -04003930hash-balance-factor <factor>
3931 Specify the balancing factor for bounded-load consistent hashing
3932 May be used in sections : defaults | frontend | listen | backend
3933 yes | no | no | yes
3934 Arguments :
3935 <factor> is the control for the maximum number of concurrent requests to
3936 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01003937 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04003938
3939 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
3940 enables an algorithm that prevents any one server from getting too many
3941 requests at once, even if some hash buckets receive many more requests than
3942 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
3943 <factor> is a percentage greater than 100. For example, if <factor> is 150,
3944 then no server will be allowed to have a load more than 1.5 times the average.
3945 If server weights are used, they will be respected.
3946
3947 If the first-choice server is disqualified, the algorithm will choose another
3948 server based on the request hash, until a server with additional capacity is
3949 found. A higher <factor> allows more imbalance between the servers, while a
3950 lower <factor> means that more servers will be checked on average, affecting
3951 performance. Reasonable values are from 125 to 200.
3952
Willy Tarreau760e81d2018-05-03 07:20:40 +02003953 This setting is also used by "balance random" which internally relies on the
3954 consistent hashing mechanism.
3955
Andrew Rodland17be45e2016-10-25 17:04:12 -04003956 See also : "balance" and "hash-type".
3957
3958
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003959hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003960 Specify a method to use for mapping hashes to servers
3961 May be used in sections : defaults | frontend | listen | backend
3962 yes | no | yes | yes
3963 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003964 <method> is the method used to select a server from the hash computed by
3965 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003966
Bhaskar98634f02013-10-29 23:30:51 -04003967 map-based the hash table is a static array containing all alive servers.
3968 The hashes will be very smooth, will consider weights, but
3969 will be static in that weight changes while a server is up
3970 will be ignored. This means that there will be no slow start.
3971 Also, since a server is selected by its position in the array,
3972 most mappings are changed when the server count changes. This
3973 means that when a server goes up or down, or when a server is
3974 added to a farm, most connections will be redistributed to
3975 different servers. This can be inconvenient with caches for
3976 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003977
Bhaskar98634f02013-10-29 23:30:51 -04003978 consistent the hash table is a tree filled with many occurrences of each
3979 server. The hash key is looked up in the tree and the closest
3980 server is chosen. This hash is dynamic, it supports changing
3981 weights while the servers are up, so it is compatible with the
3982 slow start feature. It has the advantage that when a server
3983 goes up or down, only its associations are moved. When a
3984 server is added to the farm, only a few part of the mappings
3985 are redistributed, making it an ideal method for caches.
3986 However, due to its principle, the distribution will never be
3987 very smooth and it may sometimes be necessary to adjust a
3988 server's weight or its ID to get a more balanced distribution.
3989 In order to get the same distribution on multiple load
3990 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003991 same IDs. Note: consistent hash uses sdbm and avalanche if no
3992 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003993
3994 <function> is the hash function to be used :
3995
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003996 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003997 reimplementation of ndbm) database library. It was found to do
3998 well in scrambling bits, causing better distribution of the keys
3999 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004000 function with good distribution, unless the total server weight
4001 is a multiple of 64, in which case applying the avalanche
4002 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04004003
4004 djb2 this function was first proposed by Dan Bernstein many years ago
4005 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004006 function provides a better distribution than sdbm. It generally
4007 works well with text-based inputs though it can perform extremely
4008 poorly with numeric-only input or when the total server weight is
4009 a multiple of 33, unless the avalanche modifier is also used.
4010
Willy Tarreaua0f42712013-11-14 14:30:35 +01004011 wt6 this function was designed for haproxy while testing other
4012 functions in the past. It is not as smooth as the other ones, but
4013 is much less sensible to the input data set or to the number of
4014 servers. It can make sense as an alternative to sdbm+avalanche or
4015 djb2+avalanche for consistent hashing or when hashing on numeric
4016 data such as a source IP address or a visitor identifier in a URL
4017 parameter.
4018
Willy Tarreau324f07f2015-01-20 19:44:50 +01004019 crc32 this is the most common CRC32 implementation as used in Ethernet,
4020 gzip, PNG, etc. It is slower than the other ones but may provide
4021 a better distribution or less predictable results especially when
4022 used on strings.
4023
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05004024 <modifier> indicates an optional method applied after hashing the key :
4025
4026 avalanche This directive indicates that the result from the hash
4027 function above should not be used in its raw form but that
4028 a 4-byte full avalanche hash must be applied first. The
4029 purpose of this step is to mix the resulting bits from the
4030 previous hash in order to avoid any undesired effect when
4031 the input contains some limited values or when the number of
4032 servers is a multiple of one of the hash's components (64
4033 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
4034 result less predictable, but it's also not as smooth as when
4035 using the original function. Some testing might be needed
4036 with some workloads. This hash is one of the many proposed
4037 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004038
Bhaskar98634f02013-10-29 23:30:51 -04004039 The default hash type is "map-based" and is recommended for most usages. The
4040 default function is "sdbm", the selection of a function should be based on
4041 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004042
Andrew Rodland17be45e2016-10-25 17:04:12 -04004043 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02004044
4045
Willy Tarreau0ba27502007-12-24 16:55:16 +01004046http-check disable-on-404
4047 Enable a maintenance mode upon HTTP/404 response to health-checks
4048 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01004049 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01004050 Arguments : none
4051
4052 When this option is set, a server which returns an HTTP code 404 will be
4053 excluded from further load-balancing, but will still receive persistent
4054 connections. This provides a very convenient method for Web administrators
4055 to perform a graceful shutdown of their servers. It is also important to note
4056 that a server which is detected as failed while it was in this mode will not
4057 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
4058 will immediately be reinserted into the farm. The status on the stats page
4059 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01004060 option only works in conjunction with the "httpchk" option. If this option
4061 is used with "http-check expect", then it has precedence over it so that 404
4062 responses will still be considered as soft-stop.
4063
4064 See also : "option httpchk", "http-check expect"
4065
4066
4067http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004068 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01004069 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02004070 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01004071 Arguments :
4072 <match> is a keyword indicating how to look for a specific pattern in the
4073 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004074 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01004075 exclamation mark ("!") to negate the match. Spaces are allowed
4076 between the exclamation mark and the keyword. See below for more
4077 details on the supported keywords.
4078
4079 <pattern> is the pattern to look for. It may be a string or a regular
4080 expression. If the pattern contains spaces, they must be escaped
4081 with the usual backslash ('\').
4082
4083 By default, "option httpchk" considers that response statuses 2xx and 3xx
4084 are valid, and that others are invalid. When "http-check expect" is used,
4085 it defines what is considered valid or invalid. Only one "http-check"
4086 statement is supported in a backend. If a server fails to respond or times
4087 out, the check obviously fails. The available matches are :
4088
4089 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004090 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004091 response's status code is exactly this string. If the
4092 "status" keyword is prefixed with "!", then the response
4093 will be considered invalid if the status code matches.
4094
4095 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004096 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004097 response's status code matches the expression. If the
4098 "rstatus" keyword is prefixed with "!", then the response
4099 will be considered invalid if the status code matches.
4100 This is mostly used to check for multiple codes.
4101
4102 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004103 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004104 response's body contains this exact string. If the
4105 "string" keyword is prefixed with "!", then the response
4106 will be considered invalid if the body contains this
4107 string. This can be used to look for a mandatory word at
4108 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004109 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01004110 trace).
4111
4112 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004113 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004114 response's body matches this expression. If the "rstring"
4115 keyword is prefixed with "!", then the response will be
4116 considered invalid if the body matches the expression.
4117 This can be used to look for a mandatory word at the end
4118 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01004119 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01004120
4121 It is important to note that the responses will be limited to a certain size
4122 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
4123 Thus, too large responses may not contain the mandatory pattern when using
4124 "string" or "rstring". If a large response is absolutely required, it is
4125 possible to change the default max size by setting the global variable.
4126 However, it is worth keeping in mind that parsing very large responses can
4127 waste some CPU cycles, especially when regular expressions are used, and that
4128 it is always better to focus the checks on smaller resources.
4129
Cyril Bonté32602d22015-01-30 00:07:07 +01004130 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
4131 will automatically append a "Connection: close" header, meaning that this
4132 header should not be present in the request provided by "option httpchk".
4133
Willy Tarreaubd741542010-03-16 18:46:54 +01004134 Last, if "http-check expect" is combined with "http-check disable-on-404",
4135 then this last one has precedence when the server responds with 404.
4136
4137 Examples :
4138 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004139 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01004140
4141 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004142 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01004143
4144 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004145 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01004146
4147 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004148 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01004149
Willy Tarreaubd741542010-03-16 18:46:54 +01004150 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004151
4152
Willy Tarreauef781042010-01-27 11:53:01 +01004153http-check send-state
4154 Enable emission of a state header with HTTP health checks
4155 May be used in sections : defaults | frontend | listen | backend
4156 yes | no | yes | yes
4157 Arguments : none
4158
4159 When this option is set, haproxy will systematically send a special header
4160 "X-Haproxy-Server-State" with a list of parameters indicating to each server
4161 how they are seen by haproxy. This can be used for instance when a server is
4162 manipulated without access to haproxy and the operator needs to know whether
4163 haproxy still sees it up or not, or if the server is the last one in a farm.
4164
4165 The header is composed of fields delimited by semi-colons, the first of which
4166 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
4167 checks on the total number before transition, just as appears in the stats
4168 interface. Next headers are in the form "<variable>=<value>", indicating in
4169 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08004170 - a variable "address", containing the address of the backend server.
4171 This corresponds to the <address> field in the server declaration. For
4172 unix domain sockets, it will read "unix".
4173
4174 - a variable "port", containing the port of the backend server. This
4175 corresponds to the <port> field in the server declaration. For unix
4176 domain sockets, it will read "unix".
4177
Willy Tarreauef781042010-01-27 11:53:01 +01004178 - a variable "name", containing the name of the backend followed by a slash
4179 ("/") then the name of the server. This can be used when a server is
4180 checked in multiple backends.
4181
4182 - a variable "node" containing the name of the haproxy node, as set in the
4183 global "node" variable, otherwise the system's hostname if unspecified.
4184
4185 - a variable "weight" indicating the weight of the server, a slash ("/")
4186 and the total weight of the farm (just counting usable servers). This
4187 helps to know if other servers are available to handle the load when this
4188 one fails.
4189
4190 - a variable "scur" indicating the current number of concurrent connections
4191 on the server, followed by a slash ("/") then the total number of
4192 connections on all servers of the same backend.
4193
4194 - a variable "qcur" indicating the current number of requests in the
4195 server's queue.
4196
4197 Example of a header received by the application server :
4198 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
4199 scur=13/22; qcur=0
4200
4201 See also : "option httpchk", "http-check disable-on-404"
4202
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004203
4204http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004205 Access control for Layer 7 requests
4206
4207 May be used in sections: defaults | frontend | listen | backend
4208 no | yes | yes | yes
4209
Willy Tarreau20b0de52012-12-24 15:45:22 +01004210 The http-request statement defines a set of rules which apply to layer 7
4211 processing. The rules are evaluated in their declaration order when they are
4212 met in a frontend, listen or backend section. Any rule may optionally be
4213 followed by an ACL-based condition, in which case it will only be evaluated
4214 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004215
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004216 The first keyword is the rule's action. The supported actions are described
4217 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004218
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004219 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004220
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004221 It is important to know that http-request rules are processed very early in
4222 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
4223 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
4224 visible by almost all further ACL rules.
Willy Tarreau53275e82017-11-24 07:52:01 +01004225
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004226 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
4227 in newer versions (>= 1.5). But if you need to use regular expression to
4228 delete headers, you can still use "reqdel". Also please use
4229 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Willy Tarreauccbcc372012-12-27 12:37:57 +01004230
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004231 Example:
4232 acl nagios src 192.168.129.3
4233 acl local_net src 192.168.0.0/16
4234 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01004235
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004236 http-request allow if nagios
4237 http-request allow if local_net auth_ok
4238 http-request auth realm Gimme if local_net auth_ok
4239 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01004240
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004241 Example:
4242 acl key req.hdr(X-Add-Acl-Key) -m found
4243 acl add path /addacl
4244 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01004245
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004246 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01004247
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004248 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4249 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004250
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004251 Example:
4252 acl value req.hdr(X-Value) -m found
4253 acl setmap path /setmap
4254 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004255
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004256 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004257
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004258 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4259 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004260
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004261 See also : "stats http-request", section 3.4 about userlists and section 7
4262 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004263
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004264http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004265
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004266 This is used to add a new entry into an ACL. The ACL must be loaded from a
4267 file (even a dummy empty file). The file name of the ACL to be updated is
4268 passed between parentheses. It takes one argument: <key fmt>, which follows
4269 log-format rules, to collect content of the new entry. It performs a lookup
4270 in the ACL before insertion, to avoid duplicated (or more) values. This
4271 lookup is done by a linear search and can be expensive with large lists!
4272 It is the equivalent of the "add acl" command from the stats socket, but can
4273 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06004274
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004275http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004276
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004277 This appends an HTTP header field whose name is specified in <name> and
4278 whose value is defined by <fmt> which follows the log-format rules (see
4279 Custom Log Format in section 8.2.4). This is particularly useful to pass
4280 connection-specific information to the server (e.g. the client's SSL
4281 certificate), or to combine several headers into one. This rule is not
4282 final, so it is possible to add other similar rules. Note that header
4283 addition is performed immediately, so one rule might reuse the resulting
4284 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004285
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004286http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004287
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004288 This stops the evaluation of the rules and lets the request pass the check.
4289 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06004290
Sasha Pachev218f0642014-06-16 12:05:59 -06004291
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004292http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004293
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004294 This stops the evaluation of the rules and immediately responds with an
4295 HTTP 401 or 407 error code to invite the user to present a valid user name
4296 and password. No further "http-request" rules are evaluated. An optional
4297 "realm" parameter is supported, it sets the authentication realm that is
4298 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06004299
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004300 Example:
4301 acl auth_ok http_auth_group(L1) G1
4302 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06004303
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004304http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004305
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004306 See section 10.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004307
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004308http-request capture <sample> [ len <length> | id <id> ]
4309 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004310
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004311 This captures sample expression <sample> from the request buffer, and
4312 converts it to a string of at most <len> characters. The resulting string is
4313 stored into the next request "capture" slot, so it will possibly appear next
4314 to some captured HTTP headers. It will then automatically appear in the logs,
4315 and it will be possible to extract it using sample fetch rules to feed it
4316 into headers or anything. The length should be limited given that this size
4317 will be allocated for each capture during the whole session life.
4318 Please check section 7.3 (Fetching samples) and "capture request header" for
4319 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004320
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004321 If the keyword "id" is used instead of "len", the action tries to store the
4322 captured string in a previously declared capture slot. This is useful to run
4323 captures in backends. The slot id can be declared by a previous directive
4324 "http-request capture" or with the "declare capture" keyword. If the slot
4325 <id> doesn't exist, then HAProxy fails parsing the configuration to prevent
4326 unexpected behavior at run time.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004327
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004328http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004329
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004330 This is used to delete an entry from an ACL. The ACL must be loaded from a
4331 file (even a dummy empty file). The file name of the ACL to be updated is
4332 passed between parentheses. It takes one argument: <key fmt>, which follows
4333 log-format rules, to collect content of the entry to delete.
4334 It is the equivalent of the "del acl" command from the stats socket, but can
4335 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004336
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004337http-request del-header <name> [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004338
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004339 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004340
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004341http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004342
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004343 This is used to delete an entry from a MAP. The MAP must be loaded from a
4344 file (even a dummy empty file). The file name of the MAP to be updated is
4345 passed between parentheses. It takes one argument: <key fmt>, which follows
4346 log-format rules, to collect content of the entry to delete.
4347 It takes one argument: "file name" It is the equivalent of the "del map"
4348 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02004349
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004350http-request deny [deny_status <status>] [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04004351
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004352 This stops the evaluation of the rules and immediately rejects the request
4353 and emits an HTTP 403 error, or optionally the status code specified as an
4354 argument to "deny_status". The list of permitted status codes is limited to
4355 those that can be overridden by the "errorfile" directive.
4356 No further "http-request" rules are evaluated.
Patrick Hemmer268a7072018-05-11 12:52:31 -04004357
Olivier Houchard602bf7d2019-05-10 13:59:15 +02004358http-request disable-l7-retry [ { if | unless } <condition> ]
4359 This disables any attempt to retry the request if it fails for any other
4360 reason than a connection failure. This can be useful for example to make
4361 sure POST requests aren't retried on failure.
4362
Baptiste Assmann333939c2019-01-21 08:34:50 +01004363http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
4364
4365 This action performs a DNS resolution of the output of <expr> and stores
4366 the result in the variable <var>. It uses the DNS resolvers section
4367 pointed by <resolvers>.
4368 It is possible to choose a resolution preference using the optional
4369 arguments 'ipv4' or 'ipv6'.
4370 When performing the DNS resolution, the client side connection is on
4371 pause waiting till the end of the resolution.
4372 If an IP address can be found, it is stored into <var>. If any kind of
4373 error occurs, then <var> is not set.
4374 One can use this action to discover a server IP address at run time and
4375 based on information found in the request (IE a Host header).
4376 If this action is used to find the server's IP address (using the
4377 "set-dst" action), then the server IP address in the backend must be set
4378 to 0.0.0.0.
4379
4380 Example:
4381 resolvers mydns
4382 nameserver local 127.0.0.53:53
4383 nameserver google 8.8.8.8:53
4384 timeout retry 1s
4385 hold valid 10s
4386 hold nx 3s
4387 hold other 3s
4388 hold obsolete 0s
4389 accepted_payload_size 8192
4390
4391 frontend fe
4392 bind 10.42.0.1:80
4393 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
4394 http-request capture var(txn.myip) len 40
4395
4396 # return 503 when the variable is not set,
4397 # which mean DNS resolution error
4398 use_backend b_503 unless { var(txn.myip) -m found }
4399
4400 default_backend be
4401
4402 backend b_503
4403 # dummy backend used to return 503.
4404 # one can use the errorfile directive to send a nice
4405 # 503 error page to end users
4406
4407 backend be
4408 # rule to prevent HAProxy from reconnecting to services
4409 # on the local network (forged DNS name used to scan the network)
4410 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
4411 http-request set-dst var(txn.myip)
4412 server clear 0.0.0.0:0
4413
4414 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
4415 be used to scan the network or worst won't loop over itself...
4416
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004417http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
4418
4419 This is used to build an HTTP 103 Early Hints response prior to any other one.
4420 This appends an HTTP header field to this response whose name is specified in
4421 <name> and whose value is defined by <fmt> which follows the log-format rules
4422 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01004423 to the client some Link headers to preload resources required to render the
4424 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004425
4426 See RFC 8297 for more information.
4427
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004428http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004429
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004430 This performs an HTTP redirection based on a redirect rule. This is exactly
4431 the same as the "redirect" statement except that it inserts a redirect rule
4432 which can be processed in the middle of other "http-request" rules and that
4433 these rules use the "log-format" strings. See the "redirect" keyword for the
4434 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004435
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004436http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004437
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004438 This stops the evaluation of the rules and immediately closes the connection
4439 without sending any response. It acts similarly to the
4440 "tcp-request content reject" rules. It can be useful to force an immediate
4441 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004442
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004443http-request replace-header <name> <match-regex> <replace-fmt>
4444 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02004445
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004446 This matches the regular expression in all occurrences of header field
4447 <name> according to <match-regex>, and replaces them with the
4448 <replace-fmt> argument. Format characters are allowed in replace-fmt and
4449 work like in <fmt> arguments in "http-request add-header". The match is
4450 only case-sensitive. It is important to understand that this action only
4451 considers whole header lines, regardless of the number of values they may
4452 contain. This usage is suited to headers naturally containing commas in
4453 their value, such as If-Modified-Since and so on.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004454
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004455 Example:
4456 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
William Lallemand86d0df02017-11-24 21:36:45 +01004457
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004458 # applied to:
4459 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
Willy Tarreau09448f72014-06-25 18:12:15 +02004460
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004461 # outputs:
4462 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
Willy Tarreau09448f72014-06-25 18:12:15 +02004463
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004464 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02004465
Willy Tarreau33810222019-06-12 17:44:02 +02004466http-request replace-uri <match-regex> <replace-fmt>
4467 [ { if | unless } <condition> ]
4468
4469 This matches the regular expression in the URI part of the request
4470 according to <match-regex>, and replaces it with the <replace-fmt>
4471 argument. Standard back-references using the backslash ('\') followed by a
4472 number are supported. The <fmt> field is interpreted as a log-format string
4473 so it may contain special expressions just like the <fmt> argument passed
4474 to "http-request set-uri". The match is exclusively case-sensitive. Any
4475 optional scheme, authority or query string are considered in the matching
4476 part of the URI. It is worth noting that regular expressions may be more
4477 expensive to evaluate than certain ACLs, so rare replacements may benefit
4478 from a condition to avoid performing the evaluation at all if it does not
4479 match.
4480
4481 Example:
4482 # prefix /foo : turn /bar?q=1 into /foo/bar?q=1 :
4483 http-request replace-uri (.*) /foo\1
4484
4485 # suffix /foo : turn /bar?q=1 into /bar/foo?q=1 :
4486 http-request replace-uri ([^?]*)(\?(.*))? \1/foo\2
4487
4488 # strip /foo : turn /foo/bar?q=1 into /bar?q=1
4489 http-request replace-uri /foo/(.*) /\1
4490 # or more efficient if only some requests match :
4491 http-request replace-uri /foo/(.*) /\1 if { url_beg /foo/ }
4492
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004493http-request replace-value <name> <match-regex> <replace-fmt>
4494 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02004495
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004496 This works like "replace-header" except that it matches the regex against
4497 every comma-delimited value of the header field <name> instead of the
4498 entire header. This is suited for all headers which are allowed to carry
4499 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02004500
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004501 Example:
4502 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004503
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004504 # applied to:
4505 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004506
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004507 # outputs:
4508 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004509
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004510http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4511http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004512
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004513 This actions increments the GPC0 or GPC1 counter according with the sticky
4514 counter designated by <sc-id>. If an error occurs, this action silently fails
4515 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004516
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004517http-request sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004518
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004519 This action sets the GPT0 tag according to the sticky counter designated by
4520 <sc-id> and the value of <int>. The expected result is a boolean. If an error
4521 occurs, this action silently fails and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004522
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004523http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004524
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004525 This is used to set the destination IP address to the value of specified
4526 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
4527 but provides the correct IP in a HTTP header; or you want to mask the IP for
4528 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
4529 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004530
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004531 Arguments:
4532 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4533 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004534
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004535 Example:
4536 http-request set-dst hdr(x-dst)
4537 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01004538
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004539 When possible, set-dst preserves the original destination port as long as the
4540 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004541
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004542http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004543
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004544 This is used to set the destination port address to the value of specified
4545 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
4546 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004547
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004548 Arguments:
4549 <expr> Is a standard HAProxy expression formed by a sample-fetch
4550 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004551
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004552 Example:
4553 http-request set-dst-port hdr(x-port)
4554 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004555
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004556 When possible, set-dst-port preserves the original destination address as
4557 long as the address family supports a port, otherwise it forces the
4558 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004559
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004560http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004561
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004562 This does the same as "http-request add-header" except that the header name
4563 is first removed if it existed. This is useful when passing security
4564 information to the server, where the header must not be manipulated by
4565 external users. Note that the new value is computed before the removal so it
4566 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02004567
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004568 Example:
4569 http-request set-header X-Haproxy-Current-Date %T
4570 http-request set-header X-SSL %[ssl_fc]
4571 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
4572 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4573 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4574 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4575 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4576 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4577 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02004578
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004579http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004580
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004581 This is used to change the log level of the current request when a certain
4582 condition is met. Valid levels are the 8 syslog levels (see the "log"
4583 keyword) plus the special level "silent" which disables logging for this
4584 request. This rule is not final so the last matching rule wins. This rule
4585 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004586
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004587http-request set-map(<file-name>) <key fmt> <value fmt>
4588 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004589
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004590 This is used to add a new entry into a MAP. The MAP must be loaded from a
4591 file (even a dummy empty file). The file name of the MAP to be updated is
4592 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
4593 log-format rules, used to collect MAP key, and <value fmt>, which follows
4594 log-format rules, used to collect content for the new entry.
4595 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4596 more) values. This lookup is done by a linear search and can be expensive
4597 with large lists! It is the equivalent of the "set map" command from the
4598 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004599
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004600http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004601
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004602 This is used to set the Netfilter MARK on all packets sent to the client to
4603 the value passed in <mark> on platforms which support it. This value is an
4604 unsigned 32 bit value which can be matched by netfilter and by the routing
4605 table. It can be expressed both in decimal or hexadecimal format (prefixed by
4606 "0x"). This can be useful to force certain packets to take a different route
4607 (for example a cheaper network path for bulk downloads). This works on Linux
4608 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02004609
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004610http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004611
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004612 This rewrites the request method with the result of the evaluation of format
4613 string <fmt>. There should be very few valid reasons for having to do so as
4614 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004615
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004616http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004617
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004618 This sets the "nice" factor of the current request being processed. It only
4619 has effect against the other requests being processed at the same time.
4620 The default value is 0, unless altered by the "nice" setting on the "bind"
4621 line. The accepted range is -1024..1024. The higher the value, the nicest
4622 the request will be. Lower values will make the request more important than
4623 other ones. This can be useful to improve the speed of some requests, or
4624 lower the priority of non-important requests. Using this setting without
4625 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004626
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004627http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02004628
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004629 This rewrites the request path with the result of the evaluation of format
4630 string <fmt>. The query string, if any, is left intact. If a scheme and
4631 authority is found before the path, they are left intact as well. If the
4632 request doesn't have a path ("*"), this one is replaced with the format.
4633 This can be used to prepend a directory component in front of a path for
4634 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02004635
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004636 Example :
4637 # prepend the host name before the path
4638 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004639
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004640http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02004641
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004642 This is used to set the queue priority class of the current request.
4643 The value must be a sample expression which converts to an integer in the
4644 range -2047..2047. Results outside this range will be truncated.
4645 The priority class determines the order in which queued requests are
4646 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004647
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004648http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004649
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004650 This is used to set the queue priority timestamp offset of the current
4651 request. The value must be a sample expression which converts to an integer
4652 in the range -524287..524287. Results outside this range will be truncated.
4653 When a request is queued, it is ordered first by the priority class, then by
4654 the current timestamp adjusted by the given offset in milliseconds. Lower
4655 values have higher priority.
4656 Note that the resulting timestamp is is only tracked with enough precision
4657 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
4658 adjusted timestamp exceeds this value, it will be misidentified as highest
4659 priority. Thus it is important to set "timeout queue" to a value, where when
4660 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004661
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004662http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004663
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004664 This rewrites the request's query string which appears after the first
4665 question mark ("?") with the result of the evaluation of format string <fmt>.
4666 The part prior to the question mark is left intact. If the request doesn't
4667 contain a question mark and the new value is not empty, then one is added at
4668 the end of the URI, followed by the new value. If a question mark was
4669 present, it will never be removed even if the value is empty. This can be
4670 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004671
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004672 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004673
4674 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004675 # replace "%3D" with "=" in the query string
4676 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004677
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004678http-request set-src <expr> [ { if | unless } <condition> ]
4679 This is used to set the source IP address to the value of specified
4680 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
4681 provides the correct IP in a HTTP header; or you want to mask source IP for
4682 privacy.
4683
4684 Arguments :
4685 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4686 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004687
Cyril Bonté78caf842010-03-10 22:41:43 +01004688 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004689 http-request set-src hdr(x-forwarded-for)
4690 http-request set-src src,ipmask(24)
4691
4692 When possible, set-src preserves the original source port as long as the
4693 address family allows it, otherwise the source port is set to 0.
4694
4695http-request set-src-port <expr> [ { if | unless } <condition> ]
4696
4697 This is used to set the source port address to the value of specified
4698 expression.
4699
4700 Arguments:
4701 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4702 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004703
Willy Tarreau20b0de52012-12-24 15:45:22 +01004704 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004705 http-request set-src-port hdr(x-port)
4706 http-request set-src-port int(4000)
4707
4708 When possible, set-src-port preserves the original source address as long as
4709 the address family supports a port, otherwise it forces the source address to
4710 IPv4 "0.0.0.0" before rewriting the port.
4711
4712http-request set-tos <tos> [ { if | unless } <condition> ]
4713
4714 This is used to set the TOS or DSCP field value of packets sent to the client
4715 to the value passed in <tos> on platforms which support this. This value
4716 represents the whole 8 bits of the IP TOS field, and can be expressed both in
4717 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
4718 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
4719 be used to adjust some routing behavior on border routers based on some
4720 information from the request.
4721
4722 See RFC 2474, 2597, 3260 and 4594 for more information.
4723
4724http-request set-uri <fmt> [ { if | unless } <condition> ]
4725
4726 This rewrites the request URI with the result of the evaluation of format
4727 string <fmt>. The scheme, authority, path and query string are all replaced
4728 at once. This can be used to rewrite hosts in front of proxies, or to
4729 perform complex modifications to the URI such as moving parts between the
4730 path and the query string.
4731 See also "http-request set-path" and "http-request set-query".
4732
4733http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
4734
4735 This is used to set the contents of a variable. The variable is declared
4736 inline.
4737
4738 Arguments:
4739 <var-name> The name of the variable starts with an indication about its
4740 scope. The scopes allowed are:
4741 "proc" : the variable is shared with the whole process
4742 "sess" : the variable is shared with the whole session
4743 "txn" : the variable is shared with the transaction
4744 (request and response)
4745 "req" : the variable is shared only during request
4746 processing
4747 "res" : the variable is shared only during response
4748 processing
4749 This prefix is followed by a name. The separator is a '.'.
4750 The name may only contain characters 'a-z', 'A-Z', '0-9'
4751 and '_'.
4752
4753 <expr> Is a standard HAProxy expression formed by a sample-fetch
4754 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004755
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004756 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004757 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004758
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004759http-request send-spoe-group <engine-name> <group-name>
4760 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004761
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004762 This action is used to trigger sending of a group of SPOE messages. To do so,
4763 the SPOE engine used to send messages must be defined, as well as the SPOE
4764 group to send. Of course, the SPOE engine must refer to an existing SPOE
4765 filter. If not engine name is provided on the SPOE filter line, the SPOE
4766 agent name must be used.
4767
4768 Arguments:
4769 <engine-name> The SPOE engine name.
4770
4771 <group-name> The SPOE group name as specified in the engine
4772 configuration.
4773
4774http-request silent-drop [ { if | unless } <condition> ]
4775
4776 This stops the evaluation of the rules and makes the client-facing connection
4777 suddenly disappear using a system-dependent way that tries to prevent the
4778 client from being notified. The effect it then that the client still sees an
4779 established connection while there's none on HAProxy. The purpose is to
4780 achieve a comparable effect to "tarpit" except that it doesn't use any local
4781 resource at all on the machine running HAProxy. It can resist much higher
4782 loads than "tarpit", and slow down stronger attackers. It is important to
4783 understand the impact of using this mechanism. All stateful equipment placed
4784 between the client and HAProxy (firewalls, proxies, load balancers) will also
4785 keep the established connection for a long time and may suffer from this
4786 action.
4787 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
4788 option is used to block the emission of a TCP reset. On other systems, the
4789 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
4790 router, though it's still delivered to local networks. Do not use it unless
4791 you fully understand how it works.
4792
4793http-request tarpit [deny_status <status>] [ { if | unless } <condition> ]
4794
4795 This stops the evaluation of the rules and immediately blocks the request
4796 without responding for a delay specified by "timeout tarpit" or
4797 "timeout connect" if the former is not set. After that delay, if the client
4798 is still connected, an HTTP error 500 (or optionally the status code
4799 specified as an argument to "deny_status") is returned so that the client
4800 does not suspect it has been tarpitted. Logs will report the flags "PT".
4801 The goal of the tarpit rule is to slow down robots during an attack when
4802 they're limited on the number of concurrent requests. It can be very
4803 efficient against very dumb robots, and will significantly reduce the load
4804 on firewalls compared to a "deny" rule. But when facing "correctly"
4805 developed robots, it can make things worse by forcing haproxy and the front
4806 firewall to support insane number of concurrent connections.
4807 See also the "silent-drop" action.
4808
4809http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
4810http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
4811http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
4812
4813 This enables tracking of sticky counters from current request. These rules do
4814 not stop evaluation and do not change default action. The number of counters
4815 that may be simultaneously tracked by the same connection is set in
4816 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
4817 so the track-sc number is between 0 and (MAX_SESS_STCKTR-1). The first
4818 "track-sc0" rule executed enables tracking of the counters of the specified
4819 table as the first set. The first "track-sc1" rule executed enables tracking
4820 of the counters of the specified table as the second set. The first
4821 "track-sc2" rule executed enables tracking of the counters of the specified
4822 table as the third set. It is a recommended practice to use the first set of
4823 counters for the per-frontend counters and the second set for the per-backend
4824 ones. But this is just a guideline, all may be used everywhere.
4825
4826 Arguments :
4827 <key> is mandatory, and is a sample expression rule as described in
4828 section 7.3. It describes what elements of the incoming request or
4829 connection will be analyzed, extracted, combined, and used to
4830 select which table entry to update the counters.
4831
4832 <table> is an optional table to be used instead of the default one, which
4833 is the stick-table declared in the current proxy. All the counters
4834 for the matches and updates for the key will then be performed in
4835 that table until the session ends.
4836
4837 Once a "track-sc*" rule is executed, the key is looked up in the table and if
4838 it is not found, an entry is allocated for it. Then a pointer to that entry
4839 is kept during all the session's life, and this entry's counters are updated
4840 as often as possible, every time the session's counters are updated, and also
4841 systematically when the session ends. Counters are only updated for events
4842 that happen after the tracking has been started. As an exception, connection
4843 counters and request counters are systematically updated so that they reflect
4844 useful information.
4845
4846 If the entry tracks concurrent connection counters, one connection is counted
4847 for as long as the entry is tracked, and the entry will not expire during
4848 that time. Tracking counters also provides a performance advantage over just
4849 checking the keys, because only one table lookup is performed for all ACL
4850 checks that make use of it.
4851
4852http-request unset-var(<var-name>) [ { if | unless } <condition> ]
4853
4854 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004855
4856 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004857 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004858
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004859http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004860
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004861 This will delay the processing of the request until the SSL handshake
4862 happened. This is mostly useful to delay processing early data until we're
4863 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004864
Willy Tarreauef781042010-01-27 11:53:01 +01004865
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004866http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004867 Access control for Layer 7 responses
4868
4869 May be used in sections: defaults | frontend | listen | backend
4870 no | yes | yes | yes
4871
4872 The http-response statement defines a set of rules which apply to layer 7
4873 processing. The rules are evaluated in their declaration order when they are
4874 met in a frontend, listen or backend section. Any rule may optionally be
4875 followed by an ACL-based condition, in which case it will only be evaluated
4876 if the condition is true. Since these rules apply on responses, the backend
4877 rules are applied first, followed by the frontend's rules.
4878
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004879 The first keyword is the rule's action. The supported actions are described
4880 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004881
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004882 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004883
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004884 It is important to know that http-response rules are processed very early in
4885 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
4886 headers added by "add-header"/"set-header" are visible by almost all further
4887 ACL rules.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004888
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004889 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4890 in newer versions (>= 1.5). But if you need to use regular expression to
4891 delete headers, you can still use "rspdel". Also please use
4892 "http-response deny" instead of "rspdeny".
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004893
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004894 Example:
4895 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004896
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004897 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06004898
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004899 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4900 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06004901
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004902 Example:
4903 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06004904
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004905 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004906
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004907 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4908 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06004909
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004910 See also : "http-request", section 3.4 about userlists and section 7 about
4911 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004912
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004913http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004914
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004915 This is used to add a new entry into an ACL. The ACL must be loaded from a
4916 file (even a dummy empty file). The file name of the ACL to be updated is
4917 passed between parentheses. It takes one argument: <key fmt>, which follows
4918 log-format rules, to collect content of the new entry. It performs a lookup
4919 in the ACL before insertion, to avoid duplicated (or more) values.
4920 This lookup is done by a linear search and can be expensive with large lists!
4921 It is the equivalent of the "add acl" command from the stats socket, but can
4922 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06004923
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004924http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004925
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004926 This appends an HTTP header field whose name is specified in <name> and whose
4927 value is defined by <fmt> which follows the log-format rules (see Custom Log
4928 Format in section 8.2.4). This may be used to send a cookie to a client for
4929 example, or to pass some internal information.
4930 This rule is not final, so it is possible to add other similar rules.
4931 Note that header addition is performed immediately, so one rule might reuse
4932 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004933
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004934http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004935
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004936 This stops the evaluation of the rules and lets the response pass the check.
4937 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06004938
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004939http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004940
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004941 See section 10.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06004942
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004943http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004944
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004945 This captures sample expression <sample> from the response buffer, and
4946 converts it to a string. The resulting string is stored into the next request
4947 "capture" slot, so it will possibly appear next to some captured HTTP
4948 headers. It will then automatically appear in the logs, and it will be
4949 possible to extract it using sample fetch rules to feed it into headers or
4950 anything. Please check section 7.3 (Fetching samples) and
4951 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004952
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004953 The keyword "id" is the id of the capture slot which is used for storing the
4954 string. The capture slot must be defined in an associated frontend.
4955 This is useful to run captures in backends. The slot id can be declared by a
4956 previous directive "http-response capture" or with the "declare capture"
4957 keyword.
4958 If the slot <id> doesn't exist, then HAProxy fails parsing the configuration
4959 to prevent unexpected behavior at run time.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004960
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004961http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004962
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004963 This is used to delete an entry from an ACL. The ACL must be loaded from a
4964 file (even a dummy empty file). The file name of the ACL to be updated is
4965 passed between parentheses. It takes one argument: <key fmt>, which follows
4966 log-format rules, to collect content of the entry to delete.
4967 It is the equivalent of the "del acl" command from the stats socket, but can
4968 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004969
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004970http-response del-header <name> [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004971
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004972 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004973
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004974http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02004975
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004976 This is used to delete an entry from a MAP. The MAP must be loaded from a
4977 file (even a dummy empty file). The file name of the MAP to be updated is
4978 passed between parentheses. It takes one argument: <key fmt>, which follows
4979 log-format rules, to collect content of the entry to delete.
4980 It takes one argument: "file name" It is the equivalent of the "del map"
4981 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004982
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004983http-response deny [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004984
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004985 This stops the evaluation of the rules and immediately rejects the response
4986 and emits an HTTP 502 error. No further "http-response" rules are evaluated.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004987
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004988http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004989
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004990 This performs an HTTP redirection based on a redirect rule.
4991 This supports a format string similarly to "http-request redirect" rules,
4992 with the exception that only the "location" type of redirect is possible on
4993 the response. See the "redirect" keyword for the rule's syntax. When a
4994 redirect rule is applied during a response, connections to the server are
4995 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004996
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004997http-response replace-header <name> <regex-match> <replace-fmt>
4998 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004999
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005000 This matches the regular expression in all occurrences of header field <name>
5001 according to <match-regex>, and replaces them with the <replace-fmt> argument.
5002 Format characters are allowed in replace-fmt and work like in <fmt> arguments
5003 in "add-header". The match is only case-sensitive. It is important to
5004 understand that this action only considers whole header lines, regardless of
5005 the number of values they may contain. This usage is suited to headers
5006 naturally containing commas in their value, such as Set-Cookie, Expires and
5007 so on.
William Lallemand86d0df02017-11-24 21:36:45 +01005008
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005009 Example:
5010 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02005011
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005012 # applied to:
5013 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005014
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005015 # outputs:
5016 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 +02005017
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005018 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005019
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005020http-response replace-value <name> <regex-match> <replace-fmt>
5021 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005022
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005023 This works like "replace-header" except that it matches the regex against
5024 every comma-delimited value of the header field <name> instead of the entire
5025 header. This is suited for all headers which are allowed to carry more than
5026 one value. An example could be the Accept header.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02005027
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005028 Example:
5029 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005030
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005031 # applied to:
5032 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01005033
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005034 # outputs:
5035 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01005036
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005037http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
5038http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08005039
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005040 This action increments the GPC0 or GPC1 counter according with the sticky
5041 counter designated by <sc-id>. If an error occurs, this action silently fails
5042 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02005043
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005044http-response sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02005045
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005046 This action sets the GPT0 tag according to the sticky counter designated by
5047 <sc-id> and the value of <int>. The expected result is a boolean. If an error
5048 occurs, this action silently fails and the actions evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01005049
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005050http-response send-spoe-group [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02005051
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005052 This action is used to trigger sending of a group of SPOE messages. To do so,
5053 the SPOE engine used to send messages must be defined, as well as the SPOE
5054 group to send. Of course, the SPOE engine must refer to an existing SPOE
5055 filter. If not engine name is provided on the SPOE filter line, the SPOE
5056 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005057
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005058 Arguments:
5059 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005060
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005061 <group-name> The SPOE group name as specified in the engine
5062 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02005063
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005064http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005065
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005066 This does the same as "add-header" except that the header name is first
5067 removed if it existed. This is useful when passing security information to
5068 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005069
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005070http-response set-log-level <level> [ { if | unless } <condition> ]
5071
5072 This is used to change the log level of the current request when a certain
5073 condition is met. Valid levels are the 8 syslog levels (see the "log"
5074 keyword) plus the special level "silent" which disables logging for this
5075 request. This rule is not final so the last matching rule wins. This rule can
5076 be useful to disable health checks coming from another equipment.
5077
5078http-response set-map(<file-name>) <key fmt> <value fmt>
5079
5080 This is used to add a new entry into a MAP. The MAP must be loaded from a
5081 file (even a dummy empty file). The file name of the MAP to be updated is
5082 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5083 log-format rules, used to collect MAP key, and <value fmt>, which follows
5084 log-format rules, used to collect content for the new entry. It performs a
5085 lookup in the MAP before insertion, to avoid duplicated (or more) values.
5086 This lookup is done by a linear search and can be expensive with large lists!
5087 It is the equivalent of the "set map" command from the stats socket, but can
5088 be triggered by an HTTP response.
5089
5090http-response set-mark <mark> [ { if | unless } <condition> ]
5091
5092 This is used to set the Netfilter MARK on all packets sent to the client to
5093 the value passed in <mark> on platforms which support it. This value is an
5094 unsigned 32 bit value which can be matched by netfilter and by the routing
5095 table. It can be expressed both in decimal or hexadecimal format (prefixed
5096 by "0x"). This can be useful to force certain packets to take a different
5097 route (for example a cheaper network path for bulk downloads). This works on
5098 Linux kernels 2.6.32 and above and requires admin privileges.
5099
5100http-response set-nice <nice> [ { if | unless } <condition> ]
5101
5102 This sets the "nice" factor of the current request being processed.
5103 It only has effect against the other requests being processed at the same
5104 time. The default value is 0, unless altered by the "nice" setting on the
5105 "bind" line. The accepted range is -1024..1024. The higher the value, the
5106 nicest the request will be. Lower values will make the request more important
5107 than other ones. This can be useful to improve the speed of some requests, or
5108 lower the priority of non-important requests. Using this setting without
5109 prior experimentation can cause some major slowdown.
5110
5111http-response set-status <status> [reason <str>]
5112 [ { if | unless } <condition> ]
5113
5114 This replaces the response status code with <status> which must be an integer
5115 between 100 and 999. Optionally, a custom reason text can be provided defined
5116 by <str>, or the default reason for the specified code will be used as a
5117 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005118
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005119 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005120 # return "431 Request Header Fields Too Large"
5121 http-response set-status 431
5122 # return "503 Slow Down", custom reason
5123 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005124
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005125http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005126
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005127 This is used to set the TOS or DSCP field value of packets sent to the client
5128 to the value passed in <tos> on platforms which support this.
5129 This value represents the whole 8 bits of the IP TOS field, and can be
5130 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
5131 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
5132 always 0. This can be used to adjust some routing behavior on border routers
5133 based on some information from the request.
5134
5135 See RFC 2474, 2597, 3260 and 4594 for more information.
5136
5137http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5138
5139 This is used to set the contents of a variable. The variable is declared
5140 inline.
5141
5142 Arguments:
5143 <var-name> The name of the variable starts with an indication about its
5144 scope. The scopes allowed are:
5145 "proc" : the variable is shared with the whole process
5146 "sess" : the variable is shared with the whole session
5147 "txn" : the variable is shared with the transaction
5148 (request and response)
5149 "req" : the variable is shared only during request
5150 processing
5151 "res" : the variable is shared only during response
5152 processing
5153 This prefix is followed by a name. The separator is a '.'.
5154 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5155 and '_'.
5156
5157 <expr> Is a standard HAProxy expression formed by a sample-fetch
5158 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005159
5160 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005161 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005162
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005163http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005164
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005165 This stops the evaluation of the rules and makes the client-facing connection
5166 suddenly disappear using a system-dependent way that tries to prevent the
5167 client from being notified. The effect it then that the client still sees an
5168 established connection while there's none on HAProxy. The purpose is to
5169 achieve a comparable effect to "tarpit" except that it doesn't use any local
5170 resource at all on the machine running HAProxy. It can resist much higher
5171 loads than "tarpit", and slow down stronger attackers. It is important to
5172 understand the impact of using this mechanism. All stateful equipment placed
5173 between the client and HAProxy (firewalls, proxies, load balancers) will also
5174 keep the established connection for a long time and may suffer from this
5175 action.
5176 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5177 option is used to block the emission of a TCP reset. On other systems, the
5178 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5179 router, though it's still delivered to local networks. Do not use it unless
5180 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005181
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005182http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5183http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5184http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005185
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005186 This enables tracking of sticky counters from current response. Please refer
5187 to "http-request track-sc" for a complete description. The only difference
5188 from "http-request track-sc" is the <key> sample expression can only make use
5189 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
5190 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
5191 supported, haproxy will fail and warn while parsing the config.
5192
5193http-response unset-var(<var-name>) [ { if | unless } <condition> ]
5194
5195 This is used to unset a variable. See "http-response set-var" for details
5196 about <var-name>.
5197
5198 Example:
5199 http-response unset-var(sess.last_redir)
5200
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02005201
Willy Tarreau30631952015-08-06 15:05:24 +02005202http-reuse { never | safe | aggressive | always }
5203 Declare how idle HTTP connections may be shared between requests
5204
5205 May be used in sections: defaults | frontend | listen | backend
5206 yes | no | yes | yes
5207
5208 By default, a connection established between haproxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01005209 which is considered safe for reuse is moved back to the server's idle
5210 connections pool so that any other request can make use of it. This is the
5211 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02005212
5213 The argument indicates the desired connection reuse strategy :
5214
Olivier Houchard86006a52018-12-14 19:37:49 +01005215 - "never" : idle connections are never shared between sessions. This mode
5216 may be enforced to cancel a different strategy inherited from
5217 a defaults section or for troubleshooting. For example, if an
5218 old bogus application considers that multiple requests over
5219 the same connection come from the same client and it is not
5220 possible to fix the application, it may be desirable to
5221 disable connection sharing in a single backend. An example of
5222 such an application could be an old haproxy using cookie
5223 insertion in tunnel mode and not checking any request past the
5224 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02005225
Olivier Houchard86006a52018-12-14 19:37:49 +01005226 - "safe" : this is the default and the recommended strategy. The first
5227 request of a session is always sent over its own connection,
5228 and only subsequent requests may be dispatched over other
5229 existing connections. This ensures that in case the server
5230 closes the connection when the request is being sent, the
5231 browser can decide to silently retry it. Since it is exactly
5232 equivalent to regular keep-alive, there should be no side
5233 effects.
Willy Tarreau30631952015-08-06 15:05:24 +02005234
5235 - "aggressive" : this mode may be useful in webservices environments where
5236 all servers are not necessarily known and where it would be
5237 appreciable to deliver most first requests over existing
5238 connections. In this case, first requests are only delivered
5239 over existing connections that have been reused at least once,
5240 proving that the server correctly supports connection reuse.
5241 It should only be used when it's sure that the client can
5242 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02005243 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02005244 downsides of rare connection failures.
5245
5246 - "always" : this mode is only recommended when the path to the server is
5247 known for never breaking existing connections quickly after
5248 releasing them. It allows the first request of a session to be
5249 sent to an existing connection. This can provide a significant
5250 performance increase over the "safe" strategy when the backend
5251 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005252 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02005253 sharing. It is recommended that the "http-keep-alive" timeout
5254 remains low in this mode so that no dead connections remain
5255 usable. In most cases, this will lead to the same performance
5256 gains as "aggressive" but with more risks. It should only be
5257 used when it improves the situation over "aggressive".
5258
5259 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005260 connection properties and compatibility. Specifically :
5261 - connections made with "usesrc" followed by a client-dependent value
5262 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005263
5264 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01005265 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005266
Lukas Tribusfd9b68c2018-10-27 20:06:59 +02005267 - connections with certain bogus authentication schemes (relying on the
5268 connection) like NTLM are detected, marked private and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005269
5270 No connection pool is involved, once a session dies, the last idle connection
5271 it was attached to is deleted at the same time. This ensures that connections
5272 may not last after all sessions are closed.
5273
5274 Note: connection reuse improves the accuracy of the "server maxconn" setting,
5275 because almost no new connection will be established while idle connections
5276 remain available. This is particularly true with the "always" strategy.
5277
5278 See also : "option http-keep-alive", "server maxconn"
5279
5280
Mark Lamourinec2247f02012-01-04 13:02:01 -05005281http-send-name-header [<header>]
5282 Add the server name to a request. Use the header string given by <header>
5283
5284 May be used in sections: defaults | frontend | listen | backend
5285 yes | no | yes | yes
5286
5287 Arguments :
5288
5289 <header> The header string to use to send the server name
5290
5291 The "http-send-name-header" statement causes the name of the target
Davor Ocelice9ed2812017-12-25 17:49:28 +01005292 server to be added to the headers of an HTTP request. The name
Mark Lamourinec2247f02012-01-04 13:02:01 -05005293 is added with the header string proved.
5294
5295 See also : "server"
5296
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005297id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02005298 Set a persistent ID to a proxy.
5299 May be used in sections : defaults | frontend | listen | backend
5300 no | yes | yes | yes
5301 Arguments : none
5302
5303 Set a persistent ID for the proxy. This ID must be unique and positive.
5304 An unused ID will automatically be assigned if unset. The first assigned
5305 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005306
5307
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005308ignore-persist { if | unless } <condition>
5309 Declare a condition to ignore persistence
5310 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005311 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005312
5313 By default, when cookie persistence is enabled, every requests containing
5314 the cookie are unconditionally persistent (assuming the target server is up
5315 and running).
5316
5317 The "ignore-persist" statement allows one to declare various ACL-based
5318 conditions which, when met, will cause a request to ignore persistence.
5319 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005320 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005321 persistence for a specific User-Agent (for example, some web crawler bots).
5322
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005323 The persistence is ignored when an "if" condition is met, or unless an
5324 "unless" condition is met.
5325
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005326 Example:
5327 acl url_static path_beg /static /images /img /css
5328 acl url_static path_end .gif .png .jpg .css .js
5329 ignore-persist if url_static
5330
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005331 See also : "force-persist", "cookie", and section 7 about ACL usage.
5332
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005333load-server-state-from-file { global | local | none }
5334 Allow seamless reload of HAProxy
5335 May be used in sections: defaults | frontend | listen | backend
5336 yes | no | yes | yes
5337
5338 This directive points HAProxy to a file where server state from previous
5339 running process has been saved. That way, when starting up, before handling
5340 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01005341 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005342 to tell haproxy which file to use. For now, only 2 arguments to either prevent
5343 loading state or load states from a file containing all backends and servers.
5344 The state file can be generated by running the command "show servers state"
5345 over the stats socket and redirect output.
5346
Davor Ocelice9ed2812017-12-25 17:49:28 +01005347 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005348 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02005349 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005350
5351 Arguments:
5352 global load the content of the file pointed by the global directive
5353 named "server-state-file".
5354
5355 local load the content of the file pointed by the directive
5356 "server-state-file-name" if set. If not set, then the backend
5357 name is used as a file name.
5358
5359 none don't load any stat for this backend
5360
5361 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01005362 - server's IP address is preserved across reloads by default, but the
5363 order can be changed thanks to the server's "init-addr" setting. This
5364 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01005365 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01005366 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005367
5368 - server's weight is applied from previous running process unless it has
5369 has changed between previous and new configuration files.
5370
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005371 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005372
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005373 global
5374 stats socket /tmp/socket
5375 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005376
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005377 defaults
5378 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005379
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005380 backend bk
5381 server s1 127.0.0.1:22 check weight 11
5382 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005383
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005384
5385 Then one can run :
5386
5387 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
5388
5389 Content of the file /tmp/server_state would be like this:
5390
5391 1
5392 # <field names skipped for the doc example>
5393 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5394 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5395
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005396 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005397
5398 global
5399 stats socket /tmp/socket
5400 server-state-base /etc/haproxy/states
5401
5402 defaults
5403 load-server-state-from-file local
5404
5405 backend bk
5406 server s1 127.0.0.1:22 check weight 11
5407 server s2 127.0.0.1:22 check weight 12
5408
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005409
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005410 Then one can run :
5411
5412 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
5413
5414 Content of the file /etc/haproxy/states/bk would be like this:
5415
5416 1
5417 # <field names skipped for the doc example>
5418 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5419 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5420
5421 See also: "server-state-file", "server-state-file-name", and
5422 "show servers state"
5423
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005424
Willy Tarreau2769aa02007-12-27 18:26:09 +01005425log global
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005426log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
5427 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02005428no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01005429 Enable per-instance logging of events and traffic.
5430 May be used in sections : defaults | frontend | listen | backend
5431 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02005432
5433 Prefix :
5434 no should be used when the logger list must be flushed. For example,
5435 if you don't want to inherit from the default logger list. This
5436 prefix does not allow arguments.
5437
Willy Tarreau2769aa02007-12-27 18:26:09 +01005438 Arguments :
5439 global should be used when the instance's logging parameters are the
5440 same as the global ones. This is the most common usage. "global"
5441 replaces <address>, <facility> and <level> with those of the log
5442 entries found in the "global" section. Only one "log global"
5443 statement may be used per instance, and this form takes no other
5444 parameter.
5445
5446 <address> indicates where to send the logs. It takes the same format as
5447 for the "global" section's logs, and can be one of :
5448
5449 - An IPv4 address optionally followed by a colon (':') and a UDP
5450 port. If no port is specified, 514 is used by default (the
5451 standard syslog port).
5452
David du Colombier24bb5f52011-03-17 10:40:23 +01005453 - An IPv6 address followed by a colon (':') and optionally a UDP
5454 port. If no port is specified, 514 is used by default (the
5455 standard syslog port).
5456
Willy Tarreau2769aa02007-12-27 18:26:09 +01005457 - A filesystem path to a UNIX domain socket, keeping in mind
5458 considerations for chroot (be sure the path is accessible
5459 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01005460 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005461
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005462 - A file descriptor number in the form "fd@<number>", which may
5463 point to a pipe, terminal, or socket. In this case unbuffered
5464 logs are used and one writev() call per log is performed. This
5465 is a bit expensive but acceptable for most workloads. Messages
5466 sent this way will not be truncated but may be dropped, in
5467 which case the DroppedLogs counter will be incremented. The
5468 writev() call is atomic even on pipes for messages up to
5469 PIPE_BUF size, which POSIX recommends to be at least 512 and
5470 which is 4096 bytes on most modern operating systems. Any
5471 larger message may be interleaved with messages from other
5472 processes. Exceptionally for debugging purposes the file
5473 descriptor may also be directed to a file, but doing so will
5474 significantly slow haproxy down as non-blocking calls will be
5475 ignored. Also there will be no way to purge nor rotate this
5476 file without restarting the process. Note that the configured
5477 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01005478 with a TCP syslog server. See also the "short" and "raw"
5479 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005480
5481 - "stdout" / "stderr", which are respectively aliases for "fd@1"
5482 and "fd@2", see above.
5483
5484 You may want to reference some environment variables in the
5485 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01005486
Willy Tarreau18324f52014-06-27 18:10:07 +02005487 <length> is an optional maximum line length. Log lines larger than this
5488 value will be truncated before being sent. The reason is that
5489 syslog servers act differently on log line length. All servers
5490 support the default value of 1024, but some servers simply drop
5491 larger lines while others do log them. If a server supports long
5492 lines, it may make sense to set this value here in order to avoid
5493 truncating long lines. Similarly, if a server drops long lines,
5494 it is preferable to truncate them before sending them. Accepted
5495 values are 80 to 65535 inclusive. The default value of 1024 is
5496 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01005497 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02005498
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005499 <ranges> A list of comma-separated ranges to identify the logs to sample.
5500 This is used to balance the load of the logs to send to the log
5501 server. The limits of the ranges cannot be null. They are numbered
5502 from 1. The size or period (in number of logs) of the sample must
5503 be set with <sample_size> parameter.
5504
5505 <sample_size>
5506 The size of the sample in number of logs to consider when balancing
5507 their logging loads. It is used to balance the load of the logs to
5508 send to the syslog server. This size must be greater or equal to the
5509 maximum of the high limits of the ranges.
5510 (see also <ranges> parameter).
5511
Willy Tarreauadb345d2018-11-12 07:56:13 +01005512 <format> is the log format used when generating syslog messages. It may be
5513 one of the following :
5514
5515 rfc3164 The RFC3164 syslog message format. This is the default.
5516 (https://tools.ietf.org/html/rfc3164)
5517
5518 rfc5424 The RFC5424 syslog message format.
5519 (https://tools.ietf.org/html/rfc5424)
5520
Willy Tarreaue8746a02018-11-12 08:45:00 +01005521 short A message containing only a level between angle brackets such as
5522 '<3>', followed by the text. The PID, date, time, process name
5523 and system name are omitted. This is designed to be used with a
5524 local log server. This format is compatible with what the
5525 systemd logger consumes.
5526
Willy Tarreauc1b06452018-11-12 11:57:56 +01005527 raw A message containing only the text. The level, PID, date, time,
5528 process name and system name are omitted. This is designed to
5529 be used in containers or during development, where the severity
5530 only depends on the file descriptor used (stdout/stderr).
5531
Willy Tarreau2769aa02007-12-27 18:26:09 +01005532 <facility> must be one of the 24 standard syslog facilities :
5533
Willy Tarreaue8746a02018-11-12 08:45:00 +01005534 kern user mail daemon auth syslog lpr news
5535 uucp cron auth2 ftp ntp audit alert cron2
5536 local0 local1 local2 local3 local4 local5 local6 local7
5537
Willy Tarreauc1b06452018-11-12 11:57:56 +01005538 Note that the facility is ignored for the "short" and "raw"
5539 formats, but still required as a positional field. It is
5540 recommended to use "daemon" in this case to make it clear that
5541 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005542
5543 <level> is optional and can be specified to filter outgoing messages. By
5544 default, all messages are sent. If a level is specified, only
5545 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005546 will be sent. An optional minimum level can be specified. If it
5547 is set, logs emitted with a more severe level than this one will
5548 be capped to this level. This is used to avoid sending "emerg"
5549 messages on all terminals on some default syslog configurations.
5550 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005551
5552 emerg alert crit err warning notice info debug
5553
William Lallemand0f99e342011-10-12 17:50:54 +02005554 It is important to keep in mind that it is the frontend which decides what to
5555 log from a connection, and that in case of content switching, the log entries
5556 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005557
5558 However, backend log declaration define how and where servers status changes
5559 will be logged. Level "notice" will be used to indicate a server going up,
5560 "warning" will be used for termination signals and definitive service
5561 termination, and "alert" will be used for when a server goes down.
5562
5563 Note : According to RFC3164, messages are truncated to 1024 bytes before
5564 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005565
5566 Example :
5567 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01005568 log stdout format short daemon # send log to systemd
5569 log stdout format raw daemon # send everything to stdout
5570 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005571 log 127.0.0.1:514 local0 notice # only send important events
5572 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02005573 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01005574
Willy Tarreau2769aa02007-12-27 18:26:09 +01005575
William Lallemand48940402012-01-30 16:47:22 +01005576log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005577 Specifies the log format string to use for traffic logs
5578 May be used in sections: defaults | frontend | listen | backend
5579 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01005580
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005581 This directive specifies the log format string that will be used for all logs
5582 resulting from traffic passing through the frontend using this line. If the
5583 directive is used in a defaults section, all subsequent frontends will use
5584 the same log format. Please see section 8.2.4 which covers the log format
5585 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01005586
Guillaume de Lafond29f45602017-03-31 19:52:15 +02005587 "log-format" directive overrides previous "option tcplog", "log-format" and
5588 "option httplog" directives.
5589
Dragan Dosen7ad31542015-09-28 17:16:47 +02005590log-format-sd <string>
5591 Specifies the RFC5424 structured-data log format string
5592 May be used in sections: defaults | frontend | listen | backend
5593 yes | yes | yes | no
5594
5595 This directive specifies the RFC5424 structured-data log format string that
5596 will be used for all logs resulting from traffic passing through the frontend
5597 using this line. If the directive is used in a defaults section, all
5598 subsequent frontends will use the same log format. Please see section 8.2.4
5599 which covers the log format string in depth.
5600
5601 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
5602 about the RFC5424 structured-data part.
5603
5604 Note : This log format string will be used only for loggers that have set
5605 log format to "rfc5424".
5606
5607 Example :
5608 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
5609
5610
Willy Tarreau094af4e2015-01-07 15:03:42 +01005611log-tag <string>
5612 Specifies the log tag to use for all outgoing logs
5613 May be used in sections: defaults | frontend | listen | backend
5614 yes | yes | yes | yes
5615
5616 Sets the tag field in the syslog header to this string. It defaults to the
5617 log-tag set in the global section, otherwise the program name as launched
5618 from the command line, which usually is "haproxy". Sometimes it can be useful
5619 to differentiate between multiple processes running on the same host, or to
5620 differentiate customer instances running in the same process. In the backend,
5621 logs about servers up/down will use this tag. As a hint, it can be convenient
5622 to set a log-tag related to a hosted customer in a defaults section then put
5623 all the frontends and backends for that customer, then start another customer
5624 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005625
Willy Tarreauc35362a2014-04-25 13:58:37 +02005626max-keep-alive-queue <value>
5627 Set the maximum server queue size for maintaining keep-alive connections
5628 May be used in sections: defaults | frontend | listen | backend
5629 yes | no | yes | yes
5630
5631 HTTP keep-alive tries to reuse the same server connection whenever possible,
5632 but sometimes it can be counter-productive, for example if a server has a lot
5633 of connections while other ones are idle. This is especially true for static
5634 servers.
5635
5636 The purpose of this setting is to set a threshold on the number of queued
5637 connections at which haproxy stops trying to reuse the same server and prefers
5638 to find another one. The default value, -1, means there is no limit. A value
5639 of zero means that keep-alive requests will never be queued. For very close
5640 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01005641 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02005642 use a value of 10 or less). For remote servers suffering from a high latency,
5643 higher values might be needed to cover for the latency and/or the cost of
5644 picking a different server.
5645
5646 Note that this has no impact on responses which are maintained to the same
5647 server consecutively to a 401 response. They will still go to the same server
5648 even if they have to be queued.
5649
5650 See also : "option http-server-close", "option prefer-last-server", server
5651 "maxconn" and cookie persistence.
5652
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01005653max-session-srv-conns <nb>
5654 Set the maximum number of outgoing connections we can keep idling for a given
5655 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
5656 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02005657
Willy Tarreau2769aa02007-12-27 18:26:09 +01005658maxconn <conns>
5659 Fix the maximum number of concurrent connections on a frontend
5660 May be used in sections : defaults | frontend | listen | backend
5661 yes | yes | yes | no
5662 Arguments :
5663 <conns> is the maximum number of concurrent connections the frontend will
5664 accept to serve. Excess connections will be queued by the system
5665 in the socket's listen queue and will be served once a connection
5666 closes.
5667
5668 If the system supports it, it can be useful on big sites to raise this limit
5669 very high so that haproxy manages connection queues, instead of leaving the
5670 clients with unanswered connection attempts. This value should not exceed the
5671 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01005672 of tune.bufsize (16kB by default) each, as well as some other data resulting
5673 in about 33 kB of RAM being consumed per established connection. That means
5674 that a medium system equipped with 1GB of RAM can withstand around
5675 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005676
5677 Also, when <conns> is set to large values, it is possible that the servers
5678 are not sized to accept such loads, and for this reason it is generally wise
5679 to assign them some reasonable connection limits.
5680
Willy Tarreauc8d5b952019-02-27 17:25:52 +01005681 When this value is set to zero, which is the default, the global "maxconn"
5682 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02005683
Willy Tarreau2769aa02007-12-27 18:26:09 +01005684 See also : "server", global section's "maxconn", "fullconn"
5685
5686
5687mode { tcp|http|health }
5688 Set the running mode or protocol of the instance
5689 May be used in sections : defaults | frontend | listen | backend
5690 yes | yes | yes | yes
5691 Arguments :
5692 tcp The instance will work in pure TCP mode. A full-duplex connection
5693 will be established between clients and servers, and no layer 7
5694 examination will be performed. This is the default mode. It
5695 should be used for SSL, SSH, SMTP, ...
5696
5697 http The instance will work in HTTP mode. The client request will be
5698 analyzed in depth before connecting to any server. Any request
5699 which is not RFC-compliant will be rejected. Layer 7 filtering,
5700 processing and switching will be possible. This is the mode which
5701 brings HAProxy most of its value.
5702
5703 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02005704 to incoming connections and close the connection. Alternatively,
5705 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
5706 instead. Nothing will be logged in either case. This mode is used
5707 to reply to external components health checks. This mode is
5708 deprecated and should not be used anymore as it is possible to do
5709 the same and even better by combining TCP or HTTP modes with the
5710 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005711
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005712 When doing content switching, it is mandatory that the frontend and the
5713 backend are in the same mode (generally HTTP), otherwise the configuration
5714 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005715
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005716 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005717 defaults http_instances
5718 mode http
5719
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005720 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005721
Willy Tarreau0ba27502007-12-24 16:55:16 +01005722
Cyril Bontéf0c60612010-02-06 14:44:47 +01005723monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01005724 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005725 May be used in sections : defaults | frontend | listen | backend
5726 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01005727 Arguments :
5728 if <cond> the monitor request will fail if the condition is satisfied,
5729 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005730 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01005731 are met, for instance a low number of servers both in a
5732 backend and its backup.
5733
5734 unless <cond> the monitor request will succeed only if the condition is
5735 satisfied, and will fail otherwise. Such a condition may be
5736 based on a test on the presence of a minimum number of active
5737 servers in a list of backends.
5738
5739 This statement adds a condition which can force the response to a monitor
5740 request to report a failure. By default, when an external component queries
5741 the URI dedicated to monitoring, a 200 response is returned. When one of the
5742 conditions above is met, haproxy will return 503 instead of 200. This is
5743 very useful to report a site failure to an external component which may base
5744 routing advertisements between multiple sites on the availability reported by
5745 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005746 criterion. Note that "monitor fail" only works in HTTP mode. Both status
5747 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005748
5749 Example:
5750 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01005751 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01005752 acl site_dead nbsrv(dynamic) lt 2
5753 acl site_dead nbsrv(static) lt 2
5754 monitor-uri /site_alive
5755 monitor fail if site_dead
5756
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005757 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005758
5759
5760monitor-net <source>
5761 Declare a source network which is limited to monitor requests
5762 May be used in sections : defaults | frontend | listen | backend
5763 yes | yes | yes | no
5764 Arguments :
5765 <source> is the source IPv4 address or network which will only be able to
5766 get monitor responses to any request. It can be either an IPv4
5767 address, a host name, or an address followed by a slash ('/')
5768 followed by a mask.
5769
5770 In TCP mode, any connection coming from a source matching <source> will cause
5771 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005772 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01005773 forwarding the connection to a remote server.
5774
5775 In HTTP mode, a connection coming from a source matching <source> will be
5776 accepted, the following response will be sent without waiting for a request,
5777 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
5778 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02005779 running without forwarding the request to a backend server. Note that this
5780 response is sent in raw format, without any transformation. This is important
5781 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005782
Willy Tarreau82569f92012-09-27 23:48:56 +02005783 Monitor requests are processed very early, just after tcp-request connection
5784 ACLs which are the only ones able to block them. These connections are short
5785 lived and never wait for any data from the client. They cannot be logged, and
5786 it is the intended purpose. They are only used to report HAProxy's health to
5787 an upper component, nothing more. Please note that "monitor fail" rules do
5788 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005789
Willy Tarreau95cd2832010-03-04 23:36:33 +01005790 Last, please note that only one "monitor-net" statement can be specified in
5791 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005792
Willy Tarreau2769aa02007-12-27 18:26:09 +01005793 Example :
5794 # addresses .252 and .253 are just probing us.
5795 frontend www
5796 monitor-net 192.168.0.252/31
5797
5798 See also : "monitor fail", "monitor-uri"
5799
5800
5801monitor-uri <uri>
5802 Intercept a URI used by external components' monitor requests
5803 May be used in sections : defaults | frontend | listen | backend
5804 yes | yes | yes | no
5805 Arguments :
5806 <uri> is the exact URI which we want to intercept to return HAProxy's
5807 health status instead of forwarding the request.
5808
5809 When an HTTP request referencing <uri> will be received on a frontend,
5810 HAProxy will not forward it nor log it, but instead will return either
5811 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
5812 conditions defined with "monitor fail". This is normally enough for any
5813 front-end HTTP probe to detect that the service is UP and running without
5814 forwarding the request to a backend server. Note that the HTTP method, the
5815 version and all headers are ignored, but the request must at least be valid
5816 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
5817
Willy Tarreau721d8e02017-12-01 18:25:08 +01005818 Monitor requests are processed very early, just after the request is parsed
5819 and even before any "http-request" or "block" rulesets. The only rulesets
5820 applied before are the tcp-request ones. They cannot be logged either, and it
5821 is the intended purpose. They are only used to report HAProxy's health to an
5822 upper component, nothing more. However, it is possible to add any number of
5823 conditions using "monitor fail" and ACLs so that the result can be adjusted
5824 to whatever check can be imagined (most often the number of available servers
5825 in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005826
5827 Example :
5828 # Use /haproxy_test to report haproxy's status
5829 frontend www
5830 mode http
5831 monitor-uri /haproxy_test
5832
5833 See also : "monitor fail", "monitor-net"
5834
Willy Tarreau0ba27502007-12-24 16:55:16 +01005835
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005836option abortonclose
5837no option abortonclose
5838 Enable or disable early dropping of aborted requests pending in queues.
5839 May be used in sections : defaults | frontend | listen | backend
5840 yes | no | yes | yes
5841 Arguments : none
5842
5843 In presence of very high loads, the servers will take some time to respond.
5844 The per-instance connection queue will inflate, and the response time will
5845 increase respective to the size of the queue times the average per-session
5846 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01005847 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005848 the queue, and slowing down other users, and the servers as well, because the
5849 request will eventually be served, then aborted at the first error
5850 encountered while delivering the response.
5851
5852 As there is no way to distinguish between a full STOP and a simple output
5853 close on the client side, HTTP agents should be conservative and consider
5854 that the client might only have closed its output channel while waiting for
5855 the response. However, this introduces risks of congestion when lots of users
5856 do the same, and is completely useless nowadays because probably no client at
5857 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01005858 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005859 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01005860 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005861 of being the single component to break rare but valid traffic is extremely
5862 low, which adds to the temptation to be able to abort a session early while
5863 still not served and not pollute the servers.
5864
Davor Ocelice9ed2812017-12-25 17:49:28 +01005865 In HAProxy, the user can choose the desired behavior using the option
5866 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005867 compliant and aborted requests will be served. But when the option is
5868 specified, a session with an incoming channel closed will be aborted while
5869 it is still possible, either pending in the queue for a connection slot, or
5870 during the connection establishment if the server has not yet acknowledged
5871 the connection request. This considerably reduces the queue size and the load
5872 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005873 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005874
5875 If this option has been enabled in a "defaults" section, it can be disabled
5876 in a specific instance by prepending the "no" keyword before it.
5877
5878 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
5879
5880
Willy Tarreau4076a152009-04-02 15:18:36 +02005881option accept-invalid-http-request
5882no option accept-invalid-http-request
5883 Enable or disable relaxing of HTTP request parsing
5884 May be used in sections : defaults | frontend | listen | backend
5885 yes | yes | yes | no
5886 Arguments : none
5887
Willy Tarreau91852eb2015-05-01 13:26:00 +02005888 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005889 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005890 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005891 forbidden characters are essentially used to build attacks exploiting server
5892 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5893 server will emit invalid header names for whatever reason (configuration,
5894 implementation) and the issue will not be immediately fixed. In such a case,
5895 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01005896 even if that does not make sense, by specifying this option. Similarly, the
5897 list of characters allowed to appear in a URI is well defined by RFC3986, and
5898 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
5899 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01005900 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02005901 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02005902 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
5903 to pass through (no version specified) and multiple digits for both the major
5904 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005905
5906 This option should never be enabled by default as it hides application bugs
5907 and open security breaches. It should only be deployed after a problem has
5908 been confirmed.
5909
5910 When this option is enabled, erroneous header names will still be accepted in
5911 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01005912 analysis using the "show errors" request on the UNIX stats socket. Similarly,
5913 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02005914 also helps confirming that the issue has been solved.
5915
5916 If this option has been enabled in a "defaults" section, it can be disabled
5917 in a specific instance by prepending the "no" keyword before it.
5918
5919 See also : "option accept-invalid-http-response" and "show errors" on the
5920 stats socket.
5921
5922
5923option accept-invalid-http-response
5924no option accept-invalid-http-response
5925 Enable or disable relaxing of HTTP response parsing
5926 May be used in sections : defaults | frontend | listen | backend
5927 yes | no | yes | yes
5928 Arguments : none
5929
Willy Tarreau91852eb2015-05-01 13:26:00 +02005930 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005931 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005932 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005933 forbidden characters are essentially used to build attacks exploiting server
5934 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5935 server will emit invalid header names for whatever reason (configuration,
5936 implementation) and the issue will not be immediately fixed. In such a case,
5937 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02005938 even if that does not make sense, by specifying this option. This option also
5939 relaxes the test on the HTTP version format, it allows multiple digits for
5940 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005941
5942 This option should never be enabled by default as it hides application bugs
5943 and open security breaches. It should only be deployed after a problem has
5944 been confirmed.
5945
5946 When this option is enabled, erroneous header names will still be accepted in
5947 responses, but the complete response will be captured in order to permit
5948 later analysis using the "show errors" request on the UNIX stats socket.
5949 Doing this also helps confirming that the issue has been solved.
5950
5951 If this option has been enabled in a "defaults" section, it can be disabled
5952 in a specific instance by prepending the "no" keyword before it.
5953
5954 See also : "option accept-invalid-http-request" and "show errors" on the
5955 stats socket.
5956
5957
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005958option allbackups
5959no option allbackups
5960 Use either all backup servers at a time or only the first one
5961 May be used in sections : defaults | frontend | listen | backend
5962 yes | no | yes | yes
5963 Arguments : none
5964
5965 By default, the first operational backup server gets all traffic when normal
5966 servers are all down. Sometimes, it may be preferred to use multiple backups
5967 at once, because one will not be enough. When "option allbackups" is enabled,
5968 the load balancing will be performed among all backup servers when all normal
5969 ones are unavailable. The same load balancing algorithm will be used and the
5970 servers' weights will be respected. Thus, there will not be any priority
5971 order between the backup servers anymore.
5972
5973 This option is mostly used with static server farms dedicated to return a
5974 "sorry" page when an application is completely offline.
5975
5976 If this option has been enabled in a "defaults" section, it can be disabled
5977 in a specific instance by prepending the "no" keyword before it.
5978
5979
5980option checkcache
5981no option checkcache
Godbach7056a352013-12-11 20:01:07 +08005982 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005983 May be used in sections : defaults | frontend | listen | backend
5984 yes | no | yes | yes
5985 Arguments : none
5986
5987 Some high-level frameworks set application cookies everywhere and do not
5988 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005989 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005990 high risk of session crossing or stealing between users traversing the same
5991 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005992 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005993
5994 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005995 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01005996 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005997 response to check if there's a risk of caching a cookie on a client-side
5998 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01005999 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01006000 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01006001 - all those with a return code other than 200, 203, 204, 206, 300, 301,
6002 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01006003 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01006004 - all those that result from a request using a method other than GET, HEAD,
6005 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01006006 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006007 - those with a 'Pragma: no-cache' header
6008 - those with a 'Cache-control: private' header
6009 - those with a 'Cache-control: no-store' header
6010 - those with a 'Cache-control: max-age=0' header
6011 - those with a 'Cache-control: s-maxage=0' header
6012 - those with a 'Cache-control: no-cache' header
6013 - those with a 'Cache-control: no-cache="set-cookie"' header
6014 - those with a 'Cache-control: no-cache="set-cookie,' header
6015 (allowing other fields after set-cookie)
6016
6017 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01006018 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006019 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006020 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006021 that admins are informed that there's something to be fixed.
6022
6023 Due to the high impact on the application, the application should be tested
6024 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006025 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01006026 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006027
6028 If this option has been enabled in a "defaults" section, it can be disabled
6029 in a specific instance by prepending the "no" keyword before it.
6030
6031
6032option clitcpka
6033no option clitcpka
6034 Enable or disable the sending of TCP keepalive packets on the client side
6035 May be used in sections : defaults | frontend | listen | backend
6036 yes | yes | yes | no
6037 Arguments : none
6038
6039 When there is a firewall or any session-aware component between a client and
6040 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01006041 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006042 components decides to expire a session which has remained idle for too long.
6043
6044 Enabling socket-level TCP keep-alives makes the system regularly send packets
6045 to the other end of the connection, leaving it active. The delay between
6046 keep-alive probes is controlled by the system only and depends both on the
6047 operating system and its tuning parameters.
6048
6049 It is important to understand that keep-alive packets are neither emitted nor
6050 received at the application level. It is only the network stacks which sees
6051 them. For this reason, even if one side of the proxy already uses keep-alives
6052 to maintain its connection alive, those keep-alive packets will not be
6053 forwarded to the other side of the proxy.
6054
6055 Please note that this has nothing to do with HTTP keep-alive.
6056
6057 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
6058 client side of a connection, which should help when session expirations are
6059 noticed between HAProxy and a client.
6060
6061 If this option has been enabled in a "defaults" section, it can be disabled
6062 in a specific instance by prepending the "no" keyword before it.
6063
6064 See also : "option srvtcpka", "option tcpka"
6065
6066
Willy Tarreau0ba27502007-12-24 16:55:16 +01006067option contstats
6068 Enable continuous traffic statistics updates
6069 May be used in sections : defaults | frontend | listen | backend
6070 yes | yes | yes | no
6071 Arguments : none
6072
6073 By default, counters used for statistics calculation are incremented
6074 only when a session finishes. It works quite well when serving small
6075 objects, but with big ones (for example large images or archives) or
6076 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01006077 a hedgehog. With this option enabled counters get incremented frequently
6078 along the session, typically every 5 seconds, which is often enough to
6079 produce clean graphs. Recounting touches a hotpath directly so it is not
6080 not enabled by default, as it can cause a lot of wakeups for very large
6081 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006082
6083
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006084option dontlog-normal
6085no option dontlog-normal
6086 Enable or disable logging of normal, successful connections
6087 May be used in sections : defaults | frontend | listen | backend
6088 yes | yes | yes | no
6089 Arguments : none
6090
6091 There are large sites dealing with several thousand connections per second
6092 and for which logging is a major pain. Some of them are even forced to turn
6093 logs off and cannot debug production issues. Setting this option ensures that
6094 normal connections, those which experience no error, no timeout, no retry nor
6095 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
6096 mode, the response status code is checked and return codes 5xx will still be
6097 logged.
6098
6099 It is strongly discouraged to use this option as most of the time, the key to
6100 complex issues is in the normal logs which will not be logged here. If you
6101 need to separate logs, see the "log-separate-errors" option instead.
6102
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006103 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006104 logging.
6105
6106
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006107option dontlognull
6108no option dontlognull
6109 Enable or disable logging of null connections
6110 May be used in sections : defaults | frontend | listen | backend
6111 yes | yes | yes | no
6112 Arguments : none
6113
6114 In certain environments, there are components which will regularly connect to
6115 various systems to ensure that they are still alive. It can be the case from
6116 another load balancer as well as from monitoring systems. By default, even a
6117 simple port probe or scan will produce a log. If those connections pollute
6118 the logs too much, it is possible to enable option "dontlognull" to indicate
6119 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02006120 which typically corresponds to those probes. Note that errors will still be
6121 returned to the client and accounted for in the stats. If this is not what is
6122 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006123
6124 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01006125 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006126 would not be logged.
6127
6128 If this option has been enabled in a "defaults" section, it can be disabled
6129 in a specific instance by prepending the "no" keyword before it.
6130
Willy Tarreau0f228a02015-05-01 15:37:53 +02006131 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
6132 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006133
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006134
Willy Tarreau87cf5142011-08-19 22:57:24 +02006135option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006136 Enable insertion of the X-Forwarded-For header to requests sent to servers
6137 May be used in sections : defaults | frontend | listen | backend
6138 yes | yes | yes | yes
6139 Arguments :
6140 <network> is an optional argument used to disable this option for sources
6141 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02006142 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01006143 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006144
6145 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
6146 their client address. This is sometimes annoying when the client's IP address
6147 is expected in server logs. To solve this problem, the well-known HTTP header
6148 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
6149 This header contains a value representing the client's IP address. Since this
6150 header is always appended at the end of the existing header list, the server
6151 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02006152 the server's manual to find how to enable use of this standard header. Note
6153 that only the last occurrence of the header must be used, since it is really
6154 possible that the client has already brought one.
6155
Willy Tarreaud72758d2010-01-12 10:42:19 +01006156 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02006157 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01006158 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01006159 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006160 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02006161 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01006162
6163 Sometimes, a same HAProxy instance may be shared between a direct client
6164 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6165 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6166 header for a known source address or network by adding the "except" keyword
6167 followed by the network address. In this case, any source IP matching the
6168 network will not cause an addition of this header. Most common uses are with
6169 private networks or 127.0.0.1.
6170
Willy Tarreau87cf5142011-08-19 22:57:24 +02006171 Alternatively, the keyword "if-none" states that the header will only be
6172 added if it is not present. This should only be used in perfectly trusted
6173 environment, as this might cause a security issue if headers reaching haproxy
6174 are under the control of the end-user.
6175
Willy Tarreauc27debf2008-01-06 08:57:02 +01006176 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02006177 least one of them uses it, the header will be added. Note that the backend's
6178 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02006179 both are defined. In the case of the "if-none" argument, if at least one of
6180 the frontend or the backend does not specify it, it wants the addition to be
6181 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006182
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02006183 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01006184 # Public HTTP address also used by stunnel on the same machine
6185 frontend www
6186 mode http
6187 option forwardfor except 127.0.0.1 # stunnel already adds the header
6188
Ross Westaf72a1d2008-08-03 10:51:45 +02006189 # Those servers want the IP Address in X-Client
6190 backend www
6191 mode http
6192 option forwardfor header X-Client
6193
Willy Tarreau87cf5142011-08-19 22:57:24 +02006194 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006195 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01006196
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006197
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006198option http-buffer-request
6199no option http-buffer-request
6200 Enable or disable waiting for whole HTTP request body before proceeding
6201 May be used in sections : defaults | frontend | listen | backend
6202 yes | yes | yes | yes
6203 Arguments : none
6204
6205 It is sometimes desirable to wait for the body of an HTTP request before
6206 taking a decision. This is what is being done by "balance url_param" for
6207 example. The first use case is to buffer requests from slow clients before
6208 connecting to the server. Another use case consists in taking the routing
6209 decision based on the request body's contents. This option placed in a
6210 frontend or backend forces the HTTP processing to wait until either the whole
6211 body is received, or the request buffer is full, or the first chunk is
6212 complete in case of chunked encoding. It can have undesired side effects with
Tim Düsterhus4896c442016-11-29 02:15:19 +01006213 some applications abusing HTTP by expecting unbuffered transmissions between
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006214 the frontend and the backend, so this should definitely not be used by
6215 default.
6216
Baptiste Assmanneccdf432015-10-28 13:49:01 +01006217 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006218
6219
Willy Tarreau0f228a02015-05-01 15:37:53 +02006220option http-ignore-probes
6221no option http-ignore-probes
6222 Enable or disable logging of null connections and request timeouts
6223 May be used in sections : defaults | frontend | listen | backend
6224 yes | yes | yes | no
6225 Arguments : none
6226
6227 Recently some browsers started to implement a "pre-connect" feature
6228 consisting in speculatively connecting to some recently visited web sites
6229 just in case the user would like to visit them. This results in many
6230 connections being established to web sites, which end up in 408 Request
6231 Timeout if the timeout strikes first, or 400 Bad Request when the browser
6232 decides to close them first. These ones pollute the log and feed the error
6233 counters. There was already "option dontlognull" but it's insufficient in
6234 this case. Instead, this option does the following things :
6235 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01006236 was received over a connection before it was closed;
6237 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02006238 - prevent any error counter from being incremented
6239
6240 That way the empty connection is silently ignored. Note that it is better
6241 not to use this unless it is clear that it is needed, because it will hide
6242 real problems. The most common reason for not receiving a request and seeing
6243 a 408 is due to an MTU inconsistency between the client and an intermediary
6244 element such as a VPN, which blocks too large packets. These issues are
6245 generally seen with POST requests as well as GET with large cookies. The logs
6246 are often the only way to detect them.
6247
6248 If this option has been enabled in a "defaults" section, it can be disabled
6249 in a specific instance by prepending the "no" keyword before it.
6250
6251 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
6252
6253
Willy Tarreau16bfb022010-01-16 19:48:41 +01006254option http-keep-alive
6255no option http-keep-alive
6256 Enable or disable HTTP keep-alive from client to server
6257 May be used in sections : defaults | frontend | listen | backend
6258 yes | yes | yes | yes
6259 Arguments : none
6260
Willy Tarreau70dffda2014-01-30 03:07:23 +01006261 By default HAProxy operates in keep-alive mode with regards to persistent
6262 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02006263 leaves the connection idle on both sides between the end of a response and
6264 the start of a new request. This mode may be changed by several options such
6265 as "option http-server-close", "option httpclose" or "option http-tunnel".
6266 This option allows to set back the keep-alive mode, which can be useful when
6267 another mode was used in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01006268
6269 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
6270 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01006271 network) and the fastest session reuse on the server side at the expense
6272 of maintaining idle connections to the servers. In general, it is possible
6273 with this option to achieve approximately twice the request rate that the
6274 "http-server-close" option achieves on small objects. There are mainly two
6275 situations where this option may be useful :
6276
6277 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01006278 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01006279
6280 - when the cost of establishing the connection to the server is significant
6281 compared to the cost of retrieving the associated object from the server.
6282
6283 This last case can happen when the server is a fast static server of cache.
6284 In this case, the server will need to be properly tuned to support high enough
6285 connection counts because connections will last until the client sends another
6286 request.
6287
6288 If the client request has to go to another backend or another server due to
6289 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01006290 immediately be closed and a new one re-opened. Option "prefer-last-server" is
6291 available to try optimize server selection so that if the server currently
6292 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01006293
Willy Tarreau16bfb022010-01-16 19:48:41 +01006294 At the moment, logs will not indicate whether requests came from the same
6295 session or not. The accept date reported in the logs corresponds to the end
6296 of the previous request, and the request time corresponds to the time spent
6297 waiting for a new request. The keep-alive request time is still bound to the
6298 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6299 not set.
6300
Cyril Bonté653dcd62014-02-20 00:13:15 +01006301 This option disables and replaces any previous "option httpclose", "option
Christopher Faulet315b39c2018-09-21 16:26:19 +02006302 http-server-close" or "option http-tunnel". When backend and frontend options
6303 differ, all of these 4 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006304
Christopher Faulet315b39c2018-09-21 16:26:19 +02006305 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01006306 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01006307 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006308
6309
Willy Tarreau96e31212011-05-30 18:10:30 +02006310option http-no-delay
6311no option http-no-delay
6312 Instruct the system to favor low interactive delays over performance in HTTP
6313 May be used in sections : defaults | frontend | listen | backend
6314 yes | yes | yes | yes
6315 Arguments : none
6316
6317 In HTTP, each payload is unidirectional and has no notion of interactivity.
6318 Any agent is expected to queue data somewhat for a reasonably low delay.
6319 There are some very rare server-to-server applications that abuse the HTTP
6320 protocol and expect the payload phase to be highly interactive, with many
6321 interleaved data chunks in both directions within a single request. This is
6322 absolutely not supported by the HTTP specification and will not work across
6323 most proxies or servers. When such applications attempt to do this through
6324 haproxy, it works but they will experience high delays due to the network
6325 optimizations which favor performance by instructing the system to wait for
6326 enough data to be available in order to only send full packets. Typical
6327 delays are around 200 ms per round trip. Note that this only happens with
6328 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
6329 affected.
6330
6331 When "option http-no-delay" is present in either the frontend or the backend
6332 used by a connection, all such optimizations will be disabled in order to
6333 make the exchanges as fast as possible. Of course this offers no guarantee on
6334 the functionality, as it may break at any other place. But if it works via
6335 HAProxy, it will work as fast as possible. This option should never be used
6336 by default, and should never be used at all unless such a buggy application
6337 is discovered. The impact of using this option is an increase of bandwidth
6338 usage and CPU usage, which may significantly lower performance in high
6339 latency environments.
6340
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006341 See also : "option http-buffer-request"
6342
Willy Tarreau96e31212011-05-30 18:10:30 +02006343
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006344option http-pretend-keepalive
6345no option http-pretend-keepalive
6346 Define whether haproxy will announce keepalive to the server or not
6347 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02006348 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006349 Arguments : none
6350
Christopher Faulet315b39c2018-09-21 16:26:19 +02006351 When running with "option http-server-close" or "option httpclose", haproxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006352 adds a "Connection: close" header to the request forwarded to the server.
6353 Unfortunately, when some servers see this header, they automatically refrain
6354 from using the chunked encoding for responses of unknown length, while this
6355 is totally unrelated. The immediate effect is that this prevents haproxy from
6356 maintaining the client connection alive. A second effect is that a client or
6357 a cache could receive an incomplete response without being aware of it, and
6358 consider the response complete.
6359
6360 By setting "option http-pretend-keepalive", haproxy will make the server
6361 believe it will keep the connection alive. The server will then not fall back
6362 to the abnormal undesired above. When haproxy gets the whole response, it
6363 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02006364 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006365 connection is correctly closed on the server side.
6366
6367 It is recommended not to enable this option by default, because most servers
6368 will more efficiently close the connection themselves after the last packet,
6369 and release its buffers slightly earlier. Also, the added packet on the
6370 network could slightly reduce the overall peak performance. However it is
6371 worth noting that when this option is enabled, haproxy will have slightly
6372 less work to do. So if haproxy is the bottleneck on the whole architecture,
6373 enabling this option might save a few CPU cycles.
6374
Christopher Faulet98db9762018-09-21 10:25:19 +02006375 This option may be set in backend and listen sections. Using it in a frontend
6376 section will be ignored and a warning will be reported during startup. It is
6377 a backend related option, so there is no real reason to set it on a
6378 frontend. This option may be combined with "option httpclose", which will
6379 cause keepalive to be announced to the server and close to be announced to
6380 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006381
6382 If this option has been enabled in a "defaults" section, it can be disabled
6383 in a specific instance by prepending the "no" keyword before it.
6384
Christopher Faulet315b39c2018-09-21 16:26:19 +02006385 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01006386 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006387
Willy Tarreauc27debf2008-01-06 08:57:02 +01006388
Willy Tarreaub608feb2010-01-02 22:47:18 +01006389option http-server-close
6390no option http-server-close
6391 Enable or disable HTTP connection closing on the server side
6392 May be used in sections : defaults | frontend | listen | backend
6393 yes | yes | yes | yes
6394 Arguments : none
6395
Willy Tarreau70dffda2014-01-30 03:07:23 +01006396 By default HAProxy operates in keep-alive mode with regards to persistent
6397 connections: for each connection it processes each request and response, and
6398 leaves the connection idle on both sides between the end of a response and
6399 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006400 as "option http-server-close", "option httpclose" or "option http-tunnel".
6401 Setting "option http-server-close" enables HTTP connection-close mode on the
6402 server side while keeping the ability to support HTTP keep-alive and
6403 pipelining on the client side. This provides the lowest latency on the client
6404 side (slow network) and the fastest session reuse on the server side to save
6405 server resources, similarly to "option httpclose". It also permits
6406 non-keepalive capable servers to be served in keep-alive mode to the clients
6407 if they conform to the requirements of RFC7230. Please note that some servers
6408 do not always conform to those requirements when they see "Connection: close"
6409 in the request. The effect will be that keep-alive will never be used. A
6410 workaround consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006411
6412 At the moment, logs will not indicate whether requests came from the same
6413 session or not. The accept date reported in the logs corresponds to the end
6414 of the previous request, and the request time corresponds to the time spent
6415 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01006416 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6417 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006418
6419 This option may be set both in a frontend and in a backend. It is enabled if
6420 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet315b39c2018-09-21 16:26:19 +02006421 It disables and replaces any previous "option httpclose", "option http-tunnel"
6422 or "option http-keep-alive". Please check section 4 ("Proxies") to see how
6423 this option combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006424
6425 If this option has been enabled in a "defaults" section, it can be disabled
6426 in a specific instance by prepending the "no" keyword before it.
6427
Christopher Faulet315b39c2018-09-21 16:26:19 +02006428 See also : "option httpclose", "option http-pretend-keepalive",
6429 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006430
6431
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006432option http-tunnel (deprecated)
6433no option http-tunnel (deprecated)
6434 Disable or enable HTTP connection processing after first transaction.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006435 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet4212a302018-09-21 10:42:19 +02006436 yes | yes | yes | no
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006437 Arguments : none
6438
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006439 Warning : Because it cannot work in HTTP/2, this option is deprecated and it
6440 is only supported on legacy HTTP frontends. In HTX, it is ignored and a
6441 warning is emitted during HAProxy startup.
6442
Willy Tarreau70dffda2014-01-30 03:07:23 +01006443 By default HAProxy operates in keep-alive mode with regards to persistent
6444 connections: for each connection it processes each request and response, and
6445 leaves the connection idle on both sides between the end of a response and
6446 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006447 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006448
6449 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006450 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01006451 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
6452 is normally not needed anymore unless in very specific cases such as when
6453 using an in-house protocol that looks like HTTP but is not compatible, or
6454 just to log one request per client in order to reduce log size. Note that
6455 everything which works at the HTTP level, including header parsing/addition,
6456 cookie processing or content switching will only work for the first request
6457 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006458
Christopher Faulet4212a302018-09-21 10:42:19 +02006459 This option may be set on frontend and listen sections. Using it on a backend
6460 section will be ignored and a warning will be reported during the startup. It
6461 is a frontend related option, so there is no real reason to set it on a
6462 backend.
6463
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006464 If this option has been enabled in a "defaults" section, it can be disabled
6465 in a specific instance by prepending the "no" keyword before it.
6466
Christopher Faulet315b39c2018-09-21 16:26:19 +02006467 See also : "option httpclose", "option http-server-close",
6468 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006469
6470
Willy Tarreau88d349d2010-01-25 12:15:43 +01006471option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01006472no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01006473 Make use of non-standard Proxy-Connection header instead of Connection
6474 May be used in sections : defaults | frontend | listen | backend
6475 yes | yes | yes | no
6476 Arguments : none
6477
Lukas Tribus23953682017-04-28 13:24:30 +00006478 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01006479 Connection header to indicate their wish of persistent or non-persistent
6480 connections, both browsers and proxies ignore this header for proxied
6481 connections and make use of the undocumented, non-standard Proxy-Connection
6482 header instead. The issue begins when trying to put a load balancer between
6483 browsers and such proxies, because there will be a difference between what
6484 haproxy understands and what the client and the proxy agree on.
6485
6486 By setting this option in a frontend, haproxy can automatically switch to use
6487 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006488 defined here as one where the URI begins with neither a '/' nor a '*'. This
6489 is incompatible with the HTTP tunnel mode. Note that this option can only be
6490 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01006491
Willy Tarreau844a7e72010-01-31 21:46:18 +01006492 Also, when this option is set, a request which requires authentication will
6493 automatically switch to use proxy authentication headers if it is itself a
6494 proxied request. That makes it possible to check or enforce authentication in
6495 front of an existing proxy.
6496
Willy Tarreau88d349d2010-01-25 12:15:43 +01006497 This option should normally never be used, except in front of a proxy.
6498
Christopher Faulet315b39c2018-09-21 16:26:19 +02006499 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01006500
6501
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006502option http-use-htx
6503no option http-use-htx
6504 Switch to the new HTX internal representation for HTTP protocol elements
6505 May be used in sections : defaults | frontend | listen | backend
6506 yes | yes | yes | yes
6507 Arguments : none
6508
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006509 Historically, the HTTP protocol is processed as-is. Inserting, deleting, or
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006510 modifying a header field requires to rewrite the affected part in the buffer
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006511 and to move the buffer's tail accordingly. This mode is known as the legacy
6512 HTTP mode. Since this principle has deep roots in haproxy, the HTTP/2
6513 protocol is converted to HTTP/1.1 before being processed this way. It also
6514 results in the inability to establish HTTP/2 connections to servers because
6515 of the loss of HTTP/2 semantics in the HTTP/1 representation.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006516
6517 HTX is the name of a totally new native internal representation for the HTTP
6518 protocol, that is agnostic to the version and aims at preserving semantics
6519 all along the chain. It relies on a fast parsing, tokenizing and indexing of
6520 the protocol elements so that no more memory moves are necessary and that
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006521 most elements are directly accessed. It supports using either HTTP/1 or
6522 HTTP/2 on any side regardless of the other side's version. It also supports
6523 upgrades from TCP to HTTP and implicit ones from HTTP/1 to HTTP/2 (matching
6524 the HTTP/2 preface).
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006525
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006526 This option indicates that HTX needs to be used. Since the version 2.0-dev3,
6527 the HTX is the default mode. To switch back on the legacy HTTP mode, the
6528 option must be explicitly disabled using the "no" prefix. For prior versions,
6529 the feature has incomplete functional coverage, so it is not enabled by
6530 default.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006531
6532 See also : "mode http"
6533
6534
Willy Tarreaud63335a2010-02-26 12:56:52 +01006535option httpchk
6536option httpchk <uri>
6537option httpchk <method> <uri>
6538option httpchk <method> <uri> <version>
6539 Enable HTTP protocol to check on the servers health
6540 May be used in sections : defaults | frontend | listen | backend
6541 yes | no | yes | yes
6542 Arguments :
6543 <method> is the optional HTTP method used with the requests. When not set,
6544 the "OPTIONS" method is used, as it generally requires low server
6545 processing and is easy to filter out from the logs. Any method
6546 may be used, though it is not recommended to invent non-standard
6547 ones.
6548
6549 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
6550 which is accessible by default on almost any server, but may be
6551 changed to any other URI. Query strings are permitted.
6552
6553 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
6554 but some servers might behave incorrectly in HTTP 1.0, so turning
6555 it to HTTP/1.1 may sometimes help. Note that the Host field is
6556 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
6557 after "\r\n" following the version string.
6558
6559 By default, server health checks only consist in trying to establish a TCP
6560 connection. When "option httpchk" is specified, a complete HTTP request is
6561 sent once the TCP connection is established, and responses 2xx and 3xx are
6562 considered valid, while all other ones indicate a server failure, including
6563 the lack of any response.
6564
6565 The port and interval are specified in the server configuration.
6566
6567 This option does not necessarily require an HTTP backend, it also works with
6568 plain TCP backends. This is particularly useful to check simple scripts bound
6569 to some dedicated ports using the inetd daemon.
6570
6571 Examples :
6572 # Relay HTTPS traffic to Apache instance and check service availability
6573 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
6574 backend https_relay
6575 mode tcp
6576 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
6577 server apache1 192.168.1.1:443 check port 80
6578
Simon Hormanafc47ee2013-11-25 10:46:35 +09006579 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
6580 "option pgsql-check", "http-check" and the "check", "port" and
6581 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006582
6583
Willy Tarreauc27debf2008-01-06 08:57:02 +01006584option httpclose
6585no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02006586 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01006587 May be used in sections : defaults | frontend | listen | backend
6588 yes | yes | yes | yes
6589 Arguments : none
6590
Willy Tarreau70dffda2014-01-30 03:07:23 +01006591 By default HAProxy operates in keep-alive mode with regards to persistent
6592 connections: for each connection it processes each request and response, and
6593 leaves the connection idle on both sides between the end of a response and
6594 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006595 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006596
Christopher Faulet315b39c2018-09-21 16:26:19 +02006597 If "option httpclose" is set, HAProxy will close connections with the server
6598 and the client as soon as the request and the response are received. It will
6599 alos check if a "Connection: close" header is already set in each direction,
6600 and will add one if missing. Any "Connection" header different from "close"
6601 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006602
Christopher Faulet315b39c2018-09-21 16:26:19 +02006603 This option may also be combined with "option http-pretend-keepalive", which
6604 will disable sending of the "Connection: close" header, but will still cause
6605 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006606
6607 This option may be set both in a frontend and in a backend. It is enabled if
6608 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01006609 It disables and replaces any previous "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006610 "option http-keep-alive" or "option http-tunnel". Please check section 4
6611 ("Proxies") to see how this option combines with others when frontend and
6612 backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006613
6614 If this option has been enabled in a "defaults" section, it can be disabled
6615 in a specific instance by prepending the "no" keyword before it.
6616
Christopher Faulet315b39c2018-09-21 16:26:19 +02006617 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01006618
6619
Emeric Brun3a058f32009-06-30 18:26:00 +02006620option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006621 Enable logging of HTTP request, session state and timers
6622 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006623 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02006624 Arguments :
6625 clf if the "clf" argument is added, then the output format will be
6626 the CLF format instead of HAProxy's default HTTP format. You can
6627 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01006628 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02006629 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006630
6631 By default, the log output format is very poor, as it only contains the
6632 source and destination addresses, and the instance name. By specifying
6633 "option httplog", each log line turns into a much richer format including,
6634 but not limited to, the HTTP request, the connection timers, the session
6635 status, the connections numbers, the captured headers and cookies, the
6636 frontend, backend and server name, and of course the source address and
6637 ports.
6638
PiBa-NLbd556bf2014-12-11 21:31:54 +01006639 Specifying only "option httplog" will automatically clear the 'clf' mode
6640 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02006641
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006642 "option httplog" overrides any previous "log-format" directive.
6643
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006644 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006645
Willy Tarreau55165fe2009-05-10 12:02:55 +02006646
6647option http_proxy
6648no option http_proxy
6649 Enable or disable plain HTTP proxy mode
6650 May be used in sections : defaults | frontend | listen | backend
6651 yes | yes | yes | yes
6652 Arguments : none
6653
6654 It sometimes happens that people need a pure HTTP proxy which understands
6655 basic proxy requests without caching nor any fancy feature. In this case,
6656 it may be worth setting up an HAProxy instance with the "option http_proxy"
6657 set. In this mode, no server is declared, and the connection is forwarded to
6658 the IP address and port found in the URL after the "http://" scheme.
6659
6660 No host address resolution is performed, so this only works when pure IP
6661 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006662 it will probably be used only by experts who know they need exactly it. This
6663 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02006664
6665 If this option has been enabled in a "defaults" section, it can be disabled
6666 in a specific instance by prepending the "no" keyword before it.
6667
6668 Example :
6669 # this backend understands HTTP proxy requests and forwards them directly.
6670 backend direct_forward
6671 option httpclose
6672 option http_proxy
6673
6674 See also : "option httpclose"
6675
Willy Tarreau211ad242009-10-03 21:45:07 +02006676
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006677option independent-streams
6678no option independent-streams
6679 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006680 May be used in sections : defaults | frontend | listen | backend
6681 yes | yes | yes | yes
6682 Arguments : none
6683
6684 By default, when data is sent over a socket, both the write timeout and the
6685 read timeout for that socket are refreshed, because we consider that there is
6686 activity on that socket, and we have no other means of guessing if we should
6687 receive data or not.
6688
Davor Ocelice9ed2812017-12-25 17:49:28 +01006689 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006690 exists a situation where it is desirable to disable it, and only refresh the
6691 read timeout if there are incoming data. This happens on sessions with large
6692 timeouts and low amounts of exchanged data such as telnet session. If the
6693 server suddenly disappears, the output data accumulates in the system's
6694 socket buffers, both timeouts are correctly refreshed, and there is no way
6695 to know the server does not receive them, so we don't timeout. However, when
6696 the underlying protocol always echoes sent data, it would be enough by itself
6697 to detect the issue using the read timeout. Note that this problem does not
6698 happen with more verbose protocols because data won't accumulate long in the
6699 socket buffers.
6700
6701 When this option is set on the frontend, it will disable read timeout updates
6702 on data sent to the client. There probably is little use of this case. When
6703 the option is set on the backend, it will disable read timeout updates on
6704 data sent to the server. Doing so will typically break large HTTP posts from
6705 slow lines, so use it with caution.
6706
Lukas Tribus745f15e2018-11-08 12:41:42 +01006707 Note: older versions used to call this setting "option independant-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006708 with a spelling mistake. This spelling is still supported but
6709 deprecated.
6710
Willy Tarreauce887fd2012-05-12 12:50:00 +02006711 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006712
6713
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02006714option ldap-check
6715 Use LDAPv3 health checks for server testing
6716 May be used in sections : defaults | frontend | listen | backend
6717 yes | no | yes | yes
6718 Arguments : none
6719
6720 It is possible to test that the server correctly talks LDAPv3 instead of just
6721 testing that it accepts the TCP connection. When this option is set, an
6722 LDAPv3 anonymous simple bind message is sent to the server, and the response
6723 is analyzed to find an LDAPv3 bind response message.
6724
6725 The server is considered valid only when the LDAP response contains success
6726 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
6727
6728 Logging of bind requests is server dependent see your documentation how to
6729 configure it.
6730
6731 Example :
6732 option ldap-check
6733
6734 See also : "option httpchk"
6735
6736
Simon Horman98637e52014-06-20 12:30:16 +09006737option external-check
6738 Use external processes for server health checks
6739 May be used in sections : defaults | frontend | listen | backend
6740 yes | no | yes | yes
6741
6742 It is possible to test the health of a server using an external command.
6743 This is achieved by running the executable set using "external-check
6744 command".
6745
6746 Requires the "external-check" global to be set.
6747
6748 See also : "external-check", "external-check command", "external-check path"
6749
6750
Willy Tarreau211ad242009-10-03 21:45:07 +02006751option log-health-checks
6752no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02006753 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02006754 May be used in sections : defaults | frontend | listen | backend
6755 yes | no | yes | yes
6756 Arguments : none
6757
Willy Tarreaubef1b322014-05-13 21:01:39 +02006758 By default, failed health check are logged if server is UP and successful
6759 health checks are logged if server is DOWN, so the amount of additional
6760 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02006761
Willy Tarreaubef1b322014-05-13 21:01:39 +02006762 When this option is enabled, any change of the health check status or to
6763 the server's health will be logged, so that it becomes possible to know
6764 that a server was failing occasional checks before crashing, or exactly when
6765 it failed to respond a valid HTTP status, then when the port started to
6766 reject connections, then when the server stopped responding at all.
6767
Davor Ocelice9ed2812017-12-25 17:49:28 +01006768 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02006769 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02006770
Willy Tarreaubef1b322014-05-13 21:01:39 +02006771 See also: "option httpchk", "option ldap-check", "option mysql-check",
6772 "option pgsql-check", "option redis-check", "option smtpchk",
6773 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02006774
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006775
6776option log-separate-errors
6777no option log-separate-errors
6778 Change log level for non-completely successful connections
6779 May be used in sections : defaults | frontend | listen | backend
6780 yes | yes | yes | no
6781 Arguments : none
6782
6783 Sometimes looking for errors in logs is not easy. This option makes haproxy
6784 raise the level of logs containing potentially interesting information such
6785 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
6786 level changes from "info" to "err". This makes it possible to log them
6787 separately to a different file with most syslog daemons. Be careful not to
6788 remove them from the original file, otherwise you would lose ordering which
6789 provides very important information.
6790
6791 Using this option, large sites dealing with several thousand connections per
6792 second may log normal traffic to a rotating buffer and only archive smaller
6793 error logs.
6794
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006795 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006796 logging.
6797
Willy Tarreauc27debf2008-01-06 08:57:02 +01006798
6799option logasap
6800no option logasap
6801 Enable or disable early logging of HTTP requests
6802 May be used in sections : defaults | frontend | listen | backend
6803 yes | yes | yes | no
6804 Arguments : none
6805
6806 By default, HTTP requests are logged upon termination so that the total
6807 transfer time and the number of bytes appear in the logs. When large objects
6808 are being transferred, it may take a while before the request appears in the
6809 logs. Using "option logasap", the request gets logged as soon as the server
6810 sends the complete headers. The only missing information in the logs will be
6811 the total number of bytes which will indicate everything except the amount
6812 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006813 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01006814 "Content-Length" response header so that the logs at least indicate how many
6815 bytes are expected to be transferred.
6816
Willy Tarreaucc6c8912009-02-22 10:53:55 +01006817 Examples :
6818 listen http_proxy 0.0.0.0:80
6819 mode http
6820 option httplog
6821 option logasap
6822 log 192.168.2.200 local3
6823
6824 >>> Feb 6 12:14:14 localhost \
6825 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
6826 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
6827 "GET /image.iso HTTP/1.0"
6828
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006829 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01006830 logging.
6831
6832
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006833option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006834 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006835 May be used in sections : defaults | frontend | listen | backend
6836 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006837 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006838 <username> This is the username which will be used when connecting to MySQL
6839 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006840 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006841
6842 If you specify a username, the check consists of sending two MySQL packet,
6843 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01006844 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006845 Error packet. It is a basic but useful test which does not produce error nor
6846 aborted connect on the server. However, it requires adding an authorization
6847 in the MySQL table, like this :
6848
6849 USE mysql;
6850 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
6851 FLUSH PRIVILEGES;
6852
6853 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006854 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006855 Error packet, we don't send anything in this mode. It was reported that it
6856 can generate lockout if check is too frequent and/or if there is not enough
6857 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
6858 value as if a connection is established successfully within fewer than MySQL
6859 "max_connect_errors" attempts after a previous connection was interrupted,
6860 the error count for the host is cleared to zero. If HAProxy's server get
6861 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
6862
6863 Remember that this does not check database presence nor database consistency.
6864 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006865
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02006866 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006867
6868 Most often, an incoming MySQL server needs to see the client's IP address for
6869 various purposes, including IP privilege matching and connection logging.
6870 When possible, it is often wise to masquerade the client's IP address when
6871 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006872 which requires the transparent proxy feature to be compiled in, and the MySQL
6873 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006874
6875 See also: "option httpchk"
6876
6877
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006878option nolinger
6879no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006880 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006881 May be used in sections: defaults | frontend | listen | backend
6882 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006883 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006884
Davor Ocelice9ed2812017-12-25 17:49:28 +01006885 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006886 physically disconnected), the session timeouts triggers and the session is
6887 closed. But it will remain in FIN_WAIT1 state for some time in the system,
6888 using some resources and possibly limiting the ability to establish newer
6889 connections.
6890
6891 When this happens, it is possible to activate "option nolinger" which forces
6892 the system to immediately remove any socket's pending data on close. Thus,
6893 the session is instantly purged from the system's tables. This usually has
6894 side effects such as increased number of TCP resets due to old retransmits
6895 getting immediately rejected. Some firewalls may sometimes complain about
6896 this too.
6897
6898 For this reason, it is not recommended to use this option when not absolutely
6899 needed. You know that you need it when you have thousands of FIN_WAIT1
6900 sessions on your system (TIME_WAIT ones do not count).
6901
6902 This option may be used both on frontends and backends, depending on the side
6903 where it is required. Use it on the frontend for clients, and on the backend
6904 for servers.
6905
6906 If this option has been enabled in a "defaults" section, it can be disabled
6907 in a specific instance by prepending the "no" keyword before it.
6908
6909
Willy Tarreau55165fe2009-05-10 12:02:55 +02006910option originalto [ except <network> ] [ header <name> ]
6911 Enable insertion of the X-Original-To header to requests sent to servers
6912 May be used in sections : defaults | frontend | listen | backend
6913 yes | yes | yes | yes
6914 Arguments :
6915 <network> is an optional argument used to disable this option for sources
6916 matching <network>
6917 <name> an optional argument to specify a different "X-Original-To"
6918 header name.
6919
6920 Since HAProxy can work in transparent mode, every request from a client can
6921 be redirected to the proxy and HAProxy itself can proxy every request to a
6922 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
6923 be lost. This is annoying when you want access rules based on destination ip
6924 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
6925 added by HAProxy to all requests sent to the server. This header contains a
6926 value representing the original destination IP address. Since this must be
6927 configured to always use the last occurrence of this header only. Note that
6928 only the last occurrence of the header must be used, since it is really
6929 possible that the client has already brought one.
6930
6931 The keyword "header" may be used to supply a different header name to replace
6932 the default "X-Original-To". This can be useful where you might already
6933 have a "X-Original-To" header from a different application, and you need
6934 preserve it. Also if your backend server doesn't use the "X-Original-To"
6935 header and requires different one.
6936
6937 Sometimes, a same HAProxy instance may be shared between a direct client
6938 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6939 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6940 header for a known source address or network by adding the "except" keyword
6941 followed by the network address. In this case, any source IP matching the
6942 network will not cause an addition of this header. Most common uses are with
6943 private networks or 127.0.0.1.
6944
6945 This option may be specified either in the frontend or in the backend. If at
6946 least one of them uses it, the header will be added. Note that the backend's
6947 setting of the header subargument takes precedence over the frontend's if
6948 both are defined.
6949
Willy Tarreau55165fe2009-05-10 12:02:55 +02006950 Examples :
6951 # Original Destination address
6952 frontend www
6953 mode http
6954 option originalto except 127.0.0.1
6955
6956 # Those servers want the IP Address in X-Client-Dst
6957 backend www
6958 mode http
6959 option originalto header X-Client-Dst
6960
Christopher Faulet315b39c2018-09-21 16:26:19 +02006961 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02006962
6963
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006964option persist
6965no option persist
6966 Enable or disable forced persistence on down servers
6967 May be used in sections: defaults | frontend | listen | backend
6968 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006969 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006970
6971 When an HTTP request reaches a backend with a cookie which references a dead
6972 server, by default it is redispatched to another server. It is possible to
6973 force the request to be sent to the dead server first using "option persist"
6974 if absolutely needed. A common use case is when servers are under extreme
6975 load and spend their time flapping. In this case, the users would still be
6976 directed to the server they opened the session on, in the hope they would be
6977 correctly served. It is recommended to use "option redispatch" in conjunction
6978 with this option so that in the event it would not be possible to connect to
6979 the server at all (server definitely dead), the client would finally be
6980 redirected to another valid server.
6981
6982 If this option has been enabled in a "defaults" section, it can be disabled
6983 in a specific instance by prepending the "no" keyword before it.
6984
Willy Tarreau4de91492010-01-22 19:10:05 +01006985 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006986
6987
Willy Tarreau0c122822013-12-15 18:49:01 +01006988option pgsql-check [ user <username> ]
6989 Use PostgreSQL health checks for server testing
6990 May be used in sections : defaults | frontend | listen | backend
6991 yes | no | yes | yes
6992 Arguments :
6993 <username> This is the username which will be used when connecting to
6994 PostgreSQL server.
6995
6996 The check sends a PostgreSQL StartupMessage and waits for either
6997 Authentication request or ErrorResponse message. It is a basic but useful
6998 test which does not produce error nor aborted connect on the server.
6999 This check is identical with the "mysql-check".
7000
7001 See also: "option httpchk"
7002
7003
Willy Tarreau9420b122013-12-15 18:58:25 +01007004option prefer-last-server
7005no option prefer-last-server
7006 Allow multiple load balanced requests to remain on the same server
7007 May be used in sections: defaults | frontend | listen | backend
7008 yes | no | yes | yes
7009 Arguments : none
7010
7011 When the load balancing algorithm in use is not deterministic, and a previous
7012 request was sent to a server to which haproxy still holds a connection, it is
7013 sometimes desirable that subsequent requests on a same session go to the same
7014 server as much as possible. Note that this is different from persistence, as
7015 we only indicate a preference which haproxy tries to apply without any form
7016 of warranty. The real use is for keep-alive connections sent to servers. When
7017 this option is used, haproxy will try to reuse the same connection that is
7018 attached to the server instead of rebalancing to another server, causing a
7019 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01007020 not make much sense to use this in combination with hashing algorithms. Note,
7021 haproxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02007022 to a proxy which sends a 407 (authentication required), when the load
7023 balancing algorithm is not deterministic. This is mandatory for use with the
7024 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01007025 troubleshooting some faulty applications. Option prefer-last-server might be
7026 desirable in these environments as well, to avoid redistributing the traffic
7027 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01007028
7029 If this option has been enabled in a "defaults" section, it can be disabled
7030 in a specific instance by prepending the "no" keyword before it.
7031
7032 See also: "option http-keep-alive"
7033
7034
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007035option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07007036option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007037no option redispatch
7038 Enable or disable session redistribution in case of connection failure
7039 May be used in sections: defaults | frontend | listen | backend
7040 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07007041 Arguments :
7042 <interval> The optional integer value that controls how often redispatches
7043 occur when retrying connections. Positive value P indicates a
7044 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01007045 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07007046 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007047 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07007048 positive value of 1 would indicate a redispatch on every retry,
7049 and a positive value of 3 would indicate a redispatch on every
7050 third retry. You can disable redispatches with a value of 0.
7051
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007052
7053 In HTTP mode, if a server designated by a cookie is down, clients may
7054 definitely stick to it because they cannot flush the cookie, so they will not
7055 be able to access the service anymore.
7056
Willy Tarreau59884a62019-01-02 14:48:31 +01007057 Specifying "option redispatch" will allow the proxy to break cookie or
7058 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007059
Joseph Lynch726ab712015-05-11 23:25:34 -07007060 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007061 connection failures. Of course, it requires having "retries" set to a nonzero
7062 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007063
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007064 This form is the preferred form, which replaces both the "redispatch" and
7065 "redisp" keywords.
7066
7067 If this option has been enabled in a "defaults" section, it can be disabled
7068 in a specific instance by prepending the "no" keyword before it.
7069
Willy Tarreau4de91492010-01-22 19:10:05 +01007070 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007071
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007072
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007073option redis-check
7074 Use redis health checks for server testing
7075 May be used in sections : defaults | frontend | listen | backend
7076 yes | no | yes | yes
7077 Arguments : none
7078
7079 It is possible to test that the server correctly talks REDIS protocol instead
7080 of just testing that it accepts the TCP connection. When this option is set,
7081 a PING redis command is sent to the server, and the response is analyzed to
7082 find the "+PONG" response message.
7083
7084 Example :
7085 option redis-check
7086
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007087 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007088
7089
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007090option smtpchk
7091option smtpchk <hello> <domain>
7092 Use SMTP health checks for server testing
7093 May be used in sections : defaults | frontend | listen | backend
7094 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01007095 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007096 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02007097 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007098 values will be turned into the default command ("HELO").
7099
7100 <domain> is the domain name to present to the server. It may only be
7101 specified (and is mandatory) if the hello command has been
7102 specified. By default, "localhost" is used.
7103
7104 When "option smtpchk" is set, the health checks will consist in TCP
7105 connections followed by an SMTP command. By default, this command is
7106 "HELO localhost". The server's return code is analyzed and only return codes
7107 starting with a "2" will be considered as valid. All other responses,
7108 including a lack of response will constitute an error and will indicate a
7109 dead server.
7110
7111 This test is meant to be used with SMTP servers or relays. Depending on the
7112 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01007113 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007114 25 is often easier than adjusting the configuration.
7115
7116 Most often, an incoming SMTP server needs to see the client's IP address for
7117 various purposes, including spam filtering, anti-spoofing and logging. When
7118 possible, it is often wise to masquerade the client's IP address when
7119 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007120 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007121
7122 Example :
7123 option smtpchk HELO mydomain.org
7124
7125 See also : "option httpchk", "source"
7126
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007127
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02007128option socket-stats
7129no option socket-stats
7130
7131 Enable or disable collecting & providing separate statistics for each socket.
7132 May be used in sections : defaults | frontend | listen | backend
7133 yes | yes | yes | no
7134
7135 Arguments : none
7136
7137
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007138option splice-auto
7139no option splice-auto
7140 Enable or disable automatic kernel acceleration on sockets in both directions
7141 May be used in sections : defaults | frontend | listen | backend
7142 yes | yes | yes | yes
7143 Arguments : none
7144
7145 When this option is enabled either on a frontend or on a backend, haproxy
7146 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007147 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007148 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007149 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007150 are not much aggressive in order to limit excessive use of splicing. This
7151 option requires splicing to be enabled at compile time, and may be globally
7152 disabled with the global option "nosplice". Since splice uses pipes, using it
7153 requires that there are enough spare pipes.
7154
7155 Important note: kernel-based TCP splicing is a Linux-specific feature which
7156 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
7157 transfer data between sockets without copying these data to user-space, thus
7158 providing noticeable performance gains and CPU cycles savings. Since many
7159 early implementations are buggy, corrupt data and/or are inefficient, this
7160 feature is not enabled by default, and it should be used with extreme care.
7161 While it is not possible to detect the correctness of an implementation,
7162 2.6.29 is the first version offering a properly working implementation. In
7163 case of doubt, splicing may be globally disabled using the global "nosplice"
7164 keyword.
7165
7166 Example :
7167 option splice-auto
7168
7169 If this option has been enabled in a "defaults" section, it can be disabled
7170 in a specific instance by prepending the "no" keyword before it.
7171
7172 See also : "option splice-request", "option splice-response", and global
7173 options "nosplice" and "maxpipes"
7174
7175
7176option splice-request
7177no option splice-request
7178 Enable or disable automatic kernel acceleration on sockets for requests
7179 May be used in sections : defaults | frontend | listen | backend
7180 yes | yes | yes | yes
7181 Arguments : none
7182
7183 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007184 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007185 the client to the server. It might still use the recv/send scheme if there
7186 are no spare pipes left. This option requires splicing to be enabled at
7187 compile time, and may be globally disabled with the global option "nosplice".
7188 Since splice uses pipes, using it requires that there are enough spare pipes.
7189
7190 Important note: see "option splice-auto" for usage limitations.
7191
7192 Example :
7193 option splice-request
7194
7195 If this option has been enabled in a "defaults" section, it can be disabled
7196 in a specific instance by prepending the "no" keyword before it.
7197
7198 See also : "option splice-auto", "option splice-response", and global options
7199 "nosplice" and "maxpipes"
7200
7201
7202option splice-response
7203no option splice-response
7204 Enable or disable automatic kernel acceleration on sockets for responses
7205 May be used in sections : defaults | frontend | listen | backend
7206 yes | yes | yes | yes
7207 Arguments : none
7208
7209 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007210 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007211 the server to the client. It might still use the recv/send scheme if there
7212 are no spare pipes left. This option requires splicing to be enabled at
7213 compile time, and may be globally disabled with the global option "nosplice".
7214 Since splice uses pipes, using it requires that there are enough spare pipes.
7215
7216 Important note: see "option splice-auto" for usage limitations.
7217
7218 Example :
7219 option splice-response
7220
7221 If this option has been enabled in a "defaults" section, it can be disabled
7222 in a specific instance by prepending the "no" keyword before it.
7223
7224 See also : "option splice-auto", "option splice-request", and global options
7225 "nosplice" and "maxpipes"
7226
7227
Christopher Fauletba7bc162016-11-07 21:07:38 +01007228option spop-check
7229 Use SPOP health checks for server testing
7230 May be used in sections : defaults | frontend | listen | backend
7231 no | no | no | yes
7232 Arguments : none
7233
7234 It is possible to test that the server correctly talks SPOP protocol instead
7235 of just testing that it accepts the TCP connection. When this option is set,
7236 a HELLO handshake is performed between HAProxy and the server, and the
7237 response is analyzed to check no error is reported.
7238
7239 Example :
7240 option spop-check
7241
7242 See also : "option httpchk"
7243
7244
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007245option srvtcpka
7246no option srvtcpka
7247 Enable or disable the sending of TCP keepalive packets on the server side
7248 May be used in sections : defaults | frontend | listen | backend
7249 yes | no | yes | yes
7250 Arguments : none
7251
7252 When there is a firewall or any session-aware component between a client and
7253 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007254 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007255 components decides to expire a session which has remained idle for too long.
7256
7257 Enabling socket-level TCP keep-alives makes the system regularly send packets
7258 to the other end of the connection, leaving it active. The delay between
7259 keep-alive probes is controlled by the system only and depends both on the
7260 operating system and its tuning parameters.
7261
7262 It is important to understand that keep-alive packets are neither emitted nor
7263 received at the application level. It is only the network stacks which sees
7264 them. For this reason, even if one side of the proxy already uses keep-alives
7265 to maintain its connection alive, those keep-alive packets will not be
7266 forwarded to the other side of the proxy.
7267
7268 Please note that this has nothing to do with HTTP keep-alive.
7269
7270 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
7271 server side of a connection, which should help when session expirations are
7272 noticed between HAProxy and a server.
7273
7274 If this option has been enabled in a "defaults" section, it can be disabled
7275 in a specific instance by prepending the "no" keyword before it.
7276
7277 See also : "option clitcpka", "option tcpka"
7278
7279
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007280option ssl-hello-chk
7281 Use SSLv3 client hello health checks for server testing
7282 May be used in sections : defaults | frontend | listen | backend
7283 yes | no | yes | yes
7284 Arguments : none
7285
7286 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
7287 possible to test that the server correctly talks SSL instead of just testing
7288 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
7289 SSLv3 client hello messages are sent once the connection is established to
7290 the server, and the response is analyzed to find an SSL server hello message.
7291 The server is considered valid only when the response contains this server
7292 hello message.
7293
7294 All servers tested till there correctly reply to SSLv3 client hello messages,
7295 and most servers tested do not even log the requests containing only hello
7296 messages, which is appreciable.
7297
Willy Tarreau763a95b2012-10-04 23:15:39 +02007298 Note that this check works even when SSL support was not built into haproxy
7299 because it forges the SSL message. When SSL support is available, it is best
7300 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007301
Willy Tarreau763a95b2012-10-04 23:15:39 +02007302 See also: "option httpchk", "check-ssl"
7303
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007304
Willy Tarreaued179852013-12-16 01:07:00 +01007305option tcp-check
7306 Perform health checks using tcp-check send/expect sequences
7307 May be used in sections: defaults | frontend | listen | backend
7308 yes | no | yes | yes
7309
7310 This health check method is intended to be combined with "tcp-check" command
7311 lists in order to support send/expect types of health check sequences.
7312
7313 TCP checks currently support 4 modes of operations :
7314 - no "tcp-check" directive : the health check only consists in a connection
7315 attempt, which remains the default mode.
7316
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007317 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01007318 used to send a string along with a connection opening. With some
7319 protocols, it helps sending a "QUIT" message for example that prevents
7320 the server from logging a connection error for each health check. The
7321 check result will still be based on the ability to open the connection
7322 only.
7323
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007324 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01007325 The connection is opened and haproxy waits for the server to present some
7326 contents which must validate some rules. The check result will be based
7327 on the matching between the contents and the rules. This is suited for
7328 POP, IMAP, SMTP, FTP, SSH, TELNET.
7329
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007330 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007331 used to test a hello-type protocol. HAProxy sends a message, the server
7332 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007333 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01007334 suited for protocols which require a binding or a request/response model.
7335 LDAP, MySQL, Redis and SSL are example of such protocols, though they
7336 already all have their dedicated checks with a deeper understanding of
7337 the respective protocols.
7338 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01007339 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01007340
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007341 A fifth mode can be used to insert comments in different steps of the
7342 script.
7343
7344 For each tcp-check rule you create, you can add a "comment" directive,
7345 followed by a string. This string will be reported in the log and stderr
7346 in debug mode. It is useful to make user-friendly error reporting.
7347 The "comment" is of course optional.
7348
7349
Willy Tarreaued179852013-12-16 01:07:00 +01007350 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007351 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007352 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007353 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007354
Davor Ocelice9ed2812017-12-25 17:49:28 +01007355 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007356 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007357 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007358
7359 # look for the redis master server after ensuring it speaks well
7360 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01007361 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01007362 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007363 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007364 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02007365 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007366 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01007367 tcp-check send info\ replication\r\n
7368 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007369 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007370 tcp-check send QUIT\r\n
7371 tcp-check expect string +OK
7372
Davor Ocelice9ed2812017-12-25 17:49:28 +01007373 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01007374 (send many headers before analyzing)
7375 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007376 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01007377 tcp-check send HEAD\ /\ HTTP/1.1\r\n
7378 tcp-check send Host:\ www.mydomain.com\r\n
7379 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
7380 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007381 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01007382
7383
7384 See also : "tcp-check expect", "tcp-check send"
7385
7386
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007387option tcp-smart-accept
7388no option tcp-smart-accept
7389 Enable or disable the saving of one ACK packet during the accept sequence
7390 May be used in sections : defaults | frontend | listen | backend
7391 yes | yes | yes | no
7392 Arguments : none
7393
7394 When an HTTP connection request comes in, the system acknowledges it on
7395 behalf of HAProxy, then the client immediately sends its request, and the
7396 system acknowledges it too while it is notifying HAProxy about the new
7397 connection. HAProxy then reads the request and responds. This means that we
7398 have one TCP ACK sent by the system for nothing, because the request could
7399 very well be acknowledged by HAProxy when it sends its response.
7400
7401 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
7402 sending this useless ACK on platforms which support it (currently at least
7403 Linux). It must not cause any problem, because the system will send it anyway
7404 after 40 ms if the response takes more time than expected to come.
7405
7406 During complex network debugging sessions, it may be desirable to disable
7407 this optimization because delayed ACKs can make troubleshooting more complex
7408 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007409 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007410
7411 It is also possible to force it for non-HTTP proxies by simply specifying
7412 "option tcp-smart-accept". For instance, it can make sense with some services
7413 such as SMTP where the server speaks first.
7414
7415 It is recommended to avoid forcing this option in a defaults section. In case
7416 of doubt, consider setting it back to automatic values by prepending the
7417 "default" keyword before it, or disabling it using the "no" keyword.
7418
Willy Tarreaud88edf22009-06-14 15:48:17 +02007419 See also : "option tcp-smart-connect"
7420
7421
7422option tcp-smart-connect
7423no option tcp-smart-connect
7424 Enable or disable the saving of one ACK packet during the connect sequence
7425 May be used in sections : defaults | frontend | listen | backend
7426 yes | no | yes | yes
7427 Arguments : none
7428
7429 On certain systems (at least Linux), HAProxy can ask the kernel not to
7430 immediately send an empty ACK upon a connection request, but to directly
7431 send the buffer request instead. This saves one packet on the network and
7432 thus boosts performance. It can also be useful for some servers, because they
7433 immediately get the request along with the incoming connection.
7434
7435 This feature is enabled when "option tcp-smart-connect" is set in a backend.
7436 It is not enabled by default because it makes network troubleshooting more
7437 complex.
7438
7439 It only makes sense to enable it with protocols where the client speaks first
7440 such as HTTP. In other situations, if there is no data to send in place of
7441 the ACK, a normal ACK is sent.
7442
7443 If this option has been enabled in a "defaults" section, it can be disabled
7444 in a specific instance by prepending the "no" keyword before it.
7445
7446 See also : "option tcp-smart-accept"
7447
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007448
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007449option tcpka
7450 Enable or disable the sending of TCP keepalive packets on both sides
7451 May be used in sections : defaults | frontend | listen | backend
7452 yes | yes | yes | yes
7453 Arguments : none
7454
7455 When there is a firewall or any session-aware component between a client and
7456 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007457 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007458 components decides to expire a session which has remained idle for too long.
7459
7460 Enabling socket-level TCP keep-alives makes the system regularly send packets
7461 to the other end of the connection, leaving it active. The delay between
7462 keep-alive probes is controlled by the system only and depends both on the
7463 operating system and its tuning parameters.
7464
7465 It is important to understand that keep-alive packets are neither emitted nor
7466 received at the application level. It is only the network stacks which sees
7467 them. For this reason, even if one side of the proxy already uses keep-alives
7468 to maintain its connection alive, those keep-alive packets will not be
7469 forwarded to the other side of the proxy.
7470
7471 Please note that this has nothing to do with HTTP keep-alive.
7472
7473 Using option "tcpka" enables the emission of TCP keep-alive probes on both
7474 the client and server sides of a connection. Note that this is meaningful
7475 only in "defaults" or "listen" sections. If this option is used in a
7476 frontend, only the client side will get keep-alives, and if this option is
7477 used in a backend, only the server side will get keep-alives. For this
7478 reason, it is strongly recommended to explicitly use "option clitcpka" and
7479 "option srvtcpka" when the configuration is split between frontends and
7480 backends.
7481
7482 See also : "option clitcpka", "option srvtcpka"
7483
Willy Tarreau844e3c52008-01-11 16:28:18 +01007484
7485option tcplog
7486 Enable advanced logging of TCP connections with session state and timers
7487 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01007488 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01007489 Arguments : none
7490
7491 By default, the log output format is very poor, as it only contains the
7492 source and destination addresses, and the instance name. By specifying
7493 "option tcplog", each log line turns into a much richer format including, but
7494 not limited to, the connection timers, the session status, the connections
7495 numbers, the frontend, backend and server name, and of course the source
7496 address and ports. This option is useful for pure TCP proxies in order to
7497 find which of the client or server disconnects or times out. For normal HTTP
7498 proxies, it's better to use "option httplog" which is even more complete.
7499
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007500 "option tcplog" overrides any previous "log-format" directive.
7501
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007502 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007503
7504
Willy Tarreau844e3c52008-01-11 16:28:18 +01007505option transparent
7506no option transparent
7507 Enable client-side transparent proxying
7508 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007509 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007510 Arguments : none
7511
7512 This option was introduced in order to provide layer 7 persistence to layer 3
7513 load balancers. The idea is to use the OS's ability to redirect an incoming
7514 connection for a remote address to a local process (here HAProxy), and let
7515 this process know what address was initially requested. When this option is
7516 used, sessions without cookies will be forwarded to the original destination
7517 IP address of the incoming request (which should match that of another
7518 equipment), while requests with cookies will still be forwarded to the
7519 appropriate server.
7520
7521 Note that contrary to a common belief, this option does NOT make HAProxy
7522 present the client's IP to the server when establishing the connection.
7523
Willy Tarreaua1146052011-03-01 09:51:54 +01007524 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007525 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007526
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007527
Simon Horman98637e52014-06-20 12:30:16 +09007528external-check command <command>
7529 Executable to run when performing an external-check
7530 May be used in sections : defaults | frontend | listen | backend
7531 yes | no | yes | yes
7532
7533 Arguments :
7534 <command> is the external command to run
7535
Simon Horman98637e52014-06-20 12:30:16 +09007536 The arguments passed to the to the command are:
7537
Cyril Bonté777be862014-12-02 21:21:35 +01007538 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09007539
Cyril Bonté777be862014-12-02 21:21:35 +01007540 The <proxy_address> and <proxy_port> are derived from the first listener
7541 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
7542 listener the proxy_address will be the path of the socket and the
7543 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
7544 possible to determine a listener, and both <proxy_address> and <proxy_port>
7545 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09007546
Cyril Bonté72cda2a2014-12-27 22:28:39 +01007547 Some values are also provided through environment variables.
7548
7549 Environment variables :
7550 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
7551 applicable, for example in a "backend" section).
7552
7553 HAPROXY_PROXY_ID The backend id.
7554
7555 HAPROXY_PROXY_NAME The backend name.
7556
7557 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
7558 applicable, for example in a "backend" section or
7559 for a UNIX socket).
7560
7561 HAPROXY_SERVER_ADDR The server address.
7562
7563 HAPROXY_SERVER_CURCONN The current number of connections on the server.
7564
7565 HAPROXY_SERVER_ID The server id.
7566
7567 HAPROXY_SERVER_MAXCONN The server max connections.
7568
7569 HAPROXY_SERVER_NAME The server name.
7570
7571 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
7572 socket).
7573
7574 PATH The PATH environment variable used when executing
7575 the command may be set using "external-check path".
7576
William Lallemand4d03e432019-06-14 15:35:37 +02007577 See also "2.3. Environment variables" for other variables.
7578
Simon Horman98637e52014-06-20 12:30:16 +09007579 If the command executed and exits with a zero status then the check is
7580 considered to have passed, otherwise the check is considered to have
7581 failed.
7582
7583 Example :
7584 external-check command /bin/true
7585
7586 See also : "external-check", "option external-check", "external-check path"
7587
7588
7589external-check path <path>
7590 The value of the PATH environment variable used when running an external-check
7591 May be used in sections : defaults | frontend | listen | backend
7592 yes | no | yes | yes
7593
7594 Arguments :
7595 <path> is the path used when executing external command to run
7596
7597 The default path is "".
7598
7599 Example :
7600 external-check path "/usr/bin:/bin"
7601
7602 See also : "external-check", "option external-check",
7603 "external-check command"
7604
7605
Emeric Brun647caf12009-06-30 17:57:00 +02007606persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02007607persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02007608 Enable RDP cookie-based persistence
7609 May be used in sections : defaults | frontend | listen | backend
7610 yes | no | yes | yes
7611 Arguments :
7612 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02007613 default cookie name "msts" will be used. There currently is no
7614 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02007615
7616 This statement enables persistence based on an RDP cookie. The RDP cookie
7617 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01007618 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02007619 and if an RDP cookie is found, it is decoded. If it matches a known server
7620 which is still UP (or if "option persist" is set), then the connection is
7621 forwarded to this server.
7622
7623 Note that this only makes sense in a TCP backend, but for this to work, the
7624 frontend must have waited long enough to ensure that an RDP cookie is present
7625 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007626 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02007627 a single "listen" section.
7628
Willy Tarreau61e28f22010-05-16 22:31:05 +02007629 Also, it is important to understand that the terminal server will emit this
7630 RDP cookie only if it is configured for "token redirection mode", which means
7631 that the "IP address redirection" option is disabled.
7632
Emeric Brun647caf12009-06-30 17:57:00 +02007633 Example :
7634 listen tse-farm
7635 bind :3389
7636 # wait up to 5s for an RDP cookie in the request
7637 tcp-request inspect-delay 5s
7638 tcp-request content accept if RDP_COOKIE
7639 # apply RDP cookie persistence
7640 persist rdp-cookie
7641 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02007642 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02007643 balance rdp-cookie
7644 server srv1 1.1.1.1:3389
7645 server srv2 1.1.1.2:3389
7646
Simon Hormanab814e02011-06-24 14:50:20 +09007647 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
7648 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02007649
7650
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007651rate-limit sessions <rate>
7652 Set a limit on the number of new sessions accepted per second on a frontend
7653 May be used in sections : defaults | frontend | listen | backend
7654 yes | yes | yes | no
7655 Arguments :
7656 <rate> The <rate> parameter is an integer designating the maximum number
7657 of new sessions per second to accept on the frontend.
7658
7659 When the frontend reaches the specified number of new sessions per second, it
7660 stops accepting new connections until the rate drops below the limit again.
7661 During this time, the pending sessions will be kept in the socket's backlog
7662 (in system buffers) and haproxy will not even be aware that sessions are
7663 pending. When applying very low limit on a highly loaded service, it may make
7664 sense to increase the socket's backlog using the "backlog" keyword.
7665
7666 This feature is particularly efficient at blocking connection-based attacks
7667 or service abuse on fragile servers. Since the session rate is measured every
7668 millisecond, it is extremely accurate. Also, the limit applies immediately,
7669 no delay is needed at all to detect the threshold.
7670
7671 Example : limit the connection rate on SMTP to 10 per second max
7672 listen smtp
7673 mode tcp
7674 bind :25
7675 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02007676 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007677
Willy Tarreaua17c2d92011-07-25 08:16:20 +02007678 Note : when the maximum rate is reached, the frontend's status is not changed
7679 but its sockets appear as "WAITING" in the statistics if the
7680 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007681
7682 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
7683
7684
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007685redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
7686redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
7687redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007688 Return an HTTP redirection if/unless a condition is matched
7689 May be used in sections : defaults | frontend | listen | backend
7690 no | yes | yes | yes
7691
7692 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01007693 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007694
Willy Tarreau0140f252008-11-19 21:07:09 +01007695 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007696 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007697 the HTTP "Location" header. When used in an "http-request" rule,
7698 <loc> value follows the log-format rules and can include some
7699 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007700
7701 <pfx> With "redirect prefix", the "Location" header is built from the
7702 concatenation of <pfx> and the complete URI path, including the
7703 query string, unless the "drop-query" option is specified (see
7704 below). As a special case, if <pfx> equals exactly "/", then
7705 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007706 redirect to the same URL (for instance, to insert a cookie). When
7707 used in an "http-request" rule, <pfx> value follows the log-format
7708 rules and can include some dynamic values (see Custom Log Format
7709 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007710
7711 <sch> With "redirect scheme", then the "Location" header is built by
7712 concatenating <sch> with "://" then the first occurrence of the
7713 "Host" header, and then the URI path, including the query string
7714 unless the "drop-query" option is specified (see below). If no
7715 path is found or if the path is "*", then "/" is used instead. If
7716 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007717 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007718 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007719 HTTPS. When used in an "http-request" rule, <sch> value follows
7720 the log-format rules and can include some dynamic values (see
7721 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01007722
7723 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007724 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
7725 with 302 used by default if no code is specified. 301 means
7726 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02007727 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007728 cache the redirection. 303 is equivalent to 302 except that the
7729 browser will fetch the location with a GET method. 307 is just
7730 like 302 but makes it clear that the same method must be reused.
7731 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01007732
7733 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007734 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01007735
7736 - "drop-query"
7737 When this keyword is used in a prefix-based redirection, then the
7738 location will be set without any possible query-string, which is useful
7739 for directing users to a non-secure page for instance. It has no effect
7740 with a location-type redirect.
7741
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007742 - "append-slash"
7743 This keyword may be used in conjunction with "drop-query" to redirect
7744 users who use a URL not ending with a '/' to the same one with the '/'.
7745 It can be useful to ensure that search engines will only see one URL.
7746 For this, a return code 301 is preferred.
7747
Willy Tarreau0140f252008-11-19 21:07:09 +01007748 - "set-cookie NAME[=value]"
7749 A "Set-Cookie" header will be added with NAME (and optionally "=value")
7750 to the response. This is sometimes used to indicate that a user has
7751 been seen, for instance to protect against some types of DoS. No other
7752 cookie option is added, so the cookie will be a session cookie. Note
7753 that for a browser, a sole cookie name without an equal sign is
7754 different from a cookie with an equal sign.
7755
7756 - "clear-cookie NAME[=]"
7757 A "Set-Cookie" header will be added with NAME (and optionally "="), but
7758 with the "Max-Age" attribute set to zero. This will tell the browser to
7759 delete this cookie. It is useful for instance on logout pages. It is
7760 important to note that clearing the cookie "NAME" will not remove a
7761 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
7762 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007763
7764 Example: move the login URL only to HTTPS.
7765 acl clear dst_port 80
7766 acl secure dst_port 8080
7767 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01007768 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01007769 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01007770 acl cookie_set hdr_sub(cookie) SEEN=1
7771
7772 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01007773 redirect prefix https://mysite.com if login_page !secure
7774 redirect prefix http://mysite.com drop-query if login_page !uid_given
7775 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01007776 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007777
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007778 Example: send redirects for request for articles without a '/'.
7779 acl missing_slash path_reg ^/article/[^/]*$
7780 redirect code 301 prefix / drop-query append-slash if missing_slash
7781
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007782 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01007783 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007784
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007785 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02007786 http-request redirect code 301 location \
7787 http://www.%[hdr(host)]%[capture.req.uri] \
7788 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007789
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007790 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007791
7792
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007793redisp (deprecated)
7794redispatch (deprecated)
7795 Enable or disable session redistribution in case of connection failure
7796 May be used in sections: defaults | frontend | listen | backend
7797 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007798 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007799
7800 In HTTP mode, if a server designated by a cookie is down, clients may
7801 definitely stick to it because they cannot flush the cookie, so they will not
7802 be able to access the service anymore.
7803
7804 Specifying "redispatch" will allow the proxy to break their persistence and
7805 redistribute them to a working server.
7806
7807 It also allows to retry last connection to another server in case of multiple
7808 connection failures. Of course, it requires having "retries" set to a nonzero
7809 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007810
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007811 This form is deprecated, do not use it in any new configuration, use the new
7812 "option redispatch" instead.
7813
7814 See also : "option redispatch"
7815
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007816
Willy Tarreau96d51952019-05-22 20:34:35 +02007817reqadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007818 Add a header at the end of the HTTP request
7819 May be used in sections : defaults | frontend | listen | backend
7820 no | yes | yes | yes
7821 Arguments :
7822 <string> is the complete line to be added. Any space or known delimiter
7823 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007824 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007825
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007826 <cond> is an optional matching condition built from ACLs. It makes it
7827 possible to ignore this rule when other conditions are not met.
7828
Willy Tarreau303c0352008-01-17 19:01:39 +01007829 A new line consisting in <string> followed by a line feed will be added after
7830 the last header of an HTTP request.
7831
7832 Header transformations only apply to traffic which passes through HAProxy,
7833 and not to traffic generated by HAProxy, such as health-checks or error
7834 responses.
7835
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007836 Example : add "X-Proto: SSL" to requests coming via port 81
7837 acl is-ssl dst_port 81
7838 reqadd X-Proto:\ SSL if is-ssl
7839
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007840 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
7841 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007842
7843
Willy Tarreau96d51952019-05-22 20:34:35 +02007844reqallow <search> [{if | unless} <cond>] (deprecated)
7845reqiallow <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007846 Definitely allow an HTTP request if a line matches a regular expression
7847 May be used in sections : defaults | frontend | listen | backend
7848 no | yes | yes | yes
7849 Arguments :
7850 <search> is the regular expression applied to HTTP headers and to the
7851 request line. This is an extended regular expression. Parenthesis
7852 grouping is supported and no preliminary backslash is required.
7853 Any space or known delimiter must be escaped using a backslash
7854 ('\'). The pattern applies to a full line at a time. The
7855 "reqallow" keyword strictly matches case while "reqiallow"
7856 ignores case.
7857
Willy Tarreau5321c422010-01-28 20:35:13 +01007858 <cond> is an optional matching condition built from ACLs. It makes it
7859 possible to ignore this rule when other conditions are not met.
7860
Willy Tarreau303c0352008-01-17 19:01:39 +01007861 A request containing any line which matches extended regular expression
7862 <search> will mark the request as allowed, even if any later test would
7863 result in a deny. The test applies both to the request line and to request
7864 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007865 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007866
7867 It is easier, faster and more powerful to use ACLs to write access policies.
7868 Reqdeny, reqallow and reqpass should be avoided in new designs.
7869
7870 Example :
7871 # allow www.* but refuse *.local
7872 reqiallow ^Host:\ www\.
7873 reqideny ^Host:\ .*\.local
7874
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007875 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
7876 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007877
7878
Willy Tarreau96d51952019-05-22 20:34:35 +02007879reqdel <search> [{if | unless} <cond>] (deprecated)
7880reqidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007881 Delete all headers matching a regular expression in an HTTP request
7882 May be used in sections : defaults | frontend | listen | backend
7883 no | yes | yes | yes
7884 Arguments :
7885 <search> is the regular expression applied to HTTP headers and to the
7886 request line. This is an extended regular expression. Parenthesis
7887 grouping is supported and no preliminary backslash is required.
7888 Any space or known delimiter must be escaped using a backslash
7889 ('\'). The pattern applies to a full line at a time. The "reqdel"
7890 keyword strictly matches case while "reqidel" ignores case.
7891
Willy Tarreau5321c422010-01-28 20:35:13 +01007892 <cond> is an optional matching condition built from ACLs. It makes it
7893 possible to ignore this rule when other conditions are not met.
7894
Willy Tarreau303c0352008-01-17 19:01:39 +01007895 Any header line matching extended regular expression <search> in the request
7896 will be completely deleted. Most common use of this is to remove unwanted
7897 and/or dangerous headers or cookies from a request before passing it to the
7898 next servers.
7899
7900 Header transformations only apply to traffic which passes through HAProxy,
7901 and not to traffic generated by HAProxy, such as health-checks or error
7902 responses. Keep in mind that header names are not case-sensitive.
7903
7904 Example :
7905 # remove X-Forwarded-For header and SERVER cookie
7906 reqidel ^X-Forwarded-For:.*
7907 reqidel ^Cookie:.*SERVER=
7908
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007909 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
7910 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007911
7912
Willy Tarreau96d51952019-05-22 20:34:35 +02007913reqdeny <search> [{if | unless} <cond>] (deprecated)
7914reqideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007915 Deny an HTTP request if a line matches a regular expression
7916 May be used in sections : defaults | frontend | listen | backend
7917 no | yes | yes | yes
7918 Arguments :
7919 <search> is the regular expression applied to HTTP headers and to the
7920 request line. This is an extended regular expression. Parenthesis
7921 grouping is supported and no preliminary backslash is required.
7922 Any space or known delimiter must be escaped using a backslash
7923 ('\'). The pattern applies to a full line at a time. The
7924 "reqdeny" keyword strictly matches case while "reqideny" ignores
7925 case.
7926
Willy Tarreau5321c422010-01-28 20:35:13 +01007927 <cond> is an optional matching condition built from ACLs. It makes it
7928 possible to ignore this rule when other conditions are not met.
7929
Willy Tarreau303c0352008-01-17 19:01:39 +01007930 A request containing any line which matches extended regular expression
7931 <search> will mark the request as denied, even if any later test would
7932 result in an allow. The test applies both to the request line and to request
7933 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007934 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007935
Willy Tarreauced27012008-01-17 20:35:34 +01007936 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007937 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01007938 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01007939
Willy Tarreau303c0352008-01-17 19:01:39 +01007940 It is easier, faster and more powerful to use ACLs to write access policies.
7941 Reqdeny, reqallow and reqpass should be avoided in new designs.
7942
7943 Example :
7944 # refuse *.local, then allow www.*
7945 reqideny ^Host:\ .*\.local
7946 reqiallow ^Host:\ www\.
7947
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007948 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
7949 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007950
7951
Willy Tarreau96d51952019-05-22 20:34:35 +02007952reqpass <search> [{if | unless} <cond>] (deprecated)
7953reqipass <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007954 Ignore any HTTP request line matching a regular expression in next rules
7955 May be used in sections : defaults | frontend | listen | backend
7956 no | yes | yes | yes
7957 Arguments :
7958 <search> is the regular expression applied to HTTP headers and to the
7959 request line. This is an extended regular expression. Parenthesis
7960 grouping is supported and no preliminary backslash is required.
7961 Any space or known delimiter must be escaped using a backslash
7962 ('\'). The pattern applies to a full line at a time. The
7963 "reqpass" keyword strictly matches case while "reqipass" ignores
7964 case.
7965
Willy Tarreau5321c422010-01-28 20:35:13 +01007966 <cond> is an optional matching condition built from ACLs. It makes it
7967 possible to ignore this rule when other conditions are not met.
7968
Willy Tarreau303c0352008-01-17 19:01:39 +01007969 A request containing any line which matches extended regular expression
7970 <search> will skip next rules, without assigning any deny or allow verdict.
7971 The test applies both to the request line and to request headers. Keep in
7972 mind that URLs in request line are case-sensitive while header names are not.
7973
7974 It is easier, faster and more powerful to use ACLs to write access policies.
7975 Reqdeny, reqallow and reqpass should be avoided in new designs.
7976
7977 Example :
7978 # refuse *.local, then allow www.*, but ignore "www.private.local"
7979 reqipass ^Host:\ www.private\.local
7980 reqideny ^Host:\ .*\.local
7981 reqiallow ^Host:\ www\.
7982
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007983 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
7984 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007985
7986
Willy Tarreau96d51952019-05-22 20:34:35 +02007987reqrep <search> <string> [{if | unless} <cond>] (deprecated)
7988reqirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007989 Replace a regular expression with a string in an HTTP request line
7990 May be used in sections : defaults | frontend | listen | backend
7991 no | yes | yes | yes
7992 Arguments :
7993 <search> is the regular expression applied to HTTP headers and to the
7994 request line. This is an extended regular expression. Parenthesis
7995 grouping is supported and no preliminary backslash is required.
7996 Any space or known delimiter must be escaped using a backslash
7997 ('\'). The pattern applies to a full line at a time. The "reqrep"
7998 keyword strictly matches case while "reqirep" ignores case.
7999
8000 <string> is the complete line to be added. Any space or known delimiter
8001 must be escaped using a backslash ('\'). References to matched
8002 pattern groups are possible using the common \N form, with N
8003 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008004 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008005
Willy Tarreau5321c422010-01-28 20:35:13 +01008006 <cond> is an optional matching condition built from ACLs. It makes it
8007 possible to ignore this rule when other conditions are not met.
8008
Willy Tarreau303c0352008-01-17 19:01:39 +01008009 Any line matching extended regular expression <search> in the request (both
8010 the request line and header lines) will be completely replaced with <string>.
8011 Most common use of this is to rewrite URLs or domain names in "Host" headers.
8012
8013 Header transformations only apply to traffic which passes through HAProxy,
8014 and not to traffic generated by HAProxy, such as health-checks or error
8015 responses. Note that for increased readability, it is suggested to add enough
8016 spaces between the request and the response. Keep in mind that URLs in
8017 request line are case-sensitive while header names are not.
8018
8019 Example :
8020 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04008021 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01008022 # replace "www.mydomain.com" with "www" in the host name.
8023 reqirep ^Host:\ www.mydomain.com Host:\ www
8024
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008025 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
8026 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008027
8028
Willy Tarreau96d51952019-05-22 20:34:35 +02008029reqtarpit <search> [{if | unless} <cond>] (deprecated)
8030reqitarpit <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008031 Tarpit an HTTP request containing a line matching a regular expression
8032 May be used in sections : defaults | frontend | listen | backend
8033 no | yes | yes | yes
8034 Arguments :
8035 <search> is the regular expression applied to HTTP headers and to the
8036 request line. This is an extended regular expression. Parenthesis
8037 grouping is supported and no preliminary backslash is required.
8038 Any space or known delimiter must be escaped using a backslash
8039 ('\'). The pattern applies to a full line at a time. The
8040 "reqtarpit" keyword strictly matches case while "reqitarpit"
8041 ignores case.
8042
Willy Tarreau5321c422010-01-28 20:35:13 +01008043 <cond> is an optional matching condition built from ACLs. It makes it
8044 possible to ignore this rule when other conditions are not met.
8045
Willy Tarreau303c0352008-01-17 19:01:39 +01008046 A request containing any line which matches extended regular expression
8047 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01008048 be kept open for a pre-defined time, then will return an HTTP error 500 so
8049 that the attacker does not suspect it has been tarpitted. The status 500 will
8050 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01008051 delay is defined by "timeout tarpit", or "timeout connect" if the former is
8052 not set.
8053
8054 The goal of the tarpit is to slow down robots attacking servers with
8055 identifiable requests. Many robots limit their outgoing number of connections
8056 and stay connected waiting for a reply which can take several minutes to
8057 come. Depending on the environment and attack, it may be particularly
8058 efficient at reducing the load on the network and firewalls.
8059
Willy Tarreau5321c422010-01-28 20:35:13 +01008060 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01008061 # ignore user-agents reporting any flavor of "Mozilla" or "MSIE", but
Willy Tarreau303c0352008-01-17 19:01:39 +01008062 # block all others.
8063 reqipass ^User-Agent:\.*(Mozilla|MSIE)
8064 reqitarpit ^User-Agent:
8065
Willy Tarreau5321c422010-01-28 20:35:13 +01008066 # block bad guys
8067 acl badguys src 10.1.0.3 172.16.13.20/28
8068 reqitarpit . if badguys
8069
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008070 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
8071 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008072
8073
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008074retries <value>
8075 Set the number of retries to perform on a server after a connection failure
8076 May be used in sections: defaults | frontend | listen | backend
8077 yes | no | yes | yes
8078 Arguments :
8079 <value> is the number of times a connection attempt should be retried on
8080 a server when a connection either is refused or times out. The
8081 default value is 3.
8082
8083 It is important to understand that this value applies to the number of
8084 connection attempts, not full requests. When a connection has effectively
8085 been established to a server, there will be no more retry.
8086
8087 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07008088 a turn-around timer of min("timeout connect", one second) is applied before
8089 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008090
8091 When "option redispatch" is set, the last retry may be performed on another
8092 server even if a cookie references a different server.
8093
8094 See also : "option redispatch"
8095
8096
Olivier Houcharda254a372019-04-05 15:30:12 +02008097retry-on [list of keywords]
8098 Specify when to attempt to automatically retry a failed request
8099 May be used in sections: defaults | frontend | listen | backend
8100 yes | no | yes | yes
8101 Arguments :
8102 <keywords> is a list of keywords or HTTP status codes, each representing a
8103 type of failure event on which an attempt to retry the request
8104 is desired. Please read the notes at the bottom before changing
8105 this setting. The following keywords are supported :
8106
8107 none never retry
8108
8109 conn-failure retry when the connection or the SSL handshake failed
8110 and the request could not be sent. This is the default.
8111
8112 empty-response retry when the server connection was closed after part
8113 of the request was sent, and nothing was received from
8114 the server. This type of failure may be caused by the
8115 request timeout on the server side, poor network
8116 condition, or a server crash or restart while
8117 processing the request.
8118
Olivier Houcharde3249a92019-05-03 23:01:47 +02008119 junk-response retry when the server returned something not looking
8120 like a complete HTTP response. This includes partial
8121 responses headers as well as non-HTTP contents. It
8122 usually is a bad idea to retry on such events, which
8123 may be caused a configuration issue (wrong server port)
8124 or by the request being harmful to the server (buffer
8125 overflow attack for example).
8126
Olivier Houcharda254a372019-04-05 15:30:12 +02008127 response-timeout the server timeout stroke while waiting for the server
8128 to respond to the request. This may be caused by poor
8129 network condition, the reuse of an idle connection
8130 which has expired on the path, or by the request being
8131 extremely expensive to process. It generally is a bad
8132 idea to retry on such events on servers dealing with
8133 heavy database processing (full scans, etc) as it may
8134 amplify denial of service attacks.
8135
Olivier Houchard865d8392019-05-03 22:46:27 +02008136 0rtt-rejected retry requests which were sent over early data and were
8137 rejected by the server. These requests are generally
8138 considered to be safe to retry.
8139
Olivier Houcharda254a372019-04-05 15:30:12 +02008140 <status> any HTTP status code among "404" (Not Found), "408"
8141 (Request Timeout), "425" (Too Early), "500" (Server
8142 Error), "501" (Not Implemented), "502" (Bad Gateway),
8143 "503" (Service Unavailable), "504" (Gateway Timeout).
8144
Olivier Houchardddf0e032019-05-10 18:05:40 +02008145 all-retryable-errors
8146 retry request for any error that are considered
8147 retryable. This currently activates "conn-failure",
8148 "empty-response", "junk-response", "response-timeout",
8149 "0rtt-rejected", "500", "502", "503", and "504".
8150
Olivier Houcharda254a372019-04-05 15:30:12 +02008151 Using this directive replaces any previous settings with the new ones; it is
8152 not cumulative.
8153
8154 Please note that using anything other than "none" and "conn-failure" requires
8155 to allocate a buffer and copy the whole request into it, so it has memory and
8156 performance impacts. Requests not fitting in a single buffer will never be
8157 retried (see the global tune.bufsize setting).
8158
8159 You have to make sure the application has a replay protection mechanism built
8160 in such as a unique transaction IDs passed in requests, or that replaying the
8161 same request has no consequence, or it is very dangerous to use any retry-on
8162 value beside "conn-failure" and "none". Static file servers and caches are
8163 generally considered safe against any type of retry. Using a status code can
8164 be useful to quickly leave a server showing an abnormal behavior (out of
8165 memory, file system issues, etc), but in this case it may be a good idea to
8166 immediately redispatch the connection to another server (please see "option
8167 redispatch" for this). Last, it is important to understand that most causes
8168 of failures are the requests themselves and that retrying a request causing a
8169 server to misbehave will often make the situation even worse for this server,
8170 or for the whole service in case of redispatch.
8171
8172 Unless you know exactly how the application deals with replayed requests, you
8173 should not use this directive.
8174
8175 The default is "conn-failure".
8176
8177 See also: "retries", "option redispatch", "tune.bufsize"
8178
Willy Tarreau96d51952019-05-22 20:34:35 +02008179rspadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008180 Add a header at the end of the HTTP response
8181 May be used in sections : defaults | frontend | listen | backend
8182 no | yes | yes | yes
8183 Arguments :
8184 <string> is the complete line to be added. Any space or known delimiter
8185 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008186 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008187
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008188 <cond> is an optional matching condition built from ACLs. It makes it
8189 possible to ignore this rule when other conditions are not met.
8190
Willy Tarreau303c0352008-01-17 19:01:39 +01008191 A new line consisting in <string> followed by a line feed will be added after
8192 the last header of an HTTP response.
8193
8194 Header transformations only apply to traffic which passes through HAProxy,
8195 and not to traffic generated by HAProxy, such as health-checks or error
8196 responses.
8197
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008198 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
8199 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008200
8201
Willy Tarreau96d51952019-05-22 20:34:35 +02008202rspdel <search> [{if | unless} <cond>] (deprecated)
8203rspidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008204 Delete all headers matching a regular expression in an HTTP response
8205 May be used in sections : defaults | frontend | listen | backend
8206 no | yes | yes | yes
8207 Arguments :
8208 <search> is the regular expression applied to HTTP headers and to the
8209 response line. This is an extended regular expression, so
8210 parenthesis grouping is supported and no preliminary backslash
8211 is required. Any space or known delimiter must be escaped using
8212 a backslash ('\'). The pattern applies to a full line at a time.
8213 The "rspdel" keyword strictly matches case while "rspidel"
8214 ignores case.
8215
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008216 <cond> is an optional matching condition built from ACLs. It makes it
8217 possible to ignore this rule when other conditions are not met.
8218
Willy Tarreau303c0352008-01-17 19:01:39 +01008219 Any header line matching extended regular expression <search> in the response
8220 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008221 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01008222 client.
8223
8224 Header transformations only apply to traffic which passes through HAProxy,
8225 and not to traffic generated by HAProxy, such as health-checks or error
8226 responses. Keep in mind that header names are not case-sensitive.
8227
8228 Example :
8229 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02008230 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01008231
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008232 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
8233 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008234
8235
Willy Tarreau96d51952019-05-22 20:34:35 +02008236rspdeny <search> [{if | unless} <cond>] (deprecated)
8237rspideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008238 Block an HTTP response if a line matches a regular expression
8239 May be used in sections : defaults | frontend | listen | backend
8240 no | yes | yes | yes
8241 Arguments :
8242 <search> is the regular expression applied to HTTP headers and to the
8243 response line. This is an extended regular expression, so
8244 parenthesis grouping is supported and no preliminary backslash
8245 is required. Any space or known delimiter must be escaped using
8246 a backslash ('\'). The pattern applies to a full line at a time.
8247 The "rspdeny" keyword strictly matches case while "rspideny"
8248 ignores case.
8249
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008250 <cond> is an optional matching condition built from ACLs. It makes it
8251 possible to ignore this rule when other conditions are not met.
8252
Willy Tarreau303c0352008-01-17 19:01:39 +01008253 A response containing any line which matches extended regular expression
8254 <search> will mark the request as denied. The test applies both to the
8255 response line and to response headers. Keep in mind that header names are not
8256 case-sensitive.
8257
8258 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01008259 block the response before it reaches the client. If a response is denied, it
8260 will be replaced with an HTTP 502 error so that the client never retrieves
8261 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01008262
8263 It is easier, faster and more powerful to use ACLs to write access policies.
8264 Rspdeny should be avoided in new designs.
8265
8266 Example :
8267 # Ensure that no content type matching ms-word will leak
8268 rspideny ^Content-type:\.*/ms-word
8269
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008270 See also: "reqdeny", "acl", "block", "http-response", section 6 about
8271 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008272
8273
Willy Tarreau96d51952019-05-22 20:34:35 +02008274rsprep <search> <string> [{if | unless} <cond>] (deprecated)
8275rspirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008276 Replace a regular expression with a string in an HTTP response line
8277 May be used in sections : defaults | frontend | listen | backend
8278 no | yes | yes | yes
8279 Arguments :
8280 <search> is the regular expression applied to HTTP headers and to the
8281 response line. This is an extended regular expression, so
8282 parenthesis grouping is supported and no preliminary backslash
8283 is required. Any space or known delimiter must be escaped using
8284 a backslash ('\'). The pattern applies to a full line at a time.
8285 The "rsprep" keyword strictly matches case while "rspirep"
8286 ignores case.
8287
8288 <string> is the complete line to be added. Any space or known delimiter
8289 must be escaped using a backslash ('\'). References to matched
8290 pattern groups are possible using the common \N form, with N
8291 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008292 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008293
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008294 <cond> is an optional matching condition built from ACLs. It makes it
8295 possible to ignore this rule when other conditions are not met.
8296
Willy Tarreau303c0352008-01-17 19:01:39 +01008297 Any line matching extended regular expression <search> in the response (both
8298 the response line and header lines) will be completely replaced with
8299 <string>. Most common use of this is to rewrite Location headers.
8300
8301 Header transformations only apply to traffic which passes through HAProxy,
8302 and not to traffic generated by HAProxy, such as health-checks or error
8303 responses. Note that for increased readability, it is suggested to add enough
8304 spaces between the request and the response. Keep in mind that header names
8305 are not case-sensitive.
8306
8307 Example :
8308 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
8309 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
8310
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008311 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
8312 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008313
8314
David du Colombier486df472011-03-17 10:40:26 +01008315server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008316 Declare a server in a backend
8317 May be used in sections : defaults | frontend | listen | backend
8318 no | no | yes | yes
8319 Arguments :
8320 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01008321 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05008322 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008323
David du Colombier486df472011-03-17 10:40:26 +01008324 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
8325 resolvable hostname is supported, but this name will be resolved
8326 during start-up. Address "0.0.0.0" or "*" has a special meaning.
8327 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02008328 address as the one from the client connection. This is useful in
8329 transparent proxy architectures where the client's connection is
8330 intercepted and haproxy must forward to the original destination
8331 address. This is more or less what the "transparent" keyword does
8332 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01008333 to report statistics. Optionally, an address family prefix may be
8334 used before the address to force the family regardless of the
8335 address format, which can be useful to specify a path to a unix
8336 socket with no slash ('/'). Currently supported prefixes are :
8337 - 'ipv4@' -> address is always IPv4
8338 - 'ipv6@' -> address is always IPv6
8339 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008340 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +02008341 - 'sockpair@' -> address is the FD of a connected unix
8342 socket or of a socketpair. During a connection, the
8343 backend creates a pair of connected sockets, and passes
8344 one of them over the FD. The bind part will use the
8345 received socket as the client FD. Should be used
8346 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02008347 You may want to reference some environment variables in the
8348 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01008349 variables. The "init-addr" setting can be used to modify the way
8350 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008351
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008352 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008353 be sent to this port. If unset, the same port the client
8354 connected to will be used. The port may also be prefixed by a "+"
8355 or a "-". In this case, the server's port will be determined by
8356 adding this value to the client's port.
8357
8358 <param*> is a list of parameters for this server. The "server" keywords
8359 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008360 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008361
8362 Examples :
8363 server first 10.1.1.1:1080 cookie first check inter 1000
8364 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01008365 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02008366 server backup "${SRV_BACKUP}:1080" backup
8367 server www1_dc1 "${LAN_DC1}.101:80"
8368 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008369
Willy Tarreau55dcaf62015-09-27 15:03:15 +02008370 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
8371 sun_path length is used for the address length. Some other programs
8372 such as socat use the string length only by default. Pass the option
8373 ",unix-tightsocklen=0" to any abstract socket definition in socat to
8374 make it compatible with HAProxy's.
8375
Mark Lamourinec2247f02012-01-04 13:02:01 -05008376 See also: "default-server", "http-send-name-header" and section 5 about
8377 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008378
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02008379server-state-file-name [<file>]
8380 Set the server state file to read, load and apply to servers available in
8381 this backend. It only applies when the directive "load-server-state-from-file"
8382 is set to "local". When <file> is not provided or if this directive is not
8383 set, then backend name is used. If <file> starts with a slash '/', then it is
8384 considered as an absolute path. Otherwise, <file> is concatenated to the
8385 global directive "server-state-file-base".
8386
8387 Example: the minimal configuration below would make HAProxy look for the
8388 state server file '/etc/haproxy/states/bk':
8389
8390 global
8391 server-state-file-base /etc/haproxy/states
8392
8393 backend bk
8394 load-server-state-from-file
8395
8396 See also: "server-state-file-base", "load-server-state-from-file", and
8397 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008398
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02008399server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
8400 Set a template to initialize servers with shared parameters.
8401 The names of these servers are built from <prefix> and <num | range> parameters.
8402 May be used in sections : defaults | frontend | listen | backend
8403 no | no | yes | yes
8404
8405 Arguments:
8406 <prefix> A prefix for the server names to be built.
8407
8408 <num | range>
8409 If <num> is provided, this template initializes <num> servers
8410 with 1 up to <num> as server name suffixes. A range of numbers
8411 <num_low>-<num_high> may also be used to use <num_low> up to
8412 <num_high> as server name suffixes.
8413
8414 <fqdn> A FQDN for all the servers this template initializes.
8415
8416 <port> Same meaning as "server" <port> argument (see "server" keyword).
8417
8418 <params*>
8419 Remaining server parameters among all those supported by "server"
8420 keyword.
8421
8422 Examples:
8423 # Initializes 3 servers with srv1, srv2 and srv3 as names,
8424 # google.com as FQDN, and health-check enabled.
8425 server-template srv 1-3 google.com:80 check
8426
8427 # or
8428 server-template srv 3 google.com:80 check
8429
8430 # would be equivalent to:
8431 server srv1 google.com:80 check
8432 server srv2 google.com:80 check
8433 server srv3 google.com:80 check
8434
8435
8436
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008437source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008438source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008439source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008440 Set the source address for outgoing connections
8441 May be used in sections : defaults | frontend | listen | backend
8442 yes | no | yes | yes
8443 Arguments :
8444 <addr> is the IPv4 address HAProxy will bind to before connecting to a
8445 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01008446
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008447 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01008448 the most appropriate address to reach its destination. Optionally
8449 an address family prefix may be used before the address to force
8450 the family regardless of the address format, which can be useful
8451 to specify a path to a unix socket with no slash ('/'). Currently
8452 supported prefixes are :
8453 - 'ipv4@' -> address is always IPv4
8454 - 'ipv6@' -> address is always IPv6
8455 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008456 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02008457 You may want to reference some environment variables in the
8458 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008459
8460 <port> is an optional port. It is normally not needed but may be useful
8461 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008462 the system will select a free port. Note that port ranges are not
8463 supported in the backend. If you want to force port ranges, you
8464 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008465
8466 <addr2> is the IP address to present to the server when connections are
8467 forwarded in full transparent proxy mode. This is currently only
8468 supported on some patched Linux kernels. When this address is
8469 specified, clients connecting to the server will be presented
8470 with this address, while health checks will still use the address
8471 <addr>.
8472
8473 <port2> is the optional port to present to the server when connections
8474 are forwarded in full transparent proxy mode (see <addr2> above).
8475 The default value of zero means the system will select a free
8476 port.
8477
Willy Tarreaubce70882009-09-07 11:51:47 +02008478 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
8479 This is the name of a comma-separated header list which can
8480 contain multiple IP addresses. By default, the last occurrence is
8481 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01008482 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02008483 by previous proxy, typically Stunnel. In order to use another
8484 occurrence from the last one, please see the <occ> parameter
8485 below. When the header (or occurrence) is not found, no binding
8486 is performed so that the proxy's default IP address is used. Also
8487 keep in mind that the header name is case insensitive, as for any
8488 HTTP header.
8489
8490 <occ> is the occurrence number of a value to be used in a multi-value
8491 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008492 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02008493 address. Positive values indicate a position from the first
8494 occurrence, 1 being the first one. Negative values indicate
8495 positions relative to the last one, -1 being the last one. This
8496 is helpful for situations where an X-Forwarded-For header is set
8497 at the entry point of an infrastructure and must be used several
8498 proxy layers away. When this value is not specified, -1 is
8499 assumed. Passing a zero here disables the feature.
8500
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008501 <name> is an optional interface name to which to bind to for outgoing
8502 traffic. On systems supporting this features (currently, only
8503 Linux), this allows one to bind all traffic to the server to
8504 this interface even if it is not the one the system would select
8505 based on routing tables. This should be used with extreme care.
8506 Note that using this option requires root privileges.
8507
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008508 The "source" keyword is useful in complex environments where a specific
8509 address only is allowed to connect to the servers. It may be needed when a
8510 private address must be used through a public gateway for instance, and it is
8511 known that the system cannot determine the adequate source address by itself.
8512
8513 An extension which is available on certain patched Linux kernels may be used
8514 through the "usesrc" optional keyword. It makes it possible to connect to the
8515 servers with an IP address which does not belong to the system itself. This
8516 is called "full transparent proxy mode". For this to work, the destination
8517 servers have to route their traffic back to this address through the machine
8518 running HAProxy, and IP forwarding must generally be enabled on this machine.
8519
8520 In this "full transparent proxy" mode, it is possible to force a specific IP
8521 address to be presented to the servers. This is not much used in fact. A more
8522 common use is to tell HAProxy to present the client's IP address. For this,
8523 there are two methods :
8524
8525 - present the client's IP and port addresses. This is the most transparent
8526 mode, but it can cause problems when IP connection tracking is enabled on
8527 the machine, because a same connection may be seen twice with different
8528 states. However, this solution presents the huge advantage of not
8529 limiting the system to the 64k outgoing address+port couples, because all
8530 of the client ranges may be used.
8531
8532 - present only the client's IP address and select a spare port. This
8533 solution is still quite elegant but slightly less transparent (downstream
8534 firewalls logs will not match upstream's). It also presents the downside
8535 of limiting the number of concurrent connections to the usual 64k ports.
8536 However, since the upstream and downstream ports are different, local IP
8537 connection tracking on the machine will not be upset by the reuse of the
8538 same session.
8539
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008540 This option sets the default source for all servers in the backend. It may
8541 also be specified in a "defaults" section. Finer source address specification
8542 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008543 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008544
Baptiste Assmann91bd3372015-07-17 21:59:42 +02008545 In order to work, "usesrc" requires root privileges.
8546
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008547 Examples :
8548 backend private
8549 # Connect to the servers using our 192.168.1.200 source address
8550 source 192.168.1.200
8551
8552 backend transparent_ssl1
8553 # Connect to the SSL farm from the client's source address
8554 source 192.168.1.200 usesrc clientip
8555
8556 backend transparent_ssl2
8557 # Connect to the SSL farm from the client's source address and port
8558 # not recommended if IP conntrack is present on the local machine.
8559 source 192.168.1.200 usesrc client
8560
8561 backend transparent_ssl3
8562 # Connect to the SSL farm from the client's source address. It
8563 # is more conntrack-friendly.
8564 source 192.168.1.200 usesrc clientip
8565
8566 backend transparent_smtp
8567 # Connect to the SMTP farm from the client's source address/port
8568 # with Tproxy version 4.
8569 source 0.0.0.0 usesrc clientip
8570
Willy Tarreaubce70882009-09-07 11:51:47 +02008571 backend transparent_http
8572 # Connect to the servers using the client's IP as seen by previous
8573 # proxy.
8574 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
8575
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008576 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008577 the Linux kernel on www.balabit.com, the "bind" keyword.
8578
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01008579
Willy Tarreau844e3c52008-01-11 16:28:18 +01008580srvtimeout <timeout> (deprecated)
8581 Set the maximum inactivity time on the server side.
8582 May be used in sections : defaults | frontend | listen | backend
8583 yes | no | yes | yes
8584 Arguments :
8585 <timeout> is the timeout value specified in milliseconds by default, but
8586 can be in any other unit if the number is suffixed by the unit,
8587 as explained at the top of this document.
8588
8589 The inactivity timeout applies when the server is expected to acknowledge or
8590 send data. In HTTP mode, this timeout is particularly important to consider
8591 during the first phase of the server's response, when it has to send the
8592 headers, as it directly represents the server's processing time for the
8593 request. To find out what value to put there, it's often good to start with
8594 what would be considered as unacceptable response times, then check the logs
8595 to observe the response time distribution, and adjust the value accordingly.
8596
8597 The value is specified in milliseconds by default, but can be in any other
8598 unit if the number is suffixed by the unit, as specified at the top of this
8599 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
8600 recommended that the client timeout remains equal to the server timeout in
8601 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008602 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01008603 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +01008604 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01008605
8606 This parameter is specific to backends, but can be specified once for all in
8607 "defaults" sections. This is in fact one of the easiest solutions not to
8608 forget about it. An unspecified timeout results in an infinite timeout, which
8609 is not recommended. Such a usage is accepted and works but reports a warning
8610 during startup because it may results in accumulation of expired sessions in
8611 the system if the system's timeouts are not configured either.
8612
8613 This parameter is provided for compatibility but is currently deprecated.
8614 Please use "timeout server" instead.
8615
Willy Tarreauce887fd2012-05-12 12:50:00 +02008616 See also : "timeout server", "timeout tunnel", "timeout client" and
8617 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008618
8619
Cyril Bonté66c327d2010-10-12 00:14:37 +02008620stats admin { if | unless } <cond>
8621 Enable statistics admin level if/unless a condition is matched
8622 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008623 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008624
8625 This statement enables the statistics admin level if/unless a condition is
8626 matched.
8627
8628 The admin level allows to enable/disable servers from the web interface. By
8629 default, statistics page is read-only for security reasons.
8630
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008631 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8632 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008633 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008634
Cyril Bonté23b39d92011-02-10 22:54:44 +01008635 Currently, the POST request is limited to the buffer size minus the reserved
8636 buffer space, which means that if the list of servers is too long, the
8637 request won't be processed. It is recommended to alter few servers at a
8638 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008639
8640 Example :
8641 # statistics admin level only for localhost
8642 backend stats_localhost
8643 stats enable
8644 stats admin if LOCALHOST
8645
8646 Example :
8647 # statistics admin level always enabled because of the authentication
8648 backend stats_auth
8649 stats enable
8650 stats auth admin:AdMiN123
8651 stats admin if TRUE
8652
8653 Example :
8654 # statistics admin level depends on the authenticated user
8655 userlist stats-auth
8656 group admin users admin
8657 user admin insecure-password AdMiN123
8658 group readonly users haproxy
8659 user haproxy insecure-password haproxy
8660
8661 backend stats_auth
8662 stats enable
8663 acl AUTH http_auth(stats-auth)
8664 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8665 stats http-request auth unless AUTH
8666 stats admin if AUTH_ADMIN
8667
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008668 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8669 "bind-process", section 3.4 about userlists and section 7 about
8670 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008671
8672
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008673stats auth <user>:<passwd>
8674 Enable statistics with authentication and grant access to an account
8675 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008676 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008677 Arguments :
8678 <user> is a user name to grant access to
8679
8680 <passwd> is the cleartext password associated to this user
8681
8682 This statement enables statistics with default settings, and restricts access
8683 to declared users only. It may be repeated as many times as necessary to
8684 allow as many users as desired. When a user tries to access the statistics
8685 without a valid account, a "401 Forbidden" response will be returned so that
8686 the browser asks the user to provide a valid user and password. The real
8687 which will be returned to the browser is configurable using "stats realm".
8688
8689 Since the authentication method is HTTP Basic Authentication, the passwords
8690 circulate in cleartext on the network. Thus, it was decided that the
8691 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008692 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008693
8694 It is also possible to reduce the scope of the proxies which appear in the
8695 report using "stats scope".
8696
8697 Though this statement alone is enough to enable statistics reporting, it is
8698 recommended to set all other settings in order to avoid relying on default
8699 unobvious parameters.
8700
8701 Example :
8702 # public access (limited to this backend only)
8703 backend public_www
8704 server srv1 192.168.0.1:80
8705 stats enable
8706 stats hide-version
8707 stats scope .
8708 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008709 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008710 stats auth admin1:AdMiN123
8711 stats auth admin2:AdMiN321
8712
8713 # internal monitoring access (unlimited)
8714 backend private_monitoring
8715 stats enable
8716 stats uri /admin?stats
8717 stats refresh 5s
8718
8719 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8720
8721
8722stats enable
8723 Enable statistics reporting with default settings
8724 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008725 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008726 Arguments : none
8727
8728 This statement enables statistics reporting with default settings defined
8729 at build time. Unless stated otherwise, these settings are used :
8730 - stats uri : /haproxy?stats
8731 - stats realm : "HAProxy Statistics"
8732 - stats auth : no authentication
8733 - stats scope : no restriction
8734
8735 Though this statement alone is enough to enable statistics reporting, it is
8736 recommended to set all other settings in order to avoid relying on default
8737 unobvious parameters.
8738
8739 Example :
8740 # public access (limited to this backend only)
8741 backend public_www
8742 server srv1 192.168.0.1:80
8743 stats enable
8744 stats hide-version
8745 stats scope .
8746 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008747 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008748 stats auth admin1:AdMiN123
8749 stats auth admin2:AdMiN321
8750
8751 # internal monitoring access (unlimited)
8752 backend private_monitoring
8753 stats enable
8754 stats uri /admin?stats
8755 stats refresh 5s
8756
8757 See also : "stats auth", "stats realm", "stats uri"
8758
8759
Willy Tarreaud63335a2010-02-26 12:56:52 +01008760stats hide-version
8761 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008762 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008763 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008764 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008765
Willy Tarreaud63335a2010-02-26 12:56:52 +01008766 By default, the stats page reports some useful status information along with
8767 the statistics. Among them is HAProxy's version. However, it is generally
8768 considered dangerous to report precise version to anyone, as it can help them
8769 target known weaknesses with specific attacks. The "stats hide-version"
8770 statement removes the version from the statistics report. This is recommended
8771 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008772
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008773 Though this statement alone is enough to enable statistics reporting, it is
8774 recommended to set all other settings in order to avoid relying on default
8775 unobvious parameters.
8776
Willy Tarreaud63335a2010-02-26 12:56:52 +01008777 Example :
8778 # public access (limited to this backend only)
8779 backend public_www
8780 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008781 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008782 stats hide-version
8783 stats scope .
8784 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008785 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008786 stats auth admin1:AdMiN123
8787 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008788
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008789 # internal monitoring access (unlimited)
8790 backend private_monitoring
8791 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008792 stats uri /admin?stats
8793 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008794
Willy Tarreaud63335a2010-02-26 12:56:52 +01008795 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008796
Willy Tarreau983e01e2010-01-11 18:42:06 +01008797
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008798stats http-request { allow | deny | auth [realm <realm>] }
8799 [ { if | unless } <condition> ]
8800 Access control for statistics
8801
8802 May be used in sections: defaults | frontend | listen | backend
8803 no | no | yes | yes
8804
8805 As "http-request", these set of options allow to fine control access to
8806 statistics. Each option may be followed by if/unless and acl.
8807 First option with matched condition (or option without condition) is final.
8808 For "deny" a 403 error will be returned, for "allow" normal processing is
8809 performed, for "auth" a 401/407 error code is returned so the client
8810 should be asked to enter a username and password.
8811
8812 There is no fixed limit to the number of http-request statements per
8813 instance.
8814
8815 See also : "http-request", section 3.4 about userlists and section 7
8816 about ACL usage.
8817
8818
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008819stats realm <realm>
8820 Enable statistics and set authentication realm
8821 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008822 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008823 Arguments :
8824 <realm> is the name of the HTTP Basic Authentication realm reported to
8825 the browser. The browser uses it to display it in the pop-up
8826 inviting the user to enter a valid username and password.
8827
8828 The realm is read as a single word, so any spaces in it should be escaped
8829 using a backslash ('\').
8830
8831 This statement is useful only in conjunction with "stats auth" since it is
8832 only related to authentication.
8833
8834 Though this statement alone is enough to enable statistics reporting, it is
8835 recommended to set all other settings in order to avoid relying on default
8836 unobvious parameters.
8837
8838 Example :
8839 # public access (limited to this backend only)
8840 backend public_www
8841 server srv1 192.168.0.1:80
8842 stats enable
8843 stats hide-version
8844 stats scope .
8845 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008846 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008847 stats auth admin1:AdMiN123
8848 stats auth admin2:AdMiN321
8849
8850 # internal monitoring access (unlimited)
8851 backend private_monitoring
8852 stats enable
8853 stats uri /admin?stats
8854 stats refresh 5s
8855
8856 See also : "stats auth", "stats enable", "stats uri"
8857
8858
8859stats refresh <delay>
8860 Enable statistics with automatic refresh
8861 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008862 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008863 Arguments :
8864 <delay> is the suggested refresh delay, specified in seconds, which will
8865 be returned to the browser consulting the report page. While the
8866 browser is free to apply any delay, it will generally respect it
8867 and refresh the page this every seconds. The refresh interval may
8868 be specified in any other non-default time unit, by suffixing the
8869 unit after the value, as explained at the top of this document.
8870
8871 This statement is useful on monitoring displays with a permanent page
8872 reporting the load balancer's activity. When set, the HTML report page will
8873 include a link "refresh"/"stop refresh" so that the user can select whether
8874 he wants automatic refresh of the page or not.
8875
8876 Though this statement alone is enough to enable statistics reporting, it is
8877 recommended to set all other settings in order to avoid relying on default
8878 unobvious parameters.
8879
8880 Example :
8881 # public access (limited to this backend only)
8882 backend public_www
8883 server srv1 192.168.0.1:80
8884 stats enable
8885 stats hide-version
8886 stats scope .
8887 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008888 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008889 stats auth admin1:AdMiN123
8890 stats auth admin2:AdMiN321
8891
8892 # internal monitoring access (unlimited)
8893 backend private_monitoring
8894 stats enable
8895 stats uri /admin?stats
8896 stats refresh 5s
8897
8898 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8899
8900
8901stats scope { <name> | "." }
8902 Enable statistics and limit access scope
8903 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008904 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008905 Arguments :
8906 <name> is the name of a listen, frontend or backend section to be
8907 reported. The special name "." (a single dot) designates the
8908 section in which the statement appears.
8909
8910 When this statement is specified, only the sections enumerated with this
8911 statement will appear in the report. All other ones will be hidden. This
8912 statement may appear as many times as needed if multiple sections need to be
8913 reported. Please note that the name checking is performed as simple string
8914 comparisons, and that it is never checked that a give section name really
8915 exists.
8916
8917 Though this statement alone is enough to enable statistics reporting, it is
8918 recommended to set all other settings in order to avoid relying on default
8919 unobvious parameters.
8920
8921 Example :
8922 # public access (limited to this backend only)
8923 backend public_www
8924 server srv1 192.168.0.1:80
8925 stats enable
8926 stats hide-version
8927 stats scope .
8928 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008929 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008930 stats auth admin1:AdMiN123
8931 stats auth admin2:AdMiN321
8932
8933 # internal monitoring access (unlimited)
8934 backend private_monitoring
8935 stats enable
8936 stats uri /admin?stats
8937 stats refresh 5s
8938
8939 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8940
Willy Tarreaud63335a2010-02-26 12:56:52 +01008941
Willy Tarreauc9705a12010-07-27 20:05:50 +02008942stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01008943 Enable reporting of a description on the statistics page.
8944 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008945 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008946
Willy Tarreauc9705a12010-07-27 20:05:50 +02008947 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01008948 description from global section is automatically used instead.
8949
8950 This statement is useful for users that offer shared services to their
8951 customers, where node or description should be different for each customer.
8952
8953 Though this statement alone is enough to enable statistics reporting, it is
8954 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008955 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008956
8957 Example :
8958 # internal monitoring access (unlimited)
8959 backend private_monitoring
8960 stats enable
8961 stats show-desc Master node for Europe, Asia, Africa
8962 stats uri /admin?stats
8963 stats refresh 5s
8964
8965 See also: "show-node", "stats enable", "stats uri" and "description" in
8966 global section.
8967
8968
8969stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02008970 Enable reporting additional information on the statistics page
8971 May be used in sections : defaults | frontend | listen | backend
8972 yes | yes | yes | yes
8973 Arguments : none
8974
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008975 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01008976 - cap: capabilities (proxy)
8977 - mode: one of tcp, http or health (proxy)
8978 - id: SNMP ID (proxy, socket, server)
8979 - IP (socket, server)
8980 - cookie (backend, server)
8981
8982 Though this statement alone is enough to enable statistics reporting, it is
8983 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008984 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008985
8986 See also: "stats enable", "stats uri".
8987
8988
8989stats show-node [ <name> ]
8990 Enable reporting of a host name on the statistics page.
8991 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008992 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008993 Arguments:
8994 <name> is an optional name to be reported. If unspecified, the
8995 node name from global section is automatically used instead.
8996
8997 This statement is useful for users that offer shared services to their
8998 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01008999 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01009000
9001 Though this statement alone is enough to enable statistics reporting, it is
9002 recommended to set all other settings in order to avoid relying on default
9003 unobvious parameters.
9004
9005 Example:
9006 # internal monitoring access (unlimited)
9007 backend private_monitoring
9008 stats enable
9009 stats show-node Europe-1
9010 stats uri /admin?stats
9011 stats refresh 5s
9012
9013 See also: "show-desc", "stats enable", "stats uri", and "node" in global
9014 section.
9015
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009016
9017stats uri <prefix>
9018 Enable statistics and define the URI prefix to access them
9019 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02009020 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009021 Arguments :
9022 <prefix> is the prefix of any URI which will be redirected to stats. This
9023 prefix may contain a question mark ('?') to indicate part of a
9024 query string.
9025
9026 The statistics URI is intercepted on the relayed traffic, so it appears as a
9027 page within the normal application. It is strongly advised to ensure that the
9028 selected URI will never appear in the application, otherwise it will never be
9029 possible to reach it in the application.
9030
9031 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009032 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009033 It is generally a good idea to include a question mark in the URI so that
9034 intermediate proxies refrain from caching the results. Also, since any string
9035 beginning with the prefix will be accepted as a stats request, the question
9036 mark helps ensuring that no valid URI will begin with the same words.
9037
9038 It is sometimes very convenient to use "/" as the URI prefix, and put that
9039 statement in a "listen" instance of its own. That makes it easy to dedicate
9040 an address or a port to statistics only.
9041
9042 Though this statement alone is enough to enable statistics reporting, it is
9043 recommended to set all other settings in order to avoid relying on default
9044 unobvious parameters.
9045
9046 Example :
9047 # public access (limited to this backend only)
9048 backend public_www
9049 server srv1 192.168.0.1:80
9050 stats enable
9051 stats hide-version
9052 stats scope .
9053 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01009054 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009055 stats auth admin1:AdMiN123
9056 stats auth admin2:AdMiN321
9057
9058 # internal monitoring access (unlimited)
9059 backend private_monitoring
9060 stats enable
9061 stats uri /admin?stats
9062 stats refresh 5s
9063
9064 See also : "stats auth", "stats enable", "stats realm"
9065
9066
Willy Tarreaud63335a2010-02-26 12:56:52 +01009067stick match <pattern> [table <table>] [{if | unless} <cond>]
9068 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01009069 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01009070 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009071
9072 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009073 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009074 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009075 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009076 stickiness table. This rule is mandatory.
9077
9078 <table> is an optional stickiness table name. If unspecified, the same
9079 backend's table is used. A stickiness table is declared using
9080 the "stick-table" statement.
9081
9082 <cond> is an optional matching condition. It makes it possible to match
9083 on a certain criterion only when other conditions are met (or
9084 not met). For instance, it could be used to match on a source IP
9085 address except when a request passes through a known proxy, in
9086 which case we'd match on a header containing that IP address.
9087
9088 Some protocols or applications require complex stickiness rules and cannot
9089 always simply rely on cookies nor hashing. The "stick match" statement
9090 describes a rule to extract the stickiness criterion from an incoming request
9091 or connection. See section 7 for a complete list of possible patterns and
9092 transformation rules.
9093
9094 The table has to be declared using the "stick-table" statement. It must be of
9095 a type compatible with the pattern. By default it is the one which is present
9096 in the same backend. It is possible to share a table with other backends by
9097 referencing it using the "table" keyword. If another table is referenced,
9098 the server's ID inside the backends are used. By default, all server IDs
9099 start at 1 in each backend, so the server ordering is enough. But in case of
9100 doubt, it is highly recommended to force server IDs using their "id" setting.
9101
9102 It is possible to restrict the conditions where a "stick match" statement
9103 will apply, using "if" or "unless" followed by a condition. See section 7 for
9104 ACL based conditions.
9105
9106 There is no limit on the number of "stick match" statements. The first that
9107 applies and matches will cause the request to be directed to the same server
9108 as was used for the request which created the entry. That way, multiple
9109 matches can be used as fallbacks.
9110
9111 The stick rules are checked after the persistence cookies, so they will not
9112 affect stickiness if a cookie has already been used to select a server. That
9113 way, it becomes very easy to insert cookies and match on IP addresses in
9114 order to maintain stickiness between HTTP and HTTPS.
9115
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009116 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9117 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009118 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009119
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009120 Example :
9121 # forward SMTP users to the same server they just used for POP in the
9122 # last 30 minutes
9123 backend pop
9124 mode tcp
9125 balance roundrobin
9126 stick store-request src
9127 stick-table type ip size 200k expire 30m
9128 server s1 192.168.1.1:110
9129 server s2 192.168.1.1:110
9130
9131 backend smtp
9132 mode tcp
9133 balance roundrobin
9134 stick match src table pop
9135 server s1 192.168.1.1:25
9136 server s2 192.168.1.1:25
9137
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009138 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009139 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009140
9141
9142stick on <pattern> [table <table>] [{if | unless} <condition>]
9143 Define a request pattern to associate a user to a server
9144 May be used in sections : defaults | frontend | listen | backend
9145 no | no | yes | yes
9146
9147 Note : This form is exactly equivalent to "stick match" followed by
9148 "stick store-request", all with the same arguments. Please refer
9149 to both keywords for details. It is only provided as a convenience
9150 for writing more maintainable configurations.
9151
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009152 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9153 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009154 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009155
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009156 Examples :
9157 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01009158 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009159
9160 # ...is strictly equivalent to this one :
9161 stick match src table pop if !localhost
9162 stick store-request src table pop if !localhost
9163
9164
9165 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
9166 # well as HTTP without cookie. Share the same table between both accesses.
9167 backend http
9168 mode http
9169 balance roundrobin
9170 stick on src table https
9171 cookie SRV insert indirect nocache
9172 server s1 192.168.1.1:80 cookie s1
9173 server s2 192.168.1.1:80 cookie s2
9174
9175 backend https
9176 mode tcp
9177 balance roundrobin
9178 stick-table type ip size 200k expire 30m
9179 stick on src
9180 server s1 192.168.1.1:443
9181 server s2 192.168.1.1:443
9182
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009183 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009184
9185
9186stick store-request <pattern> [table <table>] [{if | unless} <condition>]
9187 Define a request pattern used to create an entry in a stickiness table
9188 May be used in sections : defaults | frontend | listen | backend
9189 no | no | yes | yes
9190
9191 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009192 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009193 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009194 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009195 server is selected.
9196
9197 <table> is an optional stickiness table name. If unspecified, the same
9198 backend's table is used. A stickiness table is declared using
9199 the "stick-table" statement.
9200
9201 <cond> is an optional storage condition. It makes it possible to store
9202 certain criteria only when some conditions are met (or not met).
9203 For instance, it could be used to store the source IP address
9204 except when the request passes through a known proxy, in which
9205 case we'd store a converted form of a header containing that IP
9206 address.
9207
9208 Some protocols or applications require complex stickiness rules and cannot
9209 always simply rely on cookies nor hashing. The "stick store-request" statement
9210 describes a rule to decide what to extract from the request and when to do
9211 it, in order to store it into a stickiness table for further requests to
9212 match it using the "stick match" statement. Obviously the extracted part must
9213 make sense and have a chance to be matched in a further request. Storing a
9214 client's IP address for instance often makes sense. Storing an ID found in a
9215 URL parameter also makes sense. Storing a source port will almost never make
9216 any sense because it will be randomly matched. See section 7 for a complete
9217 list of possible patterns and transformation rules.
9218
9219 The table has to be declared using the "stick-table" statement. It must be of
9220 a type compatible with the pattern. By default it is the one which is present
9221 in the same backend. It is possible to share a table with other backends by
9222 referencing it using the "table" keyword. If another table is referenced,
9223 the server's ID inside the backends are used. By default, all server IDs
9224 start at 1 in each backend, so the server ordering is enough. But in case of
9225 doubt, it is highly recommended to force server IDs using their "id" setting.
9226
9227 It is possible to restrict the conditions where a "stick store-request"
9228 statement will apply, using "if" or "unless" followed by a condition. This
9229 condition will be evaluated while parsing the request, so any criteria can be
9230 used. See section 7 for ACL based conditions.
9231
9232 There is no limit on the number of "stick store-request" statements, but
9233 there is a limit of 8 simultaneous stores per request or response. This
9234 makes it possible to store up to 8 criteria, all extracted from either the
9235 request or the response, regardless of the number of rules. Only the 8 first
9236 ones which match will be kept. Using this, it is possible to feed multiple
9237 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009238 another protocol or access method. Using multiple store-request rules with
9239 the same table is possible and may be used to find the best criterion to rely
9240 on, by arranging the rules by decreasing preference order. Only the first
9241 extracted criterion for a given table will be stored. All subsequent store-
9242 request rules referencing the same table will be skipped and their ACLs will
9243 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009244
9245 The "store-request" rules are evaluated once the server connection has been
9246 established, so that the table will contain the real server that processed
9247 the request.
9248
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009249 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9250 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009251 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009252
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009253 Example :
9254 # forward SMTP users to the same server they just used for POP in the
9255 # last 30 minutes
9256 backend pop
9257 mode tcp
9258 balance roundrobin
9259 stick store-request src
9260 stick-table type ip size 200k expire 30m
9261 server s1 192.168.1.1:110
9262 server s2 192.168.1.1:110
9263
9264 backend smtp
9265 mode tcp
9266 balance roundrobin
9267 stick match src table pop
9268 server s1 192.168.1.1:25
9269 server s2 192.168.1.1:25
9270
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009271 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009272 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009273
9274
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009275stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02009276 size <size> [expire <expire>] [nopurge] [peers <peersect>]
9277 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08009278 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009279 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009280 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009281
9282 Arguments :
9283 ip a table declared with "type ip" will only store IPv4 addresses.
9284 This form is very compact (about 50 bytes per entry) and allows
9285 very fast entry lookup and stores with almost no overhead. This
9286 is mainly used to store client source IP addresses.
9287
David du Colombier9a6d3c92011-03-17 10:40:24 +01009288 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
9289 This form is very compact (about 60 bytes per entry) and allows
9290 very fast entry lookup and stores with almost no overhead. This
9291 is mainly used to store client source IP addresses.
9292
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009293 integer a table declared with "type integer" will store 32bit integers
9294 which can represent a client identifier found in a request for
9295 instance.
9296
9297 string a table declared with "type string" will store substrings of up
9298 to <len> characters. If the string provided by the pattern
9299 extractor is larger than <len>, it will be truncated before
9300 being stored. During matching, at most <len> characters will be
9301 compared between the string in the table and the extracted
9302 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009303 to 32 characters.
9304
9305 binary a table declared with "type binary" will store binary blocks
9306 of <len> bytes. If the block provided by the pattern
9307 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02009308 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009309 is shorter than <len>, it will be padded by 0. When not
9310 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009311
9312 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009313 "string" type table (See type "string" above). Or the number
9314 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009315 changing this parameter as memory usage will proportionally
9316 increase.
9317
9318 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01009319 value directly impacts memory usage. Count approximately
9320 50 bytes per entry, plus the size of a string if any. The size
9321 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009322
9323 [nopurge] indicates that we refuse to purge older entries when the table
9324 is full. When not specified and the table is full when haproxy
9325 wants to store an entry in it, it will flush a few of the oldest
9326 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009327 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009328 be desirable to refuse new entries instead of purging the older
9329 ones. That may be the case when the amount of data to store is
9330 far above the hardware limits and we prefer not to offer access
9331 to new clients than to reject the ones already connected. When
9332 using this parameter, be sure to properly set the "expire"
9333 parameter (see below).
9334
Emeric Brunf099e792010-09-27 12:05:28 +02009335 <peersect> is the name of the peers section to use for replication. Entries
9336 which associate keys to server IDs are kept synchronized with
9337 the remote peers declared in this section. All entries are also
9338 automatically learned from the local peer (old process) during a
9339 soft restart.
9340
Willy Tarreau1abc6732015-05-01 19:21:02 +02009341 NOTE : each peers section may be referenced only by tables
9342 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009343
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009344 <expire> defines the maximum duration of an entry in the table since it
9345 was last created, refreshed or matched. The expiration delay is
9346 defined using the standard time format, similarly as the various
9347 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009348 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02009349 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009350 be removed once full. Be sure not to use the "nopurge" parameter
9351 if not expiration delay is specified.
9352
Willy Tarreau08d5f982010-06-06 13:34:54 +02009353 <data_type> is used to store additional information in the stick-table. This
9354 may be used by ACLs in order to control various criteria related
9355 to the activity of the client matching the stick-table. For each
9356 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02009357 that the additional data can fit. Several data types may be
9358 stored with an entry. Multiple data types may be specified after
9359 the "store" keyword, as a comma-separated list. Alternatively,
9360 it is possible to repeat the "store" keyword followed by one or
9361 several data types. Except for the "server_id" type which is
9362 automatically detected and enabled, all data types must be
9363 explicitly declared to be stored. If an ACL references a data
9364 type which is not stored, the ACL will simply not match. Some
9365 data types require an argument which must be passed just after
9366 the type between parenthesis. See below for the supported data
9367 types and their arguments.
9368
9369 The data types that can be stored with an entry are the following :
9370 - server_id : this is an integer which holds the numeric ID of the server a
9371 request was assigned to. It is used by the "stick match", "stick store",
9372 and "stick on" rules. It is automatically enabled when referenced.
9373
9374 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
9375 integer which may be used for anything. Most of the time it will be used
9376 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01009377 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009378
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009379 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
9380 over a period. It is a positive 32-bit integer integer which may be used
9381 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01009382 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009383 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01009384 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009385
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009386 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
9387 integer which may be used for anything. Most of the time it will be used
9388 to put a special tag on some entries, for instance to note that a
9389 specific behavior was detected and must be known for future matches.
9390
9391 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
9392 over a period. It is a positive 32-bit integer integer which may be used
9393 for anything. Just like <gpc1>, it counts events, but instead of keeping
9394 a cumulative number, it maintains the rate at which the counter is
9395 incremented. Most of the time it will be used to measure the frequency of
9396 occurrence of certain events (e.g. requests to a specific URL).
9397
Willy Tarreauc9705a12010-07-27 20:05:50 +02009398 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
9399 the absolute number of connections received from clients which matched
9400 this entry. It does not mean the connections were accepted, just that
9401 they were received.
9402
9403 - conn_cur : Current Connections. It is a positive 32-bit integer which
9404 stores the concurrent connection counts for the entry. It is incremented
9405 once an incoming connection matches the entry, and decremented once the
9406 connection leaves. That way it is possible to know at any time the exact
9407 number of concurrent connections for an entry.
9408
9409 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9410 integer parameter <period> which indicates in milliseconds the length
9411 of the period over which the average is measured. It reports the average
9412 incoming connection rate over that period, in connections per period. The
9413 result is an integer which can be matched using ACLs.
9414
9415 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
9416 the absolute number of sessions received from clients which matched this
9417 entry. A session is a connection that was accepted by the layer 4 rules.
9418
9419 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9420 integer parameter <period> which indicates in milliseconds the length
9421 of the period over which the average is measured. It reports the average
9422 incoming session rate over that period, in sessions per period. The
9423 result is an integer which can be matched using ACLs.
9424
9425 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
9426 counts the absolute number of HTTP requests received from clients which
9427 matched this entry. It does not matter whether they are valid requests or
9428 not. Note that this is different from sessions when keep-alive is used on
9429 the client side.
9430
9431 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9432 integer parameter <period> which indicates in milliseconds the length
9433 of the period over which the average is measured. It reports the average
9434 HTTP request rate over that period, in requests per period. The result is
9435 an integer which can be matched using ACLs. It does not matter whether
9436 they are valid requests or not. Note that this is different from sessions
9437 when keep-alive is used on the client side.
9438
9439 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
9440 counts the absolute number of HTTP requests errors induced by clients
9441 which matched this entry. Errors are counted on invalid and truncated
9442 requests, as well as on denied or tarpitted requests, and on failed
9443 authentications. If the server responds with 4xx, then the request is
9444 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01009445 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02009446
9447 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9448 integer parameter <period> which indicates in milliseconds the length
9449 of the period over which the average is measured. It reports the average
9450 HTTP request error rate over that period, in requests per period (see
9451 http_err_cnt above for what is accounted as an error). The result is an
9452 integer which can be matched using ACLs.
9453
9454 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009455 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02009456 which matched this entry. Headers are included in the count. This may be
9457 used to limit abuse of upload features on photo or video servers.
9458
9459 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9460 integer parameter <period> which indicates in milliseconds the length
9461 of the period over which the average is measured. It reports the average
9462 incoming bytes rate over that period, in bytes per period. It may be used
9463 to detect users which upload too much and too fast. Warning: with large
9464 uploads, it is possible that the amount of uploaded data will be counted
9465 once upon termination, thus causing spikes in the average transfer speed
9466 instead of having a smooth one. This may partially be smoothed with
9467 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
9468 recommended for better fairness.
9469
9470 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009471 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02009472 matched this entry. Headers are included in the count. This may be used
9473 to limit abuse of bots sucking the whole site.
9474
9475 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
9476 an integer parameter <period> which indicates in milliseconds the length
9477 of the period over which the average is measured. It reports the average
9478 outgoing bytes rate over that period, in bytes per period. It may be used
9479 to detect users which download too much and too fast. Warning: with large
9480 transfers, it is possible that the amount of transferred data will be
9481 counted once upon termination, thus causing spikes in the average
9482 transfer speed instead of having a smooth one. This may partially be
9483 smoothed with "option contstats" though this is not perfect yet. Use of
9484 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02009485
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009486 There is only one stick-table per proxy. At the moment of writing this doc,
9487 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009488 to be required, simply create a dummy backend with a stick-table in it and
9489 reference it.
9490
9491 It is important to understand that stickiness based on learning information
9492 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01009493 lost upon restart unless peers are properly configured to transfer such
9494 information upon restart (recommended). In general it can be good as a
9495 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009496
Willy Tarreauc9705a12010-07-27 20:05:50 +02009497 Last, memory requirements may be important when storing many data types.
9498 Indeed, storing all indicators above at once in each entry requires 116 bytes
9499 per entry, or 116 MB for a 1-million entries table. This is definitely not
9500 something that can be ignored.
9501
9502 Example:
9503 # Keep track of counters of up to 1 million IP addresses over 5 minutes
9504 # and store a general purpose counter and the average connection rate
9505 # computed over a sliding window of 30 seconds.
9506 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
9507
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009508 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01009509 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009510
9511
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009512stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01009513 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009514 May be used in sections : defaults | frontend | listen | backend
9515 no | no | yes | yes
9516
9517 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009518 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009519 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01009520 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009521 server is selected.
9522
9523 <table> is an optional stickiness table name. If unspecified, the same
9524 backend's table is used. A stickiness table is declared using
9525 the "stick-table" statement.
9526
9527 <cond> is an optional storage condition. It makes it possible to store
9528 certain criteria only when some conditions are met (or not met).
9529 For instance, it could be used to store the SSL session ID only
9530 when the response is a SSL server hello.
9531
9532 Some protocols or applications require complex stickiness rules and cannot
9533 always simply rely on cookies nor hashing. The "stick store-response"
9534 statement describes a rule to decide what to extract from the response and
9535 when to do it, in order to store it into a stickiness table for further
9536 requests to match it using the "stick match" statement. Obviously the
9537 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009538 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009539 See section 7 for a complete list of possible patterns and transformation
9540 rules.
9541
9542 The table has to be declared using the "stick-table" statement. It must be of
9543 a type compatible with the pattern. By default it is the one which is present
9544 in the same backend. It is possible to share a table with other backends by
9545 referencing it using the "table" keyword. If another table is referenced,
9546 the server's ID inside the backends are used. By default, all server IDs
9547 start at 1 in each backend, so the server ordering is enough. But in case of
9548 doubt, it is highly recommended to force server IDs using their "id" setting.
9549
9550 It is possible to restrict the conditions where a "stick store-response"
9551 statement will apply, using "if" or "unless" followed by a condition. This
9552 condition will be evaluated while parsing the response, so any criteria can
9553 be used. See section 7 for ACL based conditions.
9554
9555 There is no limit on the number of "stick store-response" statements, but
9556 there is a limit of 8 simultaneous stores per request or response. This
9557 makes it possible to store up to 8 criteria, all extracted from either the
9558 request or the response, regardless of the number of rules. Only the 8 first
9559 ones which match will be kept. Using this, it is possible to feed multiple
9560 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009561 another protocol or access method. Using multiple store-response rules with
9562 the same table is possible and may be used to find the best criterion to rely
9563 on, by arranging the rules by decreasing preference order. Only the first
9564 extracted criterion for a given table will be stored. All subsequent store-
9565 response rules referencing the same table will be skipped and their ACLs will
9566 not be evaluated. However, even if a store-request rule references a table, a
9567 store-response rule may also use the same table. This means that each table
9568 may learn exactly one element from the request and one element from the
9569 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009570
9571 The table will contain the real server that processed the request.
9572
9573 Example :
9574 # Learn SSL session ID from both request and response and create affinity.
9575 backend https
9576 mode tcp
9577 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02009578 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009579 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009580
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009581 acl clienthello req_ssl_hello_type 1
9582 acl serverhello rep_ssl_hello_type 2
9583
9584 # use tcp content accepts to detects ssl client and server hello.
9585 tcp-request inspect-delay 5s
9586 tcp-request content accept if clienthello
9587
9588 # no timeout on response inspect delay by default.
9589 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009590
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009591 # SSL session ID (SSLID) may be present on a client or server hello.
9592 # Its length is coded on 1 byte at offset 43 and its value starts
9593 # at offset 44.
9594
9595 # Match and learn on request if client hello.
9596 stick on payload_lv(43,1) if clienthello
9597
9598 # Learn on response if server hello.
9599 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02009600
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009601 server s1 192.168.1.1:443
9602 server s2 192.168.1.1:443
9603
9604 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
9605 extraction.
9606
9607
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009608tcp-check connect [params*]
9609 Opens a new connection
9610 May be used in sections: defaults | frontend | listen | backend
9611 no | no | yes | yes
9612
9613 When an application lies on more than a single TCP port or when HAProxy
9614 load-balance many services in a single backend, it makes sense to probe all
9615 the services individually before considering a server as operational.
9616
9617 When there are no TCP port configured on the server line neither server port
9618 directive, then the 'tcp-check connect port <port>' must be the first step
9619 of the sequence.
9620
9621 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9622 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9623 do.
9624
9625 Parameters :
9626 They are optional and can be used to describe how HAProxy should open and
9627 use the TCP connection.
9628
9629 port if not set, check port or server port is used.
9630 It tells HAProxy where to open the connection to.
9631 <port> must be a valid TCP port source integer, from 1 to 65535.
9632
9633 send-proxy send a PROXY protocol string
9634
9635 ssl opens a ciphered connection
9636
9637 Examples:
9638 # check HTTP and HTTPs services on a server.
9639 # first open port 80 thanks to server line port directive, then
9640 # tcp-check opens port 443, ciphered and run a request on it:
9641 option tcp-check
9642 tcp-check connect
9643 tcp-check send GET\ /\ HTTP/1.0\r\n
9644 tcp-check send Host:\ haproxy.1wt.eu\r\n
9645 tcp-check send \r\n
9646 tcp-check expect rstring (2..|3..)
9647 tcp-check connect port 443 ssl
9648 tcp-check send GET\ /\ HTTP/1.0\r\n
9649 tcp-check send Host:\ haproxy.1wt.eu\r\n
9650 tcp-check send \r\n
9651 tcp-check expect rstring (2..|3..)
9652 server www 10.0.0.1 check port 80
9653
9654 # check both POP and IMAP from a single server:
9655 option tcp-check
9656 tcp-check connect port 110
9657 tcp-check expect string +OK\ POP3\ ready
9658 tcp-check connect port 143
9659 tcp-check expect string *\ OK\ IMAP4\ ready
9660 server mail 10.0.0.1 check
9661
9662 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9663
9664
9665tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009666 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009667 May be used in sections: defaults | frontend | listen | backend
9668 no | no | yes | yes
9669
9670 Arguments :
9671 <match> is a keyword indicating how to look for a specific pattern in the
9672 response. The keyword may be one of "string", "rstring" or
9673 binary.
9674 The keyword may be preceded by an exclamation mark ("!") to negate
9675 the match. Spaces are allowed between the exclamation mark and the
9676 keyword. See below for more details on the supported keywords.
9677
9678 <pattern> is the pattern to look for. It may be a string or a regular
9679 expression. If the pattern contains spaces, they must be escaped
9680 with the usual backslash ('\').
9681 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009682 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009683 two digits will represent a byte. The hexadecimal digits may be
9684 used upper or lower case.
9685
9686
9687 The available matches are intentionally similar to their http-check cousins :
9688
9689 string <string> : test the exact string matches in the response buffer.
9690 A health check response will be considered valid if the
9691 response's buffer contains this exact string. If the
9692 "string" keyword is prefixed with "!", then the response
9693 will be considered invalid if the body contains this
9694 string. This can be used to look for a mandatory pattern
9695 in a protocol response, or to detect a failure when a
9696 specific error appears in a protocol banner.
9697
9698 rstring <regex> : test a regular expression on the response buffer.
9699 A health check response will be considered valid if the
9700 response's buffer matches this expression. If the
9701 "rstring" keyword is prefixed with "!", then the response
9702 will be considered invalid if the body matches the
9703 expression.
9704
9705 binary <hexstring> : test the exact string in its hexadecimal form matches
9706 in the response buffer. A health check response will
9707 be considered valid if the response's buffer contains
9708 this exact hexadecimal string.
9709 Purpose is to match data on binary protocols.
9710
9711 It is important to note that the responses will be limited to a certain size
9712 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9713 Thus, too large responses may not contain the mandatory pattern when using
9714 "string", "rstring" or binary. If a large response is absolutely required, it
9715 is possible to change the default max size by setting the global variable.
9716 However, it is worth keeping in mind that parsing very large responses can
9717 waste some CPU cycles, especially when regular expressions are used, and that
9718 it is always better to focus the checks on smaller resources. Also, in its
9719 current state, the check will not find any string nor regex past a null
9720 character in the response. Similarly it is not possible to request matching
9721 the null character.
9722
9723 Examples :
9724 # perform a POP check
9725 option tcp-check
9726 tcp-check expect string +OK\ POP3\ ready
9727
9728 # perform an IMAP check
9729 option tcp-check
9730 tcp-check expect string *\ OK\ IMAP4\ ready
9731
9732 # look for the redis master server
9733 option tcp-check
9734 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009735 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009736 tcp-check send info\ replication\r\n
9737 tcp-check expect string role:master
9738 tcp-check send QUIT\r\n
9739 tcp-check expect string +OK
9740
9741
9742 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9743 "tcp-check send-binary", "http-check expect", tune.chksize
9744
9745
9746tcp-check send <data>
9747 Specify a string to be sent as a question during a generic health check
9748 May be used in sections: defaults | frontend | listen | backend
9749 no | no | yes | yes
9750
9751 <data> : the data to be sent as a question during a generic health check
9752 session. For now, <data> must be a string.
9753
9754 Examples :
9755 # look for the redis master server
9756 option tcp-check
9757 tcp-check send info\ replication\r\n
9758 tcp-check expect string role:master
9759
9760 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9761 "tcp-check send-binary", tune.chksize
9762
9763
Davor Ocelice9ed2812017-12-25 17:49:28 +01009764tcp-check send-binary <hexstring>
9765 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009766 tcp health check
9767 May be used in sections: defaults | frontend | listen | backend
9768 no | no | yes | yes
9769
9770 <data> : the data to be sent as a question during a generic health check
9771 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009772 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009773 response buffer. A health check response will be considered
9774 valid if the response's buffer contains this exact
9775 hexadecimal string.
9776 Purpose is to send binary data to ask on binary protocols.
9777
9778 Examples :
9779 # redis check in binary
9780 option tcp-check
9781 tcp-check send-binary 50494e470d0a # PING\r\n
9782 tcp-check expect binary 2b504F4e47 # +PONG
9783
9784
9785 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9786 "tcp-check send", tune.chksize
9787
9788
Willy Tarreaue9656522010-08-17 15:40:09 +02009789tcp-request connection <action> [{if | unless} <condition>]
9790 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009791 May be used in sections : defaults | frontend | listen | backend
9792 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009793 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009794 <action> defines the action to perform if the condition applies. See
9795 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009796
Willy Tarreaue9656522010-08-17 15:40:09 +02009797 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009798
9799 Immediately after acceptance of a new incoming connection, it is possible to
9800 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009801 or dropped or have its counters tracked. Those conditions cannot make use of
9802 any data contents because the connection has not been read from yet, and the
9803 buffers are not yet allocated. This is used to selectively and very quickly
9804 accept or drop connections from various sources with a very low overhead. If
9805 some contents need to be inspected in order to take the decision, the
9806 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009807
Willy Tarreaue9656522010-08-17 15:40:09 +02009808 The "tcp-request connection" rules are evaluated in their exact declaration
9809 order. If no rule matches or if there is no rule, the default action is to
9810 accept the incoming connection. There is no specific limit to the number of
9811 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009812
Willy Tarreaua9083d02015-05-08 15:27:59 +02009813 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02009814 - accept :
9815 accepts the connection if the condition is true (when used with "if")
9816 or false (when used with "unless"). The first such rule executed ends
9817 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009818
Willy Tarreaue9656522010-08-17 15:40:09 +02009819 - reject :
9820 rejects the connection if the condition is true (when used with "if")
9821 or false (when used with "unless"). The first such rule executed ends
9822 the rules evaluation. Rejected connections do not even become a
9823 session, which is why they are accounted separately for in the stats,
9824 as "denied connections". They are not considered for the session
9825 rate-limit and are not logged either. The reason is that these rules
9826 should only be used to filter extremely high connection rates such as
9827 the ones encountered during a massive DDoS attack. Under these extreme
9828 conditions, the simple action of logging each event would make the
9829 system collapse and would considerably lower the filtering capacity. If
9830 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02009831 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009832
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009833 - expect-proxy layer4 :
9834 configures the client-facing connection to receive a PROXY protocol
9835 header before any byte is read from the socket. This is equivalent to
9836 having the "accept-proxy" keyword on the "bind" line, except that using
9837 the TCP rule allows the PROXY protocol to be accepted only for certain
9838 IP address ranges using an ACL. This is convenient when multiple layers
9839 of load balancers are passed through by traffic coming from public
9840 hosts.
9841
Bertrand Jacquin90759682016-06-06 15:35:39 +01009842 - expect-netscaler-cip layer4 :
9843 configures the client-facing connection to receive a NetScaler Client
9844 IP insertion protocol header before any byte is read from the socket.
9845 This is equivalent to having the "accept-netscaler-cip" keyword on the
9846 "bind" line, except that using the TCP rule allows the PROXY protocol
9847 to be accepted only for certain IP address ranges using an ACL. This
9848 is convenient when multiple layers of load balancers are passed
9849 through by traffic coming from public hosts.
9850
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009851 - capture <sample> len <length> :
9852 This only applies to "tcp-request content" rules. It captures sample
9853 expression <sample> from the request buffer, and converts it to a
9854 string of at most <len> characters. The resulting string is stored into
9855 the next request "capture" slot, so it will possibly appear next to
9856 some captured HTTP headers. It will then automatically appear in the
9857 logs, and it will be possible to extract it using sample fetch rules to
9858 feed it into headers or anything. The length should be limited given
9859 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02009860 session life. Please check section 7.3 (Fetching samples) and "capture
9861 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009862
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009863 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02009864 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009865 rules do not stop evaluation and do not change default action. The
9866 number of counters that may be simultaneously tracked by the same
9867 connection is set in MAX_SESS_STKCTR at build time (reported in
9868 haproxy -vv) whichs defaults to 3, so the track-sc number is between 0
9869 and (MAX_SESS_STCKTR-1). The first "track-sc0" rule executed enables
9870 tracking of the counters of the specified table as the first set. The
9871 first "track-sc1" rule executed enables tracking of the counters of the
9872 specified table as the second set. The first "track-sc2" rule executed
9873 enables tracking of the counters of the specified table as the third
9874 set. It is a recommended practice to use the first set of counters for
9875 the per-frontend counters and the second set for the per-backend ones.
9876 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009877
Willy Tarreaue9656522010-08-17 15:40:09 +02009878 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009879 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02009880 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01009881 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009882 and used to select which table entry to update the counters.
9883 Note that "tcp-request connection" cannot use content-based
9884 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009885
Willy Tarreaue9656522010-08-17 15:40:09 +02009886 <table> is an optional table to be used instead of the default one,
9887 which is the stick-table declared in the current proxy. All
9888 the counters for the matches and updates for the key will
9889 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009890
Willy Tarreaue9656522010-08-17 15:40:09 +02009891 Once a "track-sc*" rule is executed, the key is looked up in the table
9892 and if it is not found, an entry is allocated for it. Then a pointer to
9893 that entry is kept during all the session's life, and this entry's
9894 counters are updated as often as possible, every time the session's
9895 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009896 Counters are only updated for events that happen after the tracking has
9897 been started. For example, connection counters will not be updated when
9898 tracking layer 7 information, since the connection event happens before
9899 layer7 information is extracted.
9900
Willy Tarreaue9656522010-08-17 15:40:09 +02009901 If the entry tracks concurrent connection counters, one connection is
9902 counted for as long as the entry is tracked, and the entry will not
9903 expire during that time. Tracking counters also provides a performance
9904 advantage over just checking the keys, because only one table lookup is
9905 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009906
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009907 - sc-inc-gpc0(<sc-id>):
9908 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
9909 counter designated by <sc-id>. If an error occurs, this action silently
9910 fails and the actions evaluation continues.
9911
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009912 - sc-inc-gpc1(<sc-id>):
9913 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
9914 counter designated by <sc-id>. If an error occurs, this action silently
9915 fails and the actions evaluation continues.
9916
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009917 - sc-set-gpt0(<sc-id>) <int>:
9918 This action sets the GPT0 tag according to the sticky counter designated
9919 by <sc-id> and the value of <int>. The expected result is a boolean. If
9920 an error occurs, this action silently fails and the actions evaluation
9921 continues.
9922
William Lallemand2e785f22016-05-25 01:48:42 +02009923 - set-src <expr> :
9924 Is used to set the source IP address to the value of specified
9925 expression. Useful if you want to mask source IP for privacy.
9926 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009927 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +02009928
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009929 Arguments:
9930 <expr> Is a standard HAProxy expression formed by a sample-fetch
9931 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +02009932
9933 Example:
William Lallemand2e785f22016-05-25 01:48:42 +02009934 tcp-request connection set-src src,ipmask(24)
9935
Willy Tarreau0c630532016-10-21 17:52:58 +02009936 When possible, set-src preserves the original source port as long as the
9937 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +02009938
William Lallemand44be6402016-05-25 01:51:35 +02009939 - set-src-port <expr> :
9940 Is used to set the source port address to the value of specified
9941 expression.
9942
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009943 Arguments:
9944 <expr> Is a standard HAProxy expression formed by a sample-fetch
9945 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +02009946
9947 Example:
William Lallemand44be6402016-05-25 01:51:35 +02009948 tcp-request connection set-src-port int(4000)
9949
Willy Tarreau0c630532016-10-21 17:52:58 +02009950 When possible, set-src-port preserves the original source address as long
9951 as the address family supports a port, otherwise it forces the source
9952 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02009953
William Lallemand13e9b0c2016-05-25 02:34:07 +02009954 - set-dst <expr> :
9955 Is used to set the destination IP address to the value of specified
9956 expression. Useful if you want to mask IP for privacy in log.
9957 If you want to provide an IP from a HTTP header use "http-request
9958 set-dst". If you want to connect to the new address/port, use
9959 '0.0.0.0:0' as a server address in the backend.
9960
9961 <expr> Is a standard HAProxy expression formed by a sample-fetch
9962 followed by some converters.
9963
9964 Example:
9965
9966 tcp-request connection set-dst dst,ipmask(24)
9967 tcp-request connection set-dst ipv4(10.0.0.1)
9968
Willy Tarreau0c630532016-10-21 17:52:58 +02009969 When possible, set-dst preserves the original destination port as long as
9970 the address family allows it, otherwise the destination port is set to 0.
9971
William Lallemand13e9b0c2016-05-25 02:34:07 +02009972 - set-dst-port <expr> :
9973 Is used to set the destination port address to the value of specified
9974 expression. If you want to connect to the new address/port, use
9975 '0.0.0.0:0' as a server address in the backend.
9976
9977
9978 <expr> Is a standard HAProxy expression formed by a sample-fetch
9979 followed by some converters.
9980
9981 Example:
9982
9983 tcp-request connection set-dst-port int(4000)
9984
Willy Tarreau0c630532016-10-21 17:52:58 +02009985 When possible, set-dst-port preserves the original destination address as
9986 long as the address family supports a port, otherwise it forces the
9987 destination address to IPv4 "0.0.0.0" before rewriting the port.
9988
Willy Tarreau2d392c22015-08-24 01:43:45 +02009989 - "silent-drop" :
9990 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009991 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009992 to prevent the client from being notified. The effect it then that the
9993 client still sees an established connection while there's none on
9994 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9995 except that it doesn't use any local resource at all on the machine
9996 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009997 slow down stronger attackers. It is important to understand the impact
9998 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009999 client and HAProxy (firewalls, proxies, load balancers) will also keep
10000 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010001 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010002 TCP_REPAIR socket option is used to block the emission of a TCP
10003 reset. On other systems, the socket's TTL is reduced to 1 so that the
10004 TCP reset doesn't pass the first router, though it's still delivered to
10005 local networks. Do not use it unless you fully understand how it works.
10006
Willy Tarreaue9656522010-08-17 15:40:09 +020010007 Note that the "if/unless" condition is optional. If no condition is set on
10008 the action, it is simply performed unconditionally. That can be useful for
10009 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010010
Willy Tarreaue9656522010-08-17 15:40:09 +020010011 Example: accept all connections from white-listed hosts, reject too fast
10012 connection without counting them, and track accepted connections.
10013 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010014
Willy Tarreaue9656522010-08-17 15:40:09 +020010015 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010016 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010017 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010018
Willy Tarreaue9656522010-08-17 15:40:09 +020010019 Example: accept all connections from white-listed hosts, count all other
10020 connections and reject too fast ones. This results in abusive ones
10021 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010022
Willy Tarreaue9656522010-08-17 15:40:09 +020010023 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010024 tcp-request connection track-sc0 src
10025 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010026
Willy Tarreau4f0d9192013-06-11 20:40:55 +020010027 Example: enable the PROXY protocol for traffic coming from all known proxies.
10028
10029 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10030
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010031 See section 7 about ACL usage.
10032
Willy Tarreau4f614292016-10-21 17:49:36 +020010033 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010034
10035
Willy Tarreaue9656522010-08-17 15:40:09 +020010036tcp-request content <action> [{if | unless} <condition>]
10037 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +020010038 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010039 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +020010040 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010041 <action> defines the action to perform if the condition applies. See
10042 below.
Willy Tarreau62644772008-07-16 18:36:06 +020010043
Willy Tarreaue9656522010-08-17 15:40:09 +020010044 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +020010045
Davor Ocelice9ed2812017-12-25 17:49:28 +010010046 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +020010047 called "TCP content inspection". During this stage, ACL-based rules are
10048 evaluated every time the request contents are updated, until either an
10049 "accept" or a "reject" rule matches, or the TCP request inspection delay
10050 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +020010051
Willy Tarreaue9656522010-08-17 15:40:09 +020010052 The first difference between these rules and "tcp-request connection" rules
10053 is that "tcp-request content" rules can make use of contents to take a
10054 decision. Most often, these decisions will consider a protocol recognition or
10055 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +010010056 both frontends and backends. In case of HTTP keep-alive with the client, all
10057 tcp-request content rules are evaluated again, so haproxy keeps a record of
10058 what sticky counters were assigned by a "tcp-request connection" versus a
10059 "tcp-request content" rule, and flushes all the content-related ones after
10060 processing an HTTP request, so that they may be evaluated again by the rules
10061 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010062 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +010010063 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +020010064
Willy Tarreaue9656522010-08-17 15:40:09 +020010065 Content-based rules are evaluated in their exact declaration order. If no
10066 rule matches or if there is no rule, the default action is to accept the
10067 contents. There is no specific limit to the number of rules which may be
10068 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +020010069
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010070 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010071 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +010010072 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +020010073 - reject : the request is rejected and the connection is closed
10074 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -040010075 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010076 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010077 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010078 - sc-inc-gpc1(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +020010079 - sc-set-gpt0(<sc-id>) <int>
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010080 - set-dst <expr>
10081 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010082 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010083 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +020010084 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +010010085 - send-spoe-group <engine-name> <group-name>
Willy Tarreau62644772008-07-16 18:36:06 +020010086
Willy Tarreaue9656522010-08-17 15:40:09 +020010087 They have the same meaning as their counter-parts in "tcp-request connection"
10088 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +010010089 For "do-resolve" action, please check the "http-request do-resolve"
10090 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +020010091
Willy Tarreauf3338342014-01-28 21:40:28 +010010092 While there is nothing mandatory about it, it is recommended to use the
10093 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
10094 content" rules in the frontend, and track-sc2 for "tcp-request content"
10095 rules in the backend, because that makes the configuration more readable
10096 and easier to troubleshoot, but this is just a guideline and all counters
10097 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +020010098
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010099 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +020010100 the action, it is simply performed unconditionally. That can be useful for
10101 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +020010102
Willy Tarreaue9656522010-08-17 15:40:09 +020010103 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +020010104 rules, since HTTP-specific ACL matches are able to preliminarily parse the
10105 contents of a buffer before extracting the required data. If the buffered
10106 contents do not parse as a valid HTTP message, then the ACL does not match.
10107 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +010010108 processing, so there is no risk of parsing something differently. In an HTTP
10109 backend connected to from an HTTP frontend, it is guaranteed that HTTP
10110 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +020010111
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010112 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010113 are present when the rule is processed. The rule processing engine is able to
10114 wait until the inspect delay expires when the data to be tracked is not yet
10115 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010116
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010117 The "set-dst" and "set-dst-port" are used to set respectively the destination
10118 IP and port. More information on how to use it at "http-request set-dst".
10119
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010120 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +020010121 declared inline. For "tcp-request session" rules, only session-level
10122 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010123
Daniel Schneller0b547052016-03-21 20:46:57 +010010124 <var-name> The name of the variable starts with an indication about
10125 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010126 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010127 "sess" : the variable is shared with the whole session
10128 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010129 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010130 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010131 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010132 "res" : the variable is shared only during response
10133 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010134 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010135 The name may only contain characters 'a-z', 'A-Z', '0-9',
10136 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010137
10138 <expr> Is a standard HAProxy expression formed by a sample-fetch
10139 followed by some converters.
10140
Christopher Faulet85d79c92016-11-09 16:54:56 +010010141 The "unset-var" is used to unset a variable. See above for details about
10142 <var-name>.
10143
Patrick Hemmer268a7072018-05-11 12:52:31 -040010144 The "set-priority-class" is used to set the queue priority class of the
10145 current request. The value must be a sample expression which converts to an
10146 integer in the range -2047..2047. Results outside this range will be
10147 truncated. The priority class determines the order in which queued requests
10148 are processed. Lower values have higher priority.
10149
10150 The "set-priority-offset" is used to set the queue priority timestamp offset
10151 of the current request. The value must be a sample expression which converts
10152 to an integer in the range -524287..524287. Results outside this range will be
10153 truncated. When a request is queued, it is ordered first by the priority
10154 class, then by the current timestamp adjusted by the given offset in
10155 milliseconds. Lower values have higher priority.
10156 Note that the resulting timestamp is is only tracked with enough precision for
10157 524,287ms (8m44s287ms). If the request is queued long enough to where the
10158 adjusted timestamp exceeds this value, it will be misidentified as highest
10159 priority. Thus it is important to set "timeout queue" to a value, where when
10160 combined with the offset, does not exceed this limit.
10161
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010162 The "send-spoe-group" is used to trigger sending of a group of SPOE
10163 messages. To do so, the SPOE engine used to send messages must be defined, as
10164 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10165 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10166 the SPOE agent name must be used.
10167
10168 <engine-name> The SPOE engine name.
10169
10170 <group-name> The SPOE group name as specified in the engine configuration.
10171
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010172 Example:
10173
10174 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +010010175 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010176
Willy Tarreau62644772008-07-16 18:36:06 +020010177 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +020010178 # Accept HTTP requests containing a Host header saying "example.com"
10179 # and reject everything else.
10180 acl is_host_com hdr(Host) -i example.com
10181 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +020010182 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +020010183 tcp-request content reject
10184
10185 Example:
Willy Tarreau62644772008-07-16 18:36:06 +020010186 # reject SMTP connection if client speaks first
10187 tcp-request inspect-delay 30s
10188 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010189 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +020010190
10191 # Forward HTTPS connection only if client speaks
10192 tcp-request inspect-delay 30s
10193 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010194 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +020010195 tcp-request content reject
10196
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010197 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010198 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010199 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010200 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010201 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
10202 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010203
10204 Example:
10205 # track request counts per "base" (concatenation of Host+URL)
10206 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010207 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010208
Willy Tarreaue9656522010-08-17 15:40:09 +020010209 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010210 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +020010211
10212 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +010010213 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +020010214 # protecting all our sites
10215 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010216 tcp-request connection track-sc0 src
10217 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +020010218 ...
10219 use_backend http_dynamic if { path_end .php }
10220
10221 backend http_dynamic
10222 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010223 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +020010224 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010225 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010226 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010227 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +020010228 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +020010229
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010230 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +020010231
Jarno Huuskonen95b012b2017-04-06 13:59:14 +030010232 See also : "tcp-request connection", "tcp-request session",
10233 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +020010234
10235
10236tcp-request inspect-delay <timeout>
10237 Set the maximum allowed time to wait for data during content inspection
10238 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010239 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +020010240 Arguments :
10241 <timeout> is the timeout value specified in milliseconds by default, but
10242 can be in any other unit if the number is suffixed by the unit,
10243 as explained at the top of this document.
10244
10245 People using haproxy primarily as a TCP relay are often worried about the
10246 risk of passing any type of protocol to a server without any analysis. In
10247 order to be able to analyze the request contents, we must first withhold
10248 the data then analyze them. This statement simply enables withholding of
10249 data for at most the specified amount of time.
10250
Willy Tarreaufb356202010-08-03 14:02:05 +020010251 TCP content inspection applies very early when a connection reaches a
10252 frontend, then very early when the connection is forwarded to a backend. This
10253 means that a connection may experience a first delay in the frontend and a
10254 second delay in the backend if both have tcp-request rules.
10255
Willy Tarreau62644772008-07-16 18:36:06 +020010256 Note that when performing content inspection, haproxy will evaluate the whole
10257 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010258 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +020010259 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +010010260 contents are definitive. If no delay is set, haproxy will not wait at all
10261 and will immediately apply a verdict based on the available information.
10262 Obviously this is unlikely to be very useful and might even be racy, so such
10263 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020010264
10265 As soon as a rule matches, the request is released and continues as usual. If
10266 the timeout is reached and no rule matches, the default policy will be to let
10267 it pass through unaffected.
10268
10269 For most protocols, it is enough to set it to a few seconds, as most clients
10270 send the full request immediately upon connection. Add 3 or more seconds to
10271 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +010010272 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +010010273 before the server (e.g. SMTP), or to wait for a client to talk before passing
10274 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +020010275 least the inspection delay, otherwise it will expire first. If the client
10276 closes the connection or if the buffer is full, the delay immediately expires
10277 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +020010278
Willy Tarreau55165fe2009-05-10 12:02:55 +020010279 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +020010280 "timeout client".
10281
10282
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010283tcp-response content <action> [{if | unless} <condition>]
10284 Perform an action on a session response depending on a layer 4-7 condition
10285 May be used in sections : defaults | frontend | listen | backend
10286 no | no | yes | yes
10287 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010288 <action> defines the action to perform if the condition applies. See
10289 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010290
10291 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
10292
Davor Ocelice9ed2812017-12-25 17:49:28 +010010293 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010294 called "TCP content inspection". During this stage, ACL-based rules are
10295 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010296 "accept", "close" or a "reject" rule matches, or a TCP response inspection
10297 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010298
10299 Most often, these decisions will consider a protocol recognition or validity.
10300
10301 Content-based rules are evaluated in their exact declaration order. If no
10302 rule matches or if there is no rule, the default action is to accept the
10303 contents. There is no specific limit to the number of rules which may be
10304 inserted.
10305
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010306 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010307 - accept :
10308 accepts the response if the condition is true (when used with "if")
10309 or false (when used with "unless"). The first such rule executed ends
10310 the rules evaluation.
10311
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010312 - close :
10313 immediately closes the connection with the server if the condition is
10314 true (when used with "if"), or false (when used with "unless"). The
10315 first such rule executed ends the rules evaluation. The main purpose of
10316 this action is to force a connection to be finished between a client
10317 and a server after an exchange when the application protocol expects
10318 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010319 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010320 protocols.
10321
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010322 - reject :
10323 rejects the response if the condition is true (when used with "if")
10324 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010325 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010326
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010327 - set-var(<var-name>) <expr>
10328 Sets a variable.
10329
Christopher Faulet85d79c92016-11-09 16:54:56 +010010330 - unset-var(<var-name>)
10331 Unsets a variable.
10332
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010333 - sc-inc-gpc0(<sc-id>):
10334 This action increments the GPC0 counter according to the sticky
10335 counter designated by <sc-id>. If an error occurs, this action fails
10336 silently and the actions evaluation continues.
10337
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010338 - sc-inc-gpc1(<sc-id>):
10339 This action increments the GPC1 counter according to the sticky
10340 counter designated by <sc-id>. If an error occurs, this action fails
10341 silently and the actions evaluation continues.
10342
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010343 - sc-set-gpt0(<sc-id>) <int> :
10344 This action sets the GPT0 tag according to the sticky counter designated
10345 by <sc-id> and the value of <int>. The expected result is a boolean. If
10346 an error occurs, this action silently fails and the actions evaluation
10347 continues.
10348
Willy Tarreau2d392c22015-08-24 01:43:45 +020010349 - "silent-drop" :
10350 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010351 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010352 to prevent the client from being notified. The effect it then that the
10353 client still sees an established connection while there's none on
10354 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10355 except that it doesn't use any local resource at all on the machine
10356 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010357 slow down stronger attackers. It is important to understand the impact
10358 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010359 client and HAProxy (firewalls, proxies, load balancers) will also keep
10360 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010361 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010362 TCP_REPAIR socket option is used to block the emission of a TCP
10363 reset. On other systems, the socket's TTL is reduced to 1 so that the
10364 TCP reset doesn't pass the first router, though it's still delivered to
10365 local networks. Do not use it unless you fully understand how it works.
10366
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010367 - send-spoe-group <engine-name> <group-name>
10368 Send a group of SPOE messages.
10369
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010370 Note that the "if/unless" condition is optional. If no condition is set on
10371 the action, it is simply performed unconditionally. That can be useful for
10372 for changing the default action to a reject.
10373
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010374 It is perfectly possible to match layer 7 contents with "tcp-response
10375 content" rules, but then it is important to ensure that a full response has
10376 been buffered, otherwise no contents will match. In order to achieve this,
10377 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010378 period.
10379
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010380 The "set-var" is used to set the content of a variable. The variable is
10381 declared inline.
10382
Daniel Schneller0b547052016-03-21 20:46:57 +010010383 <var-name> The name of the variable starts with an indication about
10384 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010385 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010386 "sess" : the variable is shared with the whole session
10387 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010388 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010389 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010390 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010391 "res" : the variable is shared only during response
10392 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010393 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010394 The name may only contain characters 'a-z', 'A-Z', '0-9',
10395 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010396
10397 <expr> Is a standard HAProxy expression formed by a sample-fetch
10398 followed by some converters.
10399
10400 Example:
10401
10402 tcp-request content set-var(sess.my_var) src
10403
Christopher Faulet85d79c92016-11-09 16:54:56 +010010404 The "unset-var" is used to unset a variable. See above for details about
10405 <var-name>.
10406
10407 Example:
10408
10409 tcp-request content unset-var(sess.my_var)
10410
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010411 The "send-spoe-group" is used to trigger sending of a group of SPOE
10412 messages. To do so, the SPOE engine used to send messages must be defined, as
10413 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10414 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10415 the SPOE agent name must be used.
10416
10417 <engine-name> The SPOE engine name.
10418
10419 <group-name> The SPOE group name as specified in the engine configuration.
10420
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010421 See section 7 about ACL usage.
10422
10423 See also : "tcp-request content", "tcp-response inspect-delay"
10424
10425
Willy Tarreau4f614292016-10-21 17:49:36 +020010426tcp-request session <action> [{if | unless} <condition>]
10427 Perform an action on a validated session depending on a layer 5 condition
10428 May be used in sections : defaults | frontend | listen | backend
10429 no | yes | yes | no
10430 Arguments :
10431 <action> defines the action to perform if the condition applies. See
10432 below.
10433
10434 <condition> is a standard layer5-only ACL-based condition (see section 7).
10435
Davor Ocelice9ed2812017-12-25 17:49:28 +010010436 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020010437 it is possible to evaluate some conditions to decide whether this session
10438 must be accepted or dropped or have its counters tracked. Those conditions
10439 cannot make use of any data contents because no buffers are allocated yet and
10440 the processing cannot wait at this stage. The main use case it to copy some
10441 early information into variables (since variables are accessible in the
10442 session), or to keep track of some information collected after the handshake,
10443 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010010444 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020010445 extracted information can thus be copied to a variable or tracked using
10446 "track-sc" rules. Of course it is also possible to decide to accept/reject as
10447 with other rulesets. Most operations performed here could also be performed
10448 in "tcp-request content" rules, except that in HTTP these rules are evaluated
10449 for each new request, and that might not always be acceptable. For example a
10450 rule might increment a counter on each evaluation. It would also be possible
10451 that a country is resolved by geolocation from the source IP address,
10452 assigned to a session-wide variable, then the source address rewritten from
10453 an HTTP header for all requests. If some contents need to be inspected in
10454 order to take the decision, the "tcp-request content" statements must be used
10455 instead.
10456
10457 The "tcp-request session" rules are evaluated in their exact declaration
10458 order. If no rule matches or if there is no rule, the default action is to
10459 accept the incoming session. There is no specific limit to the number of
10460 rules which may be inserted.
10461
10462 Several types of actions are supported :
10463 - accept : the request is accepted
10464 - reject : the request is rejected and the connection is closed
10465 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
10466 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010467 - sc-inc-gpc1(<sc-id>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010468 - sc-set-gpt0(<sc-id>) <int>
10469 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010470 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010471 - silent-drop
10472
10473 These actions have the same meaning as their respective counter-parts in
10474 "tcp-request connection" and "tcp-request content", so please refer to these
10475 sections for a complete description.
10476
10477 Note that the "if/unless" condition is optional. If no condition is set on
10478 the action, it is simply performed unconditionally. That can be useful for
10479 "track-sc*" actions as well as for changing the default action to a reject.
10480
10481 Example: track the original source address by default, or the one advertised
10482 in the PROXY protocol header for connection coming from the local
10483 proxies. The first connection-level rule enables receipt of the
10484 PROXY protocol for these ones, the second rule tracks whatever
10485 address we decide to keep after optional decoding.
10486
10487 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10488 tcp-request session track-sc0 src
10489
10490 Example: accept all sessions from white-listed hosts, reject too fast
10491 sessions without counting them, and track accepted sessions.
10492 This results in session rate being capped from abusive sources.
10493
10494 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10495 tcp-request session reject if { src_sess_rate gt 10 }
10496 tcp-request session track-sc0 src
10497
10498 Example: accept all sessions from white-listed hosts, count all other
10499 sessions and reject too fast ones. This results in abusive ones
10500 being blocked as long as they don't slow down.
10501
10502 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10503 tcp-request session track-sc0 src
10504 tcp-request session reject if { sc0_sess_rate gt 10 }
10505
10506 See section 7 about ACL usage.
10507
10508 See also : "tcp-request connection", "tcp-request content", "stick-table"
10509
10510
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010511tcp-response inspect-delay <timeout>
10512 Set the maximum allowed time to wait for a response during content inspection
10513 May be used in sections : defaults | frontend | listen | backend
10514 no | no | yes | yes
10515 Arguments :
10516 <timeout> is the timeout value specified in milliseconds by default, but
10517 can be in any other unit if the number is suffixed by the unit,
10518 as explained at the top of this document.
10519
10520 See also : "tcp-response content", "tcp-request inspect-delay".
10521
10522
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010523timeout check <timeout>
10524 Set additional check timeout, but only after a connection has been already
10525 established.
10526
10527 May be used in sections: defaults | frontend | listen | backend
10528 yes | no | yes | yes
10529 Arguments:
10530 <timeout> is the timeout value specified in milliseconds by default, but
10531 can be in any other unit if the number is suffixed by the unit,
10532 as explained at the top of this document.
10533
10534 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
10535 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010010536 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010537 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010010538 (Please also note that there is no valid reason to have such long connect
10539 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
10540 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010541
10542 If "timeout check" is not set haproxy uses "inter" for complete check
10543 timeout (connect + read) exactly like all <1.3.15 version.
10544
10545 In most cases check request is much simpler and faster to handle than normal
10546 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010010547 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010548
10549 This parameter is specific to backends, but can be specified once for all in
10550 "defaults" sections. This is in fact one of the easiest solutions not to
10551 forget about it.
10552
Willy Tarreau41a340d2008-01-22 12:25:31 +010010553 See also: "timeout connect", "timeout queue", "timeout server",
10554 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010555
10556
Willy Tarreau0ba27502007-12-24 16:55:16 +010010557timeout client <timeout>
10558timeout clitimeout <timeout> (deprecated)
10559 Set the maximum inactivity time on the client side.
10560 May be used in sections : defaults | frontend | listen | backend
10561 yes | yes | yes | no
10562 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010563 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010564 can be in any other unit if the number is suffixed by the unit,
10565 as explained at the top of this document.
10566
10567 The inactivity timeout applies when the client is expected to acknowledge or
10568 send data. In HTTP mode, this timeout is particularly important to consider
10569 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010570 response while it is reading data sent by the server. That said, for the
10571 first phase, it is preferable to set the "timeout http-request" to better
10572 protect HAProxy from Slowloris like attacks. The value is specified in
10573 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010010574 suffixed by the unit, as specified at the top of this document. In TCP mode
10575 (and to a lesser extent, in HTTP mode), it is highly recommended that the
10576 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010577 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010010578 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010010579 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
10580 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020010581 which overrides "timeout client" and "timeout server" for tunnels, as well as
10582 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010583
10584 This parameter is specific to frontends, but can be specified once for all in
10585 "defaults" sections. This is in fact one of the easiest solutions not to
10586 forget about it. An unspecified timeout results in an infinite timeout, which
10587 is not recommended. Such a usage is accepted and works but reports a warning
10588 during startup because it may results in accumulation of expired sessions in
10589 the system if the system's timeouts are not configured either.
10590
Willy Tarreau95c4e142017-11-26 12:18:55 +010010591 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010010592
Willy Tarreau0ba27502007-12-24 16:55:16 +010010593 This parameter replaces the old, deprecated "clitimeout". It is recommended
10594 to use it to write new configurations. The form "timeout clitimeout" is
10595 provided only by backwards compatibility but its use is strongly discouraged.
10596
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010597 See also : "clitimeout", "timeout server", "timeout tunnel",
10598 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010599
10600
Willy Tarreau05cdd962014-05-10 14:30:07 +020010601timeout client-fin <timeout>
10602 Set the inactivity timeout on the client side for half-closed connections.
10603 May be used in sections : defaults | frontend | listen | backend
10604 yes | yes | yes | no
10605 Arguments :
10606 <timeout> is the timeout value specified in milliseconds by default, but
10607 can be in any other unit if the number is suffixed by the unit,
10608 as explained at the top of this document.
10609
10610 The inactivity timeout applies when the client is expected to acknowledge or
10611 send data while one direction is already shut down. This timeout is different
10612 from "timeout client" in that it only applies to connections which are closed
10613 in one direction. This is particularly useful to avoid keeping connections in
10614 FIN_WAIT state for too long when clients do not disconnect cleanly. This
10615 problem is particularly common long connections such as RDP or WebSocket.
10616 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010010617 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
10618 frame was sent, often indicating an expectation that the connection quickly
10619 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020010620
10621 This parameter is specific to frontends, but can be specified once for all in
10622 "defaults" sections. By default it is not set, so half-closed connections
10623 will use the other timeouts (timeout.client or timeout.tunnel).
10624
10625 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
10626
10627
Willy Tarreau0ba27502007-12-24 16:55:16 +010010628timeout connect <timeout>
10629timeout contimeout <timeout> (deprecated)
10630 Set the maximum time to wait for a connection attempt to a server to succeed.
10631 May be used in sections : defaults | frontend | listen | backend
10632 yes | no | yes | yes
10633 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010634 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010635 can be in any other unit if the number is suffixed by the unit,
10636 as explained at the top of this document.
10637
10638 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010639 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010010640 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010010641 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010642 connect timeout also presets both queue and tarpit timeouts to the same value
10643 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010644
10645 This parameter is specific to backends, but can be specified once for all in
10646 "defaults" sections. This is in fact one of the easiest solutions not to
10647 forget about it. An unspecified timeout results in an infinite timeout, which
10648 is not recommended. Such a usage is accepted and works but reports a warning
10649 during startup because it may results in accumulation of failed sessions in
10650 the system if the system's timeouts are not configured either.
10651
10652 This parameter replaces the old, deprecated "contimeout". It is recommended
10653 to use it to write new configurations. The form "timeout contimeout" is
10654 provided only by backwards compatibility but its use is strongly discouraged.
10655
Willy Tarreau41a340d2008-01-22 12:25:31 +010010656 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
10657 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010658
10659
Willy Tarreaub16a5742010-01-10 14:46:16 +010010660timeout http-keep-alive <timeout>
10661 Set the maximum allowed time to wait for a new HTTP request to appear
10662 May be used in sections : defaults | frontend | listen | backend
10663 yes | yes | yes | yes
10664 Arguments :
10665 <timeout> is the timeout value specified in milliseconds by default, but
10666 can be in any other unit if the number is suffixed by the unit,
10667 as explained at the top of this document.
10668
10669 By default, the time to wait for a new request in case of keep-alive is set
10670 by "timeout http-request". However this is not always convenient because some
10671 people want very short keep-alive timeouts in order to release connections
10672 faster, and others prefer to have larger ones but still have short timeouts
10673 once the request has started to present itself.
10674
10675 The "http-keep-alive" timeout covers these needs. It will define how long to
10676 wait for a new HTTP request to start coming after a response was sent. Once
10677 the first byte of request has been seen, the "http-request" timeout is used
10678 to wait for the complete request to come. Note that empty lines prior to a
10679 new request do not refresh the timeout and are not counted as a new request.
10680
10681 There is also another difference between the two timeouts : when a connection
10682 expires during timeout http-keep-alive, no error is returned, the connection
10683 just closes. If the connection expires in "http-request" while waiting for a
10684 connection to complete, a HTTP 408 error is returned.
10685
10686 In general it is optimal to set this value to a few tens to hundreds of
10687 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010688 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010689 1 millisecond) it will probably only accept pipelined requests but not the
10690 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010691 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010692
10693 If this parameter is not set, the "http-request" timeout applies, and if both
10694 are not set, "timeout client" still applies at the lower level. It should be
10695 set in the frontend to take effect, unless the frontend is in TCP mode, in
10696 which case the HTTP backend's timeout will be used.
10697
Willy Tarreau95c4e142017-11-26 12:18:55 +010010698 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10699 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010700 (where we only have one connection per client and a connection setup).
10701
Willy Tarreaub16a5742010-01-10 14:46:16 +010010702 See also : "timeout http-request", "timeout client".
10703
10704
Willy Tarreau036fae02008-01-06 13:24:40 +010010705timeout http-request <timeout>
10706 Set the maximum allowed time to wait for a complete HTTP request
10707 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010708 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010709 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010710 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010711 can be in any other unit if the number is suffixed by the unit,
10712 as explained at the top of this document.
10713
10714 In order to offer DoS protection, it may be required to lower the maximum
10715 accepted time to receive a complete HTTP request without affecting the client
10716 timeout. This helps protecting against established connections on which
10717 nothing is sent. The client timeout cannot offer a good protection against
10718 this abuse because it is an inactivity timeout, which means that if the
10719 attacker sends one character every now and then, the timeout will not
10720 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010721 types, the request will be aborted if it does not complete in time. When the
10722 timeout expires, an HTTP 408 response is sent to the client to inform it
10723 about the problem, and the connection is closed. The logs will report
10724 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010725 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010726 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10727 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010728
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010729 By default, this timeout only applies to the header part of the request,
10730 and not to any data. As soon as the empty line is received, this timeout is
10731 not used anymore. When combined with "option http-buffer-request", this
10732 timeout also applies to the body of the request..
10733 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010734 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010735
10736 Generally it is enough to set it to a few seconds, as most clients send the
10737 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010738 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010739 generally work on local networks as long as there are no packet losses. This
10740 will prevent people from sending bare HTTP requests using telnet.
10741
10742 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010743 chunk of the incoming request. It should be set in the frontend to take
10744 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10745 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010746
Willy Tarreau0f228a02015-05-01 15:37:53 +020010747 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010748 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010749
Willy Tarreau844e3c52008-01-11 16:28:18 +010010750
10751timeout queue <timeout>
10752 Set the maximum time to wait in the queue for a connection slot to be free
10753 May be used in sections : defaults | frontend | listen | backend
10754 yes | no | yes | yes
10755 Arguments :
10756 <timeout> is the timeout value specified in milliseconds by default, but
10757 can be in any other unit if the number is suffixed by the unit,
10758 as explained at the top of this document.
10759
10760 When a server's maxconn is reached, connections are left pending in a queue
10761 which may be server-specific or global to the backend. In order not to wait
10762 indefinitely, a timeout is applied to requests pending in the queue. If the
10763 timeout is reached, it is considered that the request will almost never be
10764 served, so it is dropped and a 503 error is returned to the client.
10765
10766 The "timeout queue" statement allows to fix the maximum time for a request to
10767 be left pending in a queue. If unspecified, the same value as the backend's
10768 connection timeout ("timeout connect") is used, for backwards compatibility
10769 with older versions with no "timeout queue" parameter.
10770
10771 See also : "timeout connect", "contimeout".
10772
10773
10774timeout server <timeout>
10775timeout srvtimeout <timeout> (deprecated)
10776 Set the maximum inactivity time on the server side.
10777 May be used in sections : defaults | frontend | listen | backend
10778 yes | no | yes | yes
10779 Arguments :
10780 <timeout> is the timeout value specified in milliseconds by default, but
10781 can be in any other unit if the number is suffixed by the unit,
10782 as explained at the top of this document.
10783
10784 The inactivity timeout applies when the server is expected to acknowledge or
10785 send data. In HTTP mode, this timeout is particularly important to consider
10786 during the first phase of the server's response, when it has to send the
10787 headers, as it directly represents the server's processing time for the
10788 request. To find out what value to put there, it's often good to start with
10789 what would be considered as unacceptable response times, then check the logs
10790 to observe the response time distribution, and adjust the value accordingly.
10791
10792 The value is specified in milliseconds by default, but can be in any other
10793 unit if the number is suffixed by the unit, as specified at the top of this
10794 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
10795 recommended that the client timeout remains equal to the server timeout in
10796 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010797 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010010798 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010010799 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
10800 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020010801 "timeout tunnel", which overrides "timeout client" and "timeout server" for
10802 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010803
10804 This parameter is specific to backends, but can be specified once for all in
10805 "defaults" sections. This is in fact one of the easiest solutions not to
10806 forget about it. An unspecified timeout results in an infinite timeout, which
10807 is not recommended. Such a usage is accepted and works but reports a warning
10808 during startup because it may results in accumulation of expired sessions in
10809 the system if the system's timeouts are not configured either.
10810
10811 This parameter replaces the old, deprecated "srvtimeout". It is recommended
10812 to use it to write new configurations. The form "timeout srvtimeout" is
10813 provided only by backwards compatibility but its use is strongly discouraged.
10814
Willy Tarreauce887fd2012-05-12 12:50:00 +020010815 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010816
Willy Tarreau05cdd962014-05-10 14:30:07 +020010817
10818timeout server-fin <timeout>
10819 Set the inactivity timeout on the server side for half-closed connections.
10820 May be used in sections : defaults | frontend | listen | backend
10821 yes | no | yes | yes
10822 Arguments :
10823 <timeout> is the timeout value specified in milliseconds by default, but
10824 can be in any other unit if the number is suffixed by the unit,
10825 as explained at the top of this document.
10826
10827 The inactivity timeout applies when the server is expected to acknowledge or
10828 send data while one direction is already shut down. This timeout is different
10829 from "timeout server" in that it only applies to connections which are closed
10830 in one direction. This is particularly useful to avoid keeping connections in
10831 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
10832 This problem is particularly common long connections such as RDP or WebSocket.
10833 Note that this timeout can override "timeout tunnel" when a connection shuts
10834 down in one direction. This setting was provided for completeness, but in most
10835 situations, it should not be needed.
10836
10837 This parameter is specific to backends, but can be specified once for all in
10838 "defaults" sections. By default it is not set, so half-closed connections
10839 will use the other timeouts (timeout.server or timeout.tunnel).
10840
10841 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
10842
Willy Tarreau844e3c52008-01-11 16:28:18 +010010843
10844timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010010845 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010010846 May be used in sections : defaults | frontend | listen | backend
10847 yes | yes | yes | yes
10848 Arguments :
10849 <timeout> is the tarpit duration specified in milliseconds by default, but
10850 can be in any other unit if the number is suffixed by the unit,
10851 as explained at the top of this document.
10852
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010853 When a connection is tarpitted using "http-request tarpit" or
10854 "reqtarpit", it is maintained open with no activity for a certain
10855 amount of time, then closed. "timeout tarpit" defines how long it will
10856 be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010857
10858 The value is specified in milliseconds by default, but can be in any other
10859 unit if the number is suffixed by the unit, as specified at the top of this
10860 document. If unspecified, the same value as the backend's connection timeout
10861 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010010862 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010863
10864 See also : "timeout connect", "contimeout".
10865
10866
Willy Tarreauce887fd2012-05-12 12:50:00 +020010867timeout tunnel <timeout>
10868 Set the maximum inactivity time on the client and server side for tunnels.
10869 May be used in sections : defaults | frontend | listen | backend
10870 yes | no | yes | yes
10871 Arguments :
10872 <timeout> is the timeout value specified in milliseconds by default, but
10873 can be in any other unit if the number is suffixed by the unit,
10874 as explained at the top of this document.
10875
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010876 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020010877 between a client and a server, and the connection remains inactive in both
10878 directions. This timeout supersedes both the client and server timeouts once
10879 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010010880 analyzer remains attached to either connection (e.g. tcp content rules are
10881 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020010882 when switching to the WebSocket protocol, or forwarding a CONNECT request
10883 to a proxy), or after the first response when no keepalive/close option is
10884 specified.
10885
Willy Tarreau05cdd962014-05-10 14:30:07 +020010886 Since this timeout is usually used in conjunction with long-lived connections,
10887 it usually is a good idea to also set "timeout client-fin" to handle the
10888 situation where a client suddenly disappears from the net and does not
10889 acknowledge a close, or sends a shutdown and does not acknowledge pending
10890 data anymore. This can happen in lossy networks where firewalls are present,
10891 and is detected by the presence of large amounts of sessions in a FIN_WAIT
10892 state.
10893
Willy Tarreauce887fd2012-05-12 12:50:00 +020010894 The value is specified in milliseconds by default, but can be in any other
10895 unit if the number is suffixed by the unit, as specified at the top of this
10896 document. Whatever the expected normal idle time, it is a good practice to
10897 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010010898 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020010899
10900 This parameter is specific to backends, but can be specified once for all in
10901 "defaults" sections. This is in fact one of the easiest solutions not to
10902 forget about it.
10903
10904 Example :
10905 defaults http
10906 option http-server-close
10907 timeout connect 5s
10908 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020010909 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020010910 timeout server 30s
10911 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
10912
Willy Tarreau05cdd962014-05-10 14:30:07 +020010913 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020010914
10915
Willy Tarreau844e3c52008-01-11 16:28:18 +010010916transparent (deprecated)
10917 Enable client-side transparent proxying
10918 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010010919 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010010920 Arguments : none
10921
10922 This keyword was introduced in order to provide layer 7 persistence to layer
10923 3 load balancers. The idea is to use the OS's ability to redirect an incoming
10924 connection for a remote address to a local process (here HAProxy), and let
10925 this process know what address was initially requested. When this option is
10926 used, sessions without cookies will be forwarded to the original destination
10927 IP address of the incoming request (which should match that of another
10928 equipment), while requests with cookies will still be forwarded to the
10929 appropriate server.
10930
10931 The "transparent" keyword is deprecated, use "option transparent" instead.
10932
10933 Note that contrary to a common belief, this option does NOT make HAProxy
10934 present the client's IP to the server when establishing the connection.
10935
Willy Tarreau844e3c52008-01-11 16:28:18 +010010936 See also: "option transparent"
10937
William Lallemanda73203e2012-03-12 12:48:57 +010010938unique-id-format <string>
10939 Generate a unique ID for each request.
10940 May be used in sections : defaults | frontend | listen | backend
10941 yes | yes | yes | no
10942 Arguments :
10943 <string> is a log-format string.
10944
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010945 This keyword creates a ID for each request using the custom log format. A
10946 unique ID is useful to trace a request passing through many components of
10947 a complex infrastructure. The newly created ID may also be logged using the
10948 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010010949
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010950 The format should be composed from elements that are guaranteed to be
10951 unique when combined together. For instance, if multiple haproxy instances
10952 are involved, it might be important to include the node name. It is often
10953 needed to log the incoming connection's source and destination addresses
10954 and ports. Note that since multiple requests may be performed over the same
10955 connection, including a request counter may help differentiate them.
10956 Similarly, a timestamp may protect against a rollover of the counter.
10957 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010010958
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010959 It is recommended to use hexadecimal notation for many fields since it
10960 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010010961
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010962 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010963
Julien Vehentf21be322014-03-07 08:27:34 -050010964 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010965
10966 will generate:
10967
10968 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10969
10970 See also: "unique-id-header"
10971
10972unique-id-header <name>
10973 Add a unique ID header in the HTTP request.
10974 May be used in sections : defaults | frontend | listen | backend
10975 yes | yes | yes | no
10976 Arguments :
10977 <name> is the name of the header.
10978
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010979 Add a unique-id header in the HTTP request sent to the server, using the
10980 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010010981
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010982 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010983
Julien Vehentf21be322014-03-07 08:27:34 -050010984 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010985 unique-id-header X-Unique-ID
10986
10987 will generate:
10988
10989 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10990
10991 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010010992
Willy Tarreauf51658d2014-04-23 01:21:56 +020010993use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010994 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010995 May be used in sections : defaults | frontend | listen | backend
10996 no | yes | yes | no
10997 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010998 <backend> is the name of a valid backend or "listen" section, or a
10999 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011000
Willy Tarreauf51658d2014-04-23 01:21:56 +020011001 <condition> is a condition composed of ACLs, as described in section 7. If
11002 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011003
11004 When doing content-switching, connections arrive on a frontend and are then
11005 dispatched to various backends depending on a number of conditions. The
11006 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011007 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011008 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020011009 source address validation) or combined with a "tcp-request" rule to wait for
11010 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010011011
11012 There may be as many "use_backend" rules as desired. All of these rules are
11013 evaluated in their declaration order, and the first one which matches will
11014 assign the backend.
11015
11016 In the first form, the backend will be used if the condition is met. In the
11017 second form, the backend will be used if the condition is not met. If no
11018 condition is valid, the backend defined with "default_backend" will be used.
11019 If no default backend is defined, either the servers in the same section are
11020 used (in case of a "listen" section) or, in case of a frontend, no server is
11021 used and a 503 service unavailable response is returned.
11022
Willy Tarreau51aecc72009-07-12 09:47:04 +020011023 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010011024 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020011025 and backend processing will immediately follow, or the backend will wait for
11026 a complete HTTP request to get in. This feature is useful when a frontend
11027 must decode several protocols on a unique port, one of them being HTTP.
11028
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011029 When <backend> is a simple name, it is resolved at configuration time, and an
11030 error is reported if the specified backend does not exist. If <backend> is
11031 a log-format string instead, no check may be done at configuration time, so
11032 the backend name is resolved dynamically at run time. If the resulting
11033 backend name does not correspond to any valid backend, no other rule is
11034 evaluated, and the default_backend directive is applied instead. Note that
11035 when using dynamic backend names, it is highly recommended to use a prefix
11036 that no other backend uses in order to ensure that an unauthorized backend
11037 cannot be forced from the request.
11038
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011039 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010011040 used to detect the association between frontends and backends to compute the
11041 backend's "fullconn" setting. This cannot be done for dynamic names.
11042
11043 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
11044 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010011045
Willy Tarreau036fae02008-01-06 13:24:40 +010011046
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011047use-server <server> if <condition>
11048use-server <server> unless <condition>
11049 Only use a specific server if/unless an ACL-based condition is matched.
11050 May be used in sections : defaults | frontend | listen | backend
11051 no | no | yes | yes
11052 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020011053 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011054
11055 <condition> is a condition composed of ACLs, as described in section 7.
11056
11057 By default, connections which arrive to a backend are load-balanced across
11058 the available servers according to the configured algorithm, unless a
11059 persistence mechanism such as a cookie is used and found in the request.
11060
11061 Sometimes it is desirable to forward a particular request to a specific
11062 server without having to declare a dedicated backend for this server. This
11063 can be achieved using the "use-server" rules. These rules are evaluated after
11064 the "redirect" rules and before evaluating cookies, and they have precedence
11065 on them. There may be as many "use-server" rules as desired. All of these
11066 rules are evaluated in their declaration order, and the first one which
11067 matches will assign the server.
11068
11069 If a rule designates a server which is down, and "option persist" is not used
11070 and no force-persist rule was validated, it is ignored and evaluation goes on
11071 with the next rules until one matches.
11072
11073 In the first form, the server will be used if the condition is met. In the
11074 second form, the server will be used if the condition is not met. If no
11075 condition is valid, the processing continues and the server will be assigned
11076 according to other persistence mechanisms.
11077
11078 Note that even if a rule is matched, cookie processing is still performed but
11079 does not assign the server. This allows prefixed cookies to have their prefix
11080 stripped.
11081
11082 The "use-server" statement works both in HTTP and TCP mode. This makes it
11083 suitable for use with content-based inspection. For instance, a server could
11084 be selected in a farm according to the TLS SNI field. And if these servers
11085 have their weight set to zero, they will not be used for other traffic.
11086
11087 Example :
11088 # intercept incoming TLS requests based on the SNI field
11089 use-server www if { req_ssl_sni -i www.example.com }
11090 server www 192.168.0.1:443 weight 0
11091 use-server mail if { req_ssl_sni -i mail.example.com }
11092 server mail 192.168.0.1:587 weight 0
11093 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000011094 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011095 # all the rest is forwarded to this server
11096 server default 192.168.0.2:443 check
11097
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011098 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011099
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011100
Davor Ocelice9ed2812017-12-25 17:49:28 +0100111015. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011102--------------------------
11103
11104The "bind", "server" and "default-server" keywords support a number of settings
11105depending on some build options and on the system HAProxy was built on. These
11106settings generally each consist in one word sometimes followed by a value,
11107written on the same line as the "bind" or "server" line. All these options are
11108described in this section.
11109
11110
111115.1. Bind options
11112-----------------
11113
11114The "bind" keyword supports a certain number of settings which are all passed
11115as arguments on the same line. The order in which those arguments appear makes
11116no importance, provided that they appear after the bind address. All of these
11117parameters are optional. Some of them consist in a single words (booleans),
11118while other ones expect a value after them. In this case, the value must be
11119provided immediately after the setting name.
11120
11121The currently supported settings are the following ones.
11122
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011123accept-netscaler-cip <magic number>
11124 Enforces the use of the NetScaler Client IP insertion protocol over any
11125 connection accepted by any of the TCP sockets declared on the same line. The
11126 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
11127 the incoming connection to be used everywhere an address is used, with the
11128 only exception of "tcp-request connection" rules which will only see the
11129 real connection address. Logs will reflect the addresses indicated in the
11130 protocol, unless it is violated, in which case the real address will still
11131 be used. This keyword combined with support from external components can be
11132 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010011133 mechanism which is not always reliable and not even always usable. See also
11134 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
11135 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011136
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011137accept-proxy
11138 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020011139 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
11140 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011141 3/4 addresses of the incoming connection to be used everywhere an address is
11142 used, with the only exception of "tcp-request connection" rules which will
11143 only see the real connection address. Logs will reflect the addresses
11144 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010011145 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011146 components can be used as an efficient and reliable alternative to the
11147 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020011148 usable. See also "tcp-request connection expect-proxy" for a finer-grained
11149 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011150
Olivier Houchardc2aae742017-09-22 18:26:28 +020011151allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010011152 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010011153 due to security considerations. Because it is vulnerable to replay attacks,
11154 you should only allow if for requests that are safe to replay, ie requests
11155 that are idempotent. You can use the "wait-for-handshake" action for any
11156 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020011157
Willy Tarreauab861d32013-04-02 02:30:41 +020011158alpn <protocols>
11159 This enables the TLS ALPN extension and advertises the specified protocol
11160 list as supported on top of ALPN. The protocol list consists in a comma-
11161 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
11162 quotes). This requires that the SSL library is build with support for TLS
11163 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010011164 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
11165 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11166 now obsolete NPN extension. At the time of writing this, most browsers still
11167 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
11168 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
11169 are expected to be supported, both versions can be advertised, in order of
11170 preference, like below :
11171
11172 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020011173
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011174backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010011175 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011176 backlog is used instead, which generally defaults to the maxconn value.
11177
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010011178curves <curves>
11179 This setting is only available when support for OpenSSL was built in. It sets
11180 the string describing the list of elliptic curves algorithms ("curve suite")
11181 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
11182 string is a colon-delimited list of curve name.
11183 Example: "X25519:P-256" (without quote)
11184 When "curves" is set, "ecdhe" parameter is ignored.
11185
Emeric Brun7fb34422012-09-28 15:26:15 +020011186ecdhe <named curve>
11187 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010011188 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
11189 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020011190
Emeric Brunfd33a262012-10-11 16:28:27 +020011191ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011192 This setting is only available when support for OpenSSL was built in. It
11193 designates a PEM file from which to load CA certificates used to verify
11194 client's certificate.
11195
Emeric Brunb6dc9342012-09-28 17:55:37 +020011196ca-ignore-err [all|<errorID>,...]
11197 This setting is only available when support for OpenSSL was built in.
11198 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
11199 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
11200 error is ignored.
11201
Christopher Faulet31af49d2015-06-09 17:29:50 +020011202ca-sign-file <cafile>
11203 This setting is only available when support for OpenSSL was built in. It
11204 designates a PEM file containing both the CA certificate and the CA private
11205 key used to create and sign server's certificates. This is a mandatory
11206 setting when the dynamic generation of certificates is enabled. See
11207 'generate-certificates' for details.
11208
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000011209ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020011210 This setting is only available when support for OpenSSL was built in. It is
11211 the CA private key passphrase. This setting is optional and used only when
11212 the dynamic generation of certificates is enabled. See
11213 'generate-certificates' for details.
11214
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011215ciphers <ciphers>
11216 This setting is only available when support for OpenSSL was built in. It sets
11217 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000011218 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011219 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011220 information and recommendations see e.g.
11221 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11222 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11223 cipher configuration, please check the "ciphersuites" keyword.
11224
11225ciphersuites <ciphersuites>
11226 This setting is only available when support for OpenSSL was built in and
11227 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
11228 the list of cipher algorithms ("cipher suite") that are negotiated during the
11229 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011230 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
11231 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011232
Emeric Brunfd33a262012-10-11 16:28:27 +020011233crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011234 This setting is only available when support for OpenSSL was built in. It
11235 designates a PEM file from which to load certificate revocation list used
11236 to verify client's certificate.
11237
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011238crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000011239 This setting is only available when support for OpenSSL was built in. It
11240 designates a PEM file containing both the required certificates and any
11241 associated private keys. This file can be built by concatenating multiple
11242 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
11243 requires an intermediate certificate, this can also be concatenated into this
11244 file.
11245
11246 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
11247 are loaded.
11248
11249 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011250 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011251 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
11252 specified multiple times in order to load certificates from multiple files or
11253 directories. The certificates will be presented to clients who provide a
11254 valid TLS Server Name Indication field matching one of their CN or alt
Davor Ocelice9ed2812017-12-25 17:49:28 +010011255 subjects. Wildcards are supported, where a wildcard character '*' is used
11256 instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011257 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000011258
11259 If no SNI is provided by the client or if the SSL library does not support
11260 TLS extensions, or if the client provides an SNI hostname which does not
11261 match any certificate, then the first loaded certificate will be presented.
11262 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011263 recommended to load the default one first as a file or to ensure that it will
11264 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000011265
Emeric Brune032bfa2012-09-28 13:01:45 +020011266 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011267
Davor Ocelice9ed2812017-12-25 17:49:28 +010011268 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000011269 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010011270 choose a web server that the CA believes requires an intermediate CA (for
11271 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000011272 others, e.g. nginx, result in a wrong bundle that will not work for some
11273 clients).
11274
Emeric Brun4147b2e2014-06-16 18:36:30 +020011275 For each PEM file, haproxy checks for the presence of file at the same path
11276 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
11277 Status Request extension (also known as "OCSP stapling") is automatically
11278 enabled. The content of this file is optional. If not empty, it must contain
11279 a valid OCSP Response in DER format. In order to be valid an OCSP Response
11280 must comply with the following rules: it has to indicate a good status,
11281 it has to be a single response for the certificate of the PEM file, and it
11282 has to be valid at the moment of addition. If these rules are not respected
11283 the OCSP Response is ignored and a warning is emitted. In order to identify
11284 which certificate an OCSP Response applies to, the issuer's certificate is
11285 necessary. If the issuer's certificate is not found in the PEM file, it will
11286 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
11287 if it exists otherwise it will fail with an error.
11288
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011289 For each PEM file, haproxy also checks for the presence of file at the same
11290 path suffixed by ".sctl". If such file is found, support for Certificate
11291 Transparency (RFC6962) TLS extension is enabled. The file must contain a
11292 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
11293 to check basic syntax, but no signatures are verified.
11294
yanbzhu6c25e9e2016-01-05 12:52:02 -050011295 There are cases where it is desirable to support multiple key types, e.g. RSA
11296 and ECDSA in the cipher suites offered to the clients. This allows clients
11297 that support EC certificates to be able to use EC ciphers, while
11298 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050011299
11300 In order to provide this functionality, multiple PEM files, each with a
11301 different key type, are required. To associate these PEM files into a
11302 "cert bundle" that is recognized by haproxy, they must be named in the
11303 following way: All PEM files that are to be bundled must have the same base
11304 name, with a suffix indicating the key type. Currently, three suffixes are
11305 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
11306 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
11307 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
11308 suffix matters. To load this bundle into haproxy, specify the base name only:
11309
11310 Example : bind :8443 ssl crt example.pem
11311
yanbzhu6c25e9e2016-01-05 12:52:02 -050011312 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050011313 a cert bundle.
11314
Davor Ocelice9ed2812017-12-25 17:49:28 +010011315 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050011316 support multiple key types. PEM files are combined based on Common Name
11317 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
11318 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
11319 entries in the certificates in that bundle, haproxy will not be able to
11320 provide multi-cert support.
11321
11322 Assuming bundle in the example above contained the following:
11323
11324 Filename | CN | SAN
11325 -------------------+-----------------+-------------------
11326 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050011327 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050011328 example.pem.ecdsa | www.example.com | ecdsa.example.com
11329 -------------------+-----------------+-------------------
11330
11331 Users connecting with an SNI of "www.example.com" will be able
11332 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
11333 "rsa.example.com" will only be able to use RSA cipher suites, and users
11334 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011335 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
11336 no need to bundle certificates. ECDSA certificate will be preferred if client
11337 support it.
yanbzhud19630c2015-12-14 15:10:25 -050011338
11339 If a directory name is given as the <cert> argument, haproxy will
11340 automatically search and load bundled files in that directory.
11341
11342 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
11343 bundling. Each certificate can have its own .ocsp and .issuer file. At this
11344 time, sctl is not supported in multi-certificate bundling.
11345
Emeric Brunb6dc9342012-09-28 17:55:37 +020011346crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000011347 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010011348 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011349 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000011350 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020011351
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011352crt-list <file>
11353 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011354 designates a list of PEM file with an optional ssl configuration and a SNI
11355 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011356
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011357 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
11358
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011359 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
11360 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011361 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011362 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011363
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020011364 Wildcards are supported in the SNI filter. Negative filter are also supported,
11365 only useful in combination with a wildcard filter to exclude a particular SNI.
11366 The certificates will be presented to clients who provide a valid TLS Server
11367 Name Indication field matching one of the SNI filters. If no SNI filter is
11368 specified, the CN and alt subjects are used. This directive may be specified
11369 multiple times. See the "crt" option for more information. The default
11370 certificate is still needed to meet OpenSSL expectations. If it is not used,
11371 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011372
yanbzhu6c25e9e2016-01-05 12:52:02 -050011373 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020011374 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011375 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
11376 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
11377 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050011378
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011379 crt-list file example:
11380 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011381 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011382 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011383 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011384
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011385defer-accept
11386 Is an optional keyword which is supported only on certain Linux kernels. It
11387 states that a connection will only be accepted once some data arrive on it,
11388 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010011389 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011390 performance by ensuring that most of the request is already available when
11391 the connection is accepted. On the other hand, it will not be able to detect
11392 connections which don't talk. It is important to note that this option is
11393 broken in all kernels up to 2.6.31, as the connection is never accepted until
11394 the client talks. This can cause issues with front firewalls which would see
11395 an established connection while the proxy will only see it in SYN_RECV. This
11396 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
11397
William Lallemandf6975e92017-05-26 17:42:10 +020011398expose-fd listeners
11399 This option is only usable with the stats socket. It gives your stats socket
11400 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020011401 During a reload with the master-worker mode, the process is automatically
11402 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011403 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020011404
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011405force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011406 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011407 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011408 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011409 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011410
11411force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011412 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011413 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011414 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011415
11416force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011417 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011418 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011419 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011420
11421force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011422 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011423 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011424 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011425
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011426force-tlsv13
11427 This option enforces use of TLSv1.3 only on SSL connections instantiated from
11428 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011429 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011430
Christopher Faulet31af49d2015-06-09 17:29:50 +020011431generate-certificates
11432 This setting is only available when support for OpenSSL was built in. It
11433 enables the dynamic SSL certificates generation. A CA certificate and its
11434 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
11435 a transparent forward proxy, SSL requests generate errors because of a common
11436 name mismatch on the certificate presented to the client. With this option
11437 enabled, HAProxy will try to forge a certificate using the SNI hostname
11438 indicated by the client. This is done only if no certificate matches the SNI
11439 hostname (see 'crt-list'). If an error occurs, the default certificate is
11440 used, else the 'strict-sni' option is set.
11441 It can also be used when HAProxy is configured as a reverse proxy to ease the
11442 deployment of an architecture with many backends.
11443
11444 Creating a SSL certificate is an expensive operation, so a LRU cache is used
11445 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010011446 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020011447 certificate is used many times.
11448
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011449gid <gid>
11450 Sets the group of the UNIX sockets to the designated system gid. It can also
11451 be set by default in the global section's "unix-bind" statement. Note that
11452 some platforms simply ignore this. This setting is equivalent to the "group"
11453 setting except that the group ID is used instead of its name. This setting is
11454 ignored by non UNIX sockets.
11455
11456group <group>
11457 Sets the group of the UNIX sockets to the designated system group. It can
11458 also be set by default in the global section's "unix-bind" statement. Note
11459 that some platforms simply ignore this. This setting is equivalent to the
11460 "gid" setting except that the group name is used instead of its gid. This
11461 setting is ignored by non UNIX sockets.
11462
11463id <id>
11464 Fixes the socket ID. By default, socket IDs are automatically assigned, but
11465 sometimes it is more convenient to fix them to ease monitoring. This value
11466 must be strictly positive and unique within the listener/frontend. This
11467 option can only be used when defining only a single socket.
11468
11469interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010011470 Restricts the socket to a specific interface. When specified, only packets
11471 received from that particular interface are processed by the socket. This is
11472 currently only supported on Linux. The interface must be a primary system
11473 interface, not an aliased interface. It is also possible to bind multiple
11474 frontends to the same address if they are bound to different interfaces. Note
11475 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010011476 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
11477 uses the same interface as inbound traffic, and its associated routing table,
11478 even if there are explicit routes through different interfaces configured.
11479 This can prove useful to address asymmetric routing issues when the same
11480 client IP addresses need to be able to reach frontends hosted on different
11481 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011482
Willy Tarreauabb175f2012-09-24 12:43:26 +020011483level <level>
11484 This setting is used with the stats sockets only to restrict the nature of
11485 the commands that can be issued on the socket. It is ignored by other
11486 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010011487 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020011488 read, and no change is allowed. It would make sense on systems where it
11489 is not easy to restrict access to the socket.
11490 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011491 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020011492 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010011493 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020011494 all counters).
11495
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020011496severity-output <format>
11497 This setting is used with the stats sockets only to configure severity
11498 level output prepended to informational feedback messages. Severity
11499 level of messages can range between 0 and 7, conforming to syslog
11500 rfc5424. Valid and successful socket commands requesting data
11501 (i.e. "show map", "get acl foo" etc.) will never have a severity level
11502 prepended. It is ignored by other sockets. <format> can be one of :
11503 - "none" (default) no severity level is prepended to feedback messages.
11504 - "number" severity level is prepended as a number.
11505 - "string" severity level is prepended as a string following the
11506 rfc5424 convention.
11507
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011508maxconn <maxconn>
11509 Limits the sockets to this number of concurrent connections. Extraneous
11510 connections will remain in the system's backlog until a connection is
11511 released. If unspecified, the limit will be the same as the frontend's
11512 maxconn. Note that in case of port ranges or multiple addresses, the same
11513 value will be applied to each socket. This setting enables different
11514 limitations on expensive sockets, for instance SSL entries which may easily
11515 eat all memory.
11516
11517mode <mode>
11518 Sets the octal mode used to define access permissions on the UNIX socket. It
11519 can also be set by default in the global section's "unix-bind" statement.
11520 Note that some platforms simply ignore this. This setting is ignored by non
11521 UNIX sockets.
11522
11523mss <maxseg>
11524 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
11525 connections. This can be used to force a lower MSS for certain specific
11526 ports, for instance for connections passing through a VPN. Note that this
11527 relies on a kernel feature which is theoretically supported under Linux but
11528 was buggy in all versions prior to 2.6.28. It may or may not work on other
11529 operating systems. It may also not change the advertised value but change the
11530 effective size of outgoing segments. The commonly advertised value for TCPv4
11531 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
11532 positive, it will be used as the advertised MSS. If it is negative, it will
11533 indicate by how much to reduce the incoming connection's advertised MSS for
11534 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
11535
11536name <name>
11537 Sets an optional name for these sockets, which will be reported on the stats
11538 page.
11539
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011540namespace <name>
11541 On Linux, it is possible to specify which network namespace a socket will
11542 belong to. This directive makes it possible to explicitly bind a listener to
11543 a namespace different from the default one. Please refer to your operating
11544 system's documentation to find more details about network namespaces.
11545
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011546nice <nice>
11547 Sets the 'niceness' of connections initiated from the socket. Value must be
11548 in the range -1024..1024 inclusive, and defaults to zero. Positive values
11549 means that such connections are more friendly to others and easily offer
11550 their place in the scheduler. On the opposite, negative values mean that
11551 connections want to run with a higher priority than others. The difference
11552 only happens under high loads when the system is close to saturation.
11553 Negative values are appropriate for low-latency or administration services,
11554 and high values are generally recommended for CPU intensive tasks such as SSL
11555 processing or bulk transfers which are less sensible to latency. For example,
11556 it may make sense to use a positive value for an SMTP socket and a negative
11557 one for an RDP socket.
11558
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011559no-ca-names
11560 This setting is only available when support for OpenSSL was built in. It
11561 prevents from send CA names in server hello message when ca-file is used.
11562
Emeric Brun9b3009b2012-10-05 11:55:06 +020011563no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011564 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011565 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011566 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011567 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011568 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
11569 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011570
Emeric Brun90ad8722012-10-02 14:00:59 +020011571no-tls-tickets
11572 This setting is only available when support for OpenSSL was built in. It
11573 disables the stateless session resumption (RFC 5077 TLS Ticket
11574 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011575 session resumption is more expensive in CPU usage. This option is also
11576 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020011577
Emeric Brun9b3009b2012-10-05 11:55:06 +020011578no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011579 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011580 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011581 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011582 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011583 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11584 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011585
Emeric Brun9b3009b2012-10-05 11:55:06 +020011586no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020011587 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011588 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011589 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011590 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011591 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11592 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011593
Emeric Brun9b3009b2012-10-05 11:55:06 +020011594no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020011595 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011596 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011597 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011598 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011599 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11600 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011601
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011602no-tlsv13
11603 This setting is only available when support for OpenSSL was built in. It
11604 disables support for TLSv1.3 on any sockets instantiated from the listener
11605 when SSL is supported. Note that SSLv2 is forced disabled in the code and
11606 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011607 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11608 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011609
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011610npn <protocols>
11611 This enables the NPN TLS extension and advertises the specified protocol list
11612 as supported on top of NPN. The protocol list consists in a comma-delimited
11613 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
11614 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020011615 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010011616 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11617 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
11618 version of OpenSSL, NPN might still be used as most clients still support it
11619 at the time of writing this. It is possible to enable both NPN and ALPN
11620 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011621
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011622prefer-client-ciphers
11623 Use the client's preference when selecting the cipher suite, by default
11624 the server's preference is enforced. This option is also available on
11625 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020011626 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
11627 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
11628 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011629
Christopher Fauletc644fa92017-11-23 22:44:11 +010011630process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010011631 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010011632 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010011633 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010011634 between the two is applied. If in the end the listener is not allowed to run
11635 on any remaining process, a warning is emitted, and the listener will either
11636 run on the first process of the listener if a single process was specified,
11637 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010011638 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010011639 connections for this listener, for the the process set. If multiple processes
11640 and threads are configured, a warning is emitted, as it either results from a
11641 configuration error or a misunderstanding of these models. For the unlikely
11642 case where several ranges are needed, this directive may be repeated.
11643 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010011644
11645 all | odd | even | number[-[number]]
11646
11647 Ranges can be partially defined. The higher bound can be omitted. In such
11648 case, it is replaced by the corresponding maximum value. The main purpose of
11649 this directive is to be used with the stats sockets and have one different
11650 socket per process. The second purpose is to have multiple bind lines sharing
11651 the same IP:port but not the same process in a listener, so that the system
11652 can distribute the incoming connections into multiple queues and allow a
11653 smoother inter-process load balancing. Currently Linux 3.9 and above is known
11654 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020011655
Christopher Fauleta717b992018-04-10 14:43:00 +020011656proto <name>
11657 Forces the multiplexer's protocol to use for the incoming connections. It
11658 must be compatible with the mode of the frontend (TCP or HTTP). It must also
11659 be usable on the frontend side. The list of available protocols is reported
11660 in haproxy -vv.
11661 Idea behind this optipon is to bypass the selection of the best multiplexer's
11662 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080011663 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020011664 h2" on the bind line.
11665
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011666ssl
11667 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011668 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011669 certificate is necessary (see "crt" above). All contents in the buffers will
11670 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011671 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11672 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011673
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011674ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11675 This option enforces use of <version> or lower on SSL connections instantiated
11676 from this listener. This option is also available on global statement
11677 "ssl-default-bind-options". See also "ssl-min-ver".
11678
11679ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11680 This option enforces use of <version> or upper on SSL connections instantiated
11681 from this listener. This option is also available on global statement
11682 "ssl-default-bind-options". See also "ssl-max-ver".
11683
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011684strict-sni
11685 This setting is only available when support for OpenSSL was built in. The
11686 SSL/TLS negotiation is allow only if the client provided an SNI which match
11687 a certificate. The default certificate is not used.
11688 See the "crt" option for more information.
11689
Willy Tarreau2af207a2015-02-04 00:45:58 +010011690tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011691 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011692 listening socket. This option is available on Linux since version 2.6.37. It
11693 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011694 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011695 useful on long-lived connections experiencing long idle periods such as
11696 remote terminals or database connection pools, where the client and server
11697 timeouts must remain high to allow a long period of idle, but where it is
11698 important to detect that the client has disappeared in order to release all
11699 resources associated with its connection (and the server's session). The
11700 argument is a delay expressed in milliseconds by default. This only works
11701 for regular TCP connections, and is ignored for other protocols.
11702
Willy Tarreau1c862c52012-10-05 16:21:00 +020011703tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011704 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011705 enables TCP Fast Open on the listening socket, which means that clients which
11706 support this feature will be able to send a request and receive a response
11707 during the 3-way handshake starting from second connection, thus saving one
11708 round-trip after the first connection. This only makes sense with protocols
11709 that use high connection rates and where each round trip matters. This can
11710 possibly cause issues with many firewalls which do not accept data on SYN
11711 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011712 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11713 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11714 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011715
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011716tls-ticket-keys <keyfile>
11717 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010011718 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
11719 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
11720 The first key determines the key length used for next keys: you can't mix
11721 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
11722 build option (default 3) and at least as many keys need to be present in
11723 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
11724 penultimate one for encryption. This enables easy key rotation by just
11725 appending new key to the file and reloading the process. Keys must be
11726 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
11727 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011728 storage such as hard drives (hint: use tmpfs and don't swap those files).
11729 Lifetime hint can be changed using tune.ssl.timeout.
11730
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011731transparent
11732 Is an optional keyword which is supported only on certain Linux kernels. It
11733 indicates that the addresses will be bound even if they do not belong to the
11734 local machine, and that packets targeting any of these addresses will be
11735 intercepted just as if the addresses were locally configured. This normally
11736 requires that IP forwarding is enabled. Caution! do not use this with the
11737 default address '*', as it would redirect any traffic for the specified port.
11738 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11739 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11740 kernel version. Some distribution kernels include backports of the feature,
11741 so check for support with your vendor.
11742
Willy Tarreau77e3af92012-11-24 15:07:23 +010011743v4v6
11744 Is an optional keyword which is supported only on most recent systems
11745 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11746 and IPv6 when it uses the default address. Doing so is sometimes necessary
11747 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011748 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011749
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011750v6only
11751 Is an optional keyword which is supported only on most recent systems
11752 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11753 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011754 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11755 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011756
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011757uid <uid>
11758 Sets the owner of the UNIX sockets to the designated system uid. It can also
11759 be set by default in the global section's "unix-bind" statement. Note that
11760 some platforms simply ignore this. This setting is equivalent to the "user"
11761 setting except that the user numeric ID is used instead of its name. This
11762 setting is ignored by non UNIX sockets.
11763
11764user <user>
11765 Sets the owner of the UNIX sockets to the designated system user. It can also
11766 be set by default in the global section's "unix-bind" statement. Note that
11767 some platforms simply ignore this. This setting is equivalent to the "uid"
11768 setting except that the user name is used instead of its uid. This setting is
11769 ignored by non UNIX sockets.
11770
Emeric Brun1a073b42012-09-28 17:07:34 +020011771verify [none|optional|required]
11772 This setting is only available when support for OpenSSL was built in. If set
11773 to 'none', client certificate is not requested. This is the default. In other
11774 cases, a client certificate is requested. If the client does not provide a
11775 certificate after the request and if 'verify' is set to 'required', then the
11776 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011777 certificate provided by the client is always verified using CAs from
11778 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11779 is aborted, regardless of the 'verify' option, unless the error code exactly
11780 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011781
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200117825.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011783------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011784
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011785The "server" and "default-server" keywords support a certain number of settings
11786which are all passed as arguments on the server line. The order in which those
11787arguments appear does not count, and they are all optional. Some of those
11788settings are single words (booleans) while others expect one or several values
11789after them. In this case, the values must immediately follow the setting name.
11790Except default-server, all those settings must be specified after the server's
11791address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011792
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011793 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011794 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020011795
Frédéric Lécailled2376272017-03-21 18:52:12 +010011796Note that all these settings are supported both by "server" and "default-server"
11797keywords, except "id" which is only supported by "server".
11798
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011799The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011800
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011801addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011802 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010011803 to send health-checks or to probe the agent-check. On some servers, it may be
11804 desirable to dedicate an IP address to specific component able to perform
11805 complex tests which are more suitable to health-checks than the application.
11806 This parameter is ignored if the "check" parameter is not set. See also the
11807 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011808
Simon Hormand60d6912013-11-25 10:46:36 +090011809agent-check
11810 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010011811 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010011812 to the port set by the "agent-port" parameter and reading an ASCII string
11813 terminated by the first '\r' or '\n' met. The string is made of a series of
11814 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090011815
Willy Tarreau81f5d942013-12-09 20:51:51 +010011816 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090011817 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020011818 weight of a server as configured when haproxy starts. Note that a zero
11819 weight is reported on the stats page as "DRAIN" since it has the same
11820 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090011821
Davor Ocelice9ed2812017-12-25 17:49:28 +010011822 - The string "maxconn:" followed by an integer (no space between). Values
11823 in this format will set the maxconn of a server. The maximum number of
11824 connections advertised needs to be multiplied by the number of load
11825 balancers and different backends that use this health check to get the
11826 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020011827
Willy Tarreau81f5d942013-12-09 20:51:51 +010011828 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011829 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090011830
Willy Tarreau81f5d942013-12-09 20:51:51 +010011831 - The word "drain". This will turn the server's administrative state to the
11832 DRAIN mode, thus it will not accept any new connections other than those
11833 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090011834
Willy Tarreau81f5d942013-12-09 20:51:51 +010011835 - The word "maint". This will turn the server's administrative state to the
11836 MAINT mode, thus it will not accept any new connections at all, and health
11837 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090011838
Willy Tarreau81f5d942013-12-09 20:51:51 +010011839 - The words "down", "failed", or "stopped", optionally followed by a
11840 description string after a sharp ('#'). All of these mark the server's
11841 operating state as DOWN, but since the word itself is reported on the stats
11842 page, the difference allows an administrator to know if the situation was
11843 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010011844 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010011845 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090011846
Willy Tarreau81f5d942013-12-09 20:51:51 +010011847 - The word "up" sets back the server's operating state as UP if health checks
11848 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090011849
Willy Tarreau81f5d942013-12-09 20:51:51 +010011850 Parameters which are not advertised by the agent are not changed. For
11851 example, an agent might be designed to monitor CPU usage and only report a
11852 relative weight and never interact with the operating status. Similarly, an
11853 agent could be designed as an end-user interface with 3 radio buttons
11854 allowing an administrator to change only the administrative state. However,
11855 it is important to consider that only the agent may revert its own actions,
11856 so if a server is set to DRAIN mode or to DOWN state using the agent, the
11857 agent must implement the other equivalent actions to bring the service into
11858 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090011859
Simon Horman2f1f9552013-11-25 10:46:37 +090011860 Failure to connect to the agent is not considered an error as connectivity
11861 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010011862 parameter. Warning though, it is not a good idea to stop an agent after it
11863 reports "down", since only an agent reporting "up" will be able to turn the
11864 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010011865 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090011866
Willy Tarreau81f5d942013-12-09 20:51:51 +010011867 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010011868 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090011869
James Brown55f9ff12015-10-21 18:19:05 -070011870agent-send <string>
11871 If this option is specified, haproxy will send the given string (verbatim)
11872 to the agent server upon connection. You could, for example, encode
11873 the backend name into this string, which would enable your agent to send
11874 different responses based on the backend. Make sure to include a '\n' if
11875 you want to terminate your request with a newline.
11876
Simon Hormand60d6912013-11-25 10:46:36 +090011877agent-inter <delay>
11878 The "agent-inter" parameter sets the interval between two agent checks
11879 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11880
11881 Just as with every other time-based parameter, it may be entered in any
11882 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
11883 parameter also serves as a timeout for agent checks "timeout check" is
11884 not set. In order to reduce "resonance" effects when multiple servers are
11885 hosted on the same hardware, the agent and health checks of all servers
11886 are started with a small time offset between them. It is also possible to
11887 add some random noise in the agent and health checks interval using the
11888 global "spread-checks" keyword. This makes sense for instance when a lot
11889 of backends use the same servers.
11890
11891 See also the "agent-check" and "agent-port" parameters.
11892
Misiek768d8602017-01-09 09:52:43 +010011893agent-addr <addr>
11894 The "agent-addr" parameter sets address for agent check.
11895
11896 You can offload agent-check to another target, so you can make single place
11897 managing status and weights of servers defined in haproxy in case you can't
11898 make self-aware and self-managing services. You can specify both IP or
11899 hostname, it will be resolved.
11900
Simon Hormand60d6912013-11-25 10:46:36 +090011901agent-port <port>
11902 The "agent-port" parameter sets the TCP port used for agent checks.
11903
11904 See also the "agent-check" and "agent-inter" parameters.
11905
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011906allow-0rtt
11907 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020011908 Note that early data will be sent only if the client used early data, or
11909 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011910
Olivier Houchardc7566002018-11-20 23:33:50 +010011911alpn <protocols>
11912 This enables the TLS ALPN extension and advertises the specified protocol
11913 list as supported on top of ALPN. The protocol list consists in a comma-
11914 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
11915 quotes). This requires that the SSL library is build with support for TLS
11916 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
11917 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
11918 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11919 now obsolete NPN extension.
11920 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
11921 be advertised, in order of preference, like below :
11922
11923 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
11924
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011925backup
11926 When "backup" is present on a server line, the server is only used in load
11927 balancing when all other non-backup servers are unavailable. Requests coming
11928 with a persistence cookie referencing the server will always be served
11929 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010011930 the "allbackups" option is set in the backend. See also the "no-backup" and
11931 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011932
Emeric Brunef42d922012-10-11 16:11:36 +020011933ca-file <cafile>
11934 This setting is only available when support for OpenSSL was built in. It
11935 designates a PEM file from which to load CA certificates used to verify
11936 server's certificate.
11937
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011938check
11939 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010011940 always considered available. If "check" is set, the server is available when
11941 accepting periodic TCP connections, to ensure that it is really able to serve
11942 requests. The default address and port to send the tests to are those of the
11943 server, and the default source is the same as the one defined in the
11944 backend. It is possible to change the address using the "addr" parameter, the
11945 port using the "port" parameter, the source address using the "source"
11946 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090011947 parameters. The request method is define in the backend using the "httpchk",
11948 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010011949 refer to those options and parameters for more information. See also
11950 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011951
Willy Tarreau6c16adc2012-10-05 00:04:16 +020011952check-send-proxy
11953 This option forces emission of a PROXY protocol line with outgoing health
11954 checks, regardless of whether the server uses send-proxy or not for the
11955 normal traffic. By default, the PROXY protocol is enabled for health checks
11956 if it is already enabled for normal traffic and if no "port" nor "addr"
11957 directive is present. However, if such a directive is present, the
11958 "check-send-proxy" option needs to be used to force the use of the
11959 protocol. See also the "send-proxy" option for more information.
11960
Olivier Houchard92150142018-12-21 19:47:01 +010011961check-alpn <protocols>
11962 Defines which protocols to advertise with ALPN. The protocol list consists in
11963 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
11964 (without quotes). If it is not set, the server ALPN is used.
11965
Jérôme Magninae9bb762018-12-09 16:08:26 +010011966check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020011967 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010011968 over SSL. It is only possible to use a string to set <sni>. If you want to
11969 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020011970
Willy Tarreau763a95b2012-10-04 23:15:39 +020011971check-ssl
11972 This option forces encryption of all health checks over SSL, regardless of
11973 whether the server uses SSL or not for the normal traffic. This is generally
11974 used when an explicit "port" or "addr" directive is specified and SSL health
11975 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011976 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020011977 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
11978 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011979 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010011980 See the "ssl" option for more information and "no-check-ssl" to disable
11981 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011982
Alexander Liu2a54bb72019-05-22 19:44:48 +080011983check-via-socks4
11984 This option enables outgoinng health checks using upstream socks4 proxy. By
11985 default, the health checks won't go through socks tunnel even it was enabled
11986 for normal traffic.
11987
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011988ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011989 This setting is only available when support for OpenSSL was built in. This
11990 option sets the string describing the list of cipher algorithms that is
11991 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011992 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
11993 information and recommendations see e.g.
11994 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11995 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11996 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011997
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011998ciphersuites <ciphersuites>
11999 This setting is only available when support for OpenSSL was built in and
12000 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
12001 describing the list of cipher algorithms that is negotiated during the TLS
12002 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000012003 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
12004 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
12005 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020012006
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012007cookie <value>
12008 The "cookie" parameter sets the cookie value assigned to the server to
12009 <value>. This value will be checked in incoming requests, and the first
12010 operational server possessing the same value will be selected. In return, in
12011 cookie insertion or rewrite modes, this value will be assigned to the cookie
12012 sent to the client. There is nothing wrong in having several servers sharing
12013 the same cookie value, and it is in fact somewhat common between normal and
12014 backup servers. See also the "cookie" keyword in backend section.
12015
Emeric Brunef42d922012-10-11 16:11:36 +020012016crl-file <crlfile>
12017 This setting is only available when support for OpenSSL was built in. It
12018 designates a PEM file from which to load certificate revocation list used
12019 to verify server's certificate.
12020
Emeric Bruna7aa3092012-10-26 12:58:00 +020012021crt <cert>
12022 This setting is only available when support for OpenSSL was built in.
12023 It designates a PEM file from which to load both a certificate and the
12024 associated private key. This file can be built by concatenating both PEM
12025 files into one. This certificate will be sent if the server send a client
12026 certificate request.
12027
Willy Tarreau96839092010-03-29 10:02:24 +020012028disabled
12029 The "disabled" keyword starts the server in the "disabled" state. That means
12030 that it is marked down in maintenance mode, and no connection other than the
12031 ones allowed by persist mode will reach it. It is very well suited to setup
12032 new servers, because normal traffic will never reach them, while it is still
12033 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012034 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012035
Frédéric Lécailled2376272017-03-21 18:52:12 +010012036enabled
12037 This option may be used as 'server' setting to reset any 'disabled'
12038 setting which would have been inherited from 'default-server' directive as
12039 default value.
12040 It may also be used as 'default-server' setting to reset any previous
12041 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020012042
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012043error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010012044 If health observing is enabled, the "error-limit" parameter specifies the
12045 number of consecutive errors that triggers event selected by the "on-error"
12046 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012047
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012048 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012049
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012050fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012051 The "fall" parameter states that a server will be considered as dead after
12052 <count> consecutive unsuccessful health checks. This value defaults to 3 if
12053 unspecified. See also the "check", "inter" and "rise" parameters.
12054
Emeric Brun8694b9a2012-10-05 14:39:07 +020012055force-sslv3
12056 This option enforces use of SSLv3 only when SSL is used to communicate with
12057 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012058 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012059 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012060
12061force-tlsv10
12062 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012063 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012064 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012065
12066force-tlsv11
12067 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012068 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012069 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012070
12071force-tlsv12
12072 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012073 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012074 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020012075
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012076force-tlsv13
12077 This option enforces use of TLSv1.3 only when SSL is used to communicate with
12078 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012079 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012080
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012081id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020012082 Set a persistent ID for the server. This ID must be positive and unique for
12083 the proxy. An unused ID will automatically be assigned if unset. The first
12084 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012085
Willy Tarreau6a031d12016-11-07 19:42:35 +010012086init-addr {last | libc | none | <ip>},[...]*
12087 Indicate in what order the server's address should be resolved upon startup
12088 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012089 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010012090 method which succeeds is used. If the end of the list is reached without
12091 finding a working method, an error is thrown. Method "last" suggests to pick
12092 the address which appears in the state file (see "server-state-file"). Method
12093 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
12094 depending on the operating system and build options). Method "none"
12095 specifically indicates that the server should start without any valid IP
12096 address in a down state. It can be useful to ignore some DNS issues upon
12097 startup, waiting for the situation to get fixed later. Finally, an IP address
12098 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012099 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010012100 server used to catch old sessions and present them with a decent error
12101 message for example. When the "first" load balancing algorithm is used, this
12102 IP address could point to a fake server used to trigger the creation of new
12103 instances on the fly. This option defaults to "last,libc" indicating that the
12104 previous address found in the state file (if any) is used first, otherwise
12105 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012106 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010012107
12108 Example:
12109 defaults
12110 # never fail on address resolution
12111 default-server init-addr last,libc,none
12112
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012113inter <delay>
12114fastinter <delay>
12115downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012116 The "inter" parameter sets the interval between two consecutive health checks
12117 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
12118 It is also possible to use "fastinter" and "downinter" to optimize delays
12119 between checks depending on the server state :
12120
Pieter Baauw44fc9df2015-09-17 21:30:46 +020012121 Server state | Interval used
12122 ----------------------------------------+----------------------------------
12123 UP 100% (non-transitional) | "inter"
12124 ----------------------------------------+----------------------------------
12125 Transitionally UP (going down "fall"), | "fastinter" if set,
12126 Transitionally DOWN (going up "rise"), | "inter" otherwise.
12127 or yet unchecked. |
12128 ----------------------------------------+----------------------------------
12129 DOWN 100% (non-transitional) | "downinter" if set,
12130 | "inter" otherwise.
12131 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010012132
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012133 Just as with every other time-based parameter, they can be entered in any
12134 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
12135 serves as a timeout for health checks sent to servers if "timeout check" is
12136 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090012137 hosted on the same hardware, the agent and health checks of all servers
12138 are started with a small time offset between them. It is also possible to
12139 add some random noise in the agent and health checks interval using the
12140 global "spread-checks" keyword. This makes sense for instance when a lot
12141 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012142
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012143maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012144 The "maxconn" parameter specifies the maximal number of concurrent
12145 connections that will be sent to this server. If the number of incoming
12146 concurrent requests goes higher than this value, they will be queued, waiting
12147 for a connection to be released. This parameter is very important as it can
12148 save fragile servers from going down under extreme loads. If a "minconn"
12149 parameter is specified, the limit becomes dynamic. The default value is "0"
12150 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
12151 the backend's "fullconn" keyword.
12152
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012153maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012154 The "maxqueue" parameter specifies the maximal number of connections which
12155 will wait in the queue for this server. If this limit is reached, next
12156 requests will be redispatched to other servers instead of indefinitely
12157 waiting to be served. This will break persistence but may allow people to
12158 quickly re-log in when the server they try to connect to is dying. The
12159 default value is "0" which means the queue is unlimited. See also the
12160 "maxconn" and "minconn" parameters.
12161
Willy Tarreau9c538e02019-01-23 10:21:49 +010012162max-reuse <count>
12163 The "max-reuse" argument indicates the HTTP connection processors that they
12164 should not reuse a server connection more than this number of times to send
12165 new requests. Permitted values are -1 (the default), which disables this
12166 limit, or any positive value. Value zero will effectively disable keep-alive.
12167 This is only used to work around certain server bugs which cause them to leak
12168 resources over time. The argument is not necessarily respected by the lower
12169 layers as there might be technical limitations making it impossible to
12170 enforce. At least HTTP/2 connections to servers will respect it.
12171
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012172minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012173 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
12174 limit following the backend's load. The server will always accept at least
12175 <minconn> connections, never more than <maxconn>, and the limit will be on
12176 the ramp between both values when the backend has less than <fullconn>
12177 concurrent connections. This makes it possible to limit the load on the
12178 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012179 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012180 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012181
Willy Tarreaud72f0f32015-10-13 14:50:22 +020012182namespace <name>
12183 On Linux, it is possible to specify which network namespace a socket will
12184 belong to. This directive makes it possible to explicitly bind a server to
12185 a namespace different from the default one. Please refer to your operating
12186 system's documentation to find more details about network namespaces.
12187
Frédéric Lécailled2376272017-03-21 18:52:12 +010012188no-agent-check
12189 This option may be used as "server" setting to reset any "agent-check"
12190 setting which would have been inherited from "default-server" directive as
12191 default value.
12192 It may also be used as "default-server" setting to reset any previous
12193 "default-server" "agent-check" setting.
12194
12195no-backup
12196 This option may be used as "server" setting to reset any "backup"
12197 setting which would have been inherited from "default-server" directive as
12198 default value.
12199 It may also be used as "default-server" setting to reset any previous
12200 "default-server" "backup" setting.
12201
12202no-check
12203 This option may be used as "server" setting to reset any "check"
12204 setting which would have been inherited from "default-server" directive as
12205 default value.
12206 It may also be used as "default-server" setting to reset any previous
12207 "default-server" "check" setting.
12208
12209no-check-ssl
12210 This option may be used as "server" setting to reset any "check-ssl"
12211 setting which would have been inherited from "default-server" directive as
12212 default value.
12213 It may also be used as "default-server" setting to reset any previous
12214 "default-server" "check-ssl" setting.
12215
Frédéric Lécailled2376272017-03-21 18:52:12 +010012216no-send-proxy
12217 This option may be used as "server" setting to reset any "send-proxy"
12218 setting which would have been inherited from "default-server" directive as
12219 default value.
12220 It may also be used as "default-server" setting to reset any previous
12221 "default-server" "send-proxy" setting.
12222
12223no-send-proxy-v2
12224 This option may be used as "server" setting to reset any "send-proxy-v2"
12225 setting which would have been inherited from "default-server" directive as
12226 default value.
12227 It may also be used as "default-server" setting to reset any previous
12228 "default-server" "send-proxy-v2" setting.
12229
12230no-send-proxy-v2-ssl
12231 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
12232 setting which would have been inherited from "default-server" directive as
12233 default value.
12234 It may also be used as "default-server" setting to reset any previous
12235 "default-server" "send-proxy-v2-ssl" setting.
12236
12237no-send-proxy-v2-ssl-cn
12238 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
12239 setting which would have been inherited from "default-server" directive as
12240 default value.
12241 It may also be used as "default-server" setting to reset any previous
12242 "default-server" "send-proxy-v2-ssl-cn" setting.
12243
12244no-ssl
12245 This option may be used as "server" setting to reset any "ssl"
12246 setting which would have been inherited from "default-server" directive as
12247 default value.
12248 It may also be used as "default-server" setting to reset any previous
12249 "default-server" "ssl" setting.
12250
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010012251no-ssl-reuse
12252 This option disables SSL session reuse when SSL is used to communicate with
12253 the server. It will force the server to perform a full handshake for every
12254 new connection. It's probably only useful for benchmarking, troubleshooting,
12255 and for paranoid users.
12256
Emeric Brun9b3009b2012-10-05 11:55:06 +020012257no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012258 This option disables support for SSLv3 when SSL is used to communicate with
12259 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012260 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012261
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012262 Supported in default-server: No
12263
Emeric Brunf9c5c472012-10-11 15:28:34 +020012264no-tls-tickets
12265 This setting is only available when support for OpenSSL was built in. It
12266 disables the stateless session resumption (RFC 5077 TLS Ticket
12267 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012268 session resumption is more expensive in CPU usage for servers. This option
12269 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010012270 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020012271
Emeric Brun9b3009b2012-10-05 11:55:06 +020012272no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020012273 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012274 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12275 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012276 often makes sense to disable it when communicating with local servers. This
12277 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012278 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012279
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012280 Supported in default-server: No
12281
Emeric Brun9b3009b2012-10-05 11:55:06 +020012282no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020012283 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012284 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12285 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012286 often makes sense to disable it when communicating with local servers. This
12287 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012288 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012289
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012290 Supported in default-server: No
12291
Emeric Brun9b3009b2012-10-05 11:55:06 +020012292no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020012293 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012294 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12295 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012296 often makes sense to disable it when communicating with local servers. This
12297 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012298 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012299
12300 Supported in default-server: No
12301
12302no-tlsv13
12303 This option disables support for TLSv1.3 when SSL is used to communicate with
12304 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12305 using any configuration option. TLSv1 is more expensive than SSLv3 so it
12306 often makes sense to disable it when communicating with local servers. This
12307 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012308 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012309
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012310 Supported in default-server: No
12311
Frédéric Lécailled2376272017-03-21 18:52:12 +010012312no-verifyhost
12313 This option may be used as "server" setting to reset any "verifyhost"
12314 setting which would have been inherited from "default-server" directive as
12315 default value.
12316 It may also be used as "default-server" setting to reset any previous
12317 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012318
Simon Hormanfa461682011-06-25 09:39:49 +090012319non-stick
12320 Never add connections allocated to this sever to a stick-table.
12321 This may be used in conjunction with backup to ensure that
12322 stick-table persistence is disabled for backup servers.
12323
Olivier Houchardc7566002018-11-20 23:33:50 +010012324npn <protocols>
12325 This enables the NPN TLS extension and advertises the specified protocol list
12326 as supported on top of NPN. The protocol list consists in a comma-delimited
12327 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
12328 This requires that the SSL library is build with support for TLS extensions
12329 enabled (check with haproxy -vv). Note that the NPN extension has been
12330 replaced with the ALPN extension (see the "alpn" keyword), though this one is
12331 only available starting with OpenSSL 1.0.2.
12332
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012333observe <mode>
12334 This option enables health adjusting based on observing communication with
12335 the server. By default this functionality is disabled and enabling it also
12336 requires to enable health checks. There are two supported modes: "layer4" and
12337 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
12338 significant. In layer7, which is only allowed for http proxies, responses
12339 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010012340 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012341
12342 See also the "check", "on-error" and "error-limit".
12343
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012344on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012345 Select what should happen when enough consecutive errors are detected.
12346 Currently, four modes are available:
12347 - fastinter: force fastinter
12348 - fail-check: simulate a failed check, also forces fastinter (default)
12349 - sudden-death: simulate a pre-fatal failed health check, one more failed
12350 check will mark a server down, forces fastinter
12351 - mark-down: mark the server immediately down and force fastinter
12352
12353 See also the "check", "observe" and "error-limit".
12354
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012355on-marked-down <action>
12356 Modify what occurs when a server is marked down.
12357 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012358 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
12359 all connections to the server are immediately terminated when the server
12360 goes down. It might be used if the health check detects more complex cases
12361 than a simple connection status, and long timeouts would cause the service
12362 to remain unresponsive for too long a time. For instance, a health check
12363 might detect that a database is stuck and that there's no chance to reuse
12364 existing connections anymore. Connections killed this way are logged with
12365 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012366
12367 Actions are disabled by default
12368
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012369on-marked-up <action>
12370 Modify what occurs when a server is marked up.
12371 Currently one action is available:
12372 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
12373 done only if the server is not in backup state and if it is not disabled
12374 (it must have an effective weight > 0). This can be used sometimes to force
12375 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012376 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
12377 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012378 with extreme care. Sessions killed because a server comes up are logged
12379 with an 'U' termination code (for "Up").
12380
12381 Actions are disabled by default
12382
Olivier Houchard006e3102018-12-10 18:30:32 +010012383pool-max-conn <max>
12384 Set the maximum number of idling connections for a server. -1 means unlimited
12385 connections, 0 means no idle connections. The default is -1. When idle
12386 connections are enabled, orphaned idle connections which do not belong to any
12387 client session anymore are moved to a dedicated pool so that they remain
12388 usable by future clients. This only applies to connections that can be shared
12389 according to the same principles as those applying to "http-reuse".
12390
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012391pool-purge-delay <delay>
12392 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010012393 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020012394 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012395
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012396port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012397 Using the "port" parameter, it becomes possible to use a different port to
12398 send health-checks. On some servers, it may be desirable to dedicate a port
12399 to a specific component able to perform complex tests which are more suitable
12400 to health-checks than the application. It is common to run a simple script in
12401 inetd for instance. This parameter is ignored if the "check" parameter is not
12402 set. See also the "addr" parameter.
12403
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020012404proto <name>
12405
12406 Forces the multiplexer's protocol to use for the outgoing connections to this
12407 server. It must be compatible with the mode of the backend (TCP or HTTP). It
12408 must also be usable on the backend side. The list of available protocols is
12409 reported in haproxy -vv.
12410 Idea behind this optipon is to bypass the selection of the best multiplexer's
12411 protocol for all connections established to this server.
12412
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012413redir <prefix>
12414 The "redir" parameter enables the redirection mode for all GET and HEAD
12415 requests addressing this server. This means that instead of having HAProxy
12416 forward the request to the server, it will send an "HTTP 302" response with
12417 the "Location" header composed of this prefix immediately followed by the
12418 requested URI beginning at the leading '/' of the path component. That means
12419 that no trailing slash should be used after <prefix>. All invalid requests
12420 will be rejected, and all non-GET or HEAD requests will be normally served by
12421 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012422 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012423 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012424 users to a remote location in case of local disaster. Main use consists in
12425 increasing bandwidth for static servers by having the clients directly
12426 connect to them. Note: never use a relative location here, it would cause a
12427 loop between the client and HAProxy!
12428
12429 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
12430
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012431rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012432 The "rise" parameter states that a server will be considered as operational
12433 after <count> consecutive successful health checks. This value defaults to 2
12434 if unspecified. See also the "check", "inter" and "fall" parameters.
12435
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012436resolve-opts <option>,<option>,...
12437 Comma separated list of options to apply to DNS resolution linked to this
12438 server.
12439
12440 Available options:
12441
12442 * allow-dup-ip
12443 By default, HAProxy prevents IP address duplication in a backend when DNS
12444 resolution at runtime is in operation.
12445 That said, for some cases, it makes sense that two servers (in the same
12446 backend, being resolved by the same FQDN) have the same IP address.
12447 For such case, simply enable this option.
12448 This is the opposite of prevent-dup-ip.
12449
12450 * prevent-dup-ip
12451 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
12452 an IP address already set to a server in the same backend and sharing the
12453 same fqdn.
12454 This is the opposite of allow-dup-ip.
12455
12456 Example:
12457 backend b_myapp
12458 default-server init-addr none resolvers dns
12459 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
12460 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
12461
12462 With the option allow-dup-ip set:
12463 * if the nameserver returns a single IP address, then both servers will use
12464 it
12465 * If the nameserver returns 2 IP addresses, then each server will pick up a
12466 different address
12467
12468 Default value: not set
12469
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012470resolve-prefer <family>
12471 When DNS resolution is enabled for a server and multiple IP addresses from
12472 different families are returned, HAProxy will prefer using an IP address
12473 from the family mentioned in the "resolve-prefer" parameter.
12474 Available families: "ipv4" and "ipv6"
12475
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020012476 Default value: ipv6
12477
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012478 Example:
12479
12480 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012481
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012482resolve-net <network>[,<network[,...]]
12483 This options prioritize th choice of an ip address matching a network. This is
12484 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010012485 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010012486 different datacenters. The latency between datacenter is not negligible, so
12487 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012488 configured network, another address is selected.
12489
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012490 Example:
12491
12492 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012493
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012494resolvers <id>
12495 Points to an existing "resolvers" section to resolve current server's
12496 hostname.
12497
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012498 Example:
12499
12500 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012501
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012502 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012503
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012504send-proxy
12505 The "send-proxy" parameter enforces use of the PROXY protocol over any
12506 connection established to this server. The PROXY protocol informs the other
12507 end about the layer 3/4 addresses of the incoming connection, so that it can
12508 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010012509 upper layer protocol. For connections accepted by an "accept-proxy" or
12510 "accept-netscaler-cip" listener, the advertised address will be used. Only
12511 TCPv4 and TCPv6 address families are supported. Other families such as
12512 Unix sockets, will report an UNKNOWN family. Servers using this option can
12513 fully be chained to another instance of haproxy listening with an
12514 "accept-proxy" setting. This setting must not be used if the server isn't
12515 aware of the protocol. When health checks are sent to the server, the PROXY
12516 protocol is automatically used when this option is set, unless there is an
12517 explicit "port" or "addr" directive, in which case an explicit
12518 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012519 See also the "no-send-proxy" option of this section and "accept-proxy" and
12520 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012521
David Safb76832014-05-08 23:42:08 -040012522send-proxy-v2
12523 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
12524 over any connection established to this server. The PROXY protocol informs
12525 the other end about the layer 3/4 addresses of the incoming connection, so
12526 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020012527 whatever the upper layer protocol. It also send ALPN information if an alpn
12528 have been negotiated. This setting must not be used if the server isn't aware
12529 of this version of the protocol. See also the "no-send-proxy-v2" option of
12530 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012531
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012532proxy-v2-options <option>[,<option>]*
12533 The "proxy-v2-options" parameter add option to send in PROXY protocol version
12534 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010012535 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
12536 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010012537 certificate, "cert-key": key algorithm of the used certificate), "authority":
12538 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010012539 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012540
David Safb76832014-05-08 23:42:08 -040012541send-proxy-v2-ssl
12542 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12543 2 over any connection established to this server. The PROXY protocol informs
12544 the other end about the layer 3/4 addresses of the incoming connection, so
12545 that it can know the client's address or the public address it accessed to,
12546 whatever the upper layer protocol. In addition, the SSL information extension
12547 of the PROXY protocol is added to the PROXY protocol header. This setting
12548 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012549 See also the "no-send-proxy-v2-ssl" option of this section and the
12550 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012551
12552send-proxy-v2-ssl-cn
12553 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12554 2 over any connection established to this server. The PROXY protocol informs
12555 the other end about the layer 3/4 addresses of the incoming connection, so
12556 that it can know the client's address or the public address it accessed to,
12557 whatever the upper layer protocol. In addition, the SSL information extension
12558 of the PROXY protocol, along along with the Common Name from the subject of
12559 the client certificate (if any), is added to the PROXY protocol header. This
12560 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012561 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
12562 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012563
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012564slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012565 The "slowstart" parameter for a server accepts a value in milliseconds which
12566 indicates after how long a server which has just come back up will run at
12567 full speed. Just as with every other time-based parameter, it can be entered
12568 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
12569 linearly from 0 to 100% during this time. The limitation applies to two
12570 parameters :
12571
12572 - maxconn: the number of connections accepted by the server will grow from 1
12573 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
12574
12575 - weight: when the backend uses a dynamic weighted algorithm, the weight
12576 grows linearly from 1 to 100%. In this case, the weight is updated at every
12577 health-check. For this reason, it is important that the "inter" parameter
12578 is smaller than the "slowstart", in order to maximize the number of steps.
12579
12580 The slowstart never applies when haproxy starts, otherwise it would cause
12581 trouble to running servers. It only applies when a server has been previously
12582 seen as failed.
12583
Willy Tarreau732eac42015-07-09 11:40:25 +020012584sni <expression>
12585 The "sni" parameter evaluates the sample fetch expression, converts it to a
12586 string and uses the result as the host name sent in the SNI TLS extension to
12587 the server. A typical use case is to send the SNI received from the client in
12588 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020012589 expression, though alternatives such as req.hdr(host) can also make sense. If
12590 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012591 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010012592 "verify" directive for more details. If you want to set a SNI for health
12593 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020012594
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012595source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020012596source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012597source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012598 The "source" parameter sets the source address which will be used when
12599 connecting to the server. It follows the exact same parameters and principle
12600 as the backend "source" keyword, except that it only applies to the server
12601 referencing it. Please consult the "source" keyword for details.
12602
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012603 Additionally, the "source" statement on a server line allows one to specify a
12604 source port range by indicating the lower and higher bounds delimited by a
12605 dash ('-'). Some operating systems might require a valid IP address when a
12606 source port range is specified. It is permitted to have the same IP/range for
12607 several servers. Doing so makes it possible to bypass the maximum of 64k
12608 total concurrent connections. The limit will then reach 64k connections per
12609 server.
12610
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000012611 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
12612 specifying the source address without port(s).
12613
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012614ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020012615 This option enables SSL ciphering on outgoing connections to the server. It
12616 is critical to verify server certificates using "verify" when using SSL to
12617 connect to servers, otherwise the communication is prone to trivial man in
12618 the-middle attacks rendering SSL useless. When this option is used, health
12619 checks are automatically sent in SSL too unless there is a "port" or an
12620 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012621 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
12622 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012623
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012624ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12625 This option enforces use of <version> or lower when SSL is used to communicate
12626 with the server. This option is also available on global statement
12627 "ssl-default-server-options". See also "ssl-min-ver".
12628
12629ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12630 This option enforces use of <version> or upper when SSL is used to communicate
12631 with the server. This option is also available on global statement
12632 "ssl-default-server-options". See also "ssl-max-ver".
12633
Frédéric Lécailled2376272017-03-21 18:52:12 +010012634ssl-reuse
12635 This option may be used as "server" setting to reset any "no-ssl-reuse"
12636 setting which would have been inherited from "default-server" directive as
12637 default value.
12638 It may also be used as "default-server" setting to reset any previous
12639 "default-server" "no-ssl-reuse" setting.
12640
12641stick
12642 This option may be used as "server" setting to reset any "non-stick"
12643 setting which would have been inherited from "default-server" directive as
12644 default value.
12645 It may also be used as "default-server" setting to reset any previous
12646 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012647
Alexander Liu2a54bb72019-05-22 19:44:48 +080012648socks4 <addr>:<port>
12649 This option enables upstream socks4 tunnel for outgoinng connections to the
12650 server. Using this option won't force the health check to go via socks4 by
12651 default. You will have to use the keyword "check-via-socks4" to enable it.
12652
Willy Tarreau163d4622015-10-13 16:16:41 +020012653tcp-ut <delay>
12654 Sets the TCP User Timeout for all outgoing connections to this server. This
12655 option is available on Linux since version 2.6.37. It allows haproxy to
12656 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012657 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020012658 long-lived connections experiencing long idle periods such as remote
12659 terminals or database connection pools, where the client and server timeouts
12660 must remain high to allow a long period of idle, but where it is important to
12661 detect that the server has disappeared in order to release all resources
12662 associated with its connection (and the client's session). One typical use
12663 case is also to force dead server connections to die when health checks are
12664 too slow or during a soft reload since health checks are then disabled. The
12665 argument is a delay expressed in milliseconds by default. This only works for
12666 regular TCP connections, and is ignored for other protocols.
12667
Willy Tarreau034c88c2017-01-23 23:36:45 +010012668tfo
12669 This option enables using TCP fast open when connecting to servers, on
12670 systems that support it (currently only the Linux kernel >= 4.11).
12671 See the "tfo" bind option for more information about TCP fast open.
12672 Please note that when using tfo, you should also use the "conn-failure",
12673 "empty-response" and "response-timeout" keywords for "retry-on", or haproxy
12674 won't be able to retry the connection on failure.
12675
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012676track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020012677 This option enables ability to set the current state of the server by tracking
12678 another one. It is possible to track a server which itself tracks another
12679 server, provided that at the end of the chain, a server has health checks
12680 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012681 used, it has to be enabled on both proxies.
12682
Frédéric Lécailled2376272017-03-21 18:52:12 +010012683tls-tickets
12684 This option may be used as "server" setting to reset any "no-tls-tickets"
12685 setting which would have been inherited from "default-server" directive as
12686 default value.
12687 It may also be used as "default-server" setting to reset any previous
12688 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012689
Emeric Brunef42d922012-10-11 16:11:36 +020012690verify [none|required]
12691 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010012692 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012693 certificate provided by the server is verified using CAs from 'ca-file' and
12694 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012695 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012696 the name passed using the "sni" directive, or if not provided, the static
12697 host name passed using the "verifyhost" directive. When no name is found, the
12698 certificate's names are ignored. For this reason, without SNI it's important
12699 to use "verifyhost". On verification failure the handshake is aborted. It is
12700 critically important to verify server certificates when using SSL to connect
12701 to servers, otherwise the communication is prone to trivial man-in-the-middle
12702 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
12703 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020012704
Evan Broderbe554312013-06-27 00:05:25 -070012705verifyhost <hostname>
12706 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012707 only takes effect if 'verify required' is also specified. This directive sets
12708 a default static hostname to check the server's certificate against when no
12709 SNI was used to connect to the server. If SNI is not used, this is the only
12710 way to enable hostname verification. This static hostname, when set, will
12711 also be used for health checks (which cannot provide an SNI value). If none
12712 of the hostnames in the certificate match the specified hostname, the
12713 handshake is aborted. The hostnames in the server-provided certificate may
12714 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070012715
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012716weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012717 The "weight" parameter is used to adjust the server's weight relative to
12718 other servers. All servers will receive a load proportional to their weight
12719 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020012720 load. The default weight is 1, and the maximal value is 256. A value of 0
12721 means the server will not participate in load-balancing but will still accept
12722 persistent connections. If this parameter is used to distribute the load
12723 according to server's capacity, it is recommended to start with values which
12724 can both grow and shrink, for instance between 10 and 100 to leave enough
12725 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012726
12727
Cyril Bonté46175dd2015-07-02 22:45:32 +0200127285.3. Server IP address resolution using DNS
12729-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012730
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012731HAProxy allows using a host name on the server line to retrieve its IP address
12732using name servers. By default, HAProxy resolves the name when parsing the
12733configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012734This is not sufficient in some cases, such as in Amazon where a server's IP
12735can change after a reboot or an ELB Virtual IP can change based on current
12736workload.
12737This chapter describes how HAProxy can be configured to process server's name
12738resolution at run time.
12739Whether run time server name resolution has been enable or not, HAProxy will
12740carry on doing the first resolution when parsing the configuration.
12741
12742
Cyril Bonté46175dd2015-07-02 22:45:32 +0200127435.3.1. Global overview
12744----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012745
12746As we've seen in introduction, name resolution in HAProxy occurs at two
12747different steps of the process life:
12748
12749 1. when starting up, HAProxy parses the server line definition and matches a
12750 host name. It uses libc functions to get the host name resolved. This
12751 resolution relies on /etc/resolv.conf file.
12752
Christopher Faulet67957bd2017-09-27 11:00:59 +020012753 2. at run time, HAProxy performs periodically name resolutions for servers
12754 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012755
12756A few other events can trigger a name resolution at run time:
12757 - when a server's health check ends up in a connection timeout: this may be
12758 because the server has a new IP address. So we need to trigger a name
12759 resolution to know this new IP.
12760
Christopher Faulet67957bd2017-09-27 11:00:59 +020012761When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012762HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020012763SRV label is specified, then the corresponding SRV records will be retrieved
12764from the DNS server, and the provided hostnames will be used. The SRV label
12765will be checked periodically, and if any server are added or removed, haproxy
12766will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020012767
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012768A few things important to notice:
12769 - all the name servers are queried in the mean time. HAProxy will process the
12770 first valid response.
12771
12772 - a resolution is considered as invalid (NX, timeout, refused), when all the
12773 servers return an error.
12774
12775
Cyril Bonté46175dd2015-07-02 22:45:32 +0200127765.3.2. The resolvers section
12777----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012778
12779This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020012780HAProxy. There can be as many as resolvers section as needed. Each section can
12781contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012782
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012783When multiple name servers are configured in a resolvers section, then HAProxy
12784uses the first valid response. In case of invalid responses, only the last one
12785is treated. Purpose is to give the chance to a slow server to deliver a valid
12786answer after a fast faulty or outdated server.
12787
12788When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020012789used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012790
Christopher Faulet67957bd2017-09-27 11:00:59 +020012791 1. HAProxy retries the same DNS query with a new query type. The A queries are
12792 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
12793 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012794
Christopher Faulet67957bd2017-09-27 11:00:59 +020012795 2. When the fallback on the query type was done (or not applicable), HAProxy
12796 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012797
Christopher Faulet67957bd2017-09-27 11:00:59 +020012798 3. HAProxy retries previous steps <resolve_retires> times. If no valid
12799 response is received after that, it stops the DNS resolution and reports
12800 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012801
Christopher Faulet67957bd2017-09-27 11:00:59 +020012802For example, with 2 name servers configured in a resolvers section, the
12803following scenarios are possible:
12804
12805 - First response is valid and is applied directly, second response is
12806 ignored
12807
12808 - First response is invalid and second one is valid, then second response is
12809 applied
12810
12811 - First response is a NX domain and second one a truncated response, then
12812 HAProxy retries the query with a new type
12813
12814 - First response is a NX domain and second one is a timeout, then HAProxy
12815 retries the query with a new type
12816
12817 - Query timed out for both name servers, then HAProxy retries it with the
12818 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012819
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012820As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
12821a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020012822<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012823
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012824
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012825resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012826 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012827
12828A resolvers section accept the following parameters:
12829
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012830accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012831 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020012832 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012833 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
12834 by RFC 6891)
12835
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020012836 Note: the maximum allowed value is 8192.
12837
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012838nameserver <id> <ip>:<port>
12839 DNS server description:
12840 <id> : label of the server, should be unique
12841 <ip> : IP address of the server
12842 <port> : port where the DNS service actually runs
12843
Ben Draut44e609b2018-05-29 15:40:08 -060012844parse-resolv-conf
12845 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
12846 list. Ordered as if each nameserver in /etc/resolv.conf was individually
12847 placed in the resolvers section in place of this directive.
12848
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012849hold <status> <period>
12850 Defines <period> during which the last name resolution should be kept based
12851 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012852 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012853 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012854 <period> : interval between two successive name resolution when the last
12855 answer was in <status>. It follows the HAProxy time format.
12856 <period> is in milliseconds by default.
12857
Baptiste Assmann686408b2017-08-18 10:15:42 +020012858 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012859
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012860resolve_retries <nb>
12861 Defines the number <nb> of queries to send to resolve a server name before
12862 giving up.
12863 Default value: 3
12864
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012865 A retry occurs on name server timeout or when the full sequence of DNS query
12866 type failover is over and we need to start up from the default ANY query
12867 type.
12868
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012869timeout <event> <time>
12870 Defines timeouts related to name resolution
12871 <event> : the event on which the <time> timeout period applies to.
12872 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012873 - resolve : default time to trigger name resolutions when no
12874 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012875 Default value: 1s
12876 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012877 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012878 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012879 <time> : time related to the event. It follows the HAProxy time format.
12880 <time> is expressed in milliseconds.
12881
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012882 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012883
12884 resolvers mydns
12885 nameserver dns1 10.0.0.1:53
12886 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060012887 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012888 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020012889 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012890 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012891 hold other 30s
12892 hold refused 30s
12893 hold nx 30s
12894 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012895 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012896 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012897
12898
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128996. HTTP header manipulation
12900---------------------------
12901
12902In HTTP mode, it is possible to rewrite, add or delete some of the request and
12903response headers based on regular expressions. It is also possible to block a
12904request or a response if a particular header matches a regular expression,
12905which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010012906against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012907
Willy Tarreau70dffda2014-01-30 03:07:23 +010012908If HAProxy encounters an "Informational Response" (status code 1xx), it is able
12909to process all rsp* rules which can allow, deny, rewrite or delete a header,
12910but it will refuse to add a header to any such messages as this is not
12911HTTP-compliant. The reason for still processing headers in such responses is to
12912stop and/or fix any possible information leak which may happen, for instance
12913because another downstream equipment would unconditionally add a header, or if
12914a server name appears there. When such messages are seen, normal processing
12915still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020012916
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012917This section covers common usage of the following keywords, described in detail
12918in section 4.2 :
12919
12920 - reqadd <string>
12921 - reqallow <search>
12922 - reqiallow <search>
12923 - reqdel <search>
12924 - reqidel <search>
12925 - reqdeny <search>
12926 - reqideny <search>
12927 - reqpass <search>
12928 - reqipass <search>
12929 - reqrep <search> <replace>
12930 - reqirep <search> <replace>
12931 - reqtarpit <search>
12932 - reqitarpit <search>
12933 - rspadd <string>
12934 - rspdel <search>
12935 - rspidel <search>
12936 - rspdeny <search>
12937 - rspideny <search>
12938 - rsprep <search> <replace>
12939 - rspirep <search> <replace>
12940
12941With all these keywords, the same conventions are used. The <search> parameter
12942is a POSIX extended regular expression (regex) which supports grouping through
12943parenthesis (without the backslash). Spaces and other delimiters must be
12944prefixed with a backslash ('\') to avoid confusion with a field delimiter.
12945Other characters may be prefixed with a backslash to change their meaning :
12946
12947 \t for a tab
12948 \r for a carriage return (CR)
12949 \n for a new line (LF)
12950 \ to mark a space and differentiate it from a delimiter
12951 \# to mark a sharp and differentiate it from a comment
12952 \\ to use a backslash in a regex
12953 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
12954 \xXX to write the ASCII hex code XX as in the C language
12955
12956The <replace> parameter contains the string to be used to replace the largest
12957portion of text matching the regex. It can make use of the special characters
12958above, and can reference a substring which is delimited by parenthesis in the
12959regex, by writing a backslash ('\') immediately followed by one digit from 0 to
129609 indicating the group position (0 designating the entire line). This practice
12961is very common to users of the "sed" program.
12962
12963The <string> parameter represents the string which will systematically be added
12964after the last header line. It can also use special character sequences above.
12965
12966Notes related to these keywords :
12967---------------------------------
12968 - these keywords are not always convenient to allow/deny based on header
12969 contents. It is strongly recommended to use ACLs with the "block" keyword
12970 instead, resulting in far more flexible and manageable rules.
12971
12972 - lines are always considered as a whole. It is not possible to reference
12973 a header name only or a value only. This is important because of the way
12974 headers are written (notably the number of spaces after the colon).
12975
12976 - the first line is always considered as a header, which makes it possible to
12977 rewrite or filter HTTP requests URIs or response codes, but in turn makes
12978 it harder to distinguish between headers and request line. The regex prefix
12979 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
12980 ^[^ \t:]*: matches any header name followed by a colon.
12981
12982 - for performances reasons, the number of characters added to a request or to
12983 a response is limited at build time to values between 1 and 4 kB. This
12984 should normally be far more than enough for most usages. If it is too short
12985 on occasional usages, it is possible to gain some space by removing some
12986 useless headers before adding new ones.
12987
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012988 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012989 without the 'i' letter except that they ignore case when matching patterns.
12990
12991 - when a request passes through a frontend then a backend, all req* rules
12992 from the frontend will be evaluated, then all req* rules from the backend
12993 will be evaluated. The reverse path is applied to responses.
12994
12995 - req* statements are applied after "block" statements, so that "block" is
12996 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010012997 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012998
12999
Willy Tarreau74ca5042013-06-11 23:12:07 +0200130007. Using ACLs and fetching samples
13001----------------------------------
13002
Davor Ocelice9ed2812017-12-25 17:49:28 +010013003HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020013004client or server information, from tables, environmental information etc...
13005The action of extracting such data is called fetching a sample. Once retrieved,
13006these samples may be used for various purposes such as a key to a stick-table,
13007but most common usages consist in matching them against predefined constant
13008data called patterns.
13009
13010
130117.1. ACL basics
13012---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013013
13014The use of Access Control Lists (ACL) provides a flexible solution to perform
13015content switching and generally to take decisions based on content extracted
13016from the request, the response or any environmental status. The principle is
13017simple :
13018
Willy Tarreau74ca5042013-06-11 23:12:07 +020013019 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013020 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020013021 - apply one or multiple pattern matching methods on this sample
13022 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013023
Willy Tarreau74ca5042013-06-11 23:12:07 +020013024The actions generally consist in blocking a request, selecting a backend, or
13025adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013026
13027In order to define a test, the "acl" keyword is used. The syntax is :
13028
Willy Tarreau74ca5042013-06-11 23:12:07 +020013029 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013030
13031This creates a new ACL <aclname> or completes an existing one with new tests.
13032Those tests apply to the portion of request/response specified in <criterion>
13033and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013034an operator which may be specified before the set of values. Optionally some
13035conversion operators may be applied to the sample, and they will be specified
13036as a comma-delimited list of keywords just after the first keyword. The values
13037are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013038
13039ACL names must be formed from upper and lower case letters, digits, '-' (dash),
13040'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
13041which means that "my_acl" and "My_Acl" are two different ACLs.
13042
13043There is no enforced limit to the number of ACLs. The unused ones do not affect
13044performance, they just consume a small amount of memory.
13045
Willy Tarreau74ca5042013-06-11 23:12:07 +020013046The criterion generally is the name of a sample fetch method, or one of its ACL
13047specific declinations. The default test method is implied by the output type of
13048this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013049methods of a same sample fetch method. The sample fetch methods are the only
13050ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013051
13052Sample fetch methods return data which can be of the following types :
13053 - boolean
13054 - integer (signed or unsigned)
13055 - IPv4 or IPv6 address
13056 - string
13057 - data block
13058
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013059Converters transform any of these data into any of these. For example, some
13060converters might convert a string to a lower-case string while other ones
13061would turn a string to an IPv4 address, or apply a netmask to an IP address.
13062The resulting sample is of the type of the last converter applied to the list,
13063which defaults to the type of the sample fetch method.
13064
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013065Each sample or converter returns data of a specific type, specified with its
13066keyword in this documentation. When an ACL is declared using a standard sample
13067fetch method, certain types automatically involved a default matching method
13068which are summarized in the table below :
13069
13070 +---------------------+-----------------+
13071 | Sample or converter | Default |
13072 | output type | matching method |
13073 +---------------------+-----------------+
13074 | boolean | bool |
13075 +---------------------+-----------------+
13076 | integer | int |
13077 +---------------------+-----------------+
13078 | ip | ip |
13079 +---------------------+-----------------+
13080 | string | str |
13081 +---------------------+-----------------+
13082 | binary | none, use "-m" |
13083 +---------------------+-----------------+
13084
13085Note that in order to match a binary samples, it is mandatory to specify a
13086matching method, see below.
13087
Willy Tarreau74ca5042013-06-11 23:12:07 +020013088The ACL engine can match these types against patterns of the following types :
13089 - boolean
13090 - integer or integer range
13091 - IP address / network
13092 - string (exact, substring, suffix, prefix, subdir, domain)
13093 - regular expression
13094 - hex block
13095
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013096The following ACL flags are currently supported :
13097
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013098 -i : ignore case during matching of all subsequent patterns.
13099 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013100 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013101 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013102 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013103 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020013104 -- : force end of flags. Useful when a string looks like one of the flags.
13105
Willy Tarreau74ca5042013-06-11 23:12:07 +020013106The "-f" flag is followed by the name of a file from which all lines will be
13107read as individual values. It is even possible to pass multiple "-f" arguments
13108if the patterns are to be loaded from multiple files. Empty lines as well as
13109lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
13110will be stripped. If it is absolutely necessary to insert a valid pattern
13111beginning with a sharp, just prefix it with a space so that it is not taken for
13112a comment. Depending on the data type and match method, haproxy may load the
13113lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
13114exact string matching. In this case, duplicates will automatically be removed.
13115
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013116The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
13117parsed as two column file. The first column contains the patterns used by the
13118ACL, and the second column contain the samples. The sample can be used later by
13119a map. This can be useful in some rare cases where an ACL would just be used to
13120check for the existence of a pattern in a map before a mapping is applied.
13121
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013122The "-u" flag forces the unique id of the ACL. This unique id is used with the
13123socket interface to identify ACL and dynamically change its values. Note that a
13124file is always identified by its name even if an id is set.
13125
Willy Tarreau74ca5042013-06-11 23:12:07 +020013126Also, note that the "-i" flag applies to subsequent entries and not to entries
13127loaded from files preceding it. For instance :
13128
13129 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
13130
13131In this example, each line of "exact-ua.lst" will be exactly matched against
13132the "user-agent" header of the request. Then each line of "generic-ua" will be
13133case-insensitively matched. Then the word "test" will be insensitively matched
13134as well.
13135
13136The "-m" flag is used to select a specific pattern matching method on the input
13137sample. All ACL-specific criteria imply a pattern matching method and generally
13138do not need this flag. However, this flag is useful with generic sample fetch
13139methods to describe how they're going to be matched against the patterns. This
13140is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010013141obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013142followed by a pattern matching method name, this method is used instead of the
13143default one for the criterion. This makes it possible to match contents in ways
13144that were not initially planned, or with sample fetch methods which return a
13145string. The matching method also affects the way the patterns are parsed.
13146
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013147The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
13148By default, if the parser cannot parse ip address it considers that the parsed
13149string is maybe a domain name and try dns resolution. The flag "-n" disable this
13150resolution. It is useful for detecting malformed ip lists. Note that if the DNS
13151server is not reachable, the haproxy configuration parsing may last many minutes
13152waiting fir the timeout. During this time no error messages are displayed. The
13153flag "-n" disable this behavior. Note also that during the runtime, this
13154function is disabled for the dynamic acl modifications.
13155
Willy Tarreau74ca5042013-06-11 23:12:07 +020013156There are some restrictions however. Not all methods can be used with all
13157sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
13158be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020013159
13160 - "found" : only check if the requested sample could be found in the stream,
13161 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020013162 to pass any pattern to avoid confusion. This matching method is
13163 particularly useful to detect presence of certain contents such
13164 as headers, cookies, etc... even if they are empty and without
13165 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013166
13167 - "bool" : check the value as a boolean. It can only be applied to fetches
13168 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013169 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013170
13171 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013172 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013173
13174 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020013175 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013176
Davor Ocelice9ed2812017-12-25 17:49:28 +010013177 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020013178 binary sequence. This may be used with binary or string samples.
13179
13180 - "len" : match the sample's length as an integer. This may be used with
13181 binary or string samples.
13182
Willy Tarreau74ca5042013-06-11 23:12:07 +020013183 - "str" : exact match : match the contents against a string. This may be
13184 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013185
Willy Tarreau74ca5042013-06-11 23:12:07 +020013186 - "sub" : substring match : check that the contents contain at least one of
13187 the provided string patterns. This may be used with binary or
13188 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013189
Willy Tarreau74ca5042013-06-11 23:12:07 +020013190 - "reg" : regex match : match the contents against a list of regular
13191 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013192
Willy Tarreau74ca5042013-06-11 23:12:07 +020013193 - "beg" : prefix match : check that the contents begin like the provided
13194 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013195
Willy Tarreau74ca5042013-06-11 23:12:07 +020013196 - "end" : suffix match : check that the contents end like the provided
13197 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013198
Willy Tarreau74ca5042013-06-11 23:12:07 +020013199 - "dir" : subdir match : check that a slash-delimited portion of the
13200 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013201 This may be used with binary or string samples.
13202
Willy Tarreau74ca5042013-06-11 23:12:07 +020013203 - "dom" : domain match : check that a dot-delimited portion of the contents
13204 exactly match one of the provided string patterns. This may be
13205 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013206
13207For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
13208request, it is possible to do :
13209
13210 acl jsess_present cook(JSESSIONID) -m found
13211
13212In order to apply a regular expression on the 500 first bytes of data in the
13213buffer, one would use the following acl :
13214
13215 acl script_tag payload(0,500) -m reg -i <script>
13216
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013217On systems where the regex library is much slower when using "-i", it is
13218possible to convert the sample to lowercase before matching, like this :
13219
13220 acl script_tag payload(0,500),lower -m reg <script>
13221
Willy Tarreau74ca5042013-06-11 23:12:07 +020013222All ACL-specific criteria imply a default matching method. Most often, these
13223criteria are composed by concatenating the name of the original sample fetch
13224method and the matching method. For example, "hdr_beg" applies the "beg" match
13225to samples retrieved using the "hdr" fetch method. Since all ACL-specific
13226criteria rely on a sample fetch method, it is always possible instead to use
13227the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013228
Willy Tarreau74ca5042013-06-11 23:12:07 +020013229If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013230the matching method is simply applied to the underlying sample fetch method.
13231For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013232
Willy Tarreau74ca5042013-06-11 23:12:07 +020013233 acl short_form hdr_beg(host) www.
13234 acl alternate1 hdr_beg(host) -m beg www.
13235 acl alternate2 hdr_dom(host) -m beg www.
13236 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013237
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013238
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013239The table below summarizes the compatibility matrix between sample or converter
13240types and the pattern types to fetch against. It indicates for each compatible
13241combination the name of the matching method to be used, surrounded with angle
13242brackets ">" and "<" when the method is the default one and will work by
13243default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013244
Willy Tarreau74ca5042013-06-11 23:12:07 +020013245 +-------------------------------------------------+
13246 | Input sample type |
13247 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013248 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013249 +----------------------+---------+---------+---------+---------+---------+
13250 | none (presence only) | found | found | found | found | found |
13251 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013252 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013253 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013254 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013255 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013256 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013257 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013258 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013259 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013260 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013261 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013262 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013263 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013264 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013265 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013266 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013267 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013268 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013269 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013270 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013271 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013272 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013273 +----------------------+---------+---------+---------+---------+---------+
13274 | hex block | | | | bin | bin |
13275 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020013276
13277
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132787.1.1. Matching booleans
13279------------------------
13280
13281In order to match a boolean, no value is needed and all values are ignored.
13282Boolean matching is used by default for all fetch methods of type "boolean".
13283When boolean matching is used, the fetched value is returned as-is, which means
13284that a boolean "true" will always match and a boolean "false" will never match.
13285
13286Boolean matching may also be enforced using "-m bool" on fetch methods which
13287return an integer value. Then, integer value 0 is converted to the boolean
13288"false" and all other values are converted to "true".
13289
Willy Tarreau6a06a402007-07-15 20:15:28 +020013290
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132917.1.2. Matching integers
13292------------------------
13293
13294Integer matching applies by default to integer fetch methods. It can also be
13295enforced on boolean fetches using "-m int". In this case, "false" is converted
13296to the integer 0, and "true" is converted to the integer 1.
13297
13298Integer matching also supports integer ranges and operators. Note that integer
13299matching only applies to positive values. A range is a value expressed with a
13300lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013301
13302For instance, "1024:65535" is a valid range to represent a range of
13303unprivileged ports, and "1024:" would also work. "0:1023" is a valid
13304representation of privileged ports, and ":1023" would also work.
13305
Willy Tarreau62644772008-07-16 18:36:06 +020013306As a special case, some ACL functions support decimal numbers which are in fact
13307two integers separated by a dot. This is used with some version checks for
13308instance. All integer properties apply to those decimal numbers, including
13309ranges and operators.
13310
Willy Tarreau6a06a402007-07-15 20:15:28 +020013311For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010013312operators with ranges does not make much sense and is strongly discouraged.
13313Similarly, it does not make much sense to perform order comparisons with a set
13314of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013315
Willy Tarreau0ba27502007-12-24 16:55:16 +010013316Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013317
13318 eq : true if the tested value equals at least one value
13319 ge : true if the tested value is greater than or equal to at least one value
13320 gt : true if the tested value is greater than at least one value
13321 le : true if the tested value is less than or equal to at least one value
13322 lt : true if the tested value is less than at least one value
13323
Willy Tarreau0ba27502007-12-24 16:55:16 +010013324For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013325
13326 acl negative-length hdr_val(content-length) lt 0
13327
Willy Tarreau62644772008-07-16 18:36:06 +020013328This one matches SSL versions between 3.0 and 3.1 (inclusive) :
13329
13330 acl sslv3 req_ssl_ver 3:3.1
13331
Willy Tarreau6a06a402007-07-15 20:15:28 +020013332
Willy Tarreau74ca5042013-06-11 23:12:07 +0200133337.1.3. Matching strings
13334-----------------------
13335
13336String matching applies to string or binary fetch methods, and exists in 6
13337different forms :
13338
13339 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013340 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013341
13342 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013343 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013344
13345 - prefix match (-m beg) : the patterns are compared with the beginning of
13346 the extracted string, and the ACL matches if any of them matches.
13347
13348 - suffix match (-m end) : the patterns are compared with the end of the
13349 extracted string, and the ACL matches if any of them matches.
13350
Baptiste Assmann33db6002016-03-06 23:32:10 +010013351 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020013352 string, delimited with slashes ("/"), and the ACL matches if any of them
13353 matches.
13354
13355 - domain match (-m dom) : the patterns are looked up inside the extracted
13356 string, delimited with dots ("."), and the ACL matches if any of them
13357 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013358
13359String matching applies to verbatim strings as they are passed, with the
13360exception of the backslash ("\") which makes it possible to escape some
13361characters such as the space. If the "-i" flag is passed before the first
13362string, then the matching will be performed ignoring the case. In order
13363to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010013364before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013365
13366
Willy Tarreau74ca5042013-06-11 23:12:07 +0200133677.1.4. Matching regular expressions (regexes)
13368---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013369
13370Just like with string matching, regex matching applies to verbatim strings as
13371they are passed, with the exception of the backslash ("\") which makes it
13372possible to escape some characters such as the space. If the "-i" flag is
13373passed before the first regex, then the matching will be performed ignoring
13374the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010013375the "--" flag before the first string. Same principle applies of course to
13376match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013377
13378
Willy Tarreau74ca5042013-06-11 23:12:07 +0200133797.1.5. Matching arbitrary data blocks
13380-------------------------------------
13381
13382It is possible to match some extracted samples against a binary block which may
13383not safely be represented as a string. For this, the patterns must be passed as
13384a series of hexadecimal digits in an even number, when the match method is set
13385to binary. Each sequence of two digits will represent a byte. The hexadecimal
13386digits may be used upper or lower case.
13387
13388Example :
13389 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
13390 acl hello payload(0,6) -m bin 48656c6c6f0a
13391
13392
133937.1.6. Matching IPv4 and IPv6 addresses
13394---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013395
13396IPv4 addresses values can be specified either as plain addresses or with a
13397netmask appended, in which case the IPv4 address matches whenever it is
13398within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010013399host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010013400difficult to read and debug configurations. If hostnames are used, you should
13401at least ensure that they are present in /etc/hosts so that the configuration
13402does not depend on any random DNS match at the moment the configuration is
13403parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013404
Daniel Schnellereba56342016-04-13 00:26:52 +020013405The dotted IPv4 address notation is supported in both regular as well as the
13406abbreviated form with all-0-octets omitted:
13407
13408 +------------------+------------------+------------------+
13409 | Example 1 | Example 2 | Example 3 |
13410 +------------------+------------------+------------------+
13411 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
13412 | 192.168.1 | 10.12 | 127.1 |
13413 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
13414 | 192.168.1/22 | 10.12/8 | 127.1/8 |
13415 +------------------+------------------+------------------+
13416
13417Notice that this is different from RFC 4632 CIDR address notation in which
13418192.168.42/24 would be equivalent to 192.168.42.0/24.
13419
Willy Tarreauceb4ac92012-04-28 00:41:46 +020013420IPv6 may be entered in their usual form, with or without a netmask appended.
13421Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
13422trouble with randomly resolved IP addresses, host names are never allowed in
13423IPv6 patterns.
13424
13425HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
13426following situations :
13427 - tested address is IPv4, pattern address is IPv4, the match applies
13428 in IPv4 using the supplied mask if any.
13429 - tested address is IPv6, pattern address is IPv6, the match applies
13430 in IPv6 using the supplied mask if any.
13431 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
13432 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
13433 ::IPV4 or ::ffff:IPV4, otherwise it fails.
13434 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
13435 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
13436 applied in IPv6 using the supplied IPv6 mask.
13437
Willy Tarreau74ca5042013-06-11 23:12:07 +020013438
134397.2. Using ACLs to form conditions
13440----------------------------------
13441
13442Some actions are only performed upon a valid condition. A condition is a
13443combination of ACLs with operators. 3 operators are supported :
13444
13445 - AND (implicit)
13446 - OR (explicit with the "or" keyword or the "||" operator)
13447 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020013448
Willy Tarreau74ca5042013-06-11 23:12:07 +020013449A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020013450
Willy Tarreau74ca5042013-06-11 23:12:07 +020013451 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020013452
Willy Tarreau74ca5042013-06-11 23:12:07 +020013453Such conditions are generally used after an "if" or "unless" statement,
13454indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020013455
Willy Tarreau74ca5042013-06-11 23:12:07 +020013456For instance, to block HTTP requests to the "*" URL with methods other than
13457"OPTIONS", as well as POST requests without content-length, and GET or HEAD
13458requests with a content-length greater than 0, and finally every request which
13459is not either GET/HEAD/POST/OPTIONS !
13460
13461 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013462 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
13463 http-request deny if METH_GET HTTP_CONTENT
13464 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020013465
13466To select a different backend for requests to static contents on the "www" site
13467and to every request on the "img", "video", "download" and "ftp" hosts :
13468
13469 acl url_static path_beg /static /images /img /css
13470 acl url_static path_end .gif .png .jpg .css .js
13471 acl host_www hdr_beg(host) -i www
13472 acl host_static hdr_beg(host) -i img. video. download. ftp.
13473
Davor Ocelice9ed2812017-12-25 17:49:28 +010013474 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020013475 # of host "www". Use backend "www" for the rest.
13476 use_backend static if host_static or host_www url_static
13477 use_backend www if host_www
13478
13479It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
13480expressions that are built on the fly without needing to be declared. They must
13481be enclosed between braces, with a space before and after each brace (because
13482the braces must be seen as independent words). Example :
13483
13484 The following rule :
13485
13486 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013487 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020013488
13489 Can also be written that way :
13490
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013491 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020013492
13493It is generally not recommended to use this construct because it's a lot easier
13494to leave errors in the configuration when written that way. However, for very
13495simple rules matching only one source IP address for instance, it can make more
13496sense to use them than to declare ACLs with random names. Another example of
13497good use is the following :
13498
13499 With named ACLs :
13500
13501 acl site_dead nbsrv(dynamic) lt 2
13502 acl site_dead nbsrv(static) lt 2
13503 monitor fail if site_dead
13504
13505 With anonymous ACLs :
13506
13507 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
13508
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013509See section 4.2 for detailed help on the "http-request deny" and "use_backend"
13510keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013511
13512
135137.3. Fetching samples
13514---------------------
13515
13516Historically, sample fetch methods were only used to retrieve data to match
13517against patterns using ACLs. With the arrival of stick-tables, a new class of
13518sample fetch methods was created, most often sharing the same syntax as their
13519ACL counterpart. These sample fetch methods are also known as "fetches". As
13520of now, ACLs and fetches have converged. All ACL fetch methods have been made
13521available as fetch methods, and ACLs may use any sample fetch method as well.
13522
13523This section details all available sample fetch methods and their output type.
13524Some sample fetch methods have deprecated aliases that are used to maintain
13525compatibility with existing configurations. They are then explicitly marked as
13526deprecated and should not be used in new setups.
13527
13528The ACL derivatives are also indicated when available, with their respective
13529matching methods. These ones all have a well defined default pattern matching
13530method, so it is never necessary (though allowed) to pass the "-m" option to
13531indicate how the sample will be matched using ACLs.
13532
13533As indicated in the sample type versus matching compatibility matrix above,
13534when using a generic sample fetch method in an ACL, the "-m" option is
13535mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
13536the same keyword exists as an ACL keyword and as a standard fetch method, the
13537ACL engine will automatically pick the ACL-only one by default.
13538
13539Some of these keywords support one or multiple mandatory arguments, and one or
13540multiple optional arguments. These arguments are strongly typed and are checked
13541when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010013542incorrect argument (e.g. an unresolved backend name). Fetch function arguments
13543are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020013544is optional, it will be indicated below between square brackets ('[ ]'). When
13545all arguments are optional, the parenthesis may be omitted.
13546
13547Thus, the syntax of a standard sample fetch method is one of the following :
13548 - name
13549 - name(arg1)
13550 - name(arg1,arg2)
13551
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013552
135537.3.1. Converters
13554-----------------
13555
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013556Sample fetch methods may be combined with transformations to be applied on top
13557of the fetched sample (also called "converters"). These combinations form what
13558is called "sample expressions" and the result is a "sample". Initially this
13559was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010013560has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013561unique-id-format, add-header, ...).
13562
13563These transformations are enumerated as a series of specific keywords after the
13564sample fetch method. These keywords may equally be appended immediately after
13565the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010013566support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013567
Willy Tarreau97707872015-01-27 15:12:13 +010013568A certain category of converters are bitwise and arithmetic operators which
13569support performing basic operations on integers. Some bitwise operations are
13570supported (and, or, xor, cpl) and some arithmetic operations are supported
13571(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
13572bool) which make it possible to report a match without having to write an ACL.
13573
Willy Tarreau74ca5042013-06-11 23:12:07 +020013574The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010013575
Ben Shillitof25e8e52016-12-02 14:25:37 +00001357651d.single(<prop>[,<prop>*])
13577 Returns values for the properties requested as a string, where values are
13578 separated by the delimiter specified with "51degrees-property-separator".
13579 The device is identified using the User-Agent header passed to the
13580 converter. The function can be passed up to five property names, and if a
13581 property name can't be found, the value "NoData" is returned.
13582
13583 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013584 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
13585 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000013586 # User-Agent passed to the converter.
13587 frontend http-in
13588 bind *:8081
13589 default_backend servers
13590 http-request set-header X-51D-DeviceTypeMobileTablet \
13591 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
13592
Willy Tarreau97707872015-01-27 15:12:13 +010013593add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013594 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013595 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013596 name. The name of the variable starts with an indication about its scope. The
13597 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013598 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013599 "sess" : the variable is shared with the whole session
13600 "txn" : the variable is shared with the transaction (request and response)
13601 "req" : the variable is shared only during request processing
13602 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013603 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013604 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013605
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010013606aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
13607 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
13608 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
13609 need to be base64 encoded and the returned result is in raw byte format.
13610 If the <aead_tag> validation fails, the converter doesn't return any data.
13611 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
13612 converter requires at least OpenSSL 1.0.1.
13613
13614 Example:
13615 http-response set-header X-Decrypted-Text %[var(txn.enc),\
13616 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
13617
Willy Tarreau97707872015-01-27 15:12:13 +010013618and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013619 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013620 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013621 numeric value or a variable name. The name of the variable starts with an
13622 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013623 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013624 "sess" : the variable is shared with the whole session
13625 "txn" : the variable is shared with the transaction (request and response)
13626 "req" : the variable is shared only during request processing
13627 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013628 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013629 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013630
Holger Just1bfc24b2017-05-06 00:56:53 +020013631b64dec
13632 Converts (decodes) a base64 encoded input string to its binary
13633 representation. It performs the inverse operation of base64().
13634
Emeric Brun53d1a982014-04-30 18:21:37 +020013635base64
13636 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010013637 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020013638 an SSL ID can be copied in a header).
13639
Willy Tarreau97707872015-01-27 15:12:13 +010013640bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013641 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013642 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013643 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013644 presence of a flag).
13645
Emeric Brun54c4ac82014-11-03 15:32:43 +010013646bytes(<offset>[,<length>])
13647 Extracts some bytes from an input binary sample. The result is a binary
13648 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010013649 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010013650
Willy Tarreau280f42b2018-02-19 15:34:12 +010013651concat([<start>],[<var>],[<end>])
13652 Concatenates up to 3 fields after the current sample which is then turned to
13653 a string. The first one, <start>, is a constant string, that will be appended
13654 immediately after the existing sample. It may be omitted if not used. The
13655 second one, <var>, is a variable name. The variable will be looked up, its
13656 contents converted to a string, and it will be appended immediately after the
13657 <first> part. If the variable is not found, nothing is appended. It may be
13658 omitted as well. The third field, <end> is a constant string that will be
13659 appended after the variable. It may also be omitted. Together, these elements
13660 allow to concatenate variables with delimiters to an existing set of
13661 variables. This can be used to build new variables made of a succession of
13662 other variables, such as colon-delimited varlues. Note that due to the config
13663 parser, it is not possible to use a comma nor a closing parenthesis as
13664 delimitors.
13665
13666 Example:
13667 tcp-request session set-var(sess.src) src
13668 tcp-request session set-var(sess.dn) ssl_c_s_dn
13669 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
13670 http-request set-header x-hap-sig %[var(txn.sig)]
13671
Willy Tarreau97707872015-01-27 15:12:13 +010013672cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013673 Takes the input value of type signed integer, applies a ones-complement
13674 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010013675
Willy Tarreau80599772015-01-20 19:35:24 +010013676crc32([<avalanche>])
13677 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
13678 hash function. Optionally, it is possible to apply a full avalanche hash
13679 function to the output if the optional <avalanche> argument equals 1. This
13680 converter uses the same functions as used by the various hash-based load
13681 balancing algorithms, so it will provide exactly the same results. It is
13682 provided for compatibility with other software which want a CRC32 to be
13683 computed on some input keys, so it follows the most common implementation as
13684 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
13685 but may provide a better or at least less predictable distribution. It must
13686 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013687 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
13688
13689crc32c([<avalanche>])
13690 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
13691 hash function. Optionally, it is possible to apply a full avalanche hash
13692 function to the output if the optional <avalanche> argument equals 1. This
13693 converter uses the same functions as described in RFC4960, Appendix B [8].
13694 It is provided for compatibility with other software which want a CRC32C to be
13695 computed on some input keys. It is slower than the other algorithms and it must
13696 not be used for security purposes as a 32-bit hash is trivial to break. See
13697 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010013698
David Carlier29b3ca32015-09-25 14:09:21 +010013699da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020013700 Asks the DeviceAtlas converter to identify the User Agent string passed on
13701 input, and to emit a string made of the concatenation of the properties
13702 enumerated in argument, delimited by the separator defined by the global
13703 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000013704 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020013705 configuration language.
13706
13707 Example:
13708 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013709 bind *:8881
13710 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000013711 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 +020013712
Thierry FOURNIER9687c772015-05-07 15:46:29 +020013713debug
13714 This converter is used as debug tool. It dumps on screen the content and the
13715 type of the input sample. The sample is returned as is on its output. This
13716 converter only exists when haproxy was built with debugging enabled.
13717
Willy Tarreau97707872015-01-27 15:12:13 +010013718div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013719 Divides the input value of type signed integer by <value>, and returns the
13720 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013721 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010013722 variable name. The name of the variable starts with an indication about its
13723 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013724 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013725 "sess" : the variable is shared with the whole session
13726 "txn" : the variable is shared with the transaction (request and response)
13727 "req" : the variable is shared only during request processing
13728 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013729 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013730 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013731
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013732djb2([<avalanche>])
13733 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
13734 hash function. Optionally, it is possible to apply a full avalanche hash
13735 function to the output if the optional <avalanche> argument equals 1. This
13736 converter uses the same functions as used by the various hash-based load
13737 balancing algorithms, so it will provide exactly the same results. It is
13738 mostly intended for debugging, but can be used as a stick-table entry to
13739 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013740 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
13741 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013742
Willy Tarreau97707872015-01-27 15:12:13 +010013743even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013744 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010013745 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
13746
Marcin Deranek9631a282018-04-16 14:30:46 +020013747field(<index>,<delimiters>[,<count>])
13748 Extracts the substring at the given index counting from the beginning
13749 (positive index) or from the end (negative index) considering given delimiters
13750 from an input string. Indexes start at 1 or -1 and delimiters are a string
13751 formatted list of chars. Optionally you can specify <count> of fields to
13752 extract (default: 1). Value of 0 indicates extraction of all remaining
13753 fields.
13754
13755 Example :
13756 str(f1_f2_f3__f5),field(5,_) # f5
13757 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
13758 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
13759 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
13760 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010013761
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013762hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010013763 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013764 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010013765 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 +020013766 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010013767
Dragan Dosen3f957b22017-10-24 09:27:34 +020013768hex2i
13769 Converts a hex string containing two hex digits per input byte to an
13770 integer. If the input value can not be converted, then zero is returned.
13771
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013772http_date([<offset>])
13773 Converts an integer supposed to contain a date since epoch to a string
13774 representing this date in a format suitable for use in HTTP header fields. If
13775 an offset value is specified, then it is a number of seconds that is added to
13776 the date before the conversion is operated. This is particularly useful to
13777 emit Date header fields, Expires values in responses when combined with a
13778 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013779
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013780in_table(<table>)
13781 Uses the string representation of the input sample to perform a look up in
13782 the specified table. If the key is not found in the table, a boolean false
13783 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010013784 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013785 or not a source IP address or an Authorization header was already seen).
13786
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013787ipmask(<mask4>, [<mask6>])
13788 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013789 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013790 table entries and as such use the same server. The mask4 can be passed in
13791 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
13792 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
13793 If no mask6 is given IPv6 addresses will fail to convert for backwards
13794 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013795
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013796json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010013797 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013798 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013799 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013800 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
13801 of errors:
13802 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
13803 bytes, ...)
13804 - invalid range (the decoded value is within a UTF-8 prohibited range),
13805 - code overlong (the value is encoded with more bytes than necessary).
13806
13807 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
13808 character is greater than 0xffff because the JSON string escape specification
13809 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
13810 in 4 variants designated by a combination of two suffix letters : "p" for
13811 "permissive" and "s" for "silently ignore". The behaviors of the decoders
13812 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013813 - "ascii" : never fails;
13814 - "utf8" : fails on any detected errors;
13815 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013816 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010013817 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013818 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
13819 characters corresponding to the other errors.
13820
13821 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010013822 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013823
13824 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013825 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013826 capture request header user-agent len 150
13827 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013828
13829 Input request from client 127.0.0.1:
13830 GET / HTTP/1.0
13831 User-Agent: Very "Ugly" UA 1/2
13832
13833 Output log:
13834 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
13835
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013836language(<value>[,<default>])
13837 Returns the value with the highest q-factor from a list as extracted from the
13838 "accept-language" header using "req.fhdr". Values with no q-factor have a
13839 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
13840 belong to the list of semi-colon delimited <values> will be considered. The
13841 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
13842 given list and a default value is provided, it is returned. Note that language
13843 names may have a variant after a dash ('-'). If this variant is present in the
13844 list, it will be matched, but if it is not, only the base language is checked.
13845 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010013846 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013847 ordering of the values in the request counts, as the first value among
13848 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020013849
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013850 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013851
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013852 # this configuration switches to the backend matching a
13853 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013854
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013855 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
13856 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
13857 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
13858 use_backend spanish if es
13859 use_backend french if fr
13860 use_backend english if en
13861 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020013862
Willy Tarreau60a2ee72017-12-15 07:13:48 +010013863length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010013864 Get the length of the string. This can only be placed after a string
13865 sample fetch function or after a transformation keyword returning a string
13866 type. The result is of type integer.
13867
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013868lower
13869 Convert a string sample to lower case. This can only be placed after a string
13870 sample fetch function or after a transformation keyword returning a string
13871 type. The result is of type string.
13872
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013873ltime(<format>[,<offset>])
13874 Converts an integer supposed to contain a date since epoch to a string
13875 representing this date in local time using a format defined by the <format>
13876 string using strftime(3). The purpose is to allow any date format to be used
13877 in logs. An optional <offset> in seconds may be applied to the input date
13878 (positive or negative). See the strftime() man page for the format supported
13879 by your operating system. See also the utime converter.
13880
13881 Example :
13882
13883 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010013884 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013885 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
13886
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013887map(<map_file>[,<default_value>])
13888map_<match_type>(<map_file>[,<default_value>])
13889map_<match_type>_<output_type>(<map_file>[,<default_value>])
13890 Search the input value from <map_file> using the <match_type> matching method,
13891 and return the associated value converted to the type <output_type>. If the
13892 input value cannot be found in the <map_file>, the converter returns the
13893 <default_value>. If the <default_value> is not set, the converter fails and
13894 acts as if no input value could be fetched. If the <match_type> is not set, it
13895 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
13896 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
13897 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013898
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013899 It is important to avoid overlapping between the keys : IP addresses and
13900 strings are stored in trees, so the first of the finest match will be used.
13901 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013902
Tim Düsterhus4896c442016-11-29 02:15:19 +010013903 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013904 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013905
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013906 input type | match method | output type str | output type int | output type ip
13907 -----------+--------------+-----------------+-----------------+---------------
13908 str | str | map_str | map_str_int | map_str_ip
13909 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020013910 str | beg | map_beg | map_beg_int | map_end_ip
13911 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013912 str | sub | map_sub | map_sub_int | map_sub_ip
13913 -----------+--------------+-----------------+-----------------+---------------
13914 str | dir | map_dir | map_dir_int | map_dir_ip
13915 -----------+--------------+-----------------+-----------------+---------------
13916 str | dom | map_dom | map_dom_int | map_dom_ip
13917 -----------+--------------+-----------------+-----------------+---------------
13918 str | end | map_end | map_end_int | map_end_ip
13919 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080013920 str | reg | map_reg | map_reg_int | map_reg_ip
13921 -----------+--------------+-----------------+-----------------+---------------
13922 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013923 -----------+--------------+-----------------+-----------------+---------------
13924 int | int | map_int | map_int_int | map_int_ip
13925 -----------+--------------+-----------------+-----------------+---------------
13926 ip | ip | map_ip | map_ip_int | map_ip_ip
13927 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013928
Thierry Fournier8feaa662016-02-10 22:55:20 +010013929 The special map called "map_regm" expect matching zone in the regular
13930 expression and modify the output replacing back reference (like "\1") by
13931 the corresponding match text.
13932
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013933 The file contains one key + value per line. Lines which start with '#' are
13934 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
13935 is then the first "word" (series of non-space/tabs characters), and the value
13936 is what follows this series of space/tab till the end of the line excluding
13937 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013938
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013939 Example :
13940
13941 # this is a comment and is ignored
13942 2.22.246.0/23 United Kingdom \n
13943 <-><-----------><--><------------><---->
13944 | | | | `- trailing spaces ignored
13945 | | | `---------- value
13946 | | `-------------------- middle spaces ignored
13947 | `---------------------------- key
13948 `------------------------------------ leading spaces ignored
13949
Willy Tarreau97707872015-01-27 15:12:13 +010013950mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013951 Divides the input value of type signed integer by <value>, and returns the
13952 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013953 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013954 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013955 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013956 "sess" : the variable is shared with the whole session
13957 "txn" : the variable is shared with the transaction (request and response)
13958 "req" : the variable is shared only during request processing
13959 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013960 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013961 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013962
13963mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013964 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020013965 the product as an signed integer. In case of overflow, the largest possible
13966 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013967 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013968 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013969 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013970 "sess" : the variable is shared with the whole session
13971 "txn" : the variable is shared with the transaction (request and response)
13972 "req" : the variable is shared only during request processing
13973 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013974 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013975 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013976
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010013977nbsrv
13978 Takes an input value of type string, interprets it as a backend name and
13979 returns the number of usable servers in that backend. Can be used in places
13980 where we want to look up a backend from a dynamic name, like a result of a
13981 map lookup.
13982
Willy Tarreau97707872015-01-27 15:12:13 +010013983neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013984 Takes the input value of type signed integer, computes the opposite value,
13985 and returns the remainder as an signed integer. 0 is identity. This operator
13986 is provided for reversed subtracts : in order to subtract the input from a
13987 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010013988
13989not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013990 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013991 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013992 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013993 absence of a flag).
13994
13995odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013996 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010013997 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
13998
13999or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014000 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014001 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010014002 numeric value or a variable name. The name of the variable starts with an
14003 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014004 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014005 "sess" : the variable is shared with the whole session
14006 "txn" : the variable is shared with the transaction (request and response)
14007 "req" : the variable is shared only during request processing
14008 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010014009 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014010 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014011
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014012protobuf(<field_number>,[<field_type>])
14013 This extracts the protocol buffers message field in raw mode of an input binary
14014 sample representation of a protocol buffer message with <field_number> as field
14015 number (dotted notation) if <field_type> is not present, or as an integer sample
14016 if this field is present (see also "ungrpc" below).
14017 The list of the authorized types is the following one: "int32", "int64", "uint32",
14018 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14019 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14020 "float" for the wire type 5. Note that "string" is considered as a length-delimited
14021 type, so it does not require any <field_type> argument to be extracted.
14022 More information may be found here about the protocol buffers message field types:
14023 https://developers.google.com/protocol-buffers/docs/encoding
14024
Willy Tarreauc4dc3502015-01-23 20:39:28 +010014025regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010014026 Applies a regex-based substitution to the input string. It does the same
14027 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
14028 default it will replace in the input string the first occurrence of the
14029 largest part matching the regular expression <regex> with the substitution
14030 string <subst>. It is possible to replace all occurrences instead by adding
14031 the flag "g" in the third argument <flags>. It is also possible to make the
14032 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
14033 string, it is made up from the concatenation of all desired flags. Thus if
14034 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
14035 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010014036 configuration parser, some characters such as closing parenthesis, closing
14037 square brackets or comma are not possible to use in the arguments. The first
14038 use of this converter is to replace certain characters or sequence of
14039 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010014040
14041 Example :
14042
14043 # de-duplicate "/" in header "x-path".
14044 # input: x-path: /////a///b/c/xzxyz/
14045 # output: x-path: /a/b/c/xzxyz/
14046 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
14047
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014048capture-req(<id>)
14049 Capture the string entry in the request slot <id> and returns the entry as
14050 is. If the slot doesn't exist, the capture fails silently.
14051
14052 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014053 "http-response capture", "capture.req.hdr" and
14054 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014055
14056capture-res(<id>)
14057 Capture the string entry in the response slot <id> and returns the entry as
14058 is. If the slot doesn't exist, the capture fails silently.
14059
14060 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020014061 "http-response capture", "capture.req.hdr" and
14062 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020014063
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014064sdbm([<avalanche>])
14065 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
14066 hash function. Optionally, it is possible to apply a full avalanche hash
14067 function to the output if the optional <avalanche> argument equals 1. This
14068 converter uses the same functions as used by the various hash-based load
14069 balancing algorithms, so it will provide exactly the same results. It is
14070 mostly intended for debugging, but can be used as a stick-table entry to
14071 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014072 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
14073 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014074
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014075set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010014076 Sets a variable with the input content and returns the content on the output
14077 as-is. The variable keeps the value and the associated input type. The name of
14078 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014079 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014080 "sess" : the variable is shared with the whole session
14081 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014082 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014083 "req" : the variable is shared only during request processing,
14084 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014085 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014086 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014087
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020014088sha1
14089 Converts a binary input sample to a SHA1 digest. The result is a binary
14090 sample with length of 20 bytes.
14091
Tim Duesterhusca097c12018-04-27 21:18:45 +020014092strcmp(<var>)
14093 Compares the contents of <var> with the input value of type string. Returns
14094 the result as a signed integer compatible with strcmp(3): 0 if both strings
14095 are identical. A value less than 0 if the left string is lexicographically
14096 smaller than the right string or if the left string is shorter. A value greater
14097 than 0 otherwise (right string greater than left string or the right string is
14098 shorter).
14099
14100 Example :
14101
14102 http-request set-var(txn.host) hdr(host)
14103 # Check whether the client is attempting domain fronting.
14104 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
14105
14106
Willy Tarreau97707872015-01-27 15:12:13 +010014107sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014108 Subtracts <value> from the input value of type signed integer, and returns
14109 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014110 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010014111 or a variable name. The name of the variable starts with an indication about
14112 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014113 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014114 "sess" : the variable is shared with the whole session
14115 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014116 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014117 "req" : the variable is shared only during request processing,
14118 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014119 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014120 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014121
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014122table_bytes_in_rate(<table>)
14123 Uses the string representation of the input sample to perform a look up in
14124 the specified table. If the key is not found in the table, integer value zero
14125 is returned. Otherwise the converter returns the average client-to-server
14126 bytes rate associated with the input sample in the designated table, measured
14127 in amount of bytes over the period configured in the table. See also the
14128 sc_bytes_in_rate sample fetch keyword.
14129
14130
14131table_bytes_out_rate(<table>)
14132 Uses the string representation of the input sample to perform a look up in
14133 the specified table. If the key is not found in the table, integer value zero
14134 is returned. Otherwise the converter returns the average server-to-client
14135 bytes rate associated with the input sample in the designated table, measured
14136 in amount of bytes over the period configured in the table. See also the
14137 sc_bytes_out_rate sample fetch keyword.
14138
14139table_conn_cnt(<table>)
14140 Uses the string representation of the input sample to perform a look up in
14141 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014142 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014143 connections associated with the input sample in the designated table. See
14144 also the sc_conn_cnt sample fetch keyword.
14145
14146table_conn_cur(<table>)
14147 Uses the string representation of the input sample to perform a look up in
14148 the specified table. If the key is not found in the table, integer value zero
14149 is returned. Otherwise the converter returns the current amount of concurrent
14150 tracked connections associated with the input sample in the designated table.
14151 See also the sc_conn_cur sample fetch keyword.
14152
14153table_conn_rate(<table>)
14154 Uses the string representation of the input sample to perform a look up in
14155 the specified table. If the key is not found in the table, integer value zero
14156 is returned. Otherwise the converter returns the average incoming connection
14157 rate associated with the input sample in the designated table. See also the
14158 sc_conn_rate sample fetch keyword.
14159
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014160table_gpt0(<table>)
14161 Uses the string representation of the input sample to perform a look up in
14162 the specified table. If the key is not found in the table, boolean value zero
14163 is returned. Otherwise the converter returns the current value of the first
14164 general purpose tag associated with the input sample in the designated table.
14165 See also the sc_get_gpt0 sample fetch keyword.
14166
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014167table_gpc0(<table>)
14168 Uses the string representation of the input sample to perform a look up in
14169 the specified table. If the key is not found in the table, integer value zero
14170 is returned. Otherwise the converter returns the current value of the first
14171 general purpose counter associated with the input sample in the designated
14172 table. See also the sc_get_gpc0 sample fetch keyword.
14173
14174table_gpc0_rate(<table>)
14175 Uses the string representation of the input sample to perform a look up in
14176 the specified table. If the key is not found in the table, integer value zero
14177 is returned. Otherwise the converter returns the frequency which the gpc0
14178 counter was incremented over the configured period in the table, associated
14179 with the input sample in the designated table. See also the sc_get_gpc0_rate
14180 sample fetch keyword.
14181
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014182table_gpc1(<table>)
14183 Uses the string representation of the input sample to perform a look up in
14184 the specified table. If the key is not found in the table, integer value zero
14185 is returned. Otherwise the converter returns the current value of the second
14186 general purpose counter associated with the input sample in the designated
14187 table. See also the sc_get_gpc1 sample fetch keyword.
14188
14189table_gpc1_rate(<table>)
14190 Uses the string representation of the input sample to perform a look up in
14191 the specified table. If the key is not found in the table, integer value zero
14192 is returned. Otherwise the converter returns the frequency which the gpc1
14193 counter was incremented over the configured period in the table, associated
14194 with the input sample in the designated table. See also the sc_get_gpc1_rate
14195 sample fetch keyword.
14196
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014197table_http_err_cnt(<table>)
14198 Uses the string representation of the input sample to perform a look up in
14199 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014200 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014201 errors associated with the input sample in the designated table. See also the
14202 sc_http_err_cnt sample fetch keyword.
14203
14204table_http_err_rate(<table>)
14205 Uses the string representation of the input sample to perform a look up in
14206 the specified table. If the key is not found in the table, integer value zero
14207 is returned. Otherwise the average rate of HTTP errors associated with the
14208 input sample in the designated table, measured in amount of errors over the
14209 period configured in the table. See also the sc_http_err_rate sample fetch
14210 keyword.
14211
14212table_http_req_cnt(<table>)
14213 Uses the string representation of the input sample to perform a look up in
14214 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014215 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014216 requests associated with the input sample in the designated table. See also
14217 the sc_http_req_cnt sample fetch keyword.
14218
14219table_http_req_rate(<table>)
14220 Uses the string representation of the input sample to perform a look up in
14221 the specified table. If the key is not found in the table, integer value zero
14222 is returned. Otherwise the average rate of HTTP requests associated with the
14223 input sample in the designated table, measured in amount of requests over the
14224 period configured in the table. See also the sc_http_req_rate sample fetch
14225 keyword.
14226
14227table_kbytes_in(<table>)
14228 Uses the string representation of the input sample to perform a look up in
14229 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014230 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014231 to-server data associated with the input sample in the designated table,
14232 measured in kilobytes. The test is currently performed on 32-bit integers,
14233 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
14234 keyword.
14235
14236table_kbytes_out(<table>)
14237 Uses the string representation of the input sample to perform a look up in
14238 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014239 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014240 to-client data associated with the input sample in the designated table,
14241 measured in kilobytes. The test is currently performed on 32-bit integers,
14242 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
14243 keyword.
14244
14245table_server_id(<table>)
14246 Uses the string representation of the input sample to perform a look up in
14247 the specified table. If the key is not found in the table, integer value zero
14248 is returned. Otherwise the converter returns the server ID associated with
14249 the input sample in the designated table. A server ID is associated to a
14250 sample by a "stick" rule when a connection to a server succeeds. A server ID
14251 zero means that no server is associated with this key.
14252
14253table_sess_cnt(<table>)
14254 Uses the string representation of the input sample to perform a look up in
14255 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014256 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014257 sessions associated with the input sample in the designated table. Note that
14258 a session here refers to an incoming connection being accepted by the
14259 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
14260 keyword.
14261
14262table_sess_rate(<table>)
14263 Uses the string representation of the input sample to perform a look up in
14264 the specified table. If the key is not found in the table, integer value zero
14265 is returned. Otherwise the converter returns the average incoming session
14266 rate associated with the input sample in the designated table. Note that a
14267 session here refers to an incoming connection being accepted by the
14268 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
14269 keyword.
14270
14271table_trackers(<table>)
14272 Uses the string representation of the input sample to perform a look up in
14273 the specified table. If the key is not found in the table, integer value zero
14274 is returned. Otherwise the converter returns the current amount of concurrent
14275 connections tracking the same key as the input sample in the designated
14276 table. It differs from table_conn_cur in that it does not rely on any stored
14277 information but on the table's reference count (the "use" value which is
14278 returned by "show table" on the CLI). This may sometimes be more suited for
14279 layer7 tracking. It can be used to tell a server how many concurrent
14280 connections there are from a given address for example. See also the
14281 sc_trackers sample fetch keyword.
14282
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014283upper
14284 Convert a string sample to upper case. This can only be placed after a string
14285 sample fetch function or after a transformation keyword returning a string
14286 type. The result is of type string.
14287
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020014288url_dec
14289 Takes an url-encoded string provided as input and returns the decoded
14290 version as output. The input and the output are of type string.
14291
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014292ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014293 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014294 sample representation of a gRPC message with <field_number> as field number
14295 (dotted notation) if <field_type> is not present, or as an integer sample if this
14296 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014297 The list of the authorized types is the following one: "int32", "int64", "uint32",
14298 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14299 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14300 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014301 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014302 More information may be found here about the protocol buffers message field types:
14303 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014304
14305 Example:
14306 // with such a protocol buffer .proto file content adapted from
14307 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
14308
14309 message Point {
14310 int32 latitude = 1;
14311 int32 longitude = 2;
14312 }
14313
14314 message PPoint {
14315 Point point = 59;
14316 }
14317
14318 message Rectangle {
14319 // One corner of the rectangle.
14320 PPoint lo = 48;
14321 // The other corner of the rectangle.
14322 PPoint hi = 49;
14323 }
14324
14325 let's say a body request is made of a "Rectangle" object value (two PPoint
14326 protocol buffers messages), the four protocol buffers fields could be
14327 extracted with these "ungrpc" directives:
14328
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014329 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
14330 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
14331 req.body,ungrpc(49.59.1,int32) # "latidude" of "hi" second PPoint
14332 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
14333
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014334 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014335
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014336 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014337
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014338 As a gRPC message is alway made of a gRPC header followed by protocol buffers
14339 messages, in the previous example the "latitude" of "lo" first PPoint
14340 could be extracted with these equivalent directives:
14341
14342 req.body,ungrpc(48.59),protobuf(1,int32)
14343 req.body,ungrpc(48),protobuf(59.1,int32)
14344 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
14345
14346 Note that the first convert must be "ungrpc", the remaining ones must be
14347 "protobuf" and only the last one may have or not a second argument to
14348 interpret the previous binary sample.
14349
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014350
Christopher Faulet85d79c92016-11-09 16:54:56 +010014351unset-var(<var name>)
14352 Unsets a variable if the input content is defined. The name of the variable
14353 starts with an indication about its scope. The scopes allowed are:
14354 "proc" : the variable is shared with the whole process
14355 "sess" : the variable is shared with the whole session
14356 "txn" : the variable is shared with the transaction (request and
14357 response),
14358 "req" : the variable is shared only during request processing,
14359 "res" : the variable is shared only during response processing.
14360 This prefix is followed by a name. The separator is a '.'. The name may only
14361 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
14362
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014363utime(<format>[,<offset>])
14364 Converts an integer supposed to contain a date since epoch to a string
14365 representing this date in UTC time using a format defined by the <format>
14366 string using strftime(3). The purpose is to allow any date format to be used
14367 in logs. An optional <offset> in seconds may be applied to the input date
14368 (positive or negative). See the strftime() man page for the format supported
14369 by your operating system. See also the ltime converter.
14370
14371 Example :
14372
14373 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014374 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014375 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
14376
Marcin Deranek9631a282018-04-16 14:30:46 +020014377word(<index>,<delimiters>[,<count>])
14378 Extracts the nth word counting from the beginning (positive index) or from
14379 the end (negative index) considering given delimiters from an input string.
14380 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
14381 Optionally you can specify <count> of words to extract (default: 1).
14382 Value of 0 indicates extraction of all remaining words.
14383
14384 Example :
14385 str(f1_f2_f3__f5),word(4,_) # f5
14386 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
14387 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
14388 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
14389 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010014390
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014391wt6([<avalanche>])
14392 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
14393 hash function. Optionally, it is possible to apply a full avalanche hash
14394 function to the output if the optional <avalanche> argument equals 1. This
14395 converter uses the same functions as used by the various hash-based load
14396 balancing algorithms, so it will provide exactly the same results. It is
14397 mostly intended for debugging, but can be used as a stick-table entry to
14398 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014399 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
14400 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014401
Willy Tarreau97707872015-01-27 15:12:13 +010014402xor(<value>)
14403 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014404 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014405 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014406 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014407 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014408 "sess" : the variable is shared with the whole session
14409 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014410 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014411 "req" : the variable is shared only during request processing,
14412 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014413 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014414 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014415
Thierry FOURNIER01e09742016-12-26 11:46:11 +010014416xxh32([<seed>])
14417 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
14418 variant of the XXHash hash function. This hash supports a seed which defaults
14419 to zero but a different value maybe passed as the <seed> argument. This hash
14420 is known to be very good and very fast so it can be used to hash URLs and/or
14421 URL parameters for use as stick-table keys to collect statistics with a low
14422 collision rate, though care must be taken as the algorithm is not considered
14423 as cryptographically secure.
14424
14425xxh64([<seed>])
14426 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
14427 variant of the XXHash hash function. This hash supports a seed which defaults
14428 to zero but a different value maybe passed as the <seed> argument. This hash
14429 is known to be very good and very fast so it can be used to hash URLs and/or
14430 URL parameters for use as stick-table keys to collect statistics with a low
14431 collision rate, though care must be taken as the algorithm is not considered
14432 as cryptographically secure.
14433
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014434
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200144357.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020014436--------------------------------------------
14437
14438A first set of sample fetch methods applies to internal information which does
14439not even relate to any client information. These ones are sometimes used with
14440"monitor-fail" directives to report an internal status to external watchers.
14441The sample fetch methods described in this section are usable anywhere.
14442
14443always_false : boolean
14444 Always returns the boolean "false" value. It may be used with ACLs as a
14445 temporary replacement for another one when adjusting configurations.
14446
14447always_true : boolean
14448 Always returns the boolean "true" value. It may be used with ACLs as a
14449 temporary replacement for another one when adjusting configurations.
14450
14451avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014452 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020014453 divided by the number of active servers. The current backend is used if no
14454 backend is specified. This is very similar to "queue" except that the size of
14455 the farm is considered, in order to give a more accurate measurement of the
14456 time it may take for a new connection to be processed. The main usage is with
14457 ACL to return a sorry page to new users when it becomes certain they will get
14458 a degraded service, or to pass to the backend servers in a header so that
14459 they decide to work in degraded mode or to disable some functions to speed up
14460 the processing a bit. Note that in the event there would not be any active
14461 server anymore, twice the number of queued connections would be considered as
14462 the measured value. This is a fair estimate, as we expect one server to get
14463 back soon anyway, but we still prefer to send new traffic to another backend
14464 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
14465 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010014466
Willy Tarreau74ca5042013-06-11 23:12:07 +020014467be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020014468 Applies to the number of currently established connections on the backend,
14469 possibly including the connection being evaluated. If no backend name is
14470 specified, the current one is used. But it is also possible to check another
14471 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014472 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
14473
14474be_conn_free([<backend>]) : integer
14475 Returns an integer value corresponding to the number of available connections
14476 across available servers in the backend. Queue slots are not included. Backup
14477 servers are also not included, unless all other servers are down. If no
14478 backend name is specified, the current one is used. But it is also possible
14479 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014480 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
14481 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014482
14483 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
14484 (meaning unlimited), then this fetch clearly does not make sense, in which
14485 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014486
Willy Tarreau74ca5042013-06-11 23:12:07 +020014487be_sess_rate([<backend>]) : integer
14488 Returns an integer value corresponding to the sessions creation rate on the
14489 backend, in number of new sessions per second. This is used with ACLs to
14490 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010014491 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014492 online dictionary). It can also be useful to add this element to logs using a
14493 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014494
14495 Example :
14496 # Redirect to an error page if the dictionary is requested too often
14497 backend dynamic
14498 mode http
14499 acl being_scanned be_sess_rate gt 100
14500 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010014501
Davor Ocelice9ed2812017-12-25 17:49:28 +010014502bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014503 Returns a binary chain. The input is the hexadecimal representation
14504 of the string.
14505
14506bool(<bool>) : bool
14507 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
14508 'false' and '0' are the same. 'true' and '1' are the same.
14509
Willy Tarreau74ca5042013-06-11 23:12:07 +020014510connslots([<backend>]) : integer
14511 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014512 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020014513 connections on all servers and the maximum queue size. This is probably only
14514 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050014515
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014516 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020014517 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014518 usage; see "use_backend" keyword) can be redirected to a different backend.
14519
Willy Tarreau55165fe2009-05-10 12:02:55 +020014520 'connslots' = number of available server connection slots, + number of
14521 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014522
Willy Tarreaua36af912009-10-10 12:02:45 +020014523 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020014524 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020014525 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020014526 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010014527 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020014528 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020014529 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014530
Willy Tarreau55165fe2009-05-10 12:02:55 +020014531 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
14532 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020014533 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020014534 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014535
Willy Tarreau70fe9442018-11-22 16:07:39 +010014536cpu_calls : integer
14537 Returns the number of calls to the task processing the stream or current
14538 request since it was allocated. This number is reset for each new request on
14539 the same connections in case of HTTP keep-alive. This value should usually be
14540 low and stable (around 2 calls for a typically simple request) but may become
14541 high if some processing (compression, caching or analysis) is performed. This
14542 is purely for performance monitoring purposes.
14543
14544cpu_ns_avg : integer
14545 Returns the average number of nanoseconds spent in each call to the task
14546 processing the stream or current request. This number is reset for each new
14547 request on the same connections in case of HTTP keep-alive. This value
14548 indicates the overall cost of processing the request or the connection for
14549 each call. There is no good nor bad value but the time spent in a call
14550 automatically causes latency for other processing (see lat_ns_avg below),
14551 and may affect other connection's apparent response time. Certain operations
14552 like compression, complex regex matching or heavy Lua operations may directly
14553 affect this value, and having it in the logs will make it easier to spot the
14554 faulty processing that needs to be fixed to recover decent performance.
14555 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
14556
14557cpu_ns_tot : integer
14558 Returns the total number of nanoseconds spent in each call to the task
14559 processing the stream or current request. This number is reset for each new
14560 request on the same connections in case of HTTP keep-alive. This value
14561 indicates the overall cost of processing the request or the connection for
14562 each call. There is no good nor bad value but the time spent in a call
14563 automatically causes latency for other processing (see lat_ns_avg below),
14564 induces CPU costs on the machine, and may affect other connection's apparent
14565 response time. Certain operations like compression, complex regex matching or
14566 heavy Lua operations may directly affect this value, and having it in the
14567 logs will make it easier to spot the faulty processing that needs to be fixed
14568 to recover decent performance. The value may be artificially high due to a
14569 high cpu_calls count, for example when processing many HTTP chunks, and for
14570 this reason it is often preferred to log cpu_ns_avg instead.
14571
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014572date([<offset>]) : integer
14573 Returns the current date as the epoch (number of seconds since 01/01/1970).
14574 If an offset value is specified, then it is a number of seconds that is added
14575 to the current date before returning the value. This is particularly useful
14576 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020014577 It is useful combined with the http_date converter.
14578
14579 Example :
14580
14581 # set an expires header to now+1 hour in every response
14582 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014583
Etienne Carrierea792a0a2018-01-17 13:43:24 +010014584date_us : integer
14585 Return the microseconds part of the date (the "second" part is returned by
14586 date sample). This sample is coherent with the date sample as it is comes
14587 from the same timeval structure.
14588
Willy Tarreaud716f9b2017-10-13 11:03:15 +020014589distcc_body(<token>[,<occ>]) : binary
14590 Parses a distcc message and returns the body associated to occurrence #<occ>
14591 of the token <token>. Occurrences start at 1, and when unspecified, any may
14592 match though in practice only the first one is checked for now. This can be
14593 used to extract file names or arguments in files built using distcc through
14594 haproxy. Please refer to distcc's protocol documentation for the complete
14595 list of supported tokens.
14596
14597distcc_param(<token>[,<occ>]) : integer
14598 Parses a distcc message and returns the parameter associated to occurrence
14599 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
14600 any may match though in practice only the first one is checked for now. This
14601 can be used to extract certain information such as the protocol version, the
14602 file size or the argument in files built using distcc through haproxy.
14603 Another use case consists in waiting for the start of the preprocessed file
14604 contents before connecting to the server to avoid keeping idle connections.
14605 Please refer to distcc's protocol documentation for the complete list of
14606 supported tokens.
14607
14608 Example :
14609 # wait up to 20s for the pre-processed file to be uploaded
14610 tcp-request inspect-delay 20s
14611 tcp-request content accept if { distcc_param(DOTI) -m found }
14612 # send large files to the big farm
14613 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
14614
Willy Tarreau595ec542013-06-12 21:34:28 +020014615env(<name>) : string
14616 Returns a string containing the value of environment variable <name>. As a
14617 reminder, environment variables are per-process and are sampled when the
14618 process starts. This can be useful to pass some information to a next hop
14619 server, or with ACLs to take specific action when the process is started a
14620 certain way.
14621
14622 Examples :
14623 # Pass the Via header to next hop with the local hostname in it
14624 http-request add-header Via 1.1\ %[env(HOSTNAME)]
14625
14626 # reject cookie-less requests when the STOP environment variable is set
14627 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
14628
Willy Tarreau74ca5042013-06-11 23:12:07 +020014629fe_conn([<frontend>]) : integer
14630 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010014631 possibly including the connection being evaluated. If no frontend name is
14632 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014633 frontend. It can be used to return a sorry page before hard-blocking, or to
14634 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010014635 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020014636 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
14637 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020014638
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020014639fe_req_rate([<frontend>]) : integer
14640 Returns an integer value corresponding to the number of HTTP requests per
14641 second sent to a frontend. This number can differ from "fe_sess_rate" in
14642 situations where client-side keep-alive is enabled.
14643
Willy Tarreau74ca5042013-06-11 23:12:07 +020014644fe_sess_rate([<frontend>]) : integer
14645 Returns an integer value corresponding to the sessions creation rate on the
14646 frontend, in number of new sessions per second. This is used with ACLs to
14647 limit the incoming session rate to an acceptable range in order to prevent
14648 abuse of service at the earliest moment, for example when combined with other
14649 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
14650 down below the limit. It can also be useful to add this element to logs using
14651 a log-format directive. See also the "rate-limit sessions" directive for use
14652 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014653
14654 Example :
14655 # This frontend limits incoming mails to 10/s with a max of 100
14656 # concurrent connections. We accept any connection below 10/s, and
14657 # force excess clients to wait for 100 ms. Since clients are limited to
14658 # 100 max, there cannot be more than 10 incoming mails per second.
14659 frontend mail
14660 bind :25
14661 mode tcp
14662 maxconn 100
14663 acl too_fast fe_sess_rate ge 10
14664 tcp-request inspect-delay 100ms
14665 tcp-request content accept if ! too_fast
14666 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010014667
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010014668hostname : string
14669 Returns the system hostname.
14670
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014671int(<integer>) : signed integer
14672 Returns a signed integer.
14673
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014674ipv4(<ipv4>) : ipv4
14675 Returns an ipv4.
14676
14677ipv6(<ipv6>) : ipv6
14678 Returns an ipv6.
14679
Willy Tarreau70fe9442018-11-22 16:07:39 +010014680lat_ns_avg : integer
14681 Returns the average number of nanoseconds spent between the moment the task
14682 handling the stream is woken up and the moment it is effectively called. This
14683 number is reset for each new request on the same connections in case of HTTP
14684 keep-alive. This value indicates the overall latency inflicted to the current
14685 request by all other requests being processed in parallel, and is a direct
14686 indicator of perceived performance due to noisy neighbours. In order to keep
14687 the value low, it is possible to reduce the scheduler's run queue depth using
14688 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14689 once using "tune.maxpollevents", to decrease the stream's nice value using
14690 the "nice" option on the "bind" lines or in the frontend, or to look for
14691 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14692 whose processing needs to be adjusted or fixed. Compression of large buffers
14693 could be a culprit, like heavy regex or long lists of regex.
14694 Note: this value is exactly lat_ns_tot divided by cpu_calls.
14695
14696lat_ns_tot : integer
14697 Returns the total number of nanoseconds spent between the moment the task
14698 handling the stream is woken up and the moment it is effectively called. This
14699 number is reset for each new request on the same connections in case of HTTP
14700 keep-alive. This value indicates the overall latency inflicted to the current
14701 request by all other requests being processed in parallel, and is a direct
14702 indicator of perceived performance due to noisy neighbours. In order to keep
14703 the value low, it is possible to reduce the scheduler's run queue depth using
14704 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14705 once using "tune.maxpollevents", to decrease the stream's nice value using
14706 the "nice" option on the "bind" lines or in the frontend, or to look for
14707 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14708 whose processing needs to be adjusted or fixed. Compression of large buffers
14709 could be a culprit, like heavy regex or long lists of regex. Note: while it
14710 may intuitively seem that the total latency adds to a transfer time, it is
14711 almost never true because while a task waits for the CPU, network buffers
14712 continue to fill up and the next call will process more at once. The value
14713 may be artificially high due to a high cpu_calls count, for example when
14714 processing many HTTP chunks, and for this reason it is often preferred to log
14715 lat_ns_avg instead, which is a more relevant performance indicator.
14716
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014717meth(<method>) : method
14718 Returns a method.
14719
Willy Tarreau0f30d262014-11-24 16:02:05 +010014720nbproc : integer
14721 Returns an integer value corresponding to the number of processes that were
14722 started (it equals the global "nbproc" setting). This is useful for logging
14723 and debugging purposes.
14724
Willy Tarreau74ca5042013-06-11 23:12:07 +020014725nbsrv([<backend>]) : integer
14726 Returns an integer value corresponding to the number of usable servers of
14727 either the current backend or the named backend. This is mostly used with
14728 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010014729 switch to an alternate backend when the number of servers is too low to
14730 to handle some load. It is useful to report a failure when combined with
14731 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014732
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040014733prio_class : integer
14734 Returns the priority class of the current session for http mode or connection
14735 for tcp mode. The value will be that set by the last call to "http-request
14736 set-priority-class" or "tcp-request content set-priority-class".
14737
14738prio_offset : integer
14739 Returns the priority offset of the current session for http mode or
14740 connection for tcp mode. The value will be that set by the last call to
14741 "http-request set-priority-offset" or "tcp-request content
14742 set-priority-offset".
14743
Willy Tarreau0f30d262014-11-24 16:02:05 +010014744proc : integer
14745 Returns an integer value corresponding to the position of the process calling
14746 the function, between 1 and global.nbproc. This is useful for logging and
14747 debugging purposes.
14748
Willy Tarreau74ca5042013-06-11 23:12:07 +020014749queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014750 Returns the total number of queued connections of the designated backend,
14751 including all the connections in server queues. If no backend name is
14752 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014753 one. This is useful with ACLs or to pass statistics to backend servers. This
14754 can be used to take actions when queuing goes above a known level, generally
14755 indicating a surge of traffic or a massive slowdown on the servers. One
14756 possible action could be to reject new users but still accept old ones. See
14757 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
14758
Willy Tarreau84310e22014-02-14 11:59:04 +010014759rand([<range>]) : integer
14760 Returns a random integer value within a range of <range> possible values,
14761 starting at zero. If the range is not specified, it defaults to 2^32, which
14762 gives numbers between 0 and 4294967295. It can be useful to pass some values
14763 needed to take some routing decisions for example, or just for debugging
14764 purposes. This random must not be used for security purposes.
14765
Willy Tarreau74ca5042013-06-11 23:12:07 +020014766srv_conn([<backend>/]<server>) : integer
14767 Returns an integer value corresponding to the number of currently established
14768 connections on the designated server, possibly including the connection being
14769 evaluated. If <backend> is omitted, then the server is looked up in the
14770 current backend. It can be used to use a specific farm when one server is
14771 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014772 connections with it. See also the "fe_conn", "be_conn", "queue", and
14773 "srv_conn_free" fetch methods.
14774
14775srv_conn_free([<backend>/]<server>) : integer
14776 Returns an integer value corresponding to the number of available connections
14777 on the designated server, possibly including the connection being evaluated.
14778 The value does not include queue slots. If <backend> is omitted, then the
14779 server is looked up in the current backend. It can be used to use a specific
14780 farm when one server is full, or to inform the server about our view of the
14781 number of active connections with it. See also the "be_conn_free" and
14782 "srv_conn" fetch methods.
14783
14784 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
14785 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014786
14787srv_is_up([<backend>/]<server>) : boolean
14788 Returns true when the designated server is UP, and false when it is either
14789 DOWN or in maintenance mode. If <backend> is omitted, then the server is
14790 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010014791 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014792 availability). Another possible use which is more of a hack consists in
14793 using dummy servers as boolean variables that can be enabled or disabled from
14794 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
14795
Willy Tarreauff2b7af2017-10-13 11:46:26 +020014796srv_queue([<backend>/]<server>) : integer
14797 Returns an integer value corresponding to the number of connections currently
14798 pending in the designated server's queue. If <backend> is omitted, then the
14799 server is looked up in the current backend. It can sometimes be used together
14800 with the "use-server" directive to force to use a known faster server when it
14801 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
14802 fetch methods.
14803
Willy Tarreau74ca5042013-06-11 23:12:07 +020014804srv_sess_rate([<backend>/]<server>) : integer
14805 Returns an integer corresponding to the sessions creation rate on the
14806 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014807 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020014808 used with ACLs but can make sense with logs too. This is used to switch to an
14809 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010014810 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020014811 overloading servers).
14812
14813 Example :
14814 # Redirect to a separate back
14815 acl srv1_full srv_sess_rate(be1/srv1) gt 50
14816 acl srv2_full srv_sess_rate(be1/srv2) gt 50
14817 use_backend be2 if srv1_full or srv2_full
14818
Willy Tarreau0f30d262014-11-24 16:02:05 +010014819stopping : boolean
14820 Returns TRUE if the process calling the function is currently stopping. This
14821 can be useful for logging, or for relaxing certain checks or helping close
14822 certain connections upon graceful shutdown.
14823
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014824str(<string>) : string
14825 Returns a string.
14826
Willy Tarreau74ca5042013-06-11 23:12:07 +020014827table_avl([<table>]) : integer
14828 Returns the total number of available entries in the current proxy's
14829 stick-table or in the designated stick-table. See also table_cnt.
14830
14831table_cnt([<table>]) : integer
14832 Returns the total number of entries currently in use in the current proxy's
14833 stick-table or in the designated stick-table. See also src_conn_cnt and
14834 table_avl for other entry counting methods.
14835
Christopher Faulet34adb2a2017-11-21 21:45:38 +010014836thread : integer
14837 Returns an integer value corresponding to the position of the thread calling
14838 the function, between 0 and (global.nbthread-1). This is useful for logging
14839 and debugging purposes.
14840
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014841var(<var-name>) : undefined
14842 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010014843 sample fetch fails. The name of the variable starts with an indication
14844 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014845 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014846 "sess" : the variable is shared with the whole session
14847 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014848 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014849 "req" : the variable is shared only during request processing,
14850 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014851 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014852 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014853
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200148547.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020014855----------------------------------
14856
14857The layer 4 usually describes just the transport layer which in haproxy is
14858closest to the connection, where no content is yet made available. The fetch
14859methods described here are usable as low as the "tcp-request connection" rule
14860sets unless they require some future information. Those generally include
14861TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014862the incoming connection. For retrieving a value from a sticky counters, the
14863counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020014864"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
14865used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
14866can be specified as the first integer argument when using the "sc_" prefix.
14867Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
14868table may be specified with the "sc*" form, in which case the currently
14869tracked key will be looked up into this alternate table instead of the table
14870currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014871
Jérôme Magnin35e53a62019-01-16 14:38:37 +010014872bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010014873 Returns the backend connection's HTTP major version encoding, which may be 1
14874 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14875 encoding and not the version present in the request header.
14876
Willy Tarreau74ca5042013-06-11 23:12:07 +020014877be_id : integer
14878 Returns an integer containing the current backend's id. It can be used in
14879 frontends with responses to check which backend processed the request.
14880
Marcin Deranekd2471c22016-12-12 14:08:05 +010014881be_name : string
14882 Returns a string containing the current backend's name. It can be used in
14883 frontends with responses to check which backend processed the request.
14884
Willy Tarreau74ca5042013-06-11 23:12:07 +020014885dst : ip
14886 This is the destination IPv4 address of the connection on the client side,
14887 which is the address the client connected to. It can be useful when running
14888 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
14889 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010014890 RFC 4291. When the incoming connection passed through address translation or
14891 redirection involving connection tracking, the original destination address
14892 before the redirection will be reported. On Linux systems, the source and
14893 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
14894 is set, because a late response may reopen a timed out connection and switch
14895 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014896
14897dst_conn : integer
14898 Returns an integer value corresponding to the number of currently established
14899 connections on the same socket including the one being evaluated. It is
14900 normally used with ACLs but can as well be used to pass the information to
14901 servers in an HTTP header or in logs. It can be used to either return a sorry
14902 page before hard-blocking, or to use a specific backend to drain new requests
14903 when the socket is considered saturated. This offers the ability to assign
14904 different limits to different listening ports or addresses. See also the
14905 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014906
Willy Tarreau16e01562016-08-09 16:46:18 +020014907dst_is_local : boolean
14908 Returns true if the destination address of the incoming connection is local
14909 to the system, or false if the address doesn't exist on the system, meaning
14910 that it was intercepted in transparent mode. It can be useful to apply
14911 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010014912 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020014913 be delivered only on this address, or SSH access could be locally redirected.
14914 Please note that the check involves a few system calls, so it's better to do
14915 it only once per connection.
14916
Willy Tarreau74ca5042013-06-11 23:12:07 +020014917dst_port : integer
14918 Returns an integer value corresponding to the destination TCP port of the
14919 connection on the client side, which is the port the client connected to.
14920 This might be used when running in transparent mode, when assigning dynamic
14921 ports to some clients for a whole application session, to stick all users to
14922 a same server, or to pass the destination port information to a server using
14923 an HTTP header.
14924
Willy Tarreau60ca10a2017-08-18 15:26:54 +020014925fc_http_major : integer
14926 Reports the front connection's HTTP major version encoding, which may be 1
14927 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14928 encoding and not on the version present in the request header.
14929
Emeric Brun4f603012017-01-05 15:11:44 +010014930fc_rcvd_proxy : boolean
14931 Returns true if the client initiated the connection with a PROXY protocol
14932 header.
14933
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020014934fc_rtt(<unit>) : integer
14935 Returns the Round Trip Time (RTT) measured by the kernel for the client
14936 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
14937 can be set to "ms" for milliseconds or "us" for microseconds. If the server
14938 connection is not established, if the connection is not TCP or if the
14939 operating system does not support TCP_INFO, for example Linux kernels before
14940 2.4, the sample fetch fails.
14941
14942fc_rttvar(<unit>) : integer
14943 Returns the Round Trip Time (RTT) variance measured by the kernel for the
14944 client connection. <unit> is facultative, by default the unit is milliseconds.
14945 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
14946 server connection is not established, if the connection is not TCP or if the
14947 operating system does not support TCP_INFO, for example Linux kernels before
14948 2.4, the sample fetch fails.
14949
Joe Williams30fcd392016-08-10 07:06:44 -070014950fc_unacked(<unit>) : integer
14951 Returns the unacked counter measured by the kernel for the client connection.
14952 If the server connection is not established, if the connection is not TCP or
14953 if the operating system does not support TCP_INFO, for example Linux kernels
14954 before 2.4, the sample fetch fails.
14955
14956fc_sacked(<unit>) : integer
14957 Returns the sacked counter measured by the kernel for the client connection.
14958 If the server connection is not established, if the connection is not TCP or
14959 if the operating system does not support TCP_INFO, for example Linux kernels
14960 before 2.4, the sample fetch fails.
14961
14962fc_retrans(<unit>) : integer
14963 Returns the retransmits counter measured by the kernel for the client
14964 connection. If the server connection is not established, if the connection is
14965 not TCP or if the operating system does not support TCP_INFO, for example
14966 Linux kernels before 2.4, the sample fetch fails.
14967
14968fc_fackets(<unit>) : integer
14969 Returns the fack counter measured by the kernel for the client
14970 connection. If the server connection is not established, if the connection is
14971 not TCP or if the operating system does not support TCP_INFO, for example
14972 Linux kernels before 2.4, the sample fetch fails.
14973
14974fc_lost(<unit>) : integer
14975 Returns the lost counter measured by the kernel for the client
14976 connection. If the server connection is not established, if the connection is
14977 not TCP or if the operating system does not support TCP_INFO, for example
14978 Linux kernels before 2.4, the sample fetch fails.
14979
14980fc_reordering(<unit>) : integer
14981 Returns the reordering counter measured by the kernel for the client
14982 connection. If the server connection is not established, if the connection is
14983 not TCP or if the operating system does not support TCP_INFO, for example
14984 Linux kernels before 2.4, the sample fetch fails.
14985
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020014986fe_defbe : string
14987 Returns a string containing the frontend's default backend name. It can be
14988 used in frontends to check which backend will handle requests by default.
14989
Willy Tarreau74ca5042013-06-11 23:12:07 +020014990fe_id : integer
14991 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010014992 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020014993 coming via a same frontend to the same server.
14994
Marcin Deranekd2471c22016-12-12 14:08:05 +010014995fe_name : string
14996 Returns a string containing the current frontend's name. It can be used in
14997 backends to check from which frontend it was called, or to stick all users
14998 coming via a same frontend to the same server.
14999
Cyril Bonté62ba8702014-04-22 23:52:25 +020015000sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015001sc0_bytes_in_rate([<table>]) : integer
15002sc1_bytes_in_rate([<table>]) : integer
15003sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015004 Returns the average client-to-server bytes rate from the currently tracked
15005 counters, measured in amount of bytes over the period configured in the
15006 table. See also src_bytes_in_rate.
15007
Cyril Bonté62ba8702014-04-22 23:52:25 +020015008sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015009sc0_bytes_out_rate([<table>]) : integer
15010sc1_bytes_out_rate([<table>]) : integer
15011sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015012 Returns the average server-to-client bytes rate from the currently tracked
15013 counters, measured in amount of bytes over the period configured in the
15014 table. See also src_bytes_out_rate.
15015
Cyril Bonté62ba8702014-04-22 23:52:25 +020015016sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015017sc0_clr_gpc0([<table>]) : integer
15018sc1_clr_gpc0([<table>]) : integer
15019sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020015020 Clears the first General Purpose Counter associated to the currently tracked
15021 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010015022 stored value is zero, so first invocation will always return zero. This is
15023 typically used as a second ACL in an expression in order to mark a connection
15024 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015025
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015026 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015027 # block if 5 consecutive requests continue to come faster than 10 sess
15028 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015029 acl abuse sc0_http_req_rate gt 10
15030 acl kill sc0_inc_gpc0 gt 5
15031 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015032 tcp-request connection accept if !abuse save
15033 tcp-request connection reject if abuse kill
15034
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015035sc_clr_gpc1(<ctr>[,<table>]) : integer
15036sc0_clr_gpc1([<table>]) : integer
15037sc1_clr_gpc1([<table>]) : integer
15038sc2_clr_gpc1([<table>]) : integer
15039 Clears the second General Purpose Counter associated to the currently tracked
15040 counters, and returns its previous value. Before the first invocation, the
15041 stored value is zero, so first invocation will always return zero. This is
15042 typically used as a second ACL in an expression in order to mark a connection
15043 when a first ACL was verified.
15044
Cyril Bonté62ba8702014-04-22 23:52:25 +020015045sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015046sc0_conn_cnt([<table>]) : integer
15047sc1_conn_cnt([<table>]) : integer
15048sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015049 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015050 counters. See also src_conn_cnt.
15051
Cyril Bonté62ba8702014-04-22 23:52:25 +020015052sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015053sc0_conn_cur([<table>]) : integer
15054sc1_conn_cur([<table>]) : integer
15055sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015056 Returns the current amount of concurrent connections tracking the same
15057 tracked counters. This number is automatically incremented when tracking
15058 begins and decremented when tracking stops. See also src_conn_cur.
15059
Cyril Bonté62ba8702014-04-22 23:52:25 +020015060sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015061sc0_conn_rate([<table>]) : integer
15062sc1_conn_rate([<table>]) : integer
15063sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015064 Returns the average connection rate from the currently tracked counters,
15065 measured in amount of connections over the period configured in the table.
15066 See also src_conn_rate.
15067
Cyril Bonté62ba8702014-04-22 23:52:25 +020015068sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015069sc0_get_gpc0([<table>]) : integer
15070sc1_get_gpc0([<table>]) : integer
15071sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015072 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015073 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015074
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015075sc_get_gpc1(<ctr>[,<table>]) : integer
15076sc0_get_gpc1([<table>]) : integer
15077sc1_get_gpc1([<table>]) : integer
15078sc2_get_gpc1([<table>]) : integer
15079 Returns the value of the second General Purpose Counter associated to the
15080 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
15081
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015082sc_get_gpt0(<ctr>[,<table>]) : integer
15083sc0_get_gpt0([<table>]) : integer
15084sc1_get_gpt0([<table>]) : integer
15085sc2_get_gpt0([<table>]) : integer
15086 Returns the value of the first General Purpose Tag associated to the
15087 currently tracked counters. See also src_get_gpt0.
15088
Cyril Bonté62ba8702014-04-22 23:52:25 +020015089sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015090sc0_gpc0_rate([<table>]) : integer
15091sc1_gpc0_rate([<table>]) : integer
15092sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015093 Returns the average increment rate of the first General Purpose Counter
15094 associated to the currently tracked counters. It reports the frequency
15095 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015096 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15097 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15098 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020015099
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015100sc_gpc1_rate(<ctr>[,<table>]) : integer
15101sc0_gpc1_rate([<table>]) : integer
15102sc1_gpc1_rate([<table>]) : integer
15103sc2_gpc1_rate([<table>]) : integer
15104 Returns the average increment rate of the second General Purpose Counter
15105 associated to the currently tracked counters. It reports the frequency
15106 which the gpc1 counter was incremented over the configured period. See also
15107 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15108 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15109 be returned, as "gpc1" only holds the event count.
15110
Cyril Bonté62ba8702014-04-22 23:52:25 +020015111sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015112sc0_http_err_cnt([<table>]) : integer
15113sc1_http_err_cnt([<table>]) : integer
15114sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015115 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015116 counters. This includes the both request errors and 4xx error responses.
15117 See also src_http_err_cnt.
15118
Cyril Bonté62ba8702014-04-22 23:52:25 +020015119sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015120sc0_http_err_rate([<table>]) : integer
15121sc1_http_err_rate([<table>]) : integer
15122sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015123 Returns the average rate of HTTP errors from the currently tracked counters,
15124 measured in amount of errors over the period configured in the table. This
15125 includes the both request errors and 4xx error responses. See also
15126 src_http_err_rate.
15127
Cyril Bonté62ba8702014-04-22 23:52:25 +020015128sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015129sc0_http_req_cnt([<table>]) : integer
15130sc1_http_req_cnt([<table>]) : integer
15131sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015132 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015133 counters. This includes every started request, valid or not. See also
15134 src_http_req_cnt.
15135
Cyril Bonté62ba8702014-04-22 23:52:25 +020015136sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015137sc0_http_req_rate([<table>]) : integer
15138sc1_http_req_rate([<table>]) : integer
15139sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015140 Returns the average rate of HTTP requests from the currently tracked
15141 counters, measured in amount of requests over the period configured in
15142 the table. This includes every started request, valid or not. See also
15143 src_http_req_rate.
15144
Cyril Bonté62ba8702014-04-22 23:52:25 +020015145sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015146sc0_inc_gpc0([<table>]) : integer
15147sc1_inc_gpc0([<table>]) : integer
15148sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015149 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010015150 tracked counters, and returns its new value. Before the first invocation,
15151 the stored value is zero, so first invocation will increase it to 1 and will
15152 return 1. This is typically used as a second ACL in an expression in order
15153 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020015154
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015155 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015156 acl abuse sc0_http_req_rate gt 10
15157 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015158 tcp-request connection reject if abuse kill
15159
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015160sc_inc_gpc1(<ctr>[,<table>]) : integer
15161sc0_inc_gpc1([<table>]) : integer
15162sc1_inc_gpc1([<table>]) : integer
15163sc2_inc_gpc1([<table>]) : integer
15164 Increments the second General Purpose Counter associated to the currently
15165 tracked counters, and returns its new value. Before the first invocation,
15166 the stored value is zero, so first invocation will increase it to 1 and will
15167 return 1. This is typically used as a second ACL in an expression in order
15168 to mark a connection when a first ACL was verified.
15169
Cyril Bonté62ba8702014-04-22 23:52:25 +020015170sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015171sc0_kbytes_in([<table>]) : integer
15172sc1_kbytes_in([<table>]) : integer
15173sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015174 Returns the total amount of client-to-server data from the currently tracked
15175 counters, measured in kilobytes. The test is currently performed on 32-bit
15176 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020015177
Cyril Bonté62ba8702014-04-22 23:52:25 +020015178sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015179sc0_kbytes_out([<table>]) : integer
15180sc1_kbytes_out([<table>]) : integer
15181sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015182 Returns the total amount of server-to-client data from the currently tracked
15183 counters, measured in kilobytes. The test is currently performed on 32-bit
15184 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020015185
Cyril Bonté62ba8702014-04-22 23:52:25 +020015186sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015187sc0_sess_cnt([<table>]) : integer
15188sc1_sess_cnt([<table>]) : integer
15189sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015190 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020015191 into sessions, which means that they were accepted by a "tcp-request
15192 connection" rule, from the currently tracked counters. A backend may count
15193 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015194 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020015195 with the client. See also src_sess_cnt.
15196
Cyril Bonté62ba8702014-04-22 23:52:25 +020015197sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015198sc0_sess_rate([<table>]) : integer
15199sc1_sess_rate([<table>]) : integer
15200sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015201 Returns the average session rate from the currently tracked counters,
15202 measured in amount of sessions over the period configured in the table. A
15203 session is a connection that got past the early "tcp-request connection"
15204 rules. A backend may count more sessions than connections because each
15205 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015206 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020015207
Cyril Bonté62ba8702014-04-22 23:52:25 +020015208sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020015209sc0_tracked([<table>]) : boolean
15210sc1_tracked([<table>]) : boolean
15211sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020015212 Returns true if the designated session counter is currently being tracked by
15213 the current session. This can be useful when deciding whether or not we want
15214 to set some values in a header passed to the server.
15215
Cyril Bonté62ba8702014-04-22 23:52:25 +020015216sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015217sc0_trackers([<table>]) : integer
15218sc1_trackers([<table>]) : integer
15219sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010015220 Returns the current amount of concurrent connections tracking the same
15221 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015222 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010015223 that it does not rely on any stored information but on the table's reference
15224 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015225 may sometimes be more suited for layer7 tracking. It can be used to tell a
15226 server how many concurrent connections there are from a given address for
15227 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010015228
Willy Tarreau74ca5042013-06-11 23:12:07 +020015229so_id : integer
15230 Returns an integer containing the current listening socket's id. It is useful
15231 in frontends involving many "bind" lines, or to stick all users coming via a
15232 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015233
Willy Tarreau74ca5042013-06-11 23:12:07 +020015234src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010015235 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020015236 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
15237 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
15238 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015239 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
15240 directive is used, it can be the address of a client behind another
15241 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015242 "tcp-request connection" which sees the real address. When the incoming
15243 connection passed through address translation or redirection involving
15244 connection tracking, the original destination address before the redirection
15245 will be reported. On Linux systems, the source and destination may seldom
15246 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
15247 response may reopen a timed out connection and switch what is believed to be
15248 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015249
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010015250 Example:
15251 # add an HTTP header in requests with the originating address' country
15252 http-request set-header X-Country %[src,map_ip(geoip.lst)]
15253
Willy Tarreau74ca5042013-06-11 23:12:07 +020015254src_bytes_in_rate([<table>]) : integer
15255 Returns the average bytes rate from the incoming connection's source address
15256 in the current proxy's stick-table or in the designated stick-table, measured
15257 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015258 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015259
Willy Tarreau74ca5042013-06-11 23:12:07 +020015260src_bytes_out_rate([<table>]) : integer
15261 Returns the average bytes rate to the incoming connection's source address in
15262 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015263 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015264 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015265
Willy Tarreau74ca5042013-06-11 23:12:07 +020015266src_clr_gpc0([<table>]) : integer
15267 Clears the first General Purpose Counter associated to the incoming
15268 connection's source address in the current proxy's stick-table or in the
15269 designated stick-table, and returns its previous value. If the address is not
15270 found, an entry is created and 0 is returned. This is typically used as a
15271 second ACL in an expression in order to mark a connection when a first ACL
15272 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015273
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015274 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015275 # block if 5 consecutive requests continue to come faster than 10 sess
15276 # per second, and reset the counter as soon as the traffic slows down.
15277 acl abuse src_http_req_rate gt 10
15278 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010015279 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015280 tcp-request connection accept if !abuse save
15281 tcp-request connection reject if abuse kill
15282
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015283src_clr_gpc1([<table>]) : integer
15284 Clears the second General Purpose Counter associated to the incoming
15285 connection's source address in the current proxy's stick-table or in the
15286 designated stick-table, and returns its previous value. If the address is not
15287 found, an entry is created and 0 is returned. This is typically used as a
15288 second ACL in an expression in order to mark a connection when a first ACL
15289 was verified.
15290
Willy Tarreau74ca5042013-06-11 23:12:07 +020015291src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015292 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020015293 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015294 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015295 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015296
Willy Tarreau74ca5042013-06-11 23:12:07 +020015297src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015298 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015299 current incoming connection's source address in the current proxy's
15300 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015301 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015302
Willy Tarreau74ca5042013-06-11 23:12:07 +020015303src_conn_rate([<table>]) : integer
15304 Returns the average connection rate from the incoming connection's source
15305 address in the current proxy's stick-table or in the designated stick-table,
15306 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015307 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015308
Willy Tarreau74ca5042013-06-11 23:12:07 +020015309src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015310 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015311 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015312 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015313 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015314
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015315src_get_gpc1([<table>]) : integer
15316 Returns the value of the second General Purpose Counter associated to the
15317 incoming connection's source address in the current proxy's stick-table or in
15318 the designated stick-table. If the address is not found, zero is returned.
15319 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
15320
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015321src_get_gpt0([<table>]) : integer
15322 Returns the value of the first General Purpose Tag associated to the
15323 incoming connection's source address in the current proxy's stick-table or in
15324 the designated stick-table. If the address is not found, zero is returned.
15325 See also sc/sc0/sc1/sc2_get_gpt0.
15326
Willy Tarreau74ca5042013-06-11 23:12:07 +020015327src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015328 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020015329 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015330 stick-table or in the designated stick-table. It reports the frequency
15331 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015332 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15333 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15334 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015335
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015336src_gpc1_rate([<table>]) : integer
15337 Returns the average increment rate of the second General Purpose Counter
15338 associated to the incoming connection's source address in the current proxy's
15339 stick-table or in the designated stick-table. It reports the frequency
15340 which the gpc1 counter was incremented over the configured period. See also
15341 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15342 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15343 be returned, as "gpc1" only holds the event count.
15344
Willy Tarreau74ca5042013-06-11 23:12:07 +020015345src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015346 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015347 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020015348 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015349 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020015350 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015351
Willy Tarreau74ca5042013-06-11 23:12:07 +020015352src_http_err_rate([<table>]) : integer
15353 Returns the average rate of HTTP errors from the incoming connection's source
15354 address in the current proxy's stick-table or in the designated stick-table,
15355 measured in amount of errors over the period configured in the table. This
15356 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015357 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015358
Willy Tarreau74ca5042013-06-11 23:12:07 +020015359src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015360 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015361 source address in the current proxy's stick-table or in the designated stick-
15362 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015363 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015364
Willy Tarreau74ca5042013-06-11 23:12:07 +020015365src_http_req_rate([<table>]) : integer
15366 Returns the average rate of HTTP requests from the incoming connection's
15367 source address in the current proxy's stick-table or in the designated stick-
15368 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020015369 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015370 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015371
Willy Tarreau74ca5042013-06-11 23:12:07 +020015372src_inc_gpc0([<table>]) : integer
15373 Increments the first General Purpose Counter associated to the incoming
15374 connection's source address in the current proxy's stick-table or in the
15375 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015376 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015377 This is typically used as a second ACL in an expression in order to mark a
15378 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020015379
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015380 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020015381 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010015382 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015383 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020015384
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015385src_inc_gpc1([<table>]) : integer
15386 Increments the second General Purpose Counter associated to the incoming
15387 connection's source address in the current proxy's stick-table or in the
15388 designated stick-table, and returns its new value. If the address is not
15389 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
15390 This is typically used as a second ACL in an expression in order to mark a
15391 connection when a first ACL was verified.
15392
Willy Tarreau16e01562016-08-09 16:46:18 +020015393src_is_local : boolean
15394 Returns true if the source address of the incoming connection is local to the
15395 system, or false if the address doesn't exist on the system, meaning that it
15396 comes from a remote machine. Note that UNIX addresses are considered local.
15397 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015398 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020015399 note that the check involves a few system calls, so it's better to do it only
15400 once per connection.
15401
Willy Tarreau74ca5042013-06-11 23:12:07 +020015402src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015403 Returns the total amount of data received from the incoming connection's
15404 source address in the current proxy's stick-table or in the designated
15405 stick-table, measured in kilobytes. If the address is not found, zero is
15406 returned. The test is currently performed on 32-bit integers, which limits
15407 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015408
Willy Tarreau74ca5042013-06-11 23:12:07 +020015409src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015410 Returns the total amount of data sent to the incoming connection's source
15411 address in the current proxy's stick-table or in the designated stick-table,
15412 measured in kilobytes. If the address is not found, zero is returned. The
15413 test is currently performed on 32-bit integers, which limits values to 4
15414 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015415
Willy Tarreau74ca5042013-06-11 23:12:07 +020015416src_port : integer
15417 Returns an integer value corresponding to the TCP source port of the
15418 connection on the client side, which is the port the client connected from.
15419 Usage of this function is very limited as modern protocols do not care much
15420 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010015421
Willy Tarreau74ca5042013-06-11 23:12:07 +020015422src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015423 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020015424 connection's source IPv4 address in the current proxy's stick-table or in the
15425 designated stick-table, that were transformed into sessions, which means that
15426 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015427 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015428
Willy Tarreau74ca5042013-06-11 23:12:07 +020015429src_sess_rate([<table>]) : integer
15430 Returns the average session rate from the incoming connection's source
15431 address in the current proxy's stick-table or in the designated stick-table,
15432 measured in amount of sessions over the period configured in the table. A
15433 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015434 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015435
Willy Tarreau74ca5042013-06-11 23:12:07 +020015436src_updt_conn_cnt([<table>]) : integer
15437 Creates or updates the entry associated to the incoming connection's source
15438 address in the current proxy's stick-table or in the designated stick-table.
15439 This table must be configured to store the "conn_cnt" data type, otherwise
15440 the match will be ignored. The current count is incremented by one, and the
15441 expiration timer refreshed. The updated count is returned, so this match
15442 can't return zero. This was used to reject service abusers based on their
15443 source address. Note: it is recommended to use the more complete "track-sc*"
15444 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015445
15446 Example :
15447 # This frontend limits incoming SSH connections to 3 per 10 second for
15448 # each source address, and rejects excess connections until a 10 second
15449 # silence is observed. At most 20 addresses are tracked.
15450 listen ssh
15451 bind :22
15452 mode tcp
15453 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020015454 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020015455 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015456 server local 127.0.0.1:22
15457
Willy Tarreau74ca5042013-06-11 23:12:07 +020015458srv_id : integer
15459 Returns an integer containing the server's id when processing the response.
15460 While it's almost only used with ACLs, it may be used for logging or
15461 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020015462
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200154637.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020015464----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020015465
Willy Tarreau74ca5042013-06-11 23:12:07 +020015466The layer 5 usually describes just the session layer which in haproxy is
15467closest to the session once all the connection handshakes are finished, but
15468when no content is yet made available. The fetch methods described here are
15469usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015470future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020015471
Ben Shillitof25e8e52016-12-02 14:25:37 +00001547251d.all(<prop>[,<prop>*]) : string
15473 Returns values for the properties requested as a string, where values are
15474 separated by the delimiter specified with "51degrees-property-separator".
15475 The device is identified using all the important HTTP headers from the
15476 request. The function can be passed up to five property names, and if a
15477 property name can't be found, the value "NoData" is returned.
15478
15479 Example :
15480 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
15481 # containing the three properties requested using all relevant headers from
15482 # the request.
15483 frontend http-in
15484 bind *:8081
15485 default_backend servers
15486 http-request set-header X-51D-DeviceTypeMobileTablet \
15487 %[51d.all(DeviceType,IsMobile,IsTablet)]
15488
Emeric Brun645ae792014-04-30 14:21:06 +020015489ssl_bc : boolean
15490 Returns true when the back connection was made via an SSL/TLS transport
15491 layer and is locally deciphered. This means the outgoing connection was made
15492 other a server with the "ssl" option.
15493
15494ssl_bc_alg_keysize : integer
15495 Returns the symmetric cipher key size supported in bits when the outgoing
15496 connection was made over an SSL/TLS transport layer.
15497
Olivier Houchard6b77f492018-11-22 18:18:29 +010015498ssl_bc_alpn : string
15499 This extracts the Application Layer Protocol Negotiation field from an
15500 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020015501 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010015502 server. The SSL library must have been built with support for TLS
15503 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15504 not advertised unless the "alpn" keyword on the "server" line specifies a
15505 protocol list. Also, nothing forces the server to pick a protocol from this
15506 list, any other one may be requested. The TLS ALPN extension is meant to
15507 replace the TLS NPN extension. See also "ssl_bc_npn".
15508
Emeric Brun645ae792014-04-30 14:21:06 +020015509ssl_bc_cipher : string
15510 Returns the name of the used cipher when the outgoing connection was made
15511 over an SSL/TLS transport layer.
15512
Patrick Hemmer65674662019-06-04 08:13:03 -040015513ssl_bc_client_random : binary
15514 Returns the client random of the back connection when the incoming connection
15515 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15516 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15517
Emeric Brun74f7ffa2018-02-19 16:14:12 +010015518ssl_bc_is_resumed : boolean
15519 Returns true when the back connection was made over an SSL/TLS transport
15520 layer and the newly created SSL session was resumed using a cached
15521 session or a TLS ticket.
15522
Olivier Houchard6b77f492018-11-22 18:18:29 +010015523ssl_bc_npn : string
15524 This extracts the Next Protocol Negotiation field from an outgoing connection
15525 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020015526 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010015527 built with support for TLS extensions enabled (check haproxy -vv). Note that
15528 the TLS NPN extension is not advertised unless the "npn" keyword on the
15529 "server" line specifies a protocol list. Also, nothing forces the server to
15530 pick a protocol from this list, any other one may be used. Please note that
15531 the TLS NPN extension was replaced with ALPN.
15532
Emeric Brun645ae792014-04-30 14:21:06 +020015533ssl_bc_protocol : string
15534 Returns the name of the used protocol when the outgoing connection was made
15535 over an SSL/TLS transport layer.
15536
Emeric Brunb73a9b02014-04-30 18:49:19 +020015537ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020015538 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015539 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15540 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020015541
Patrick Hemmer65674662019-06-04 08:13:03 -040015542ssl_bc_server_random : binary
15543 Returns the server random of the back connection when the incoming connection
15544 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15545 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15546
Emeric Brun645ae792014-04-30 14:21:06 +020015547ssl_bc_session_id : binary
15548 Returns the SSL ID of the back connection when the outgoing connection was
15549 made over an SSL/TLS transport layer. It is useful to log if we want to know
15550 if session was reused or not.
15551
Patrick Hemmere0275472018-04-28 19:15:51 -040015552ssl_bc_session_key : binary
15553 Returns the SSL session master key of the back connection when the outgoing
15554 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15555 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15556 BoringSSL.
15557
Emeric Brun645ae792014-04-30 14:21:06 +020015558ssl_bc_use_keysize : integer
15559 Returns the symmetric cipher key size used in bits when the outgoing
15560 connection was made over an SSL/TLS transport layer.
15561
Willy Tarreau74ca5042013-06-11 23:12:07 +020015562ssl_c_ca_err : integer
15563 When the incoming connection was made over an SSL/TLS transport layer,
15564 returns the ID of the first error detected during verification of the client
15565 certificate at depth > 0, or 0 if no error was encountered during this
15566 verification process. Please refer to your SSL library's documentation to
15567 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020015568
Willy Tarreau74ca5042013-06-11 23:12:07 +020015569ssl_c_ca_err_depth : integer
15570 When the incoming connection was made over an SSL/TLS transport layer,
15571 returns the depth in the CA chain of the first error detected during the
15572 verification of the client certificate. If no error is encountered, 0 is
15573 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015574
Emeric Brun43e79582014-10-29 19:03:26 +010015575ssl_c_der : binary
15576 Returns the DER formatted certificate presented by the client when the
15577 incoming connection was made over an SSL/TLS transport layer. When used for
15578 an ACL, the value(s) to match against can be passed in hexadecimal form.
15579
Willy Tarreau74ca5042013-06-11 23:12:07 +020015580ssl_c_err : integer
15581 When the incoming connection was made over an SSL/TLS transport layer,
15582 returns the ID of the first error detected during verification at depth 0, or
15583 0 if no error was encountered during this verification process. Please refer
15584 to your SSL library's documentation to find the exhaustive list of error
15585 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020015586
Willy Tarreau74ca5042013-06-11 23:12:07 +020015587ssl_c_i_dn([<entry>[,<occ>]]) : string
15588 When the incoming connection was made over an SSL/TLS transport layer,
15589 returns the full distinguished name of the issuer of the certificate
15590 presented by the client when no <entry> is specified, or the value of the
15591 first given entry found from the beginning of the DN. If a positive/negative
15592 occurrence number is specified as the optional second argument, it returns
15593 the value of the nth given entry value from the beginning/end of the DN.
15594 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
15595 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020015596
Willy Tarreau74ca5042013-06-11 23:12:07 +020015597ssl_c_key_alg : string
15598 Returns the name of the algorithm used to generate the key of the certificate
15599 presented by the client when the incoming connection was made over an SSL/TLS
15600 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020015601
Willy Tarreau74ca5042013-06-11 23:12:07 +020015602ssl_c_notafter : string
15603 Returns the end date presented by the client as a formatted string
15604 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15605 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020015606
Willy Tarreau74ca5042013-06-11 23:12:07 +020015607ssl_c_notbefore : string
15608 Returns the start date presented by the client as a formatted string
15609 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15610 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015611
Willy Tarreau74ca5042013-06-11 23:12:07 +020015612ssl_c_s_dn([<entry>[,<occ>]]) : string
15613 When the incoming connection was made over an SSL/TLS transport layer,
15614 returns the full distinguished name of the subject of the certificate
15615 presented by the client when no <entry> is specified, or the value of the
15616 first given entry found from the beginning of the DN. If a positive/negative
15617 occurrence number is specified as the optional second argument, it returns
15618 the value of the nth given entry value from the beginning/end of the DN.
15619 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
15620 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015621
Willy Tarreau74ca5042013-06-11 23:12:07 +020015622ssl_c_serial : binary
15623 Returns the serial of the certificate presented by the client when the
15624 incoming connection was made over an SSL/TLS transport layer. When used for
15625 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015626
Willy Tarreau74ca5042013-06-11 23:12:07 +020015627ssl_c_sha1 : binary
15628 Returns the SHA-1 fingerprint of the certificate presented by the client when
15629 the incoming connection was made over an SSL/TLS transport layer. This can be
15630 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015631 Note that the output is binary, so if you want to pass that signature to the
15632 server, you need to encode it in hex or base64, such as in the example below:
15633
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015634 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015635 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020015636
Willy Tarreau74ca5042013-06-11 23:12:07 +020015637ssl_c_sig_alg : string
15638 Returns the name of the algorithm used to sign the certificate presented by
15639 the client when the incoming connection was made over an SSL/TLS transport
15640 layer.
Emeric Brun87855892012-10-17 17:39:35 +020015641
Willy Tarreau74ca5042013-06-11 23:12:07 +020015642ssl_c_used : boolean
15643 Returns true if current SSL session uses a client certificate even if current
15644 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020015645
Willy Tarreau74ca5042013-06-11 23:12:07 +020015646ssl_c_verify : integer
15647 Returns the verify result error ID when the incoming connection was made over
15648 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
15649 refer to your SSL library's documentation for an exhaustive list of error
15650 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020015651
Willy Tarreau74ca5042013-06-11 23:12:07 +020015652ssl_c_version : integer
15653 Returns the version of the certificate presented by the client when the
15654 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020015655
Emeric Brun43e79582014-10-29 19:03:26 +010015656ssl_f_der : binary
15657 Returns the DER formatted certificate presented by the frontend when the
15658 incoming connection was made over an SSL/TLS transport layer. When used for
15659 an ACL, the value(s) to match against can be passed in hexadecimal form.
15660
Willy Tarreau74ca5042013-06-11 23:12:07 +020015661ssl_f_i_dn([<entry>[,<occ>]]) : string
15662 When the incoming connection was made over an SSL/TLS transport layer,
15663 returns the full distinguished name of the issuer of the certificate
15664 presented by the frontend when no <entry> is specified, or the value of the
15665 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020015666 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020015667 the value of the nth given entry value from the beginning/end of the DN.
15668 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
15669 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020015670
Willy Tarreau74ca5042013-06-11 23:12:07 +020015671ssl_f_key_alg : string
15672 Returns the name of the algorithm used to generate the key of the certificate
15673 presented by the frontend when the incoming connection was made over an
15674 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015675
Willy Tarreau74ca5042013-06-11 23:12:07 +020015676ssl_f_notafter : string
15677 Returns the end date presented by the frontend as a formatted string
15678 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15679 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015680
Willy Tarreau74ca5042013-06-11 23:12:07 +020015681ssl_f_notbefore : string
15682 Returns the start date presented by the frontend as a formatted string
15683 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15684 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020015685
Willy Tarreau74ca5042013-06-11 23:12:07 +020015686ssl_f_s_dn([<entry>[,<occ>]]) : string
15687 When the incoming connection was made over an SSL/TLS transport layer,
15688 returns the full distinguished name of the subject of the certificate
15689 presented by the frontend when no <entry> is specified, or the value of the
15690 first given entry found from the beginning of the DN. If a positive/negative
15691 occurrence number is specified as the optional second argument, it returns
15692 the value of the nth given entry value from the beginning/end of the DN.
15693 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
15694 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020015695
Willy Tarreau74ca5042013-06-11 23:12:07 +020015696ssl_f_serial : binary
15697 Returns the serial of the certificate presented by the frontend when the
15698 incoming connection was made over an SSL/TLS transport layer. When used for
15699 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020015700
Emeric Brun55f4fa82014-04-30 17:11:25 +020015701ssl_f_sha1 : binary
15702 Returns the SHA-1 fingerprint of the certificate presented by the frontend
15703 when the incoming connection was made over an SSL/TLS transport layer. This
15704 can be used to know which certificate was chosen using SNI.
15705
Willy Tarreau74ca5042013-06-11 23:12:07 +020015706ssl_f_sig_alg : string
15707 Returns the name of the algorithm used to sign the certificate presented by
15708 the frontend when the incoming connection was made over an SSL/TLS transport
15709 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015710
Willy Tarreau74ca5042013-06-11 23:12:07 +020015711ssl_f_version : integer
15712 Returns the version of the certificate presented by the frontend when the
15713 incoming connection was made over an SSL/TLS transport layer.
15714
15715ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015716 Returns true when the front connection was made via an SSL/TLS transport
15717 layer and is locally deciphered. This means it has matched a socket declared
15718 with a "bind" line having the "ssl" option.
15719
Willy Tarreau74ca5042013-06-11 23:12:07 +020015720 Example :
15721 # This passes "X-Proto: https" to servers when client connects over SSL
15722 listen http-https
15723 bind :80
15724 bind :443 ssl crt /etc/haproxy.pem
15725 http-request add-header X-Proto https if { ssl_fc }
15726
15727ssl_fc_alg_keysize : integer
15728 Returns the symmetric cipher key size supported in bits when the incoming
15729 connection was made over an SSL/TLS transport layer.
15730
15731ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015732 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020015733 incoming connection made via a TLS transport layer and locally deciphered by
15734 haproxy. The result is a string containing the protocol name advertised by
15735 the client. 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 "bind" line specifies a
15738 protocol list. Also, nothing forces the client 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_fc_npn".
15741
Willy Tarreau74ca5042013-06-11 23:12:07 +020015742ssl_fc_cipher : string
15743 Returns the name of the used cipher when the incoming connection was made
15744 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020015745
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015746ssl_fc_cipherlist_bin : binary
15747 Returns the binary form of the client hello cipher list. The maximum returned
15748 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015749 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015750
15751ssl_fc_cipherlist_hex : string
15752 Returns the binary form of the client hello cipher list encoded as
15753 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015754 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015755
15756ssl_fc_cipherlist_str : string
15757 Returns the decoded text form of the client hello cipher list. The maximum
15758 number of ciphers returned is according with the value of
15759 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010015760 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020015761 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015762
15763ssl_fc_cipherlist_xxh : integer
15764 Returns a xxh64 of the cipher list. This hash can be return only is the value
15765 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015766 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015767
Patrick Hemmer65674662019-06-04 08:13:03 -040015768ssl_fc_client_random : binary
15769 Returns the client random of the front connection when the incoming connection
15770 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15771 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15772
Willy Tarreau74ca5042013-06-11 23:12:07 +020015773ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015774 Returns true if a client certificate is present in an incoming connection over
15775 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010015776 Note: on SSL session resumption with Session ID or TLS ticket, client
15777 certificate is not present in the current connection but may be retrieved
15778 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
15779 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015780
Olivier Houchardccaa7de2017-10-02 11:51:03 +020015781ssl_fc_has_early : boolean
15782 Returns true if early data were sent, and the handshake didn't happen yet. As
15783 it has security implications, it is useful to be able to refuse those, or
15784 wait until the handshake happened.
15785
Willy Tarreau74ca5042013-06-11 23:12:07 +020015786ssl_fc_has_sni : boolean
15787 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020015788 in an incoming connection was made over an SSL/TLS transport layer. Returns
15789 true when the incoming connection presents a TLS SNI field. This requires
15790 that the SSL library is build with support for TLS extensions enabled (check
15791 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020015792
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020015793ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015794 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010015795 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
15796 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015797
Willy Tarreau74ca5042013-06-11 23:12:07 +020015798ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015799 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020015800 made via a TLS transport layer and locally deciphered by haproxy. The result
15801 is a string containing the protocol name advertised by the client. The SSL
15802 library must have been built with support for TLS extensions enabled (check
15803 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
15804 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
15805 forces the client to pick a protocol from this list, any other one may be
15806 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020015807
Willy Tarreau74ca5042013-06-11 23:12:07 +020015808ssl_fc_protocol : string
15809 Returns the name of the used protocol when the incoming connection was made
15810 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020015811
Emeric Brunb73a9b02014-04-30 18:49:19 +020015812ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040015813 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015814 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15815 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040015816
Patrick Hemmer65674662019-06-04 08:13:03 -040015817ssl_fc_server_random : binary
15818 Returns the server random of the front connection when the incoming connection
15819 was made over an SSL/TLS transport layer. It is useful to to decrypt traffic
15820 sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or BoringSSL.
15821
Willy Tarreau74ca5042013-06-11 23:12:07 +020015822ssl_fc_session_id : binary
15823 Returns the SSL ID of the front connection when the incoming connection was
15824 made over an SSL/TLS transport layer. It is useful to stick a given client to
15825 a server. It is important to note that some browsers refresh their session ID
15826 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020015827
Patrick Hemmere0275472018-04-28 19:15:51 -040015828ssl_fc_session_key : binary
15829 Returns the SSL session master key of the front connection when the incoming
15830 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15831 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15832 BoringSSL.
15833
15834
Willy Tarreau74ca5042013-06-11 23:12:07 +020015835ssl_fc_sni : string
15836 This extracts the Server Name Indication TLS extension (SNI) field from an
15837 incoming connection made via an SSL/TLS transport layer and locally
15838 deciphered by haproxy. The result (when present) typically is a string
15839 matching the HTTPS host name (253 chars or less). The SSL library must have
15840 been built with support for TLS extensions enabled (check haproxy -vv).
15841
15842 This fetch is different from "req_ssl_sni" above in that it applies to the
15843 connection being deciphered by haproxy and not to SSL contents being blindly
15844 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020015845 requires that the SSL library is build with support for TLS extensions
15846 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020015847
Willy Tarreau74ca5042013-06-11 23:12:07 +020015848 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020015849 ssl_fc_sni_end : suffix match
15850 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020015851
Willy Tarreau74ca5042013-06-11 23:12:07 +020015852ssl_fc_use_keysize : integer
15853 Returns the symmetric cipher key size used in bits when the incoming
15854 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015855
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015856
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200158577.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020015858------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015859
Willy Tarreau74ca5042013-06-11 23:12:07 +020015860Fetching samples from buffer contents is a bit different from the previous
15861sample fetches above because the sampled data are ephemeral. These data can
15862only be used when they're available and will be lost when they're forwarded.
15863For this reason, samples fetched from buffer contents during a request cannot
15864be used in a response for example. Even while the data are being fetched, they
15865can change. Sometimes it is necessary to set some delays or combine multiple
15866sample fetch methods to ensure that the expected data are complete and usable,
15867for example through TCP request content inspection. Please see the "tcp-request
15868content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020015869
Willy Tarreau74ca5042013-06-11 23:12:07 +020015870payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010015871 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015872 "stick on", "stick match"), and for "res.payload" when used in the context of
15873 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015874
Willy Tarreau74ca5042013-06-11 23:12:07 +020015875payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
15876 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010015877 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015878 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015879
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020015880req.hdrs : string
15881 Returns the current request headers as string including the last empty line
15882 separating headers from the request body. The last empty line can be used to
15883 detect a truncated header block. This sample fetch is useful for some SPOE
15884 headers analyzers and for advanced logging.
15885
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020015886req.hdrs_bin : binary
15887 Returns the current request headers contained in preparsed binary form. This
15888 is useful for offloading some processing with SPOE. Each string is described
15889 by a length followed by the number of bytes indicated in the length. The
15890 length is represented using the variable integer encoding detailed in the
15891 SPOE documentation. The end of the list is marked by a couple of empty header
15892 names and values (length of 0 for both).
15893
15894 *(<str:header-name><str:header-value>)<empty string><empty string>
15895
15896 int: refer to the SPOE documentation for the encoding
15897 str: <int:length><bytes>
15898
Willy Tarreau74ca5042013-06-11 23:12:07 +020015899req.len : integer
15900req_len : integer (deprecated)
15901 Returns an integer value corresponding to the number of bytes present in the
15902 request buffer. This is mostly used in ACL. It is important to understand
15903 that this test does not return false as long as the buffer is changing. This
15904 means that a check with equality to zero will almost always immediately match
15905 at the beginning of the session, while a test for more data will wait for
15906 that data to come in and return false only when haproxy is certain that no
15907 more data will come in. This test was designed to be used with TCP request
15908 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015909
Willy Tarreau74ca5042013-06-11 23:12:07 +020015910req.payload(<offset>,<length>) : binary
15911 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015912 in the request buffer. As a special case, if the <length> argument is zero,
15913 the the whole buffer from <offset> to the end is extracted. This can be used
15914 with ACLs in order to check for the presence of some content in a buffer at
15915 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015916
Willy Tarreau74ca5042013-06-11 23:12:07 +020015917 ACL alternatives :
15918 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015919
Willy Tarreau74ca5042013-06-11 23:12:07 +020015920req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
15921 This extracts a binary block whose size is specified at <offset1> for <length>
15922 bytes, and which starts at <offset2> if specified or just after the length in
15923 the request buffer. The <offset2> parameter also supports relative offsets if
15924 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015925
Willy Tarreau74ca5042013-06-11 23:12:07 +020015926 ACL alternatives :
15927 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015928
Willy Tarreau74ca5042013-06-11 23:12:07 +020015929 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015930
Willy Tarreau74ca5042013-06-11 23:12:07 +020015931req.proto_http : boolean
15932req_proto_http : boolean (deprecated)
15933 Returns true when data in the request buffer look like HTTP and correctly
15934 parses as such. It is the same parser as the common HTTP request parser which
15935 is used so there should be no surprises. The test does not match until the
15936 request is complete, failed or timed out. This test may be used to report the
15937 protocol in TCP logs, but the biggest use is to block TCP request analysis
15938 until a complete HTTP request is present in the buffer, for example to track
15939 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015940
Willy Tarreau74ca5042013-06-11 23:12:07 +020015941 Example:
15942 # track request counts per "base" (concatenation of Host+URL)
15943 tcp-request inspect-delay 10s
15944 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015945 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015946
Willy Tarreau74ca5042013-06-11 23:12:07 +020015947req.rdp_cookie([<name>]) : string
15948rdp_cookie([<name>]) : string (deprecated)
15949 When the request buffer looks like the RDP protocol, extracts the RDP cookie
15950 <name>, or any cookie if unspecified. The parser only checks for the first
15951 cookie, as illustrated in the RDP protocol specification. The cookie name is
15952 case insensitive. Generally the "MSTS" cookie name will be used, as it can
15953 contain the user name of the client connecting to the server if properly
15954 configured on the client. The "MSTSHASH" cookie is often used as well for
15955 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015956
Willy Tarreau74ca5042013-06-11 23:12:07 +020015957 This differs from "balance rdp-cookie" in that any balancing algorithm may be
15958 used and thus the distribution of clients to backend servers is not linked to
15959 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
15960 such as "balance roundrobin" or "balance leastconn" will lead to a more even
15961 distribution of clients to backend servers than the hash used by "balance
15962 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015963
Willy Tarreau74ca5042013-06-11 23:12:07 +020015964 ACL derivatives :
15965 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015966
Willy Tarreau74ca5042013-06-11 23:12:07 +020015967 Example :
15968 listen tse-farm
15969 bind 0.0.0.0:3389
15970 # wait up to 5s for an RDP cookie in the request
15971 tcp-request inspect-delay 5s
15972 tcp-request content accept if RDP_COOKIE
15973 # apply RDP cookie persistence
15974 persist rdp-cookie
15975 # Persist based on the mstshash cookie
15976 # This is only useful makes sense if
15977 # balance rdp-cookie is not used
15978 stick-table type string size 204800
15979 stick on req.rdp_cookie(mstshash)
15980 server srv1 1.1.1.1:3389
15981 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015982
Willy Tarreau74ca5042013-06-11 23:12:07 +020015983 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
15984 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015985
Willy Tarreau74ca5042013-06-11 23:12:07 +020015986req.rdp_cookie_cnt([name]) : integer
15987rdp_cookie_cnt([name]) : integer (deprecated)
15988 Tries to parse the request buffer as RDP protocol, then returns an integer
15989 corresponding to the number of RDP cookies found. If an optional cookie name
15990 is passed, only cookies matching this name are considered. This is mostly
15991 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015992
Willy Tarreau74ca5042013-06-11 23:12:07 +020015993 ACL derivatives :
15994 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015995
Alex Zorin4afdd132018-12-30 13:56:28 +110015996req.ssl_alpn : string
15997 Returns a string containing the values of the Application-Layer Protocol
15998 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
15999 ClientHello message. Note that this only applies to raw contents found in the
16000 request buffer and not to the contents deciphered via an SSL data layer, so
16001 this will not work with "bind" lines having the "ssl" option. This is useful
16002 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020016003 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110016004
16005 Examples :
16006 # Wait for a client hello for at most 5 seconds
16007 tcp-request inspect-delay 5s
16008 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020016009 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110016010 default_backend bk_default
16011
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016012req.ssl_ec_ext : boolean
16013 Returns a boolean identifying if client sent the Supported Elliptic Curves
16014 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020016015 message. This can be used to present ECC compatible clients with EC
16016 certificate and to use RSA for all others, on the same IP address. Note that
16017 this only applies to raw contents found in the request buffer and not to
16018 contents deciphered via an SSL data layer, so this will not work with "bind"
16019 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020016020
Willy Tarreau74ca5042013-06-11 23:12:07 +020016021req.ssl_hello_type : integer
16022req_ssl_hello_type : integer (deprecated)
16023 Returns an integer value containing the type of the SSL hello message found
16024 in the request buffer if the buffer contains data that parse as a complete
16025 SSL (v3 or superior) client hello message. Note that this only applies to raw
16026 contents found in the request buffer and not to contents deciphered via an
16027 SSL data layer, so this will not work with "bind" lines having the "ssl"
16028 option. This is mostly used in ACL to detect presence of an SSL hello message
16029 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016030
Willy Tarreau74ca5042013-06-11 23:12:07 +020016031req.ssl_sni : string
16032req_ssl_sni : string (deprecated)
16033 Returns a string containing the value of the Server Name TLS extension sent
16034 by a client in a TLS stream passing through the request buffer if the buffer
16035 contains data that parse as a complete SSL (v3 or superior) client hello
16036 message. Note that this only applies to raw contents found in the request
16037 buffer and not to contents deciphered via an SSL data layer, so this will not
16038 work with "bind" lines having the "ssl" option. SNI normally contains the
16039 name of the host the client tries to connect to (for recent browsers). SNI is
16040 useful for allowing or denying access to certain hosts when SSL/TLS is used
16041 by the client. This test was designed to be used with TCP request content
16042 inspection. If content switching is needed, it is recommended to first wait
16043 for a complete client hello (type 1), like in the example below. See also
16044 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016045
Willy Tarreau74ca5042013-06-11 23:12:07 +020016046 ACL derivatives :
16047 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016048
Willy Tarreau74ca5042013-06-11 23:12:07 +020016049 Examples :
16050 # Wait for a client hello for at most 5 seconds
16051 tcp-request inspect-delay 5s
16052 tcp-request content accept if { req_ssl_hello_type 1 }
16053 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
16054 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020016055
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053016056req.ssl_st_ext : integer
16057 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
16058 Returns 1 if the client sent SessionTicket TLS Extension
16059 Returns 2 if the client also sent non-zero length TLS SessionTicket
16060 Note that this only applies to raw contents found in the request buffer and
16061 not to contents deciphered via an SSL data layer, so this will not work with
16062 "bind" lines having the "ssl" option. This can for example be used to detect
16063 whether the client sent a SessionTicket or not and stick it accordingly, if
16064 no SessionTicket then stick on SessionID or don't stick as there's no server
16065 side state is there when SessionTickets are in use.
16066
Willy Tarreau74ca5042013-06-11 23:12:07 +020016067req.ssl_ver : integer
16068req_ssl_ver : integer (deprecated)
16069 Returns an integer value containing the version of the SSL/TLS protocol of a
16070 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
16071 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
16072 composed of the major version multiplied by 65536, added to the minor
16073 version. Note that this only applies to raw contents found in the request
16074 buffer and not to contents deciphered via an SSL data layer, so this will not
16075 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010016076 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020016077 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016078
Willy Tarreau74ca5042013-06-11 23:12:07 +020016079 ACL derivatives :
16080 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016081
Willy Tarreau47e8eba2013-09-11 23:28:46 +020016082res.len : integer
16083 Returns an integer value corresponding to the number of bytes present in the
16084 response buffer. This is mostly used in ACL. It is important to understand
16085 that this test does not return false as long as the buffer is changing. This
16086 means that a check with equality to zero will almost always immediately match
16087 at the beginning of the session, while a test for more data will wait for
16088 that data to come in and return false only when haproxy is certain that no
16089 more data will come in. This test was designed to be used with TCP response
16090 content inspection.
16091
Willy Tarreau74ca5042013-06-11 23:12:07 +020016092res.payload(<offset>,<length>) : binary
16093 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020016094 in the response buffer. As a special case, if the <length> argument is zero,
16095 the the whole buffer from <offset> to the end is extracted. This can be used
16096 with ACLs in order to check for the presence of some content in a buffer at
16097 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016098
Willy Tarreau74ca5042013-06-11 23:12:07 +020016099res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16100 This extracts a binary block whose size is specified at <offset1> for <length>
16101 bytes, and which starts at <offset2> if specified or just after the length in
16102 the response buffer. The <offset2> parameter also supports relative offsets
16103 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016104
Willy Tarreau74ca5042013-06-11 23:12:07 +020016105 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016106
Willy Tarreau971f7b62015-09-29 14:06:59 +020016107res.ssl_hello_type : integer
16108rep_ssl_hello_type : integer (deprecated)
16109 Returns an integer value containing the type of the SSL hello message found
16110 in the response buffer if the buffer contains data that parses as a complete
16111 SSL (v3 or superior) hello message. Note that this only applies to raw
16112 contents found in the response buffer and not to contents deciphered via an
16113 SSL data layer, so this will not work with "server" lines having the "ssl"
16114 option. This is mostly used in ACL to detect presence of an SSL hello message
16115 that is supposed to contain an SSL session ID usable for stickiness.
16116
Willy Tarreau74ca5042013-06-11 23:12:07 +020016117wait_end : boolean
16118 This fetch either returns true when the inspection period is over, or does
16119 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016120 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020016121 actions, such as a delayed reject for some special addresses. Since it either
16122 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016123 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020016124 "WAIT_END" is always usable without prior declaration. This test was designed
16125 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016126
Willy Tarreau74ca5042013-06-11 23:12:07 +020016127 Examples :
16128 # delay every incoming request by 2 seconds
16129 tcp-request inspect-delay 2s
16130 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010016131
Willy Tarreau74ca5042013-06-11 23:12:07 +020016132 # don't immediately tell bad guys they are rejected
16133 tcp-request inspect-delay 10s
16134 acl goodguys src 10.0.0.0/24
16135 acl badguys src 10.0.1.0/24
16136 tcp-request content accept if goodguys
16137 tcp-request content reject if badguys WAIT_END
16138 tcp-request content reject
16139
16140
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200161417.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016142--------------------------------------
16143
16144It is possible to fetch samples from HTTP contents, requests and responses.
16145This application layer is also called layer 7. It is only possible to fetch the
16146data in this section when a full HTTP request or response has been parsed from
16147its respective request or response buffer. This is always the case with all
16148HTTP specific rules and for sections running with "mode http". When using TCP
16149content inspection, it may be necessary to support an inspection delay in order
16150to let the request or response come in first. These fetches may require a bit
16151more CPU resources than the layer 4 ones, but not much since the request and
16152response are indexed.
16153
16154base : string
16155 This returns the concatenation of the first Host header and the path part of
16156 the request, which starts at the first slash and ends before the question
16157 mark. It can be useful in virtual hosted environments to detect URL abuses as
16158 well as to improve shared caches efficiency. Using this with a limited size
16159 stick table also allows one to collect statistics about most commonly
16160 requested objects by host/path. With ACLs it can allow simple content
16161 switching rules involving the host and the path at the same time, such as
16162 "www.example.com/favicon.ico". See also "path" and "uri".
16163
16164 ACL derivatives :
16165 base : exact string match
16166 base_beg : prefix match
16167 base_dir : subdir match
16168 base_dom : domain match
16169 base_end : suffix match
16170 base_len : length match
16171 base_reg : regex match
16172 base_sub : substring match
16173
16174base32 : integer
16175 This returns a 32-bit hash of the value returned by the "base" fetch method
16176 above. This is useful to track per-URL activity on high traffic sites without
16177 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016178 memory. The output type is an unsigned integer. The hash function used is
16179 SDBM with full avalanche on the output. Technically, base32 is exactly equal
16180 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020016181
16182base32+src : binary
16183 This returns the concatenation of the base32 fetch above and the src fetch
16184 below. The resulting type is of type binary, with a size of 8 or 20 bytes
16185 depending on the source address family. This can be used to track per-IP,
16186 per-URL counters.
16187
William Lallemand65ad6e12014-01-31 15:08:02 +010016188capture.req.hdr(<idx>) : string
16189 This extracts the content of the header captured by the "capture request
16190 header", idx is the position of the capture keyword in the configuration.
16191 The first entry is an index of 0. See also: "capture request header".
16192
16193capture.req.method : string
16194 This extracts the METHOD of an HTTP request. It can be used in both request
16195 and response. Unlike "method", it can be used in both request and response
16196 because it's allocated.
16197
16198capture.req.uri : string
16199 This extracts the request's URI, which starts at the first slash and ends
16200 before the first space in the request (without the host part). Unlike "path"
16201 and "url", it can be used in both request and response because it's
16202 allocated.
16203
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016204capture.req.ver : string
16205 This extracts the request's HTTP version and returns either "HTTP/1.0" or
16206 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
16207 logs because it relies on a persistent flag.
16208
William Lallemand65ad6e12014-01-31 15:08:02 +010016209capture.res.hdr(<idx>) : string
16210 This extracts the content of the header captured by the "capture response
16211 header", idx is the position of the capture keyword in the configuration.
16212 The first entry is an index of 0.
16213 See also: "capture response header"
16214
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016215capture.res.ver : string
16216 This extracts the response's HTTP version and returns either "HTTP/1.0" or
16217 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
16218 persistent flag.
16219
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016220req.body : binary
16221 This returns the HTTP request's available body as a block of data. It
16222 requires that the request body has been buffered made available using
16223 "option http-buffer-request". In case of chunked-encoded body, currently only
16224 the first chunk is analyzed.
16225
Thierry FOURNIER9826c772015-05-20 15:50:54 +020016226req.body_param([<name>) : string
16227 This fetch assumes that the body of the POST request is url-encoded. The user
16228 can check if the "content-type" contains the value
16229 "application/x-www-form-urlencoded". This extracts the first occurrence of the
16230 parameter <name> in the body, which ends before '&'. The parameter name is
16231 case-sensitive. If no name is given, any parameter will match, and the first
16232 one will be returned. The result is a string corresponding to the value of the
16233 parameter <name> as presented in the request body (no URL decoding is
16234 performed). Note that the ACL version of this fetch iterates over multiple
16235 parameters and will iteratively report all parameters values if no name is
16236 given.
16237
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016238req.body_len : integer
16239 This returns the length of the HTTP request's available body in bytes. It may
16240 be lower than the advertised length if the body is larger than the buffer. It
16241 requires that the request body has been buffered made available using
16242 "option http-buffer-request".
16243
16244req.body_size : integer
16245 This returns the advertised length of the HTTP request's body in bytes. It
16246 will represent the advertised Content-Length header, or the size of the first
16247 chunk in case of chunked encoding. In order to parse the chunks, it requires
16248 that the request body has been buffered made available using
16249 "option http-buffer-request".
16250
Willy Tarreau74ca5042013-06-11 23:12:07 +020016251req.cook([<name>]) : string
16252cook([<name>]) : string (deprecated)
16253 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16254 header line from the request, and returns its value as string. If no name is
16255 specified, the first cookie value is returned. When used with ACLs, all
16256 matching cookies are evaluated. Spaces around the name and the value are
16257 ignored as requested by the Cookie header specification (RFC6265). The cookie
16258 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
16259 well return an empty value if it is present. Use the "found" match to detect
16260 presence. Use the res.cook() variant for response cookies sent by the server.
16261
16262 ACL derivatives :
16263 cook([<name>]) : exact string match
16264 cook_beg([<name>]) : prefix match
16265 cook_dir([<name>]) : subdir match
16266 cook_dom([<name>]) : domain match
16267 cook_end([<name>]) : suffix match
16268 cook_len([<name>]) : length match
16269 cook_reg([<name>]) : regex match
16270 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016271
Willy Tarreau74ca5042013-06-11 23:12:07 +020016272req.cook_cnt([<name>]) : integer
16273cook_cnt([<name>]) : integer (deprecated)
16274 Returns an integer value representing the number of occurrences of the cookie
16275 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016276
Willy Tarreau74ca5042013-06-11 23:12:07 +020016277req.cook_val([<name>]) : integer
16278cook_val([<name>]) : integer (deprecated)
16279 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16280 header line from the request, and converts its value to an integer which is
16281 returned. If no name is specified, the first cookie value is returned. When
16282 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020016283
Willy Tarreau74ca5042013-06-11 23:12:07 +020016284cookie([<name>]) : string (deprecated)
16285 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16286 header line from the request, or a "Set-Cookie" header from the response, and
16287 returns its value as a string. A typical use is to get multiple clients
16288 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020016289 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020016290 multi-peer synchronization and state keeping across restarts. If no name is
16291 specified, the first cookie value is returned. This fetch should not be used
16292 anymore and should be replaced by req.cook() or res.cook() instead as it
16293 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016294
Willy Tarreau74ca5042013-06-11 23:12:07 +020016295hdr([<name>[,<occ>]]) : string
16296 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
16297 used on responses. Please refer to these respective fetches for more details.
16298 In case of doubt about the fetch direction, please use the explicit ones.
16299 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016300 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016301
Willy Tarreau74ca5042013-06-11 23:12:07 +020016302req.fhdr(<name>[,<occ>]) : string
16303 This extracts the last occurrence of header <name> in an HTTP request. When
16304 used from an ACL, all occurrences are iterated over until a match is found.
16305 Optionally, a specific occurrence might be specified as a position number.
16306 Positive values indicate a position from the first occurrence, with 1 being
16307 the first one. Negative values indicate positions relative to the last one,
16308 with -1 being the last one. It differs from req.hdr() in that any commas
16309 present in the value are returned and are not used as delimiters. This is
16310 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016311
Willy Tarreau74ca5042013-06-11 23:12:07 +020016312req.fhdr_cnt([<name>]) : integer
16313 Returns an integer value representing the number of occurrences of request
16314 header field name <name>, or the total number of header fields if <name> is
16315 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
16316 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016317
Willy Tarreau74ca5042013-06-11 23:12:07 +020016318req.hdr([<name>[,<occ>]]) : string
16319 This extracts the last occurrence of header <name> in an HTTP request. When
16320 used from an ACL, all occurrences are iterated over until a match is found.
16321 Optionally, a specific occurrence might be specified as a position number.
16322 Positive values indicate a position from the first occurrence, with 1 being
16323 the first one. Negative values indicate positions relative to the last one,
16324 with -1 being the last one. A typical use is with the X-Forwarded-For header
16325 once converted to IP, associated with an IP stick-table. The function
16326 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000016327 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020016328 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010016329 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010016330
Willy Tarreau74ca5042013-06-11 23:12:07 +020016331 ACL derivatives :
16332 hdr([<name>[,<occ>]]) : exact string match
16333 hdr_beg([<name>[,<occ>]]) : prefix match
16334 hdr_dir([<name>[,<occ>]]) : subdir match
16335 hdr_dom([<name>[,<occ>]]) : domain match
16336 hdr_end([<name>[,<occ>]]) : suffix match
16337 hdr_len([<name>[,<occ>]]) : length match
16338 hdr_reg([<name>[,<occ>]]) : regex match
16339 hdr_sub([<name>[,<occ>]]) : substring match
16340
16341req.hdr_cnt([<name>]) : integer
16342hdr_cnt([<header>]) : integer (deprecated)
16343 Returns an integer value representing the number of occurrences of request
16344 header field name <name>, or the total number of header field values if
16345 <name> is not specified. It is important to remember that one header line may
16346 count as several headers if it has several values. The function considers any
16347 comma as a delimiter for distinct values. If full-line headers are desired
16348 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
16349 detect presence, absence or abuse of a specific header, as well as to block
16350 request smuggling attacks by rejecting requests which contain more than one
16351 of certain headers. See "req.hdr" for more information on header matching.
16352
16353req.hdr_ip([<name>[,<occ>]]) : ip
16354hdr_ip([<name>[,<occ>]]) : ip (deprecated)
16355 This extracts the last occurrence of header <name> in an HTTP request,
16356 converts it to an IPv4 or IPv6 address and returns this address. When used
16357 with ACLs, all occurrences are checked, and if <name> is omitted, every value
16358 of every header is checked. Optionally, a specific occurrence might be
16359 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010016360 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020016361 positions relative to the last one, with -1 being the last one. A typical use
16362 is with the X-Forwarded-For and X-Client-IP headers.
16363
16364req.hdr_val([<name>[,<occ>]]) : integer
16365hdr_val([<name>[,<occ>]]) : integer (deprecated)
16366 This extracts the last occurrence of header <name> in an HTTP request, and
16367 converts it to an integer value. When used with ACLs, all occurrences are
16368 checked, and if <name> is omitted, every value of every header is checked.
16369 Optionally, a specific occurrence might be specified as a position number.
16370 Positive values indicate a position from the first occurrence, with 1 being
16371 the first one. Negative values indicate positions relative to the last one,
16372 with -1 being the last one. A typical use is with the X-Forwarded-For header.
16373
Frédéric Lécailleec891192019-02-26 15:02:35 +010016374
16375
Willy Tarreau74ca5042013-06-11 23:12:07 +020016376http_auth(<userlist>) : boolean
16377 Returns a boolean indicating whether the authentication data received from
16378 the client match a username & password stored in the specified userlist. This
16379 fetch function is not really useful outside of ACLs. Currently only http
16380 basic auth is supported.
16381
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016382http_auth_group(<userlist>) : string
16383 Returns a string corresponding to the user name found in the authentication
16384 data received from the client if both the user name and password are valid
16385 according to the specified userlist. The main purpose is to use it in ACLs
16386 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016387 This fetch function is not really useful outside of ACLs. Currently only http
16388 basic auth is supported.
16389
16390 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016391 http_auth_group(<userlist>) : group ...
16392 Returns true when the user extracted from the request and whose password is
16393 valid according to the specified userlist belongs to at least one of the
16394 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016395
16396http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020016397 Returns true when the request being processed is the first one of the
16398 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020016399 from some requests when a request is not the first one, or to help grouping
16400 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020016401
Willy Tarreau74ca5042013-06-11 23:12:07 +020016402method : integer + string
16403 Returns an integer value corresponding to the method in the HTTP request. For
16404 example, "GET" equals 1 (check sources to establish the matching). Value 9
16405 means "other method" and may be converted to a string extracted from the
16406 stream. This should not be used directly as a sample, this is only meant to
16407 be used from ACLs, which transparently convert methods from patterns to these
16408 integer + string values. Some predefined ACL already check for most common
16409 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016410
Willy Tarreau74ca5042013-06-11 23:12:07 +020016411 ACL derivatives :
16412 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016413
Willy Tarreau74ca5042013-06-11 23:12:07 +020016414 Example :
16415 # only accept GET and HEAD requests
16416 acl valid_method method GET HEAD
16417 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020016418
Willy Tarreau74ca5042013-06-11 23:12:07 +020016419path : string
16420 This extracts the request's URL path, which starts at the first slash and
16421 ends before the question mark (without the host part). A typical use is with
16422 prefetch-capable caches, and with portals which need to aggregate multiple
16423 information from databases and keep them in caches. Note that with outgoing
16424 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010016425 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020016426 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016427
Willy Tarreau74ca5042013-06-11 23:12:07 +020016428 ACL derivatives :
16429 path : exact string match
16430 path_beg : prefix match
16431 path_dir : subdir match
16432 path_dom : domain match
16433 path_end : suffix match
16434 path_len : length match
16435 path_reg : regex match
16436 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016437
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016438query : string
16439 This extracts the request's query string, which starts after the first
16440 question mark. If no question mark is present, this fetch returns nothing. If
16441 a question mark is present but nothing follows, it returns an empty string.
16442 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010016443 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016444 which stops before the question mark.
16445
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016446req.hdr_names([<delim>]) : string
16447 This builds a string made from the concatenation of all header names as they
16448 appear in the request when the rule is evaluated. The default delimiter is
16449 the comma (',') but it may be overridden as an optional argument <delim>. In
16450 this case, only the first character of <delim> is considered.
16451
Willy Tarreau74ca5042013-06-11 23:12:07 +020016452req.ver : string
16453req_ver : string (deprecated)
16454 Returns the version string from the HTTP request, for example "1.1". This can
16455 be useful for logs, but is mostly there for ACL. Some predefined ACL already
16456 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016457
Willy Tarreau74ca5042013-06-11 23:12:07 +020016458 ACL derivatives :
16459 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020016460
Willy Tarreau74ca5042013-06-11 23:12:07 +020016461res.comp : boolean
16462 Returns the boolean "true" value if the response has been compressed by
16463 HAProxy, otherwise returns boolean "false". This may be used to add
16464 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016465
Willy Tarreau74ca5042013-06-11 23:12:07 +020016466res.comp_algo : string
16467 Returns a string containing the name of the algorithm used if the response
16468 was compressed by HAProxy, for example : "deflate". This may be used to add
16469 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016470
Willy Tarreau74ca5042013-06-11 23:12:07 +020016471res.cook([<name>]) : string
16472scook([<name>]) : string (deprecated)
16473 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16474 header line from the response, and returns its value as string. If no name is
16475 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016476
Willy Tarreau74ca5042013-06-11 23:12:07 +020016477 ACL derivatives :
16478 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016479
Willy Tarreau74ca5042013-06-11 23:12:07 +020016480res.cook_cnt([<name>]) : integer
16481scook_cnt([<name>]) : integer (deprecated)
16482 Returns an integer value representing the number of occurrences of the cookie
16483 <name> in the response, or all cookies if <name> is not specified. This is
16484 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016485
Willy Tarreau74ca5042013-06-11 23:12:07 +020016486res.cook_val([<name>]) : integer
16487scook_val([<name>]) : integer (deprecated)
16488 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16489 header line from the response, and converts its value to an integer which is
16490 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016491
Willy Tarreau74ca5042013-06-11 23:12:07 +020016492res.fhdr([<name>[,<occ>]]) : string
16493 This extracts the last occurrence of header <name> in an HTTP response, or of
16494 the last header if no <name> is specified. Optionally, a specific occurrence
16495 might be specified as a position number. Positive values indicate a position
16496 from the first occurrence, with 1 being the first one. Negative values
16497 indicate positions relative to the last one, with -1 being the last one. It
16498 differs from res.hdr() in that any commas present in the value are returned
16499 and are not used as delimiters. If this is not desired, the res.hdr() fetch
16500 should be used instead. This is sometimes useful with headers such as Date or
16501 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016502
Willy Tarreau74ca5042013-06-11 23:12:07 +020016503res.fhdr_cnt([<name>]) : integer
16504 Returns an integer value representing the number of occurrences of response
16505 header field name <name>, or the total number of header fields if <name> is
16506 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
16507 the number of full line headers and does not stop on commas. If this is not
16508 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016509
Willy Tarreau74ca5042013-06-11 23:12:07 +020016510res.hdr([<name>[,<occ>]]) : string
16511shdr([<name>[,<occ>]]) : string (deprecated)
16512 This extracts the last occurrence of header <name> in an HTTP response, or of
16513 the last header if no <name> is specified. Optionally, a specific occurrence
16514 might be specified as a position number. Positive values indicate a position
16515 from the first occurrence, with 1 being the first one. Negative values
16516 indicate positions relative to the last one, with -1 being the last one. This
16517 can be useful to learn some data into a stick-table. The function considers
16518 any comma as a delimiter for distinct values. If this is not desired, the
16519 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016520
Willy Tarreau74ca5042013-06-11 23:12:07 +020016521 ACL derivatives :
16522 shdr([<name>[,<occ>]]) : exact string match
16523 shdr_beg([<name>[,<occ>]]) : prefix match
16524 shdr_dir([<name>[,<occ>]]) : subdir match
16525 shdr_dom([<name>[,<occ>]]) : domain match
16526 shdr_end([<name>[,<occ>]]) : suffix match
16527 shdr_len([<name>[,<occ>]]) : length match
16528 shdr_reg([<name>[,<occ>]]) : regex match
16529 shdr_sub([<name>[,<occ>]]) : substring match
16530
16531res.hdr_cnt([<name>]) : integer
16532shdr_cnt([<name>]) : integer (deprecated)
16533 Returns an integer value representing the number of occurrences of response
16534 header field name <name>, or the total number of header fields if <name> is
16535 not specified. The function considers any comma as a delimiter for distinct
16536 values. If this is not desired, the res.fhdr_cnt() fetch should be used
16537 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016538
Willy Tarreau74ca5042013-06-11 23:12:07 +020016539res.hdr_ip([<name>[,<occ>]]) : ip
16540shdr_ip([<name>[,<occ>]]) : ip (deprecated)
16541 This extracts the last occurrence of header <name> in an HTTP response,
16542 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
16543 specific occurrence might be specified as a position number. Positive values
16544 indicate a position from the first occurrence, with 1 being the first one.
16545 Negative values indicate positions relative to the last one, with -1 being
16546 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016547
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016548res.hdr_names([<delim>]) : string
16549 This builds a string made from the concatenation of all header names as they
16550 appear in the response when the rule is evaluated. The default delimiter is
16551 the comma (',') but it may be overridden as an optional argument <delim>. In
16552 this case, only the first character of <delim> is considered.
16553
Willy Tarreau74ca5042013-06-11 23:12:07 +020016554res.hdr_val([<name>[,<occ>]]) : integer
16555shdr_val([<name>[,<occ>]]) : integer (deprecated)
16556 This extracts the last occurrence of header <name> in an HTTP response, and
16557 converts it to an integer value. Optionally, a specific occurrence might be
16558 specified as a position number. Positive values indicate a position from the
16559 first occurrence, with 1 being the first one. Negative values indicate
16560 positions relative to the last one, with -1 being the last one. This can be
16561 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016562
Willy Tarreau74ca5042013-06-11 23:12:07 +020016563res.ver : string
16564resp_ver : string (deprecated)
16565 Returns the version string from the HTTP response, for example "1.1". This
16566 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020016567
Willy Tarreau74ca5042013-06-11 23:12:07 +020016568 ACL derivatives :
16569 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016570
Willy Tarreau74ca5042013-06-11 23:12:07 +020016571set-cookie([<name>]) : string (deprecated)
16572 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16573 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020016574 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020016575 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016576
Willy Tarreau74ca5042013-06-11 23:12:07 +020016577 This fetch function is deprecated and has been superseded by the "res.cook"
16578 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016579
Willy Tarreau74ca5042013-06-11 23:12:07 +020016580status : integer
16581 Returns an integer containing the HTTP status code in the HTTP response, for
16582 example, 302. It is mostly used within ACLs and integer ranges, for example,
16583 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016584
Thierry Fournier0e00dca2016-04-07 15:47:40 +020016585unique-id : string
16586 Returns the unique-id attached to the request. The directive
16587 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
16588 fails. Note that the unique-id is usually used with HTTP requests, however this
16589 sample fetch can be used with other protocols. Obviously, if it is used with
16590 other protocols than HTTP, the unique-id-format directive must not contain
16591 HTTP parts. See: unique-id-format and unique-id-header
16592
Willy Tarreau74ca5042013-06-11 23:12:07 +020016593url : string
16594 This extracts the request's URL as presented in the request. A typical use is
16595 with prefetch-capable caches, and with portals which need to aggregate
16596 multiple information from databases and keep them in caches. With ACLs, using
16597 "path" is preferred over using "url", because clients may send a full URL as
16598 is normally done with proxies. The only real use is to match "*" which does
16599 not match in "path", and for which there is already a predefined ACL. See
16600 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016601
Willy Tarreau74ca5042013-06-11 23:12:07 +020016602 ACL derivatives :
16603 url : exact string match
16604 url_beg : prefix match
16605 url_dir : subdir match
16606 url_dom : domain match
16607 url_end : suffix match
16608 url_len : length match
16609 url_reg : regex match
16610 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016611
Willy Tarreau74ca5042013-06-11 23:12:07 +020016612url_ip : ip
16613 This extracts the IP address from the request's URL when the host part is
16614 presented as an IP address. Its use is very limited. For instance, a
16615 monitoring system might use this field as an alternative for the source IP in
16616 order to test what path a given source address would follow, or to force an
16617 entry in a table for a given source address. With ACLs it can be used to
16618 restrict access to certain systems through a proxy, for example when combined
16619 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016620
Willy Tarreau74ca5042013-06-11 23:12:07 +020016621url_port : integer
16622 This extracts the port part from the request's URL. Note that if the port is
16623 not specified in the request, port 80 is assumed. With ACLs it can be used to
16624 restrict access to certain systems through a proxy, for example when combined
16625 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016626
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016627urlp([<name>[,<delim>]]) : string
16628url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020016629 This extracts the first occurrence of the parameter <name> in the query
16630 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016631 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
16632 any parameter will match, and the first one will be returned. The result is
16633 a string corresponding to the value of the parameter <name> as presented in
16634 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020016635 stickiness based on a client ID, to extract an application cookie passed as a
16636 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016637 this fetch iterates over multiple parameters and will iteratively report all
16638 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016639
Willy Tarreau74ca5042013-06-11 23:12:07 +020016640 ACL derivatives :
16641 urlp(<name>[,<delim>]) : exact string match
16642 urlp_beg(<name>[,<delim>]) : prefix match
16643 urlp_dir(<name>[,<delim>]) : subdir match
16644 urlp_dom(<name>[,<delim>]) : domain match
16645 urlp_end(<name>[,<delim>]) : suffix match
16646 urlp_len(<name>[,<delim>]) : length match
16647 urlp_reg(<name>[,<delim>]) : regex match
16648 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016649
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016650
Willy Tarreau74ca5042013-06-11 23:12:07 +020016651 Example :
16652 # match http://example.com/foo?PHPSESSIONID=some_id
16653 stick on urlp(PHPSESSIONID)
16654 # match http://example.com/foo;JSESSIONID=some_id
16655 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016656
Jarno Huuskonen676f6222017-03-30 09:19:45 +030016657urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020016658 See "urlp" above. This one extracts the URL parameter <name> in the request
16659 and converts it to an integer value. This can be used for session stickiness
16660 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020016661
Dragan Dosen0070cd52016-06-16 12:19:49 +020016662url32 : integer
16663 This returns a 32-bit hash of the value obtained by concatenating the first
16664 Host header and the whole URL including parameters (not only the path part of
16665 the request, as in the "base32" fetch above). This is useful to track per-URL
16666 activity. A shorter hash is stored, saving a lot of memory. The output type
16667 is an unsigned integer.
16668
16669url32+src : binary
16670 This returns the concatenation of the "url32" fetch and the "src" fetch. The
16671 resulting type is of type binary, with a size of 8 or 20 bytes depending on
16672 the source address family. This can be used to track per-IP, per-URL counters.
16673
Willy Tarreau198a7442008-01-17 12:05:32 +010016674
Willy Tarreau74ca5042013-06-11 23:12:07 +0200166757.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016676---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016677
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016678Some predefined ACLs are hard-coded so that they do not have to be declared in
16679every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020016680order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010016681
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016682ACL name Equivalent to Usage
16683---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016684FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020016685HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016686HTTP_1.0 req_ver 1.0 match HTTP version 1.0
16687HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010016688HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
16689HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
16690HTTP_URL_SLASH url_beg / match URL beginning with "/"
16691HTTP_URL_STAR url * match URL equal to "*"
16692LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016693METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016694METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016695METH_GET method GET HEAD match HTTP GET or HEAD method
16696METH_HEAD method HEAD match HTTP HEAD method
16697METH_OPTIONS method OPTIONS match HTTP OPTIONS method
16698METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016699METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016700METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020016701RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016702REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010016703TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016704WAIT_END wait_end wait for end of content analysis
16705---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016706
Willy Tarreaub937b7e2010-01-12 15:27:54 +010016707
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167088. Logging
16709----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010016710
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016711One of HAProxy's strong points certainly lies is its precise logs. It probably
16712provides the finest level of information available for such a product, which is
16713very important for troubleshooting complex environments. Standard information
16714provided in logs include client ports, TCP/HTTP state timers, precise session
16715state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016716to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016717headers.
16718
16719In order to improve administrators reactivity, it offers a great transparency
16720about encountered problems, both internal and external, and it is possible to
16721send logs to different sources at the same time with different level filters :
16722
16723 - global process-level logs (system errors, start/stop, etc..)
16724 - per-instance system and internal errors (lack of resource, bugs, ...)
16725 - per-instance external troubles (servers up/down, max connections)
16726 - per-instance activity (client connections), either at the establishment or
16727 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010016728 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060016729 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016730
16731The ability to distribute different levels of logs to different log servers
16732allow several production teams to interact and to fix their problems as soon
16733as possible. For example, the system team might monitor system-wide errors,
16734while the application team might be monitoring the up/down for their servers in
16735real time, and the security team might analyze the activity logs with one hour
16736delay.
16737
16738
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167398.1. Log levels
16740---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016741
Simon Hormandf791f52011-05-29 15:01:10 +090016742TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016743source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090016744HTTP request, HTTP return code, number of bytes transmitted, conditions
16745in which the session ended, and even exchanged cookies values. For example
16746track a particular user's problems. All messages may be sent to up to two
16747syslog servers. Check the "log" keyword in section 4.2 for more information
16748about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016749
16750
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167518.2. Log formats
16752----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016753
William Lallemand48940402012-01-30 16:47:22 +010016754HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090016755and will be detailed in the following sections. A few of them may vary
16756slightly with the configuration, due to indicators specific to certain
16757options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016758
16759 - the default format, which is very basic and very rarely used. It only
16760 provides very basic information about the incoming connection at the moment
16761 it is accepted : source IP:port, destination IP:port, and frontend-name.
16762 This mode will eventually disappear so it will not be described to great
16763 extents.
16764
16765 - the TCP format, which is more advanced. This format is enabled when "option
16766 tcplog" is set on the frontend. HAProxy will then usually wait for the
16767 connection to terminate before logging. This format provides much richer
16768 information, such as timers, connection counts, queue size, etc... This
16769 format is recommended for pure TCP proxies.
16770
16771 - the HTTP format, which is the most advanced for HTTP proxying. This format
16772 is enabled when "option httplog" is set on the frontend. It provides the
16773 same information as the TCP format with some HTTP-specific fields such as
16774 the request, the status code, and captures of headers and cookies. This
16775 format is recommended for HTTP proxies.
16776
Emeric Brun3a058f32009-06-30 18:26:00 +020016777 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
16778 fields arranged in the same order as the CLF format. In this mode, all
16779 timers, captures, flags, etc... appear one per field after the end of the
16780 common fields, in the same order they appear in the standard HTTP format.
16781
William Lallemand48940402012-01-30 16:47:22 +010016782 - the custom log format, allows you to make your own log line.
16783
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016784Next sections will go deeper into details for each of these formats. Format
16785specification will be performed on a "field" basis. Unless stated otherwise, a
16786field is a portion of text delimited by any number of spaces. Since syslog
16787servers are susceptible of inserting fields at the beginning of a line, it is
16788always assumed that the first field is the one containing the process name and
16789identifier.
16790
16791Note : Since log lines may be quite long, the log examples in sections below
16792 might be broken into multiple lines. The example log lines will be
16793 prefixed with 3 closing angle brackets ('>>>') and each time a log is
16794 broken into multiple lines, each non-final line will end with a
16795 backslash ('\') and the next line will start indented by two characters.
16796
16797
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167988.2.1. Default log format
16799-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016800
16801This format is used when no specific option is set. The log is emitted as soon
16802as the connection is accepted. One should note that this currently is the only
16803format which logs the request's destination IP and ports.
16804
16805 Example :
16806 listen www
16807 mode http
16808 log global
16809 server srv1 127.0.0.1:8000
16810
16811 >>> Feb 6 12:12:09 localhost \
16812 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
16813 (www/HTTP)
16814
16815 Field Format Extract from the example above
16816 1 process_name '[' pid ']:' haproxy[14385]:
16817 2 'Connect from' Connect from
16818 3 source_ip ':' source_port 10.0.1.2:33312
16819 4 'to' to
16820 5 destination_ip ':' destination_port 10.0.3.31:8012
16821 6 '(' frontend_name '/' mode ')' (www/HTTP)
16822
16823Detailed fields description :
16824 - "source_ip" is the IP address of the client which initiated the connection.
16825 - "source_port" is the TCP port of the client which initiated the connection.
16826 - "destination_ip" is the IP address the client connected to.
16827 - "destination_port" is the TCP port the client connected to.
16828 - "frontend_name" is the name of the frontend (or listener) which received
16829 and processed the connection.
16830 - "mode is the mode the frontend is operating (TCP or HTTP).
16831
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016832In case of a UNIX socket, the source and destination addresses are marked as
16833"unix:" and the ports reflect the internal ID of the socket which accepted the
16834connection (the same ID as reported in the stats).
16835
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016836It is advised not to use this deprecated format for newer installations as it
16837will eventually disappear.
16838
16839
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200168408.2.2. TCP log format
16841---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016842
16843The TCP format is used when "option tcplog" is specified in the frontend, and
16844is the recommended format for pure TCP proxies. It provides a lot of precious
16845information for troubleshooting. Since this format includes timers and byte
16846counts, the log is normally emitted at the end of the session. It can be
16847emitted earlier if "option logasap" is specified, which makes sense in most
16848environments with long sessions such as remote terminals. Sessions which match
16849the "monitor" rules are never logged. It is also possible not to emit logs for
16850sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016851specifying "option dontlognull" in the frontend. Successful connections will
16852not be logged if "option dontlog-normal" is specified in the frontend. A few
16853fields may slightly vary depending on some configuration options, those are
16854marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016855
16856 Example :
16857 frontend fnt
16858 mode tcp
16859 option tcplog
16860 log global
16861 default_backend bck
16862
16863 backend bck
16864 server srv1 127.0.0.1:8000
16865
16866 >>> Feb 6 12:12:56 localhost \
16867 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
16868 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
16869
16870 Field Format Extract from the example above
16871 1 process_name '[' pid ']:' haproxy[14387]:
16872 2 client_ip ':' client_port 10.0.1.2:33313
16873 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
16874 4 frontend_name fnt
16875 5 backend_name '/' server_name bck/srv1
16876 6 Tw '/' Tc '/' Tt* 0/0/5007
16877 7 bytes_read* 212
16878 8 termination_state --
16879 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
16880 10 srv_queue '/' backend_queue 0/0
16881
16882Detailed fields description :
16883 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016884 connection to haproxy. If the connection was accepted on a UNIX socket
16885 instead, the IP address would be replaced with the word "unix". Note that
16886 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016887 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016888 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016889 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016890
16891 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016892 If the connection was accepted on a UNIX socket instead, the port would be
16893 replaced with the ID of the accepting socket, which is also reported in the
16894 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016895
16896 - "accept_date" is the exact date when the connection was received by haproxy
16897 (which might be very slightly different from the date observed on the
16898 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020016899 the same date which may appear in any upstream firewall's log. When used in
16900 HTTP mode, the accept_date field will be reset to the first moment the
16901 connection is ready to receive a new request (end of previous response for
16902 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016903
16904 - "frontend_name" is the name of the frontend (or listener) which received
16905 and processed the connection.
16906
16907 - "backend_name" is the name of the backend (or listener) which was selected
16908 to manage the connection to the server. This will be the same as the
16909 frontend if no switching rule has been applied, which is common for TCP
16910 applications.
16911
16912 - "server_name" is the name of the last server to which the connection was
16913 sent, which might differ from the first one if there were connection errors
16914 and a redispatch occurred. Note that this server belongs to the backend
16915 which processed the request. If the connection was aborted before reaching
16916 a server, "<NOSRV>" is indicated instead of a server name.
16917
16918 - "Tw" is the total time in milliseconds spent waiting in the various queues.
16919 It can be "-1" if the connection was aborted before reaching the queue.
16920 See "Timers" below for more details.
16921
16922 - "Tc" is the total time in milliseconds spent waiting for the connection to
16923 establish to the final server, including retries. It can be "-1" if the
16924 connection was aborted before a connection could be established. See
16925 "Timers" below for more details.
16926
16927 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016928 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016929 "option logasap" was specified, then the time counting stops at the moment
16930 the log is emitted. In this case, a '+' sign is prepended before the value,
16931 indicating that the final one will be larger. See "Timers" below for more
16932 details.
16933
16934 - "bytes_read" is the total number of bytes transmitted from the server to
16935 the client when the log is emitted. If "option logasap" is specified, the
16936 this value will be prefixed with a '+' sign indicating that the final one
16937 may be larger. Please note that this value is a 64-bit counter, so log
16938 analysis tools must be able to handle it without overflowing.
16939
16940 - "termination_state" is the condition the session was in when the session
16941 ended. This indicates the session state, which side caused the end of
16942 session to happen, and for what reason (timeout, error, ...). The normal
16943 flags should be "--", indicating the session was closed by either end with
16944 no data remaining in buffers. See below "Session state at disconnection"
16945 for more details.
16946
16947 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016948 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016949 limits have been reached. For instance, if actconn is close to 512 when
16950 multiple connection errors occur, chances are high that the system limits
16951 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016952 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016953
16954 - "feconn" is the total number of concurrent connections on the frontend when
16955 the session was logged. It is useful to estimate the amount of resource
16956 required to sustain high loads, and to detect when the frontend's "maxconn"
16957 has been reached. Most often when this value increases by huge jumps, it is
16958 because there is congestion on the backend servers, but sometimes it can be
16959 caused by a denial of service attack.
16960
16961 - "beconn" is the total number of concurrent connections handled by the
16962 backend when the session was logged. It includes the total number of
16963 concurrent connections active on servers as well as the number of
16964 connections pending in queues. It is useful to estimate the amount of
16965 additional servers needed to support high loads for a given application.
16966 Most often when this value increases by huge jumps, it is because there is
16967 congestion on the backend servers, but sometimes it can be caused by a
16968 denial of service attack.
16969
16970 - "srv_conn" is the total number of concurrent connections still active on
16971 the server when the session was logged. It can never exceed the server's
16972 configured "maxconn" parameter. If this value is very often close or equal
16973 to the server's "maxconn", it means that traffic regulation is involved a
16974 lot, meaning that either the server's maxconn value is too low, or that
16975 there aren't enough servers to process the load with an optimal response
16976 time. When only one of the server's "srv_conn" is high, it usually means
16977 that this server has some trouble causing the connections to take longer to
16978 be processed than on other servers.
16979
16980 - "retries" is the number of connection retries experienced by this session
16981 when trying to connect to the server. It must normally be zero, unless a
16982 server is being stopped at the same moment the connection was attempted.
16983 Frequent retries generally indicate either a network problem between
16984 haproxy and the server, or a misconfigured system backlog on the server
16985 preventing new connections from being queued. This field may optionally be
16986 prefixed with a '+' sign, indicating that the session has experienced a
16987 redispatch after the maximal retry count has been reached on the initial
16988 server. In this case, the server name appearing in the log is the one the
16989 connection was redispatched to, and not the first one, though both may
16990 sometimes be the same in case of hashing for instance. So as a general rule
16991 of thumb, when a '+' is present in front of the retry count, this count
16992 should not be attributed to the logged server.
16993
16994 - "srv_queue" is the total number of requests which were processed before
16995 this one in the server queue. It is zero when the request has not gone
16996 through the server queue. It makes it possible to estimate the approximate
16997 server's response time by dividing the time spent in queue by the number of
16998 requests in the queue. It is worth noting that if a session experiences a
16999 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017000 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017001 backend queue unless a redispatch occurs.
17002
17003 - "backend_queue" is the total number of requests which were processed before
17004 this one in the backend's global queue. It is zero when the request has not
17005 gone through the global queue. It makes it possible to estimate the average
17006 queue length, which easily translates into a number of missing servers when
17007 divided by a server's "maxconn" parameter. It is worth noting that if a
17008 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017009 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017010 through both the server queue and the backend queue unless a redispatch
17011 occurs.
17012
17013
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200170148.2.3. HTTP log format
17015----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017016
17017The HTTP format is the most complete and the best suited for HTTP proxies. It
17018is enabled by when "option httplog" is specified in the frontend. It provides
17019the same level of information as the TCP format with additional features which
17020are specific to the HTTP protocol. Just like the TCP format, the log is usually
17021emitted at the end of the session, unless "option logasap" is specified, which
17022generally only makes sense for download sites. A session which matches the
17023"monitor" rules will never logged. It is also possible not to log sessions for
17024which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017025frontend. Successful connections will not be logged if "option dontlog-normal"
17026is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017027
17028Most fields are shared with the TCP log, some being different. A few fields may
17029slightly vary depending on some configuration options. Those ones are marked
17030with a star ('*') after the field name below.
17031
17032 Example :
17033 frontend http-in
17034 mode http
17035 option httplog
17036 log global
17037 default_backend bck
17038
17039 backend static
17040 server srv1 127.0.0.1:8000
17041
17042 >>> Feb 6 12:14:14 localhost \
17043 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
17044 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 +010017045 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017046
17047 Field Format Extract from the example above
17048 1 process_name '[' pid ']:' haproxy[14389]:
17049 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017050 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017051 4 frontend_name http-in
17052 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017053 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017054 7 status_code 200
17055 8 bytes_read* 2750
17056 9 captured_request_cookie -
17057 10 captured_response_cookie -
17058 11 termination_state ----
17059 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
17060 13 srv_queue '/' backend_queue 0/0
17061 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
17062 15 '{' captured_response_headers* '}' {}
17063 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010017064
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017065Detailed fields description :
17066 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017067 connection to haproxy. If the connection was accepted on a UNIX socket
17068 instead, the IP address would be replaced with the word "unix". Note that
17069 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017070 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010017071 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010017072 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017073
17074 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010017075 If the connection was accepted on a UNIX socket instead, the port would be
17076 replaced with the ID of the accepting socket, which is also reported in the
17077 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017078
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017079 - "request_date" is the exact date when the first byte of the HTTP request
17080 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017081
17082 - "frontend_name" is the name of the frontend (or listener) which received
17083 and processed the connection.
17084
17085 - "backend_name" is the name of the backend (or listener) which was selected
17086 to manage the connection to the server. This will be the same as the
17087 frontend if no switching rule has been applied.
17088
17089 - "server_name" is the name of the last server to which the connection was
17090 sent, which might differ from the first one if there were connection errors
17091 and a redispatch occurred. Note that this server belongs to the backend
17092 which processed the request. If the request was aborted before reaching a
17093 server, "<NOSRV>" is indicated instead of a server name. If the request was
17094 intercepted by the stats subsystem, "<STATS>" is indicated instead.
17095
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017096 - "TR" is the total time in milliseconds spent waiting for a full HTTP
17097 request from the client (not counting body) after the first byte was
17098 received. It can be "-1" if the connection was aborted before a complete
17099 request could be received or the a bad request was received. It should
17100 always be very small because a request generally fits in one single packet.
17101 Large times here generally indicate network issues between the client and
Willy Tarreau590a0512018-09-05 11:56:48 +020017102 haproxy or requests being typed by hand. See section 8.4 "Timing Events"
17103 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017104
17105 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17106 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020017107 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017108
17109 - "Tc" is the total time in milliseconds spent waiting for the connection to
17110 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020017111 request was aborted before a connection could be established. See section
17112 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017113
17114 - "Tr" is the total time in milliseconds spent waiting for the server to send
17115 a full HTTP response, not counting data. It can be "-1" if the request was
17116 aborted before a complete response could be received. It generally matches
17117 the server's processing time for the request, though it may be altered by
17118 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020017119 "GET" requests generally indicate an overloaded server. See section 8.4
17120 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017121
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017122 - "Ta" is the time the request remained active in haproxy, which is the total
17123 time in milliseconds elapsed between the first byte of the request was
17124 received and the last byte of response was sent. It covers all possible
17125 processing except the handshake (see Th) and idle time (see Ti). There is
17126 one exception, if "option logasap" was specified, then the time counting
17127 stops at the moment the log is emitted. In this case, a '+' sign is
17128 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020017129 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017130
17131 - "status_code" is the HTTP status code returned to the client. This status
17132 is generally set by the server, but it might also be set by haproxy when
17133 the server cannot be reached or when its response is blocked by haproxy.
17134
17135 - "bytes_read" is the total number of bytes transmitted to the client when
17136 the log is emitted. This does include HTTP headers. If "option logasap" is
17137 specified, the this value will be prefixed with a '+' sign indicating that
17138 the final one may be larger. Please note that this value is a 64-bit
17139 counter, so log analysis tools must be able to handle it without
17140 overflowing.
17141
17142 - "captured_request_cookie" is an optional "name=value" entry indicating that
17143 the client had this cookie in the request. The cookie name and its maximum
17144 length are defined by the "capture cookie" statement in the frontend
17145 configuration. The field is a single dash ('-') when the option is not
17146 set. Only one cookie may be captured, it is generally used to track session
17147 ID exchanges between a client and a server to detect session crossing
17148 between clients due to application bugs. For more details, please consult
17149 the section "Capturing HTTP headers and cookies" below.
17150
17151 - "captured_response_cookie" is an optional "name=value" entry indicating
17152 that the server has returned a cookie with its response. The cookie name
17153 and its maximum length are defined by the "capture cookie" statement in the
17154 frontend configuration. The field is a single dash ('-') when the option is
17155 not set. Only one cookie may be captured, it is generally used to track
17156 session ID exchanges between a client and a server to detect session
17157 crossing between clients due to application bugs. For more details, please
17158 consult the section "Capturing HTTP headers and cookies" below.
17159
17160 - "termination_state" is the condition the session was in when the session
17161 ended. This indicates the session state, which side caused the end of
17162 session to happen, for what reason (timeout, error, ...), just like in TCP
17163 logs, and information about persistence operations on cookies in the last
17164 two characters. The normal flags should begin with "--", indicating the
17165 session was closed by either end with no data remaining in buffers. See
17166 below "Session state at disconnection" for more details.
17167
17168 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017169 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017170 limits have been reached. For instance, if actconn is close to 512 or 1024
17171 when multiple connection errors occur, chances are high that the system
17172 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017173 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017174 system.
17175
17176 - "feconn" is the total number of concurrent connections on the frontend when
17177 the session was logged. It is useful to estimate the amount of resource
17178 required to sustain high loads, and to detect when the frontend's "maxconn"
17179 has been reached. Most often when this value increases by huge jumps, it is
17180 because there is congestion on the backend servers, but sometimes it can be
17181 caused by a denial of service attack.
17182
17183 - "beconn" is the total number of concurrent connections handled by the
17184 backend when the session was logged. It includes the total number of
17185 concurrent connections active on servers as well as the number of
17186 connections pending in queues. It is useful to estimate the amount of
17187 additional servers needed to support high loads for a given application.
17188 Most often when this value increases by huge jumps, it is because there is
17189 congestion on the backend servers, but sometimes it can be caused by a
17190 denial of service attack.
17191
17192 - "srv_conn" is the total number of concurrent connections still active on
17193 the server when the session was logged. It can never exceed the server's
17194 configured "maxconn" parameter. If this value is very often close or equal
17195 to the server's "maxconn", it means that traffic regulation is involved a
17196 lot, meaning that either the server's maxconn value is too low, or that
17197 there aren't enough servers to process the load with an optimal response
17198 time. When only one of the server's "srv_conn" is high, it usually means
17199 that this server has some trouble causing the requests to take longer to be
17200 processed than on other servers.
17201
17202 - "retries" is the number of connection retries experienced by this session
17203 when trying to connect to the server. It must normally be zero, unless a
17204 server is being stopped at the same moment the connection was attempted.
17205 Frequent retries generally indicate either a network problem between
17206 haproxy and the server, or a misconfigured system backlog on the server
17207 preventing new connections from being queued. This field may optionally be
17208 prefixed with a '+' sign, indicating that the session has experienced a
17209 redispatch after the maximal retry count has been reached on the initial
17210 server. In this case, the server name appearing in the log is the one the
17211 connection was redispatched to, and not the first one, though both may
17212 sometimes be the same in case of hashing for instance. So as a general rule
17213 of thumb, when a '+' is present in front of the retry count, this count
17214 should not be attributed to the logged server.
17215
17216 - "srv_queue" is the total number of requests which were processed before
17217 this one in the server queue. It is zero when the request has not gone
17218 through the server queue. It makes it possible to estimate the approximate
17219 server's response time by dividing the time spent in queue by the number of
17220 requests in the queue. It is worth noting that if a session experiences a
17221 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017222 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017223 backend queue unless a redispatch occurs.
17224
17225 - "backend_queue" is the total number of requests which were processed before
17226 this one in the backend's global queue. It is zero when the request has not
17227 gone through the global queue. It makes it possible to estimate the average
17228 queue length, which easily translates into a number of missing servers when
17229 divided by a server's "maxconn" parameter. It is worth noting that if a
17230 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017231 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017232 through both the server queue and the backend queue unless a redispatch
17233 occurs.
17234
17235 - "captured_request_headers" is a list of headers captured in the request due
17236 to the presence of the "capture request header" statement in the frontend.
17237 Multiple headers can be captured, they will be delimited by a vertical bar
17238 ('|'). When no capture is enabled, the braces do not appear, causing a
17239 shift of remaining fields. It is important to note that this field may
17240 contain spaces, and that using it requires a smarter log parser than when
17241 it's not used. Please consult the section "Capturing HTTP headers and
17242 cookies" below for more details.
17243
17244 - "captured_response_headers" is a list of headers captured in the response
17245 due to the presence of the "capture response header" statement in the
17246 frontend. Multiple headers can be captured, they will be delimited by a
17247 vertical bar ('|'). When no capture is enabled, the braces do not appear,
17248 causing a shift of remaining fields. It is important to note that this
17249 field may contain spaces, and that using it requires a smarter log parser
17250 than when it's not used. Please consult the section "Capturing HTTP headers
17251 and cookies" below for more details.
17252
17253 - "http_request" is the complete HTTP request line, including the method,
17254 request and HTTP version string. Non-printable characters are encoded (see
17255 below the section "Non-printable characters"). This is always the last
17256 field, and it is always delimited by quotes and is the only one which can
17257 contain quotes. If new fields are added to the log format, they will be
17258 added before this field. This field might be truncated if the request is
17259 huge and does not fit in the standard syslog buffer (1024 characters). This
17260 is the reason why this field must always remain the last one.
17261
17262
Cyril Bontédc4d9032012-04-08 21:57:39 +0200172638.2.4. Custom log format
17264------------------------
William Lallemand48940402012-01-30 16:47:22 +010017265
Willy Tarreau2beef582012-12-20 17:22:52 +010017266The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010017267mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010017268
Davor Ocelice9ed2812017-12-25 17:49:28 +010017269HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010017270Variables can take arguments using braces ('{}'), and multiple arguments are
17271separated by commas within the braces. Flags may be added or removed by
17272prefixing them with a '+' or '-' sign.
17273
17274Special variable "%o" may be used to propagate its flags to all other
17275variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010017276("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010017277
Willy Tarreauc8368452012-12-21 00:09:23 +010017278If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020017279as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010017280less common information such as the client's SSL certificate's DN, or to log
17281the key that would be used to store an entry into a stick table.
17282
William Lallemand48940402012-01-30 16:47:22 +010017283Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017284In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010017285in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010017286
Dragan Dosen835b9212016-02-12 13:23:03 +010017287Note: when using the RFC5424 syslog message format, the characters '"',
17288'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
17289https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
17290such cases, the use of the flag "E" should be considered.
17291
William Lallemand48940402012-01-30 16:47:22 +010017292Flags are :
17293 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017294 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010017295 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
17296 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010017297
17298 Example:
17299
17300 log-format %T\ %t\ Some\ Text
17301 log-format %{+Q}o\ %t\ %s\ %{-Q}r
17302
Dragan Dosen835b9212016-02-12 13:23:03 +010017303 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
17304
William Lallemand48940402012-01-30 16:47:22 +010017305At the moment, the default HTTP format is defined this way :
17306
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017307 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
17308 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010017309
William Lallemandbddd4fd2012-02-27 11:23:10 +010017310the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010017311
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017312 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
17313 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
17314 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010017315
William Lallemandbddd4fd2012-02-27 11:23:10 +010017316and the default TCP format is defined this way :
17317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017318 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
17319 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010017320
William Lallemand48940402012-01-30 16:47:22 +010017321Please refer to the table below for currently defined variables :
17322
William Lallemandbddd4fd2012-02-27 11:23:10 +010017323 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017324 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017325 +---+------+-----------------------------------------------+-------------+
17326 | | %o | special variable, apply flags on all next var | |
17327 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010017328 | | %B | bytes_read (from server to client) | numeric |
17329 | H | %CC | captured_request_cookie | string |
17330 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020017331 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017332 | H | %HM | HTTP method (ex: POST) | string |
17333 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000017334 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017335 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
17336 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010017337 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020017338 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017339 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017340 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017341 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020017342 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080017343 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017344 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
17345 | H | %Ti | idle time before the HTTP request | numeric |
17346 | H | %Tq | Th + Ti + TR | numeric |
17347 | H | %TR | time to receive the full request from 1st byte| numeric |
17348 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017349 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017350 | | %Tt | Tt | numeric |
17351 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017352 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017353 | | %ac | actconn | numeric |
17354 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017355 | | %bc | beconn (backend concurrent connections) | numeric |
17356 | | %bi | backend_source_ip (connecting address) | IP |
17357 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017358 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017359 | | %ci | client_ip (accepted address) | IP |
17360 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017361 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017362 | | %fc | feconn (frontend concurrent connections) | numeric |
17363 | | %fi | frontend_ip (accepting address) | IP |
17364 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020017365 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020017366 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017367 | | %hr | captured_request_headers default style | string |
17368 | | %hrl | captured_request_headers CLF style | string list |
17369 | | %hs | captured_response_headers default style | string |
17370 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020017371 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017372 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017373 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017374 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010017375 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017376 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017377 | | %sc | srv_conn (server concurrent connections) | numeric |
17378 | | %si | server_IP (target address) | IP |
17379 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017380 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017381 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
17382 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017383 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017384 | H | %tr | date_time of HTTP request | date |
17385 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020017386 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017387 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017388 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017389 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010017390
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017391 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010017392
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017393
173948.2.5. Error log format
17395-----------------------
17396
17397When an incoming connection fails due to an SSL handshake or an invalid PROXY
17398protocol header, haproxy will log the event using a shorter, fixed line format.
17399By default, logs are emitted at the LOG_INFO level, unless the option
17400"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017401will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017402logged if the "dontlognull" option is set.
17403
17404The format looks like this :
17405
17406 >>> Dec 3 18:27:14 localhost \
17407 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
17408 Connection error during SSL handshake
17409
17410 Field Format Extract from the example above
17411 1 process_name '[' pid ']:' haproxy[6103]:
17412 2 client_ip ':' client_port 127.0.0.1:56059
17413 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
17414 4 frontend_name "/" bind_name ":" frt/f1:
17415 5 message Connection error during SSL handshake
17416
17417These fields just provide minimal information to help debugging connection
17418failures.
17419
17420
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174218.3. Advanced logging options
17422-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017423
17424Some advanced logging options are often looked for but are not easy to find out
17425just by looking at the various options. Here is an entry point for the few
17426options which can enable better logging. Please refer to the keywords reference
17427for more information about their usage.
17428
17429
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174308.3.1. Disabling logging of external tests
17431------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017432
17433It is quite common to have some monitoring tools perform health checks on
17434haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
17435commercial load-balancer, and sometimes it will simply be a more complete
17436monitoring system such as Nagios. When the tests are very frequent, users often
17437ask how to disable logging for those checks. There are three possibilities :
17438
17439 - if connections come from everywhere and are just TCP probes, it is often
17440 desired to simply disable logging of connections without data exchange, by
17441 setting "option dontlognull" in the frontend. It also disables logging of
17442 port scans, which may or may not be desired.
17443
17444 - if the connection come from a known source network, use "monitor-net" to
17445 declare this network as monitoring only. Any host in this network will then
17446 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017447 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017448 such as other load-balancers.
17449
17450 - if the tests are performed on a known URI, use "monitor-uri" to declare
17451 this URI as dedicated to monitoring. Any host sending this request will
17452 only get the result of a health-check, and the request will not be logged.
17453
17454
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174558.3.2. Logging before waiting for the session to terminate
17456----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017457
17458The problem with logging at end of connection is that you have no clue about
17459what is happening during very long sessions, such as remote terminal sessions
17460or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010017461"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017462just before data transfer begins. This means that in case of TCP, it will still
17463log the connection status to the server, and in case of HTTP, it will log just
17464after processing the server headers. In this case, the number of bytes reported
17465is the number of header bytes sent to the client. In order to avoid confusion
17466with normal logs, the total time field and the number of bytes are prefixed
17467with a '+' sign which means that real numbers are certainly larger.
17468
17469
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174708.3.3. Raising log level upon errors
17471------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017472
17473Sometimes it is more convenient to separate normal traffic from errors logs,
17474for instance in order to ease error monitoring from log files. When the option
17475"log-separate-errors" is used, connections which experience errors, timeouts,
17476retries, redispatches or HTTP status codes 5xx will see their syslog level
17477raised from "info" to "err". This will help a syslog daemon store the log in
17478a separate file. It is very important to keep the errors in the normal traffic
17479file too, so that log ordering is not altered. You should also be careful if
17480you already have configured your syslog daemon to store all logs higher than
17481"notice" in an "admin" file, because the "err" level is higher than "notice".
17482
17483
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174848.3.4. Disabling logging of successful connections
17485--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017486
17487Although this may sound strange at first, some large sites have to deal with
17488multiple thousands of logs per second and are experiencing difficulties keeping
17489them intact for a long time or detecting errors within them. If the option
17490"dontlog-normal" is set on the frontend, all normal connections will not be
17491logged. In this regard, a normal connection is defined as one without any
17492error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
17493and a response with a status 5xx is not considered normal and will be logged
17494too. Of course, doing is is really discouraged as it will remove most of the
17495useful information from the logs. Do this only if you have no other
17496alternative.
17497
17498
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200174998.4. Timing events
17500------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017501
17502Timers provide a great help in troubleshooting network problems. All values are
17503reported in milliseconds (ms). These timers should be used in conjunction with
17504the session termination flags. In TCP mode with "option tcplog" set on the
17505frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017506mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
17507addition, three other measures are provided, "Th", "Ti", and "Tq".
17508
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010017509Timings events in HTTP mode:
17510
17511 first request 2nd request
17512 |<-------------------------------->|<-------------- ...
17513 t tr t tr ...
17514 ---|----|----|----|----|----|----|----|----|--
17515 : Th Ti TR Tw Tc Tr Td : Ti ...
17516 :<---- Tq ---->: :
17517 :<-------------- Tt -------------->:
17518 :<--------- Ta --------->:
17519
17520Timings events in TCP mode:
17521
17522 TCP session
17523 |<----------------->|
17524 t t
17525 ---|----|----|----|----|---
17526 | Th Tw Tc Td |
17527 |<------ Tt ------->|
17528
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017529 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017530 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017531 only happen once during the whole connection's lifetime. A large time here
17532 may indicate that the client only pre-established the connection without
17533 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010017534 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020017535 SSL handshake was very expensive to compute. Please note that this time is
17536 reported only before the first request, so it is safe to average it over
17537 all request to calculate the amortized value. The second and subsequent
17538 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017539
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017540 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
17541 counts between the end of the handshakes and the first byte of the HTTP
17542 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020017543 to count after the end of the transmission the previous response. When a
17544 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
17545 after the previous request. Some browsers pre-establish connections to a
17546 server in order to reduce the latency of a future request, and keep them
17547 pending until they need it. This delay will be reported as the idle time. A
17548 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017549
17550 - TR: total time to get the client request (HTTP mode only). It's the time
17551 elapsed between the first bytes received and the moment the proxy received
17552 the empty line marking the end of the HTTP headers. The value "-1"
17553 indicates that the end of headers has never been seen. This happens when
17554 the client closes prematurely or times out. This time is usually very short
17555 since most requests fit in a single packet. A large time may indicate a
17556 request typed by hand during a test.
17557
17558 - Tq: total time to get the client request from the accept date or since the
17559 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010017560 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 +020017561 returns -1 as well. This timer used to be very useful before the arrival of
17562 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
17563 it in favor of TR nowadays, as the idle time adds a lot of noise to the
17564 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017565
17566 - Tw: total time spent in the queues waiting for a connection slot. It
17567 accounts for backend queue as well as the server queues, and depends on the
17568 queue size, and the time needed for the server to complete previous
17569 requests. The value "-1" means that the request was killed before reaching
17570 the queue, which is generally what happens with invalid or denied requests.
17571
17572 - Tc: total time to establish the TCP connection to the server. It's the time
17573 elapsed between the moment the proxy sent the connection request, and the
17574 moment it was acknowledged by the server, or between the TCP SYN packet and
17575 the matching SYN/ACK packet in return. The value "-1" means that the
17576 connection never established.
17577
17578 - Tr: server response time (HTTP mode only). It's the time elapsed between
17579 the moment the TCP connection was established to the server and the moment
17580 the server sent its complete response headers. It purely shows its request
17581 processing time, without the network overhead due to the data transmission.
17582 It is worth noting that when the client has data to send to the server, for
17583 instance during a POST request, the time already runs, and this can distort
17584 apparent response time. For this reason, it's generally wise not to trust
17585 too much this field for POST requests initiated from clients behind an
17586 untrusted network. A value of "-1" here means that the last the response
17587 header (empty line) was never seen, most likely because the server timeout
17588 stroke before the server managed to process the request.
17589
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017590 - Ta: total active time for the HTTP request, between the moment the proxy
17591 received the first byte of the request header and the emission of the last
17592 byte of the response body. The exception is when the "logasap" option is
17593 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
17594 a '+' sign. From this field, we can deduce "Td", the data transmission time,
17595 by subtracting other timers when valid :
17596
17597 Td = Ta - (TR + Tw + Tc + Tr)
17598
17599 Timers with "-1" values have to be excluded from this equation. Note that
17600 "Ta" can never be negative.
17601
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017602 - Tt: total session duration time, between the moment the proxy accepted it
17603 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017604 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
17605 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017606 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017607
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017608 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017609
17610 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017611 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
17612 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017613
17614These timers provide precious indications on trouble causes. Since the TCP
17615protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
17616that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017617due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
17618"Tt" is close to a timeout value specified in the configuration, it often means
17619that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017620
17621Most common cases :
17622
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017623 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
17624 the client and the proxy. This is very rare on local networks but might
17625 happen when clients are on far remote networks and send large requests. It
17626 may happen that values larger than usual appear here without any network
17627 cause. Sometimes, during an attack or just after a resource starvation has
17628 ended, haproxy may accept thousands of connections in a few milliseconds.
17629 The time spent accepting these connections will inevitably slightly delay
17630 processing of other connections, and it can happen that request times in the
17631 order of a few tens of milliseconds are measured after a few thousands of
17632 new connections have been accepted at once. Using one of the keep-alive
17633 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020017634 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017635
17636 - If "Tc" is close to 3000, a packet has probably been lost between the
17637 server and the proxy during the server connection phase. This value should
17638 always be very low, such as 1 ms on local networks and less than a few tens
17639 of ms on remote networks.
17640
Willy Tarreau55165fe2009-05-10 12:02:55 +020017641 - If "Tr" is nearly always lower than 3000 except some rare values which seem
17642 to be the average majored by 3000, there are probably some packets lost
17643 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017644
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017645 - If "Ta" is large even for small byte counts, it generally is because
17646 neither the client nor the server decides to close the connection while
17647 haproxy is running in tunnel mode and both have agreed on a keep-alive
17648 connection mode. In order to solve this issue, it will be needed to specify
17649 one of the HTTP options to manipulate keep-alive or close options on either
17650 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
17651 important when connection regulation is used with the "maxconn" option on
17652 the servers, since no new connection will be sent to the server until
17653 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017654
17655Other noticeable HTTP log cases ('xx' means any value to be ignored) :
17656
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017657 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017658 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017659 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017660
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017661 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017662 or it aborted too early. Check the session termination flags
17663 then "timeout http-request" and "timeout client" settings.
17664
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017665 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017666 servers were out of order, because the request was invalid
17667 or forbidden by ACL rules. Check the session termination
17668 flags.
17669
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017670 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
17671 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017672 Check the session termination flags, then check the
17673 "timeout connect" setting. Note that the tarpit action might
17674 return similar-looking patterns, with "Tw" equal to the time
17675 the client connection was maintained open.
17676
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017677 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017678 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017679 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017680 termination flags, then check the "timeout server" setting.
17681
17682
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200176838.5. Session state at disconnection
17684-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017685
17686TCP and HTTP logs provide a session termination indicator in the
17687"termination_state" field, just before the number of active connections. It is
176882-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
17689each of which has a special meaning :
17690
17691 - On the first character, a code reporting the first event which caused the
17692 session to terminate :
17693
17694 C : the TCP session was unexpectedly aborted by the client.
17695
17696 S : the TCP session was unexpectedly aborted by the server, or the
17697 server explicitly refused it.
17698
17699 P : the session was prematurely aborted by the proxy, because of a
17700 connection limit enforcement, because a DENY filter was matched,
17701 because of a security check which detected and blocked a dangerous
17702 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010017703 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020017704
17705 L : the session was locally processed by haproxy and was not passed to
17706 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017707
17708 R : a resource on the proxy has been exhausted (memory, sockets, source
17709 ports, ...). Usually, this appears during the connection phase, and
17710 system logs should contain a copy of the precise error. If this
17711 happens, it must be considered as a very serious anomaly which
17712 should be fixed as soon as possible by any means.
17713
17714 I : an internal error was identified by the proxy during a self-check.
17715 This should NEVER happen, and you are encouraged to report any log
17716 containing this, because this would almost certainly be a bug. It
17717 would be wise to preventively restart the process after such an
17718 event too, in case it would be caused by memory corruption.
17719
Simon Horman752dc4a2011-06-21 14:34:59 +090017720 D : the session was killed by haproxy because the server was detected
17721 as down and was configured to kill all connections when going down.
17722
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070017723 U : the session was killed by haproxy on this backup server because an
17724 active server was detected as up and was configured to kill all
17725 backup connections when going up.
17726
Willy Tarreaua2a64e92011-09-07 23:01:56 +020017727 K : the session was actively killed by an admin operating on haproxy.
17728
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017729 c : the client-side timeout expired while waiting for the client to
17730 send or receive data.
17731
17732 s : the server-side timeout expired while waiting for the server to
17733 send or receive data.
17734
17735 - : normal session completion, both the client and the server closed
17736 with nothing left in the buffers.
17737
17738 - on the second character, the TCP or HTTP session state when it was closed :
17739
Willy Tarreauf7b30a92010-12-06 22:59:17 +010017740 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017741 (HTTP mode only). Nothing was sent to any server.
17742
17743 Q : the proxy was waiting in the QUEUE for a connection slot. This can
17744 only happen when servers have a 'maxconn' parameter set. It can
17745 also happen in the global queue after a redispatch consecutive to
17746 a failed attempt to connect to a dying server. If no redispatch is
17747 reported, then no connection attempt was made to any server.
17748
17749 C : the proxy was waiting for the CONNECTION to establish on the
17750 server. The server might at most have noticed a connection attempt.
17751
17752 H : the proxy was waiting for complete, valid response HEADERS from the
17753 server (HTTP only).
17754
17755 D : the session was in the DATA phase.
17756
17757 L : the proxy was still transmitting LAST data to the client while the
17758 server had already finished. This one is very rare as it can only
17759 happen when the client dies while receiving the last packets.
17760
17761 T : the request was tarpitted. It has been held open with the client
17762 during the whole "timeout tarpit" duration or until the client
17763 closed, both of which will be reported in the "Tw" timer.
17764
17765 - : normal session completion after end of data transfer.
17766
17767 - the third character tells whether the persistence cookie was provided by
17768 the client (only in HTTP mode) :
17769
17770 N : the client provided NO cookie. This is usually the case for new
17771 visitors, so counting the number of occurrences of this flag in the
17772 logs generally indicate a valid trend for the site frequentation.
17773
17774 I : the client provided an INVALID cookie matching no known server.
17775 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020017776 cookies between HTTP/HTTPS sites, persistence conditionally
17777 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017778
17779 D : the client provided a cookie designating a server which was DOWN,
17780 so either "option persist" was used and the client was sent to
17781 this server, or it was not set and the client was redispatched to
17782 another server.
17783
Willy Tarreau996a92c2010-10-13 19:30:47 +020017784 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017785 server.
17786
Willy Tarreau996a92c2010-10-13 19:30:47 +020017787 E : the client provided a valid cookie, but with a last date which was
17788 older than what is allowed by the "maxidle" cookie parameter, so
17789 the cookie is consider EXPIRED and is ignored. The request will be
17790 redispatched just as if there was no cookie.
17791
17792 O : the client provided a valid cookie, but with a first date which was
17793 older than what is allowed by the "maxlife" cookie parameter, so
17794 the cookie is consider too OLD and is ignored. The request will be
17795 redispatched just as if there was no cookie.
17796
Willy Tarreauc89ccb62012-04-05 21:18:22 +020017797 U : a cookie was present but was not used to select the server because
17798 some other server selection mechanism was used instead (typically a
17799 "use-server" rule).
17800
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017801 - : does not apply (no cookie set in configuration).
17802
17803 - the last character reports what operations were performed on the persistence
17804 cookie returned by the server (only in HTTP mode) :
17805
17806 N : NO cookie was provided by the server, and none was inserted either.
17807
17808 I : no cookie was provided by the server, and the proxy INSERTED one.
17809 Note that in "cookie insert" mode, if the server provides a cookie,
17810 it will still be overwritten and reported as "I" here.
17811
Willy Tarreau996a92c2010-10-13 19:30:47 +020017812 U : the proxy UPDATED the last date in the cookie that was presented by
17813 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017814 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020017815 date indicated in the cookie. If any other change happens, such as
17816 a redispatch, then the cookie will be marked as inserted instead.
17817
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017818 P : a cookie was PROVIDED by the server and transmitted as-is.
17819
17820 R : the cookie provided by the server was REWRITTEN by the proxy, which
17821 happens in "cookie rewrite" or "cookie prefix" modes.
17822
17823 D : the cookie provided by the server was DELETED by the proxy.
17824
17825 - : does not apply (no cookie set in configuration).
17826
Willy Tarreau996a92c2010-10-13 19:30:47 +020017827The combination of the two first flags gives a lot of information about what
17828was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017829helpful to detect server saturation, network troubles, local system resource
17830starvation, attacks, etc...
17831
17832The most common termination flags combinations are indicated below. They are
17833alphabetically sorted, with the lowercase set just after the upper case for
17834easier finding and understanding.
17835
17836 Flags Reason
17837
17838 -- Normal termination.
17839
17840 CC The client aborted before the connection could be established to the
17841 server. This can happen when haproxy tries to connect to a recently
17842 dead (or unchecked) server, and the client aborts while haproxy is
17843 waiting for the server to respond or for "timeout connect" to expire.
17844
17845 CD The client unexpectedly aborted during data transfer. This can be
17846 caused by a browser crash, by an intermediate equipment between the
17847 client and haproxy which decided to actively break the connection,
17848 by network routing issues between the client and haproxy, or by a
17849 keep-alive session between the server and the client terminated first
17850 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010017851
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017852 cD The client did not send nor acknowledge any data for as long as the
17853 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020017854 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017855
17856 CH The client aborted while waiting for the server to start responding.
17857 It might be the server taking too long to respond or the client
17858 clicking the 'Stop' button too fast.
17859
17860 cH The "timeout client" stroke while waiting for client data during a
17861 POST request. This is sometimes caused by too large TCP MSS values
17862 for PPPoE networks which cannot transport full-sized packets. It can
17863 also happen when client timeout is smaller than server timeout and
17864 the server takes too long to respond.
17865
17866 CQ The client aborted while its session was queued, waiting for a server
17867 with enough empty slots to accept it. It might be that either all the
17868 servers were saturated or that the assigned server was taking too
17869 long a time to respond.
17870
17871 CR The client aborted before sending a full HTTP request. Most likely
17872 the request was typed by hand using a telnet client, and aborted
17873 too early. The HTTP status code is likely a 400 here. Sometimes this
17874 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020017875 and the client. "option http-ignore-probes" can be used to ignore
17876 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017877
17878 cR The "timeout http-request" stroke before the client sent a full HTTP
17879 request. This is sometimes caused by too large TCP MSS values on the
17880 client side for PPPoE networks which cannot transport full-sized
17881 packets, or by clients sending requests by hand and not typing fast
17882 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020017883 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020017884 some browsers started to implement a "pre-connect" feature consisting
17885 in speculatively connecting to some recently visited web sites just
17886 in case the user would like to visit them. This results in many
17887 connections being established to web sites, which end up in 408
17888 Request Timeout if the timeout strikes first, or 400 Bad Request when
17889 the browser decides to close them first. These ones pollute the log
17890 and feed the error counters. Some versions of some browsers have even
17891 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010017892 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020017893 http-ignore-probes" in the frontend, resulting in connections with
17894 zero data transfer to be totally ignored. This will definitely hide
17895 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017896
17897 CT The client aborted while its session was tarpitted. It is important to
17898 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020017899 wrong tarpit rules have been written. If a lot of them happen, it
17900 might make sense to lower the "timeout tarpit" value to something
17901 closer to the average reported "Tw" timer, in order not to consume
17902 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017903
Willy Tarreau570f2212013-06-10 16:42:09 +020017904 LR The request was intercepted and locally handled by haproxy. Generally
17905 it means that this was a redirect or a stats request.
17906
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017907 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017908 the TCP connection (the proxy received a TCP RST or an ICMP message
17909 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010017910 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017911 or no ARP response on local network). When this happens in HTTP mode,
17912 the status code is likely a 502 or 503 here.
17913
17914 sC The "timeout connect" stroke before a connection to the server could
17915 complete. When this happens in HTTP mode, the status code is likely a
17916 503 or 504 here.
17917
17918 SD The connection to the server died with an error during the data
17919 transfer. This usually means that haproxy has received an RST from
17920 the server or an ICMP message from an intermediate equipment while
17921 exchanging data with the server. This can be caused by a server crash
17922 or by a network issue on an intermediate equipment.
17923
17924 sD The server did not send nor acknowledge any data for as long as the
17925 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010017926 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017927 load-balancers, ...), as well as keep-alive sessions maintained
17928 between the client and the server expiring first on haproxy.
17929
17930 SH The server aborted before sending its full HTTP response headers, or
17931 it crashed while processing the request. Since a server aborting at
17932 this moment is very rare, it would be wise to inspect its logs to
17933 control whether it crashed and why. The logged request may indicate a
17934 small set of faulty requests, demonstrating bugs in the application.
17935 Sometimes this might also be caused by an IDS killing the connection
17936 between haproxy and the server.
17937
17938 sH The "timeout server" stroke before the server could return its
17939 response headers. This is the most common anomaly, indicating too
17940 long transactions, probably caused by server or database saturation.
17941 The immediate workaround consists in increasing the "timeout server"
17942 setting, but it is important to keep in mind that the user experience
17943 will suffer from these long response times. The only long term
17944 solution is to fix the application.
17945
17946 sQ The session spent too much time in queue and has been expired. See
17947 the "timeout queue" and "timeout connect" settings to find out how to
17948 fix this if it happens too often. If it often happens massively in
17949 short periods, it may indicate general problems on the affected
17950 servers due to I/O or database congestion, or saturation caused by
17951 external attacks.
17952
17953 PC The proxy refused to establish a connection to the server because the
17954 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020017955 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017956 so that it does not happen anymore. This status is very rare and
17957 might happen when the global "ulimit-n" parameter is forced by hand.
17958
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017959 PD The proxy blocked an incorrectly formatted chunked encoded message in
17960 a request or a response, after the server has emitted its headers. In
17961 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010017962 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020017963 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017964
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017965 PH The proxy blocked the server's response, because it was invalid,
17966 incomplete, dangerous (cache control), or matched a security filter.
17967 In any case, an HTTP 502 error is sent to the client. One possible
17968 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017969 containing unauthorized characters. It is also possible but quite
17970 rare, that the proxy blocked a chunked-encoding request from the
17971 client due to an invalid syntax, before the server responded. In this
17972 case, an HTTP 400 error is sent to the client and reported in the
17973 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017974
17975 PR The proxy blocked the client's HTTP request, either because of an
17976 invalid HTTP syntax, in which case it returned an HTTP 400 error to
17977 the client, or because a deny filter matched, in which case it
17978 returned an HTTP 403 error.
17979
17980 PT The proxy blocked the client's request and has tarpitted its
17981 connection before returning it a 500 server error. Nothing was sent
17982 to the server. The connection was maintained open for as long as
17983 reported by the "Tw" timer field.
17984
17985 RC A local resource has been exhausted (memory, sockets, source ports)
17986 preventing the connection to the server from establishing. The error
17987 logs will tell precisely what was missing. This is very rare and can
17988 only be solved by proper system tuning.
17989
Willy Tarreau996a92c2010-10-13 19:30:47 +020017990The combination of the two last flags gives a lot of information about how
17991persistence was handled by the client, the server and by haproxy. This is very
17992important to troubleshoot disconnections, when users complain they have to
17993re-authenticate. The commonly encountered flags are :
17994
17995 -- Persistence cookie is not enabled.
17996
17997 NN No cookie was provided by the client, none was inserted in the
17998 response. For instance, this can be in insert mode with "postonly"
17999 set on a GET request.
18000
18001 II A cookie designating an invalid server was provided by the client,
18002 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040018003 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020018004 value can be presented by a client when no other server knows it.
18005
18006 NI No cookie was provided by the client, one was inserted in the
18007 response. This typically happens for first requests from every user
18008 in "insert" mode, which makes it an easy way to count real users.
18009
18010 VN A cookie was provided by the client, none was inserted in the
18011 response. This happens for most responses for which the client has
18012 already got a cookie.
18013
18014 VU A cookie was provided by the client, with a last visit date which is
18015 not completely up-to-date, so an updated cookie was provided in
18016 response. This can also happen if there was no date at all, or if
18017 there was a date but the "maxidle" parameter was not set, so that the
18018 cookie can be switched to unlimited time.
18019
18020 EI A cookie was provided by the client, with a last visit date which is
18021 too old for the "maxidle" parameter, so the cookie was ignored and a
18022 new cookie was inserted in the response.
18023
18024 OI A cookie was provided by the client, with a first visit date which is
18025 too old for the "maxlife" parameter, so the cookie was ignored and a
18026 new cookie was inserted in the response.
18027
18028 DI The server designated by the cookie was down, a new server was
18029 selected and a new cookie was emitted in the response.
18030
18031 VI The server designated by the cookie was not marked dead but could not
18032 be reached. A redispatch happened and selected another one, which was
18033 then advertised in the response.
18034
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018035
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180368.6. Non-printable characters
18037-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018038
18039In order not to cause trouble to log analysis tools or terminals during log
18040consulting, non-printable characters are not sent as-is into log files, but are
18041converted to the two-digits hexadecimal representation of their ASCII code,
18042prefixed by the character '#'. The only characters that can be logged without
18043being escaped are comprised between 32 and 126 (inclusive). Obviously, the
18044escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
18045is the same for the character '"' which becomes "#22", as well as '{', '|' and
18046'}' when logging headers.
18047
18048Note that the space character (' ') is not encoded in headers, which can cause
18049issues for tools relying on space count to locate fields. A typical header
18050containing spaces is "User-Agent".
18051
18052Last, it has been observed that some syslog daemons such as syslog-ng escape
18053the quote ('"') with a backslash ('\'). The reverse operation can safely be
18054performed since no quote may appear anywhere else in the logs.
18055
18056
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180578.7. Capturing HTTP cookies
18058---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018059
18060Cookie capture simplifies the tracking a complete user session. This can be
18061achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018062section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018063cookie will simultaneously be checked in the request ("Cookie:" header) and in
18064the response ("Set-Cookie:" header). The respective values will be reported in
18065the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018066locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018067not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
18068user switches to a new session for example, because the server will reassign it
18069a new cookie. It is also possible to detect if a server unexpectedly sets a
18070wrong cookie to a client, leading to session crossing.
18071
18072 Examples :
18073 # capture the first cookie whose name starts with "ASPSESSION"
18074 capture cookie ASPSESSION len 32
18075
18076 # capture the first cookie whose name is exactly "vgnvisitor"
18077 capture cookie vgnvisitor= len 32
18078
18079
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180808.8. Capturing HTTP headers
18081---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018082
18083Header captures are useful to track unique request identifiers set by an upper
18084proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
18085the response, one can search for information about the response length, how the
18086server asked the cache to behave, or an object location during a redirection.
18087
18088Header captures are performed using the "capture request header" and "capture
18089response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018090section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018091
18092It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018093time. Non-existent headers are logged as empty strings, and if one header
18094appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018095are grouped within braces '{' and '}' in the same order as they were declared,
18096and delimited with a vertical bar '|' without any space. Response headers
18097follow the same representation, but are displayed after a space following the
18098request headers block. These blocks are displayed just before the HTTP request
18099in the logs.
18100
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020018101As a special case, it is possible to specify an HTTP header capture in a TCP
18102frontend. The purpose is to enable logging of headers which will be parsed in
18103an HTTP backend if the request is then switched to this HTTP backend.
18104
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018105 Example :
18106 # This instance chains to the outgoing proxy
18107 listen proxy-out
18108 mode http
18109 option httplog
18110 option logasap
18111 log global
18112 server cache1 192.168.1.1:3128
18113
18114 # log the name of the virtual server
18115 capture request header Host len 20
18116
18117 # log the amount of data uploaded during a POST
18118 capture request header Content-Length len 10
18119
18120 # log the beginning of the referrer
18121 capture request header Referer len 20
18122
18123 # server name (useful for outgoing proxies only)
18124 capture response header Server len 20
18125
18126 # logging the content-length is useful with "option logasap"
18127 capture response header Content-Length len 10
18128
Davor Ocelice9ed2812017-12-25 17:49:28 +010018129 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018130 capture response header Cache-Control len 8
18131
18132 # the Via header will report the next proxy's name
18133 capture response header Via len 20
18134
18135 # log the URL location during a redirection
18136 capture response header Location len 20
18137
18138 >>> Aug 9 20:26:09 localhost \
18139 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
18140 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
18141 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
18142 "GET http://fr.adserver.yahoo.com/"
18143
18144 >>> Aug 9 20:30:46 localhost \
18145 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
18146 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
18147 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018148 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018149
18150 >>> Aug 9 20:30:46 localhost \
18151 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
18152 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
18153 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
18154 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018155 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018156
18157
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200181588.9. Examples of logs
18159---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018160
18161These are real-world examples of logs accompanied with an explanation. Some of
18162them have been made up by hand. The syslog part has been removed for better
18163reading. Their sole purpose is to explain how to decipher them.
18164
18165 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
18166 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
18167 "HEAD / HTTP/1.0"
18168
18169 => long request (6.5s) entered by hand through 'telnet'. The server replied
18170 in 147 ms, and the session ended normally ('----')
18171
18172 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
18173 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
18174 0/9 "HEAD / HTTP/1.0"
18175
18176 => Idem, but the request was queued in the global queue behind 9 other
18177 requests, and waited there for 1230 ms.
18178
18179 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
18180 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
18181 "GET /image.iso HTTP/1.0"
18182
18183 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018184 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018185 14 ms, 243 bytes of headers were sent to the client, and total time from
18186 accept to first data byte is 30 ms.
18187
18188 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
18189 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
18190 "GET /cgi-bin/bug.cgi? HTTP/1.0"
18191
18192 => the proxy blocked a server response either because of an "rspdeny" or
18193 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020018194 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018195 risked being cached. In this case, the response is replaced with a "502
18196 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
18197 to return the 502 and not the server.
18198
18199 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018200 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 +010018201
18202 => the client never completed its request and aborted itself ("C---") after
18203 8.5s, while the proxy was waiting for the request headers ("-R--").
18204 Nothing was sent to any server.
18205
18206 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
18207 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
18208
18209 => The client never completed its request, which was aborted by the
18210 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010018211 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018212 send a 408 return code to the client.
18213
18214 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
18215 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
18216
18217 => This log was produced with "option tcplog". The client timed out after
18218 5 seconds ("c----").
18219
18220 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
18221 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018222 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018223
18224 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018225 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018226 (config says 'retries 3'), and no redispatch (otherwise we would have
18227 seen "/+3"). Status code 503 was returned to the client. There were 115
18228 connections on this server, 202 connections on this proxy, and 205 on
18229 the global process. It is possible that the server refused the
18230 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010018231
Willy Tarreau52b2d222011-09-07 23:48:48 +020018232
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200182339. Supported filters
18234--------------------
18235
18236Here are listed officially supported filters with the list of parameters they
18237accept. Depending on compile options, some of these filters might be
18238unavailable. The list of available filters is reported in haproxy -vv.
18239
18240See also : "filter"
18241
182429.1. Trace
18243----------
18244
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018245filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018246
18247 Arguments:
18248 <name> is an arbitrary name that will be reported in
18249 messages. If no name is provided, "TRACE" is used.
18250
18251 <random-parsing> enables the random parsing of data exchanged between
18252 the client and the server. By default, this filter
18253 parses all available data. With this parameter, it
18254 only parses a random amount of the available data.
18255
Davor Ocelice9ed2812017-12-25 17:49:28 +010018256 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018257 default, this filter forwards all previously parsed
18258 data. With this parameter, it only forwards a random
18259 amount of the parsed data.
18260
Davor Ocelice9ed2812017-12-25 17:49:28 +010018261 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018262
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018263This filter can be used as a base to develop new filters. It defines all
18264callbacks and print a message on the standard error stream (stderr) with useful
18265information for all of them. It may be useful to debug the activity of other
18266filters or, quite simply, HAProxy's activity.
18267
18268Using <random-parsing> and/or <random-forwarding> parameters is a good way to
18269tests the behavior of a filter that parses data exchanged between a client and
18270a server by adding some latencies in the processing.
18271
18272
182739.2. HTTP compression
18274---------------------
18275
18276filter compression
18277
18278The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
18279keyword must still be used to enable and configure the HTTP compression. And
Christopher Faulet27d93c32018-12-15 22:32:02 +010018280when no other filter is used, it is enough. When used with the cache enabled,
18281it is also enough. In this case, the compression is always done after the
18282response is stored in the cache. But it is mandatory to explicitly use a filter
18283line to enable the HTTP compression when at least one filter other than the
18284cache is used for the same listener/frontend/backend. This is important to know
18285the filters evaluation order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018286
Christopher Faulet27d93c32018-12-15 22:32:02 +010018287See also : "compression" and section 9.4 about the cache filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018288
18289
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200182909.3. Stream Processing Offload Engine (SPOE)
18291--------------------------------------------
18292
18293filter spoe [engine <name>] config <file>
18294
18295 Arguments :
18296
18297 <name> is the engine name that will be used to find the right scope in
18298 the configuration file. If not provided, all the file will be
18299 parsed.
18300
18301 <file> is the path of the engine configuration file. This file can
18302 contain configuration of several engines. In this case, each
18303 part must be placed in its own scope.
18304
18305The Stream Processing Offload Engine (SPOE) is a filter communicating with
18306external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010018307streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018308exchanged with them are configured in dedicated files, for the main part. It
18309also requires dedicated backends, defined in HAProxy configuration.
18310
18311SPOE communicates with external components using an in-house binary protocol,
18312the Stream Processing Offload Protocol (SPOP).
18313
Tim Düsterhus4896c442016-11-29 02:15:19 +010018314For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018315"doc/SPOE.txt".
18316
18317Important note:
18318 The SPOE filter is highly experimental for now and was not heavily
18319 tested. It is really not production ready. So use it carefully.
18320
Christopher Faulet99a17a22018-12-11 09:18:27 +0100183219.4. Cache
18322----------
18323
18324filter cache <name>
18325
18326 Arguments :
18327
18328 <name> is name of the cache section this filter will use.
18329
18330The cache uses a filter to store cacheable responses. The HTTP rules
18331"cache-store" and "cache-use" must be used to define how and when to use a
18332cache. By default the correpsonding filter is implicitly defined. And when no
Christopher Faulet27d93c32018-12-15 22:32:02 +010018333other filters than cache or compression are used, it is enough. In such case,
18334the compression filter is always evaluated after the cache filter. But it is
18335mandatory to explicitly use a filter line to use a cache when at least one
18336filter other than the compression is used for the same
18337listener/frontend/backend. This is important to know the filters evaluation
18338order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018339
Christopher Faulet27d93c32018-12-15 22:32:02 +010018340See also : section 9.2 about the compression filter and section 10 about cache.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018341
William Lallemand86d0df02017-11-24 21:36:45 +01001834210. Cache
18343---------
18344
18345HAProxy provides a cache, which was designed to perform cache on small objects
18346(favicon, css...). This is a minimalist low-maintenance cache which runs in
18347RAM.
18348
18349The cache is based on a memory which is shared between processes and threads,
Cyril Bonté7b888f12017-11-26 22:24:31 +010018350this memory is split in blocks of 1k.
William Lallemand86d0df02017-11-24 21:36:45 +010018351
18352If an object is not used anymore, it can be deleted to store a new object
18353independently of its expiration date. The oldest objects are deleted first
18354when we try to allocate a new one.
18355
Cyril Bonté7b888f12017-11-26 22:24:31 +010018356The cache uses a hash of the host header and the URI as the key.
William Lallemand86d0df02017-11-24 21:36:45 +010018357
18358It's possible to view the status of a cache using the Unix socket command
18359"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
18360for more details.
18361
18362When an object is delivered from the cache, the server name in the log is
18363replaced by "<CACHE>".
18364
Cyril Bonté7b888f12017-11-26 22:24:31 +01001836510.1. Limitation
18366----------------
William Lallemand86d0df02017-11-24 21:36:45 +010018367
18368The cache won't store and won't deliver objects in these cases:
18369
18370- If the response is not a 200
18371- If the response contains a Vary header
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018372- If the Content-Length + the headers size is greater than "max-object-size"
William Lallemand86d0df02017-11-24 21:36:45 +010018373- If the response is not cacheable
18374
18375- If the request is not a GET
18376- If the HTTP version of the request is smaller than 1.1
William Lallemand8a16fe02018-05-22 11:04:33 +020018377- If the request contains an Authorization header
William Lallemand86d0df02017-11-24 21:36:45 +010018378
Christopher Faulet99a17a22018-12-11 09:18:27 +010018379Caution!: For HAProxy version prior to 1.9, due to the limitation of the
18380filters, it is not recommended to use the cache with other filters. Using them
18381can cause undefined behavior if they modify the response (compression for
18382example). For HAProxy 1.9 and greater, it is safe, for HTX proxies only (see
18383"option http-use-htx" for details).
William Lallemand86d0df02017-11-24 21:36:45 +010018384
Cyril Bonté7b888f12017-11-26 22:24:31 +01001838510.2. Setup
18386-----------
William Lallemand86d0df02017-11-24 21:36:45 +010018387
18388To setup a cache, you must define a cache section and use it in a proxy with
18389the corresponding http-request and response actions.
18390
Cyril Bonté7b888f12017-11-26 22:24:31 +01001839110.2.1. Cache section
18392---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018393
18394cache <name>
18395 Declare a cache section, allocate a shared cache memory named <name>, the
18396 size of cache is mandatory.
18397
18398total-max-size <megabytes>
Davor Ocelice9ed2812017-12-25 17:49:28 +010018399 Define the size in RAM of the cache in megabytes. This size is split in
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018400 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
William Lallemand86d0df02017-11-24 21:36:45 +010018401
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018402max-object-size <bytes>
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018403 Define the maximum size of the objects to be cached. Must not be greater than
18404 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
18405 All objects with sizes larger than "max-object-size" will not be cached.
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018406
William Lallemand86d0df02017-11-24 21:36:45 +010018407max-age <seconds>
18408 Define the maximum expiration duration. The expiration is set has the lowest
18409 value between the s-maxage or max-age (in this order) directive in the
18410 Cache-Control response header and this value. The default value is 60
18411 seconds, which means that you can't cache an object more than 60 seconds by
18412 default.
18413
Cyril Bonté7b888f12017-11-26 22:24:31 +01001841410.2.2. Proxy section
18415---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018416
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018417http-request cache-use <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018418 Try to deliver a cached object from the cache <name>. This directive is also
18419 mandatory to store the cache as it calculates the cache hash. If you want to
18420 use a condition for both storage and delivering that's a good idea to put it
18421 after this one.
18422
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018423http-response cache-store <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018424 Store an http-response within the cache. The storage of the response headers
18425 is done at this step, which means you can use others http-response actions
18426 to modify headers before or after the storage of the response. This action
18427 is responsible for the setup of the cache storage filter.
18428
18429
18430Example:
18431
18432 backend bck1
18433 mode http
18434
18435 http-request cache-use foobar
18436 http-response cache-store foobar
18437 server srv1 127.0.0.1:80
18438
18439 cache foobar
18440 total-max-size 4
18441 max-age 240
18442
Willy Tarreau0ba27502007-12-24 16:55:16 +010018443/*
18444 * Local variables:
18445 * fill-column: 79
18446 * End:
18447 */