blob: d2cdf27497321b6932e6c949b51cfe36776c0529 [file] [log] [blame]
Willy Tarreau6a06a402007-07-15 20:15:28 +02001 ----------------------
Willy Tarreau8317b282014-04-23 01:49:41 +02002 HAProxy
Willy Tarreau6a06a402007-07-15 20:15:28 +02003 Configuration Manual
4 ----------------------
Willy Tarreaufba74ea2018-12-22 11:19:45 +01005 version 2.0
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreauabc874e2019-06-02 12:06:08 +02007 2019/06/02
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
Willy Tarreauc57f0e22009-05-10 13:12:33 +020054
554. Proxies
564.1. Proxy keywords matrix
574.2. Alphabetically sorted keywords reference
58
Davor Ocelice9ed2812017-12-25 17:49:28 +0100595. Bind and server options
Willy Tarreau086fbf52012-09-24 20:34:51 +0200605.1. Bind options
615.2. Server and default-server options
Baptiste Assmann1fa66662015-04-14 00:28:47 +0200625.3. Server DNS resolution
635.3.1. Global overview
645.3.2. The resolvers section
Willy Tarreauc57f0e22009-05-10 13:12:33 +020065
666. HTTP header manipulation
67
Willy Tarreau74ca5042013-06-11 23:12:07 +0200687. Using ACLs and fetching samples
697.1. ACL basics
707.1.1. Matching booleans
717.1.2. Matching integers
727.1.3. Matching strings
737.1.4. Matching regular expressions (regexes)
747.1.5. Matching arbitrary data blocks
757.1.6. Matching IPv4 and IPv6 addresses
767.2. Using ACLs to form conditions
777.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200787.3.1. Converters
797.3.2. Fetching samples from internal states
807.3.3. Fetching samples at Layer 4
817.3.4. Fetching samples at Layer 5
827.3.5. Fetching samples from buffer contents (Layer 6)
837.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200847.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020085
868. Logging
878.1. Log levels
888.2. Log formats
898.2.1. Default log format
908.2.2. TCP log format
918.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100928.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100938.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200948.3. Advanced logging options
958.3.1. Disabling logging of external tests
968.3.2. Logging before waiting for the session to terminate
978.3.3. Raising log level upon errors
988.3.4. Disabling logging of successful connections
998.4. Timing events
1008.5. Session state at disconnection
1018.6. Non-printable characters
1028.7. Capturing HTTP cookies
1038.8. Capturing HTTP headers
1048.9. Examples of logs
105
Christopher Fauletc3fe5332016-04-07 15:30:10 +02001069. Supported filters
1079.1. Trace
1089.2. HTTP compression
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +02001099.3. Stream Processing Offload Engine (SPOE)
Christopher Faulet99a17a22018-12-11 09:18:27 +01001109.4. Cache
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200111
William Lallemand86d0df02017-11-24 21:36:45 +010011210. Cache
Cyril Bonté7b888f12017-11-26 22:24:31 +010011310.1. Limitation
11410.2. Setup
11510.2.1. Cache section
11610.2.2. Proxy section
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200117
1181. Quick reminder about HTTP
119----------------------------
120
Davor Ocelice9ed2812017-12-25 17:49:28 +0100121When HAProxy is running in HTTP mode, both the request and the response are
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200122fully analyzed and indexed, thus it becomes possible to build matching criteria
123on almost anything found in the contents.
124
125However, it is important to understand how HTTP requests and responses are
126formed, and how HAProxy decomposes them. It will then become easier to write
127correct rules and to debug existing configurations.
128
129
1301.1. The HTTP transaction model
131-------------------------------
132
133The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100134to one and only one response. Traditionally, a TCP connection is established
Davor Ocelice9ed2812017-12-25 17:49:28 +0100135from the client to the server, a request is sent by the client through the
136connection, the server responds, and the connection is closed. A new request
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200137will involve a new connection :
138
139 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
140
141In this mode, called the "HTTP close" mode, there are as many connection
142establishments as there are HTTP transactions. Since the connection is closed
143by the server after the response, the client does not need to know the content
144length.
145
146Due to the transactional nature of the protocol, it was possible to improve it
147to avoid closing a connection between two subsequent transactions. In this mode
148however, it is mandatory that the server indicates the content length for each
149response so that the client does not wait indefinitely. For this, a special
150header is used: "Content-length". This mode is called the "keep-alive" mode :
151
152 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
153
154Its advantages are a reduced latency between transactions, and less processing
155power required on the server side. It is generally better than the close mode,
156but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200157a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200158
Willy Tarreau95c4e142017-11-26 12:18:55 +0100159Another improvement in the communications is the pipelining mode. It still uses
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200160keep-alive, but the client does not wait for the first response to send the
161second request. This is useful for fetching large number of images composing a
162page :
163
164 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
165
166This can obviously have a tremendous benefit on performance because the network
167latency is eliminated between subsequent requests. Many HTTP agents do not
168correctly support pipelining since there is no way to associate a response with
169the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100170server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200171
Willy Tarreau95c4e142017-11-26 12:18:55 +0100172The next improvement is the multiplexed mode, as implemented in HTTP/2. This
173time, each transaction is assigned a single stream identifier, and all streams
174are multiplexed over an existing connection. Many requests can be sent in
175parallel by the client, and responses can arrive in any order since they also
176carry the stream identifier.
177
Willy Tarreau70dffda2014-01-30 03:07:23 +0100178By default HAProxy operates in keep-alive mode with regards to persistent
179connections: for each connection it processes each request and response, and
180leaves the connection idle on both sides between the end of a response and the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100181start of a new request. When it receives HTTP/2 connections from a client, it
182processes all the requests in parallel and leaves the connection idling,
183waiting for new requests, just as if it was a keep-alive HTTP connection.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200184
Christopher Faulet315b39c2018-09-21 16:26:19 +0200185HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +0100186 - keep alive : all requests and responses are processed (default)
187 - tunnel : only the first request and response are processed,
Christopher Faulet6c9bbb22019-03-26 21:37:23 +0100188 everything else is forwarded with no analysis (deprecated).
Willy Tarreau70dffda2014-01-30 03:07:23 +0100189 - server close : the server-facing connection is closed after the response.
Christopher Faulet315b39c2018-09-21 16:26:19 +0200190 - close : the connection is actively closed after end of response.
Willy Tarreau70dffda2014-01-30 03:07:23 +0100191
Davor Ocelice9ed2812017-12-25 17:49:28 +0100192For HTTP/2, the connection mode resembles more the "server close" mode : given
193the independence of all streams, there is currently no place to hook the idle
Willy Tarreau95c4e142017-11-26 12:18:55 +0100194server connection after a response, so it is closed after the response. HTTP/2
195is only supported for incoming connections, not on connections going to
196servers.
197
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200198
1991.2. HTTP request
200-----------------
201
202First, let's consider this HTTP request :
203
204 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100205 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200206 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
207 2 Host: www.mydomain.com
208 3 User-agent: my small browser
209 4 Accept: image/jpeg, image/gif
210 5 Accept: image/png
211
212
2131.2.1. The Request line
214-----------------------
215
216Line 1 is the "request line". It is always composed of 3 fields :
217
218 - a METHOD : GET
219 - a URI : /serv/login.php?lang=en&profile=2
220 - a version tag : HTTP/1.1
221
222All of them are delimited by what the standard calls LWS (linear white spaces),
223which are commonly spaces, but can also be tabs or line feeds/carriage returns
224followed by spaces/tabs. The method itself cannot contain any colon (':') and
225is limited to alphabetic letters. All those various combinations make it
226desirable that HAProxy performs the splitting itself rather than leaving it to
227the user to write a complex or inaccurate regular expression.
228
229The URI itself can have several forms :
230
231 - A "relative URI" :
232
233 /serv/login.php?lang=en&profile=2
234
235 It is a complete URL without the host part. This is generally what is
236 received by servers, reverse proxies and transparent proxies.
237
238 - An "absolute URI", also called a "URL" :
239
240 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
241
242 It is composed of a "scheme" (the protocol name followed by '://'), a host
243 name or address, optionally a colon (':') followed by a port number, then
244 a relative URI beginning at the first slash ('/') after the address part.
245 This is generally what proxies receive, but a server supporting HTTP/1.1
246 must accept this form too.
247
248 - a star ('*') : this form is only accepted in association with the OPTIONS
249 method and is not relayable. It is used to inquiry a next hop's
250 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100251
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200252 - an address:port combination : 192.168.0.12:80
253 This is used with the CONNECT method, which is used to establish TCP
254 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
255 other protocols too.
256
257In a relative URI, two sub-parts are identified. The part before the question
258mark is called the "path". It is typically the relative path to static objects
259on the server. The part after the question mark is called the "query string".
260It is mostly used with GET requests sent to dynamic scripts and is very
261specific to the language, framework or application in use.
262
Willy Tarreau95c4e142017-11-26 12:18:55 +0100263HTTP/2 doesn't convey a version information with the request, so the version is
Davor Ocelice9ed2812017-12-25 17:49:28 +0100264assumed to be the same as the one of the underlying protocol (i.e. "HTTP/2").
Willy Tarreau95c4e142017-11-26 12:18:55 +0100265However, haproxy natively processes HTTP/1.x requests and headers, so requests
266received over an HTTP/2 connection are transcoded to HTTP/1.1 before being
267processed. This explains why they still appear as "HTTP/1.1" in haproxy's logs
268as well as in server logs.
269
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200270
2711.2.2. The request headers
272--------------------------
273
274The headers start at the second line. They are composed of a name at the
275beginning of the line, immediately followed by a colon (':'). Traditionally,
276an LWS is added after the colon but that's not required. Then come the values.
277Multiple identical headers may be folded into one single line, delimiting the
278values with commas, provided that their order is respected. This is commonly
279encountered in the "Cookie:" field. A header may span over multiple lines if
280the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
281define a total of 3 values for the "Accept:" header.
282
Davor Ocelice9ed2812017-12-25 17:49:28 +0100283Contrary to a common misconception, header names are not case-sensitive, and
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200284their values are not either if they refer to other header names (such as the
Willy Tarreau95c4e142017-11-26 12:18:55 +0100285"Connection:" header). In HTTP/2, header names are always sent in lower case,
286as can be seen when running in debug mode.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200287
288The end of the headers is indicated by the first empty line. People often say
289that it's a double line feed, which is not exact, even if a double line feed
290is one valid form of empty line.
291
292Fortunately, HAProxy takes care of all these complex combinations when indexing
293headers, checking values and counting them, so there is no reason to worry
294about the way they could be written, but it is important not to accuse an
295application of being buggy if it does unusual, valid things.
296
297Important note:
Lukas Tribus23953682017-04-28 13:24:30 +0000298 As suggested by RFC7231, HAProxy normalizes headers by replacing line breaks
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200299 in the middle of headers by LWS in order to join multi-line headers. This
300 is necessary for proper analysis and helps less capable HTTP parsers to work
301 correctly and not to be fooled by such complex constructs.
302
303
3041.3. HTTP response
305------------------
306
307An HTTP response looks very much like an HTTP request. Both are called HTTP
308messages. Let's consider this HTTP response :
309
310 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100311 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200312 1 HTTP/1.1 200 OK
313 2 Content-length: 350
314 3 Content-Type: text/html
315
Willy Tarreau816b9792009-09-15 21:25:21 +0200316As a special case, HTTP supports so called "Informational responses" as status
317codes 1xx. These messages are special in that they don't convey any part of the
318response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100319continue to post its request for instance. In the case of a status 100 response
320the requested information will be carried by the next non-100 response message
321following the informational one. This implies that multiple responses may be
322sent to a single request, and that this only works when keep-alive is enabled
323(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
324correctly forward and skip them, and only process the next non-100 response. As
325such, these messages are neither logged nor transformed, unless explicitly
326state otherwise. Status 101 messages indicate that the protocol is changing
327over the same connection and that haproxy must switch to tunnel mode, just as
328if a CONNECT had occurred. Then the Upgrade header would contain additional
329information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200330
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200331
Davor Ocelice9ed2812017-12-25 17:49:28 +01003321.3.1. The response line
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200333------------------------
334
335Line 1 is the "response line". It is always composed of 3 fields :
336
337 - a version tag : HTTP/1.1
338 - a status code : 200
339 - a reason : OK
340
341The status code is always 3-digit. The first digit indicates a general status :
Davor Ocelice9ed2812017-12-25 17:49:28 +0100342 - 1xx = informational message to be skipped (e.g. 100, 101)
343 - 2xx = OK, content is following (e.g. 200, 206)
344 - 3xx = OK, no content following (e.g. 302, 304)
345 - 4xx = error caused by the client (e.g. 401, 403, 404)
346 - 5xx = error caused by the server (e.g. 500, 502, 503)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200347
Lukas Tribus23953682017-04-28 13:24:30 +0000348Please refer to RFC7231 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100349"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200350found there, but it's a common practice to respect the well-established
351messages. It can be composed of one or multiple words, such as "OK", "Found",
352or "Authentication Required".
353
Davor Ocelice9ed2812017-12-25 17:49:28 +0100354HAProxy may emit the following status codes by itself :
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200355
356 Code When / reason
357 200 access to stats page, and when replying to monitoring requests
358 301 when performing a redirection, depending on the configured code
359 302 when performing a redirection, depending on the configured code
360 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100361 307 when performing a redirection, depending on the configured code
362 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200363 400 for an invalid or too large request
364 401 when an authentication is required to perform the action (when
365 accessing the stats page)
366 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
367 408 when the request timeout strikes before the request is complete
368 500 when haproxy encounters an unrecoverable internal error, such as a
369 memory allocation failure, which should never happen
370 502 when the server returns an empty, invalid or incomplete response, or
371 when an "rspdeny" filter blocks the response.
372 503 when no server was available to handle the request, or in response to
373 monitoring requests which match the "monitor fail" condition
374 504 when the response timeout strikes before the server responds
375
376The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3774.2).
378
379
3801.3.2. The response headers
381---------------------------
382
383Response headers work exactly like request headers, and as such, HAProxy uses
384the same parsing function for both. Please refer to paragraph 1.2.2 for more
385details.
386
387
3882. Configuring HAProxy
389----------------------
390
3912.1. Configuration file format
392------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200393
394HAProxy's configuration process involves 3 major sources of parameters :
395
396 - the arguments from the command-line, which always take precedence
397 - the "global" section, which sets process-wide parameters
398 - the proxies sections which can take form of "defaults", "listen",
399 "frontend" and "backend".
400
Willy Tarreau0ba27502007-12-24 16:55:16 +0100401The configuration file syntax consists in lines beginning with a keyword
402referenced in this manual, optionally followed by one or several parameters
William Lallemandf9873ba2015-05-05 17:37:14 +0200403delimited by spaces.
Willy Tarreau0ba27502007-12-24 16:55:16 +0100404
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200405
William Lallemandf9873ba2015-05-05 17:37:14 +02004062.2. Quoting and escaping
407-------------------------
408
409HAProxy's configuration introduces a quoting and escaping system similar to
410many programming languages. The configuration file supports 3 types: escaping
411with a backslash, weak quoting with double quotes, and strong quoting with
412single quotes.
413
414If spaces have to be entered in strings, then they must be escaped by preceding
415them by a backslash ('\') or by quoting them. Backslashes also have to be
416escaped by doubling or strong quoting them.
417
418Escaping is achieved by preceding a special character by a backslash ('\'):
419
420 \ to mark a space and differentiate it from a delimiter
421 \# to mark a hash and differentiate it from a comment
422 \\ to use a backslash
423 \' to use a single quote and differentiate it from strong quoting
424 \" to use a double quote and differentiate it from weak quoting
425
426Weak quoting is achieved by using double quotes (""). Weak quoting prevents
427the interpretation of:
428
429 space as a parameter separator
430 ' single quote as a strong quoting delimiter
431 # hash as a comment start
432
William Lallemandb2f07452015-05-12 14:27:13 +0200433Weak quoting permits the interpretation of variables, if you want to use a non
434-interpreted dollar within a double quoted string, you should escape it with a
435backslash ("\$"), it does not work outside weak quoting.
436
437Interpretation of escaping and special characters are not prevented by weak
William Lallemandf9873ba2015-05-05 17:37:14 +0200438quoting.
439
440Strong quoting is achieved by using single quotes (''). Inside single quotes,
441nothing is interpreted, it's the efficient way to quote regexes.
442
443Quoted and escaped strings are replaced in memory by their interpreted
444equivalent, it allows you to perform concatenation.
445
446 Example:
447 # those are equivalents:
448 log-format %{+Q}o\ %t\ %s\ %{-Q}r
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
454 # those are equivalents:
455 reqrep "^([^\ :]*)\ /static/(.*)" \1\ /\2
456 reqrep "^([^ :]*)\ /static/(.*)" '\1 /\2'
457 reqrep "^([^ :]*)\ /static/(.*)" "\1 /\2"
458 reqrep "^([^ :]*)\ /static/(.*)" "\1\ /\2"
459
460
William Lallemandb2f07452015-05-12 14:27:13 +02004612.3. Environment variables
462--------------------------
463
464HAProxy's configuration supports environment variables. Those variables are
465interpreted only within double quotes. Variables are expanded during the
466configuration parsing. Variable names must be preceded by a dollar ("$") and
467optionally enclosed with braces ("{}") similarly to what is done in Bourne
468shell. Variable names can contain alphanumerical characters or the character
469underscore ("_") but should not start with a digit.
470
471 Example:
472
473 bind "fd@${FD_APP1}"
474
475 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
476
477 user "$HAPROXY_USER"
478
William Lallemanddaf4cd22018-04-17 16:46:13 +0200479A special variable $HAPROXY_LOCALPEER is defined at the startup of the process
480which contains the name of the local peer. (See "-L" in the management guide.)
481
William Lallemandb2f07452015-05-12 14:27:13 +0200482
4832.4. Time format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200484----------------
485
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100486Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100487values are generally expressed in milliseconds (unless explicitly stated
488otherwise) but may be expressed in any other unit by suffixing the unit to the
489numeric value. It is important to consider this because it will not be repeated
490for every keyword. Supported units are :
491
492 - us : microseconds. 1 microsecond = 1/1000000 second
493 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
494 - s : seconds. 1s = 1000ms
495 - m : minutes. 1m = 60s = 60000ms
496 - h : hours. 1h = 60m = 3600s = 3600000ms
497 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
498
499
Lukas Tribusaa83a312017-03-21 09:25:09 +00005002.5. Examples
Patrick Mezard35da19c2010-06-12 17:02:47 +0200501-------------
502
503 # Simple configuration for an HTTP proxy listening on port 80 on all
504 # interfaces and forwarding requests to a single backend "servers" with a
505 # single server "server1" listening on 127.0.0.1:8000
506 global
507 daemon
508 maxconn 256
509
510 defaults
511 mode http
512 timeout connect 5000ms
513 timeout client 50000ms
514 timeout server 50000ms
515
516 frontend http-in
517 bind *:80
518 default_backend servers
519
520 backend servers
521 server server1 127.0.0.1:8000 maxconn 32
522
523
524 # The same configuration defined with a single listen block. Shorter but
525 # less expressive, especially in HTTP mode.
526 global
527 daemon
528 maxconn 256
529
530 defaults
531 mode http
532 timeout connect 5000ms
533 timeout client 50000ms
534 timeout server 50000ms
535
536 listen http-in
537 bind *:80
538 server server1 127.0.0.1:8000 maxconn 32
539
540
541Assuming haproxy is in $PATH, test these configurations in a shell with:
542
Willy Tarreauccb289d2010-12-11 20:19:38 +0100543 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200544
545
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005463. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200547--------------------
548
549Parameters in the "global" section are process-wide and often OS-specific. They
550are generally set once for all and do not need being changed once correct. Some
551of them have command-line equivalents.
552
553The following keywords are supported in the "global" section :
554
555 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200556 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200557 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200558 - crt-base
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200559 - cpu-map
Willy Tarreau6a06a402007-07-15 20:15:28 +0200560 - daemon
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200561 - description
562 - deviceatlas-json-file
563 - deviceatlas-log-level
564 - deviceatlas-separator
565 - deviceatlas-properties-cookie
Simon Horman98637e52014-06-20 12:30:16 +0900566 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200567 - gid
568 - group
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100569 - hard-stop-after
Willy Tarreau6a06a402007-07-15 20:15:28 +0200570 - log
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200571 - log-tag
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100572 - log-send-hostname
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200573 - lua-load
William Lallemand27edc4b2019-05-07 17:49:33 +0200574 - mworker-max-reloads
Willy Tarreau6a06a402007-07-15 20:15:28 +0200575 - nbproc
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200576 - nbthread
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200577 - node
Willy Tarreau6a06a402007-07-15 20:15:28 +0200578 - pidfile
Willy Tarreau1d549722016-02-16 12:41:57 +0100579 - presetenv
580 - resetenv
Willy Tarreau6a06a402007-07-15 20:15:28 +0200581 - uid
582 - ulimit-n
583 - user
Willy Tarreau636848a2019-04-15 19:38:50 +0200584 - set-dumpable
Willy Tarreau1d549722016-02-16 12:41:57 +0100585 - setenv
Willy Tarreaufbee7132007-10-18 13:53:22 +0200586 - stats
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200587 - ssl-default-bind-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200588 - ssl-default-bind-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200589 - ssl-default-bind-options
590 - ssl-default-server-ciphers
Dirkjan Bussink415150f2018-09-14 11:14:21 +0200591 - ssl-default-server-ciphersuites
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200592 - ssl-default-server-options
593 - ssl-dh-param-file
Emeric Brun850efd52014-01-29 12:24:34 +0100594 - ssl-server-verify
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100595 - unix-bind
Willy Tarreau1d549722016-02-16 12:41:57 +0100596 - unsetenv
Thomas Holmesdb04f192015-05-18 13:21:39 +0100597 - 51degrees-data-file
598 - 51degrees-property-name-list
Dragan Dosen93b38d92015-06-29 16:43:25 +0200599 - 51degrees-property-separator
Dragan Dosenae6d39a2015-06-29 16:43:27 +0200600 - 51degrees-cache-size
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200601 - wurfl-data-file
602 - wurfl-information-list
603 - wurfl-information-list-separator
Willy Tarreaub3cc9f22019-04-19 16:03:32 +0200604 - wurfl-cache-size
Willy Tarreaud72758d2010-01-12 10:42:19 +0100605
Willy Tarreau6a06a402007-07-15 20:15:28 +0200606 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200607 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200608 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200609 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100610 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100611 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100612 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200613 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200614 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200615 - maxsslrate
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200616 - maxzlibmem
Willy Tarreau6a06a402007-07-15 20:15:28 +0200617 - noepoll
618 - nokqueue
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +0000619 - noevports
Willy Tarreau6a06a402007-07-15 20:15:28 +0200620 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100621 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300622 - nogetaddrinfo
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +0000623 - noreuseport
Willy Tarreau75c62c22018-11-22 11:02:09 +0100624 - profiling.tasks
Willy Tarreaufe255b72007-10-14 23:09:26 +0200625 - spread-checks
Baptiste Assmann5626f482015-08-23 10:00:10 +0200626 - server-state-base
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +0200627 - server-state-file
Grant Zhang872f9c22017-01-21 01:10:18 +0000628 - ssl-engine
Grant Zhangfa6c7ee2017-01-14 01:42:15 +0000629 - ssl-mode-async
Baptiste Assmann3493d0f2015-10-12 20:21:23 +0200630 - tune.buffers.limit
631 - tune.buffers.reserve
Willy Tarreau27a674e2009-08-17 07:23:33 +0200632 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200633 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100634 - tune.comp.maxlevel
Willy Tarreaufe20e5b2017-07-27 11:42:14 +0200635 - tune.h2.header-table-size
Willy Tarreaue6baec02017-07-27 11:45:11 +0200636 - tune.h2.initial-window-size
Willy Tarreau5242ef82017-07-27 11:47:28 +0200637 - tune.h2.max-concurrent-streams
Willy Tarreau193b8c62012-11-22 00:17:38 +0100638 - tune.http.cookielen
Stéphane Cottin23e9e932017-05-18 08:58:41 +0200639 - tune.http.logurilen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200640 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100641 - tune.idletimer
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100642 - tune.lua.forced-yield
Willy Tarreau32f61e22015-03-18 17:54:59 +0100643 - tune.lua.maxmem
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100644 - tune.lua.session-timeout
645 - tune.lua.task-timeout
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +0200646 - tune.lua.service-timeout
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100647 - tune.maxaccept
648 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200649 - tune.maxrewrite
Willy Tarreauf3045d22015-04-29 16:24:50 +0200650 - tune.pattern.cache-size
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200651 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100652 - tune.rcvbuf.client
653 - tune.rcvbuf.server
Willy Tarreaub22fc302015-12-14 12:04:35 +0100654 - tune.recv_enough
Olivier Houchard1599b802018-05-24 18:59:04 +0200655 - tune.runqueue-depth
Willy Tarreaue803de22010-01-21 17:43:04 +0100656 - tune.sndbuf.client
657 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100658 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100659 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200660 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100661 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200662 - tune.ssl.default-dh-param
Christopher Faulet31af49d2015-06-09 17:29:50 +0200663 - tune.ssl.ssl-ctx-cache-size
Thierry FOURNIER5bf77322017-02-25 12:45:22 +0100664 - tune.ssl.capture-cipherlist-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200665 - tune.vars.global-max-size
Christopher Fauletff2613e2016-11-09 11:36:17 +0100666 - tune.vars.proc-max-size
Thierry FOURNIER4834bc72015-06-06 19:29:07 +0200667 - tune.vars.reqres-max-size
668 - tune.vars.sess-max-size
669 - tune.vars.txn-max-size
William Lallemanda509e4c2012-11-07 16:54:34 +0100670 - tune.zlib.memlevel
671 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100672
Willy Tarreau6a06a402007-07-15 20:15:28 +0200673 * Debugging
674 - debug
675 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200676
677
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006783.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200679------------------------------------
680
Emeric Brunc8e8d122012-10-02 18:42:10 +0200681ca-base <dir>
682 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200683 relative path is used with "ca-file" or "crl-file" directives. Absolute
684 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200685
Willy Tarreau6a06a402007-07-15 20:15:28 +0200686chroot <jail dir>
687 Changes current directory to <jail dir> and performs a chroot() there before
688 dropping privileges. This increases the security level in case an unknown
689 vulnerability would be exploited, since it would make it very hard for the
690 attacker to exploit the system. This only works when the process is started
691 with superuser privileges. It is important to ensure that <jail_dir> is both
Davor Ocelice9ed2812017-12-25 17:49:28 +0100692 empty and non-writable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100693
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100694cpu-map [auto:]<process-set>[/<thread-set>] <cpu-set>...
695 On Linux 2.6 and above, it is possible to bind a process or a thread to a
696 specific CPU set. This means that the process or the thread will never run on
697 other CPUs. The "cpu-map" directive specifies CPU sets for process or thread
698 sets. The first argument is a process set, eventually followed by a thread
699 set. These sets have the format
700
701 all | odd | even | number[-[number]]
702
703 <number>> must be a number between 1 and 32 or 64, depending on the machine's
Davor Ocelice9ed2812017-12-25 17:49:28 +0100704 word size. Any process IDs above nbproc and any thread IDs above nbthread are
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100705 ignored. It is possible to specify a range with two such number delimited by
706 a dash ('-'). It also is possible to specify all processes at once using
Christopher Faulet1dcb9cb2017-11-22 10:24:40 +0100707 "all", only odd numbers using "odd" or even numbers using "even", just like
708 with the "bind-process" directive. The second and forthcoming arguments are
Davor Ocelice9ed2812017-12-25 17:49:28 +0100709 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 +0100710 range with two such numbers delimited by a dash ('-'). Multiple CPU numbers
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100711 or ranges may be specified, and the processes or threads will be allowed to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100712 bind to all of them. Obviously, multiple "cpu-map" directives may be
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100713 specified. Each "cpu-map" directive will replace the previous ones when they
714 overlap. A thread will be bound on the intersection of its mapping and the
715 one of the process on which it is attached. If the intersection is null, no
716 specific binding will be set for the thread.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100717
Christopher Fauletff4121f2017-11-22 16:38:49 +0100718 Ranges can be partially defined. The higher bound can be omitted. In such
719 case, it is replaced by the corresponding maximum value, 32 or 64 depending
720 on the machine's word size.
721
Christopher Faulet26028f62017-11-22 15:01:51 +0100722 The prefix "auto:" can be added before the process set to let HAProxy
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100723 automatically bind a process or a thread to a CPU by incrementing
724 process/thread and CPU sets. To be valid, both sets must have the same
725 size. No matter the declaration order of the CPU sets, it will be bound from
726 the lowest to the highest bound. Having a process and a thread range with the
727 "auto:" prefix is not supported. Only one range is supported, the other one
728 must be a fixed number.
Christopher Faulet26028f62017-11-22 15:01:51 +0100729
730 Examples:
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100731 cpu-map 1-4 0-3 # bind processes 1 to 4 on the first 4 CPUs
732
733 cpu-map 1/all 0-3 # bind all threads of the first process on the
734 # first 4 CPUs
735
736 cpu-map 1- 0- # will be replaced by "cpu-map 1-64 0-63"
737 # or "cpu-map 1-32 0-31" depending on the machine's
738 # word size.
739
Christopher Faulet26028f62017-11-22 15:01:51 +0100740 # all these lines bind the process 1 to the cpu 0, the process 2 to cpu 1
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100741 # and so on.
Christopher Faulet26028f62017-11-22 15:01:51 +0100742 cpu-map auto:1-4 0-3
743 cpu-map auto:1-4 0-1 2-3
744 cpu-map auto:1-4 3 2 1 0
745
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100746 # all these lines bind the thread 1 to the cpu 0, the thread 2 to cpu 1
747 # and so on.
748 cpu-map auto:1/1-4 0-3
749 cpu-map auto:1/1-4 0-1 2-3
750 cpu-map auto:1/1-4 3 2 1 0
751
Davor Ocelice9ed2812017-12-25 17:49:28 +0100752 # bind each process to exactly one CPU using all/odd/even keyword
Christopher Faulet26028f62017-11-22 15:01:51 +0100753 cpu-map auto:all 0-63
754 cpu-map auto:even 0-31
755 cpu-map auto:odd 32-63
756
757 # invalid cpu-map because process and CPU sets have different sizes.
758 cpu-map auto:1-4 0 # invalid
759 cpu-map auto:1 0-3 # invalid
760
Christopher Fauletcb6a9452017-11-22 16:50:41 +0100761 # invalid cpu-map because automatic binding is used with a process range
762 # and a thread range.
763 cpu-map auto:all/all 0 # invalid
764 cpu-map auto:all/1-4 0 # invalid
765 cpu-map auto:1-4/all 0 # invalid
766
Emeric Brunc8e8d122012-10-02 18:42:10 +0200767crt-base <dir>
768 Assigns a default directory to fetch SSL certificates from when a relative
769 path is used with "crtfile" directives. Absolute locations specified after
770 "crtfile" prevail and ignore "crt-base".
771
Willy Tarreau6a06a402007-07-15 20:15:28 +0200772daemon
773 Makes the process fork into background. This is the recommended mode of
774 operation. It is equivalent to the command line "-D" argument. It can be
Lukas Tribusf46bf952017-11-21 12:39:34 +0100775 disabled by the command line "-db" argument. This option is ignored in
776 systemd mode.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200777
David Carlier8167f302015-06-01 13:50:06 +0200778deviceatlas-json-file <path>
779 Sets the path of the DeviceAtlas JSON data file to be loaded by the API.
Davor Ocelice9ed2812017-12-25 17:49:28 +0100780 The path must be a valid JSON data file and accessible by HAProxy process.
David Carlier8167f302015-06-01 13:50:06 +0200781
782deviceatlas-log-level <value>
Davor Ocelice9ed2812017-12-25 17:49:28 +0100783 Sets the level of information returned by the API. This directive is
David Carlier8167f302015-06-01 13:50:06 +0200784 optional and set to 0 by default if not set.
785
786deviceatlas-separator <char>
787 Sets the character separator for the API properties results. This directive
788 is optional and set to | by default if not set.
789
Cyril Bonté0306c4a2015-10-26 22:37:38 +0100790deviceatlas-properties-cookie <name>
Cyril Bonté307ee1e2015-09-28 23:16:06 +0200791 Sets the client cookie's name used for the detection if the DeviceAtlas
792 Client-side component was used during the request. This directive is optional
793 and set to DAPROPS by default if not set.
David Carlier29b3ca32015-09-25 14:09:21 +0100794
Simon Horman98637e52014-06-20 12:30:16 +0900795external-check
796 Allows the use of an external agent to perform health checks.
797 This is disabled by default as a security precaution.
798 See "option external-check".
799
Willy Tarreau6a06a402007-07-15 20:15:28 +0200800gid <number>
801 Changes the process' group ID to <number>. It is recommended that the group
802 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
803 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100804 Note that if haproxy is started from a user having supplementary groups, it
805 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200806 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100807
Cyril Bonté203ec5a2017-03-23 22:44:13 +0100808hard-stop-after <time>
809 Defines the maximum time allowed to perform a clean soft-stop.
810
811 Arguments :
812 <time> is the maximum time (by default in milliseconds) for which the
813 instance will remain alive when a soft-stop is received via the
814 SIGUSR1 signal.
815
816 This may be used to ensure that the instance will quit even if connections
817 remain opened during a soft-stop (for example with long timeouts for a proxy
818 in tcp mode). It applies both in TCP and HTTP mode.
819
820 Example:
821 global
822 hard-stop-after 30s
823
Willy Tarreau6a06a402007-07-15 20:15:28 +0200824group <group name>
825 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
826 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100827
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200828log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
829 <facility> [max level [min level]]
Cyril Bonté3e954872018-03-20 23:30:27 +0100830 Adds a global syslog server. Several global servers can be defined. They
Davor Ocelice9ed2812017-12-25 17:49:28 +0100831 will receive logs for starts and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100832 configured with "log global".
833
834 <address> can be one of:
835
Willy Tarreau2769aa02007-12-27 18:26:09 +0100836 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100837 no port is specified, 514 is used by default (the standard syslog
838 port).
839
David du Colombier24bb5f52011-03-17 10:40:23 +0100840 - An IPv6 address followed by a colon and optionally a UDP port. If
841 no port is specified, 514 is used by default (the standard syslog
842 port).
843
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100844 - A filesystem path to a datagram UNIX domain socket, keeping in mind
Robert Tsai81ae1952007-12-05 10:47:29 +0100845 considerations for chroot (be sure the path is accessible inside
846 the chroot) and uid/gid (be sure the path is appropriately
Davor Ocelice9ed2812017-12-25 17:49:28 +0100847 writable).
Robert Tsai81ae1952007-12-05 10:47:29 +0100848
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100849 - A file descriptor number in the form "fd@<number>", which may point
850 to a pipe, terminal, or socket. In this case unbuffered logs are used
851 and one writev() call per log is performed. This is a bit expensive
852 but acceptable for most workloads. Messages sent this way will not be
853 truncated but may be dropped, in which case the DroppedLogs counter
854 will be incremented. The writev() call is atomic even on pipes for
855 messages up to PIPE_BUF size, which POSIX recommends to be at least
856 512 and which is 4096 bytes on most modern operating systems. Any
857 larger message may be interleaved with messages from other processes.
858 Exceptionally for debugging purposes the file descriptor may also be
859 directed to a file, but doing so will significantly slow haproxy down
860 as non-blocking calls will be ignored. Also there will be no way to
861 purge nor rotate this file without restarting the process. Note that
862 the configured syslog format is preserved, so the output is suitable
Willy Tarreauc1b06452018-11-12 11:57:56 +0100863 for use with a TCP syslog server. See also the "short" and "raw"
864 format below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +0100865
866 - "stdout" / "stderr", which are respectively aliases for "fd@1" and
867 "fd@2", see above.
868
William Lallemandb2f07452015-05-12 14:27:13 +0200869 You may want to reference some environment variables in the address
870 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +0100871
Willy Tarreau18324f52014-06-27 18:10:07 +0200872 <length> is an optional maximum line length. Log lines larger than this value
873 will be truncated before being sent. The reason is that syslog
874 servers act differently on log line length. All servers support the
875 default value of 1024, but some servers simply drop larger lines
876 while others do log them. If a server supports long lines, it may
877 make sense to set this value here in order to avoid truncating long
878 lines. Similarly, if a server drops long lines, it is preferable to
879 truncate them before sending them. Accepted values are 80 to 65535
880 inclusive. The default value of 1024 is generally fine for all
881 standard usages. Some specific cases of long captures or
Davor Ocelice9ed2812017-12-25 17:49:28 +0100882 JSON-formatted logs may require larger values. You may also need to
883 increase "tune.http.logurilen" if your request URIs are truncated.
Willy Tarreau18324f52014-06-27 18:10:07 +0200884
Dragan Dosen7ad31542015-09-28 17:16:47 +0200885 <format> is the log format used when generating syslog messages. It may be
886 one of the following :
887
888 rfc3164 The RFC3164 syslog message format. This is the default.
889 (https://tools.ietf.org/html/rfc3164)
890
891 rfc5424 The RFC5424 syslog message format.
892 (https://tools.ietf.org/html/rfc5424)
893
Willy Tarreaue8746a02018-11-12 08:45:00 +0100894 short A message containing only a level between angle brackets such as
895 '<3>', followed by the text. The PID, date, time, process name
896 and system name are omitted. This is designed to be used with a
897 local log server. This format is compatible with what the systemd
898 logger consumes.
899
Willy Tarreauc1b06452018-11-12 11:57:56 +0100900 raw A message containing only the text. The level, PID, date, time,
901 process name and system name are omitted. This is designed to be
902 used in containers or during development, where the severity only
903 depends on the file descriptor used (stdout/stderr).
904
Frédéric Lécailled690dfa2019-04-25 10:52:17 +0200905 <ranges> A list of comma-separated ranges to identify the logs to sample.
906 This is used to balance the load of the logs to send to the log
907 server. The limits of the ranges cannot be null. They are numbered
908 from 1. The size or period (in number of logs) of the sample must be
909 set with <sample_size> parameter.
910
911 <sample_size>
912 The size of the sample in number of logs to consider when balancing
913 their logging loads. It is used to balance the load of the logs to
914 send to the syslog server. This size must be greater or equal to the
915 maximum of the high limits of the ranges.
916 (see also <ranges> parameter).
917
Robert Tsai81ae1952007-12-05 10:47:29 +0100918 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200919
Willy Tarreaue8746a02018-11-12 08:45:00 +0100920 kern user mail daemon auth syslog lpr news
921 uucp cron auth2 ftp ntp audit alert cron2
922 local0 local1 local2 local3 local4 local5 local6 local7
923
Willy Tarreauc1b06452018-11-12 11:57:56 +0100924 Note that the facility is ignored for the "short" and "raw"
925 formats, but still required as a positional field. It is
926 recommended to use "daemon" in this case to make it clear that
927 it's only supposed to be used locally.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200928
929 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200930 all messages are sent. If a maximum level is specified, only messages with a
931 severity at least as important as this level will be sent. An optional minimum
932 level can be specified. If it is set, logs emitted with a more severe level
933 than this one will be capped to this level. This is used to avoid sending
934 "emerg" messages on all terminals on some default syslog configurations.
935 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200936
Cyril Bontédc4d9032012-04-08 21:57:39 +0200937 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200938
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100939log-send-hostname [<string>]
940 Sets the hostname field in the syslog header. If optional "string" parameter
941 is set the header is set to the string contents, otherwise uses the hostname
942 of the system. Generally used if one is not relaying logs through an
943 intermediate syslog server or for simply customizing the hostname printed in
944 the logs.
945
Kevinm48936af2010-12-22 16:08:21 +0000946log-tag <string>
947 Sets the tag field in the syslog header to this string. It defaults to the
948 program name as launched from the command line, which usually is "haproxy".
949 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100950 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000951
Thierry FOURNIER90da1912015-03-05 11:17:06 +0100952lua-load <file>
953 This global directive loads and executes a Lua file. This directive can be
954 used multiple times.
955
William Lallemand4cfede82017-11-24 22:02:34 +0100956master-worker [no-exit-on-failure]
William Lallemande202b1e2017-06-01 17:38:56 +0200957 Master-worker mode. It is equivalent to the command line "-W" argument.
958 This mode will launch a "master" which will monitor the "workers". Using
959 this mode, you can reload HAProxy directly by sending a SIGUSR2 signal to
Davor Ocelice9ed2812017-12-25 17:49:28 +0100960 the master. The master-worker mode is compatible either with the foreground
William Lallemande202b1e2017-06-01 17:38:56 +0200961 or daemon mode. It is recommended to use this mode with multiprocess and
962 systemd.
William Lallemand4cfede82017-11-24 22:02:34 +0100963 By default, if a worker exits with a bad return code, in the case of a
964 segfault for example, all workers will be killed, and the master will leave.
965 It is convenient to combine this behavior with Restart=on-failure in a
966 systemd unit file in order to relaunch the whole process. If you don't want
967 this behavior, you must use the keyword "no-exit-on-failure".
William Lallemande202b1e2017-06-01 17:38:56 +0200968
William Lallemand4cfede82017-11-24 22:02:34 +0100969 See also "-W" in the management guide.
William Lallemande202b1e2017-06-01 17:38:56 +0200970
William Lallemand27edc4b2019-05-07 17:49:33 +0200971mworker-max-reloads <number>
972 In master-worker mode, this option limits the number of time a worker can
973 survive to a reload. If the worker did not left after a reload, once its
974 number of reloads is greater than this number, the worker will receive a
975 SIGTERM. This option helps to keep under control the number of workers.
976 See also "show proc" in the Management Guide.
977
Willy Tarreau6a06a402007-07-15 20:15:28 +0200978nbproc <number>
979 Creates <number> processes when going daemon. This requires the "daemon"
980 mode. By default, only one process is created, which is the recommended mode
981 of operation. For systems limited to small sets of file descriptors per
Willy Tarreau149ab772019-01-26 14:27:06 +0100982 process, it may be needed to fork multiple daemons. When set to a value
983 larger than 1, threads are automatically disabled. USING MULTIPLE PROCESSES
Willy Tarreau1f672a82019-01-26 14:20:55 +0100984 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon" and
985 "nbthread".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200986
Christopher Fauletbe0faa22017-08-29 15:37:10 +0200987nbthread <number>
988 This setting is only available when support for threads was built in. It
Willy Tarreau26f6ae12019-02-02 12:56:15 +0100989 makes haproxy run on <number> threads. This is exclusive with "nbproc". While
990 "nbproc" historically used to be the only way to use multiple processors, it
991 also involved a number of shortcomings related to the lack of synchronization
992 between processes (health-checks, peers, stick-tables, stats, ...) which do
993 not affect threads. As such, any modern configuration is strongly encouraged
Willy Tarreau149ab772019-01-26 14:27:06 +0100994 to migrate away from "nbproc" to "nbthread". "nbthread" also works when
995 HAProxy is started in foreground. On some platforms supporting CPU affinity,
996 when nbproc is not used, the default "nbthread" value is automatically set to
997 the number of CPUs the process is bound to upon startup. This means that the
998 thread count can easily be adjusted from the calling process using commands
999 like "taskset" or "cpuset". Otherwise, this value defaults to 1. The default
1000 value is reported in the output of "haproxy -vv". See also "nbproc".
Christopher Fauletbe0faa22017-08-29 15:37:10 +02001001
Willy Tarreau6a06a402007-07-15 20:15:28 +02001002pidfile <pidfile>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001003 Writes PIDs of all daemons into file <pidfile>. This option is equivalent to
Willy Tarreau6a06a402007-07-15 20:15:28 +02001004 the "-p" command line argument. The file must be accessible to the user
1005 starting the process. See also "daemon".
1006
Willy Tarreau1d549722016-02-16 12:41:57 +01001007presetenv <name> <value>
1008 Sets environment variable <name> to value <value>. If the variable exists, it
1009 is NOT overwritten. The changes immediately take effect so that the next line
1010 in the configuration file sees the new value. See also "setenv", "resetenv",
1011 and "unsetenv".
1012
1013resetenv [<name> ...]
1014 Removes all environment variables except the ones specified in argument. It
1015 allows to use a clean controlled environment before setting new values with
1016 setenv or unsetenv. Please note that some internal functions may make use of
1017 some environment variables, such as time manipulation functions, but also
1018 OpenSSL or even external checks. This must be used with extreme care and only
1019 after complete validation. The changes immediately take effect so that the
1020 next line in the configuration file sees the new environment. See also
1021 "setenv", "presetenv", and "unsetenv".
1022
Christopher Fauletff4121f2017-11-22 16:38:49 +01001023stats bind-process [ all | odd | even | <process_num>[-[process_num>]] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +02001024 Limits the stats socket to a certain set of processes numbers. By default the
1025 stats socket is bound to all processes, causing a warning to be emitted when
1026 nbproc is greater than 1 because there is no way to select the target process
1027 when connecting. However, by using this setting, it becomes possible to pin
1028 the stats socket to a specific set of processes, typically the first one. The
1029 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001030 the number of processes used. The maximum process ID depends on the machine's
Christopher Fauletff4121f2017-11-22 16:38:49 +01001031 word size (32 or 64). Ranges can be partially defined. The higher bound can
1032 be omitted. In such case, it is replaced by the corresponding maximum
1033 value. A better option consists in using the "process" setting of the "stats
1034 socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +02001035
Baptiste Assmann5626f482015-08-23 10:00:10 +02001036server-state-base <directory>
1037 Specifies the directory prefix to be prepended in front of all servers state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001038 file names which do not start with a '/'. See also "server-state-file",
1039 "load-server-state-from-file" and "server-state-file-name".
Baptiste Assmannef1f0fc2015-08-23 10:06:39 +02001040
1041server-state-file <file>
1042 Specifies the path to the file containing state of servers. If the path starts
1043 with a slash ('/'), it is considered absolute, otherwise it is considered
1044 relative to the directory specified using "server-state-base" (if set) or to
1045 the current directory. Before reloading HAProxy, it is possible to save the
1046 servers' current state using the stats command "show servers state". The
1047 output of this command must be written in the file pointed by <file>. When
1048 starting up, before handling traffic, HAProxy will read, load and apply state
1049 for each server found in the file and available in its current running
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02001050 configuration. See also "server-state-base" and "show servers state",
1051 "load-server-state-from-file" and "server-state-file-name"
Baptiste Assmann5626f482015-08-23 10:00:10 +02001052
Willy Tarreau1d549722016-02-16 12:41:57 +01001053setenv <name> <value>
1054 Sets environment variable <name> to value <value>. If the variable exists, it
1055 is overwritten. The changes immediately take effect so that the next line in
1056 the configuration file sees the new value. See also "presetenv", "resetenv",
1057 and "unsetenv".
1058
Willy Tarreau636848a2019-04-15 19:38:50 +02001059set-dumpable
1060 This option is better left disabled by default and enabled only upon a
1061 developer's request. It has no impact on performance nor stability but will
1062 try hard to re-enable core dumps that were possibly disabled by file size
1063 limitations (ulimit -f), core size limitations (ulimit -c), or "dumpability"
1064 of a process after changing its UID/GID (such as /proc/sys/fs/suid_dumpable
1065 on Linux). Core dumps might still be limited by the current directory's
1066 permissions (check what directory the file is started from), the chroot
1067 directory's permission (it may be needed to temporarily disable the chroot
1068 directive or to move it to a dedicated writable location), or any other
1069 system-specific constraint. For example, some Linux flavours are notorious
1070 for replacing the default core file with a path to an executable not even
1071 installed on the system (check /proc/sys/kernel/core_pattern). Often, simply
1072 writing "core", "core.%p" or "/var/log/core/core.%p" addresses the issue.
1073 When trying to enable this option waiting for a rare issue to re-appear, it's
1074 often a good idea to first try to obtain such a dump by issuing, for example,
1075 "kill -11" to the haproxy process and verify that it leaves a core where
1076 expected when dying.
1077
Willy Tarreau610f04b2014-02-13 11:36:41 +01001078ssl-default-bind-ciphers <ciphers>
1079 This setting is only available when support for OpenSSL was built in. It sets
1080 the default string describing the list of cipher algorithms ("cipher suite")
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001081 that are negotiated during the SSL/TLS handshake up to TLSv1.2 for all
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001082 "bind" lines which do not explicitly define theirs. The format of the string
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001083 is defined in "man 1 ciphers" from OpenSSL man pages. For background
1084 information and recommendations see e.g.
1085 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1086 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
1087 cipher configuration, please check the "ssl-default-bind-ciphersuites" keyword.
1088 Please check the "bind" keyword for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001089
1090ssl-default-bind-ciphersuites <ciphersuites>
1091 This setting is only available when support for OpenSSL was built in and
1092 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default string
1093 describing the list of cipher algorithms ("cipher suite") that are negotiated
1094 during the TLSv1.3 handshake for all "bind" lines which do not explicitly define
1095 theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001096 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1097 cipher configuration for TLSv1.2 and earlier, please check the
1098 "ssl-default-bind-ciphers" keyword. Please check the "bind" keyword for more
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001099 information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001100
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001101ssl-default-bind-options [<option>]...
1102 This setting is only available when support for OpenSSL was built in. It sets
1103 default ssl-options to force on all "bind" lines. Please check the "bind"
1104 keyword to see available options.
1105
1106 Example:
1107 global
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +02001108 ssl-default-bind-options ssl-min-ver TLSv1.0 no-tls-tickets
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001109
Willy Tarreau610f04b2014-02-13 11:36:41 +01001110ssl-default-server-ciphers <ciphers>
1111 This setting is only available when support for OpenSSL was built in. It
1112 sets the default string describing the list of cipher algorithms that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +00001113 negotiated during the SSL/TLS handshake up to TLSv1.2 with the server,
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001114 for all "server" lines which do not explicitly define theirs. The format of
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001115 the string is defined in "man 1 ciphers" from OpenSSL man pages. For background
1116 information and recommendations see e.g.
1117 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
1118 (https://mozilla.github.io/server-side-tls/ssl-config-generator/).
1119 For TLSv1.3 cipher configuration, please check the
1120 "ssl-default-server-ciphersuites" keyword. Please check the "server" keyword
1121 for more information.
Dirkjan Bussink415150f2018-09-14 11:14:21 +02001122
1123ssl-default-server-ciphersuites <ciphersuites>
1124 This setting is only available when support for OpenSSL was built in and
1125 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the default
1126 string describing the list of cipher algorithms that are negotiated during
1127 the TLSv1.3 handshake with the server, for all "server" lines which do not
1128 explicitly define theirs. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +00001129 "man 1 ciphers" from OpenSSL man pages under the section "ciphersuites". For
1130 cipher configuration for TLSv1.2 and earlier, please check the
1131 "ssl-default-server-ciphers" keyword. Please check the "server" keyword for
1132 more information.
Willy Tarreau610f04b2014-02-13 11:36:41 +01001133
Emeric Brun2c86cbf2014-10-30 15:56:50 +01001134ssl-default-server-options [<option>]...
1135 This setting is only available when support for OpenSSL was built in. It sets
1136 default ssl-options to force on all "server" lines. Please check the "server"
1137 keyword to see available options.
1138
Remi Gacogne47783ef2015-05-29 15:53:22 +02001139ssl-dh-param-file <file>
1140 This setting is only available when support for OpenSSL was built in. It sets
1141 the default DH parameters that are used during the SSL/TLS handshake when
1142 ephemeral Diffie-Hellman (DHE) key exchange is used, for all "bind" lines
Davor Ocelice9ed2812017-12-25 17:49:28 +01001143 which do not explicitly define theirs. It will be overridden by custom DH
Remi Gacogne47783ef2015-05-29 15:53:22 +02001144 parameters found in a bind certificate file if any. If custom DH parameters
Cyril Bonté307ee1e2015-09-28 23:16:06 +02001145 are not specified either by using ssl-dh-param-file or by setting them
1146 directly in the certificate file, pre-generated DH parameters of the size
1147 specified by tune.ssl.default-dh-param will be used. Custom parameters are
1148 known to be more secure and therefore their use is recommended.
Remi Gacogne47783ef2015-05-29 15:53:22 +02001149 Custom DH parameters may be generated by using the OpenSSL command
1150 "openssl dhparam <size>", where size should be at least 2048, as 1024-bit DH
1151 parameters should not be considered secure anymore.
1152
Emeric Brun850efd52014-01-29 12:24:34 +01001153ssl-server-verify [none|required]
1154 The default behavior for SSL verify on servers side. If specified to 'none',
1155 servers certificates are not verified. The default is 'required' except if
1156 forced using cmdline option '-dV'.
1157
Willy Tarreauabb175f2012-09-24 12:43:26 +02001158stats socket [<address:port>|<path>] [param*]
1159 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
1160 Connections to this socket will return various statistics outputs and even
1161 allow some commands to be issued to change some runtime settings. Please
Willy Tarreau1af20c72017-06-23 16:01:14 +02001162 consult section 9.3 "Unix Socket commands" of Management Guide for more
Kevin Decherf949c7202015-10-13 23:26:44 +02001163 details.
Willy Tarreau6162db22009-10-10 17:13:00 +02001164
Willy Tarreauabb175f2012-09-24 12:43:26 +02001165 All parameters supported by "bind" lines are supported, for instance to
1166 restrict access to some users or their access rights. Please consult
1167 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001168
1169stats timeout <timeout, in milliseconds>
1170 The default timeout on the stats socket is set to 10 seconds. It is possible
1171 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +01001172 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +02001173
1174stats maxconn <connections>
1175 By default, the stats socket is limited to 10 concurrent connections. It is
1176 possible to change this value with "stats maxconn".
1177
Willy Tarreau6a06a402007-07-15 20:15:28 +02001178uid <number>
1179 Changes the process' user ID to <number>. It is recommended that the user ID
1180 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
1181 be started with superuser privileges in order to be able to switch to another
1182 one. See also "gid" and "user".
1183
1184ulimit-n <number>
1185 Sets the maximum number of per-process file-descriptors to <number>. By
1186 default, it is automatically computed, so it is recommended not to use this
1187 option.
1188
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001189unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
1190 [ group <group> ] [ gid <gid> ]
1191
1192 Fixes common settings to UNIX listening sockets declared in "bind" statements.
1193 This is mainly used to simplify declaration of those UNIX sockets and reduce
1194 the risk of errors, since those settings are most commonly required but are
1195 also process-specific. The <prefix> setting can be used to force all socket
1196 path to be relative to that directory. This might be needed to access another
1197 component's chroot. Note that those paths are resolved before haproxy chroots
1198 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
1199 all have the same meaning as their homonyms used by the "bind" statement. If
1200 both are specified, the "bind" statement has priority, meaning that the
1201 "unix-bind" settings may be seen as process-wide default settings.
1202
Willy Tarreau1d549722016-02-16 12:41:57 +01001203unsetenv [<name> ...]
1204 Removes environment variables specified in arguments. This can be useful to
1205 hide some sensitive information that are occasionally inherited from the
1206 user's environment during some operations. Variables which did not exist are
1207 silently ignored so that after the operation, it is certain that none of
1208 these variables remain. The changes immediately take effect so that the next
1209 line in the configuration file will not see these variables. See also
1210 "setenv", "presetenv", and "resetenv".
1211
Willy Tarreau6a06a402007-07-15 20:15:28 +02001212user <user name>
1213 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
1214 See also "uid" and "group".
1215
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02001216node <name>
1217 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
1218
1219 This statement is useful in HA configurations where two or more processes or
1220 servers share the same IP address. By setting a different node-name on all
1221 nodes, it becomes easy to immediately spot what server is handling the
1222 traffic.
1223
1224description <text>
1225 Add a text that describes the instance.
1226
1227 Please note that it is required to escape certain characters (# for example)
1228 and this text is inserted into a html page so you should avoid using
1229 "<" and ">" characters.
1230
Thomas Holmesdb04f192015-05-18 13:21:39 +0100123151degrees-data-file <file path>
1232 The path of the 51Degrees data file to provide device detection services. The
Davor Ocelice9ed2812017-12-25 17:49:28 +01001233 file should be unzipped and accessible by HAProxy with relevant permissions.
Thomas Holmesdb04f192015-05-18 13:21:39 +01001234
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001235 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001236 compiled with USE_51DEGREES.
1237
Ben Shillitof25e8e52016-12-02 14:25:37 +0000123851degrees-property-name-list [<string> ...]
Thomas Holmesdb04f192015-05-18 13:21:39 +01001239 A list of 51Degrees property names to be load from the dataset. A full list
1240 of names is available on the 51Degrees website:
1241 https://51degrees.com/resources/property-dictionary
1242
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001243 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001244 compiled with USE_51DEGREES.
1245
Dragan Dosen93b38d92015-06-29 16:43:25 +0200124651degrees-property-separator <char>
Thomas Holmesdb04f192015-05-18 13:21:39 +01001247 A char that will be appended to every property value in a response header
1248 containing 51Degrees results. If not set that will be set as ','.
1249
Dragan Dosenae6d39a2015-06-29 16:43:27 +02001250 Please note that this option is only available when haproxy has been
1251 compiled with USE_51DEGREES.
1252
125351degrees-cache-size <number>
1254 Sets the size of the 51Degrees converter cache to <number> entries. This
1255 is an LRU cache which reminds previous device detections and their results.
1256 By default, this cache is disabled.
1257
1258 Please note that this option is only available when haproxy has been
Thomas Holmesdb04f192015-05-18 13:21:39 +01001259 compiled with USE_51DEGREES.
1260
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001261wurfl-data-file <file path>
1262 The path of the WURFL data file to provide device detection services. The
1263 file should be accessible by HAProxy with relevant permissions.
1264
1265 Please note that this option is only available when haproxy has been compiled
1266 with USE_WURFL=1.
1267
1268wurfl-information-list [<capability>]*
1269 A space-delimited list of WURFL capabilities, virtual capabilities, property
1270 names we plan to use in injected headers. A full list of capability and
1271 virtual capability names is available on the Scientiamobile website :
1272
1273 https://www.scientiamobile.com/wurflCapability
1274
1275 Valid WURFL properties are:
1276 - wurfl_id Contains the device ID of the matched device.
1277
1278 - wurfl_root_id Contains the device root ID of the matched
1279 device.
1280
1281 - wurfl_isdevroot Tells if the matched device is a root device.
1282 Possible values are "TRUE" or "FALSE".
1283
1284 - wurfl_useragent The original useragent coming with this
1285 particular web request.
1286
1287 - wurfl_api_version Contains a string representing the currently
1288 used Libwurfl API version.
1289
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001290 - wurfl_info A string containing information on the parsed
1291 wurfl.xml and its full path.
1292
1293 - wurfl_last_load_time Contains the UNIX timestamp of the last time
1294 WURFL has been loaded successfully.
1295
1296 - wurfl_normalized_useragent The normalized useragent.
1297
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001298 Please note that this option is only available when haproxy has been compiled
1299 with USE_WURFL=1.
1300
1301wurfl-information-list-separator <char>
1302 A char that will be used to separate values in a response header containing
1303 WURFL results. If not set that a comma (',') will be used by default.
1304
1305 Please note that this option is only available when haproxy has been compiled
1306 with USE_WURFL=1.
1307
1308wurfl-patch-file [<file path>]
1309 A list of WURFL patch file paths. Note that patches are loaded during startup
1310 thus before the chroot.
1311
1312 Please note that this option is only available when haproxy has been compiled
1313 with USE_WURFL=1.
1314
paulborilebad132c2019-04-18 11:57:04 +02001315wurfl-cache-size <size>
1316 Sets the WURFL Useragent cache size. For faster lookups, already processed user
1317 agents are kept in a LRU cache :
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001318 - "0" : no cache is used.
paulborilebad132c2019-04-18 11:57:04 +02001319 - <size> : size of lru cache in elements.
Willy Tarreaub3cc9f22019-04-19 16:03:32 +02001320
1321 Please note that this option is only available when haproxy has been compiled
1322 with USE_WURFL=1.
1323
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013243.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +02001325-----------------------
1326
Willy Tarreaubeb859a2018-11-22 18:07:59 +01001327busy-polling
1328 In some situations, especially when dealing with low latency on processors
1329 supporting a variable frequency or when running inside virtual machines, each
1330 time the process waits for an I/O using the poller, the processor goes back
1331 to sleep or is offered to another VM for a long time, and it causes
1332 excessively high latencies. This option provides a solution preventing the
1333 processor from sleeping by always using a null timeout on the pollers. This
1334 results in a significant latency reduction (30 to 100 microseconds observed)
1335 at the expense of a risk to overheat the processor. It may even be used with
1336 threads, in which case improperly bound threads may heavily conflict,
1337 resulting in a worse performance and high values for the CPU stolen fields
1338 in "show info" output, indicating which threads are misconfigured. It is
1339 important not to let the process run on the same processor as the network
1340 interrupts when this option is used. It is also better to avoid using it on
1341 multiple CPU threads sharing the same core. This option is disabled by
1342 default. If it has been enabled, it may still be forcibly disabled by
1343 prefixing it with the "no" keyword. It is ignored by the "select" and
1344 "poll" pollers.
1345
Willy Tarreau1746eec2014-04-25 10:46:47 +02001346max-spread-checks <delay in milliseconds>
1347 By default, haproxy tries to spread the start of health checks across the
1348 smallest health check interval of all the servers in a farm. The principle is
1349 to avoid hammering services running on the same server. But when using large
1350 check intervals (10 seconds or more), the last servers in the farm take some
1351 time before starting to be tested, which can be a problem. This parameter is
1352 used to enforce an upper bound on delay between the first and the last check,
1353 even if the servers' check intervals are larger. When servers run with
1354 shorter intervals, their intervals will be respected though.
1355
Willy Tarreau6a06a402007-07-15 20:15:28 +02001356maxconn <number>
1357 Sets the maximum per-process number of concurrent connections to <number>. It
1358 is equivalent to the command-line argument "-n". Proxies will stop accepting
1359 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +02001360 automatically adjusted according to this value. See also "ulimit-n". Note:
1361 the "select" poller cannot reliably use more than 1024 file descriptors on
1362 some platforms. If your platform only supports select and reports "select
1363 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaub28f3442019-03-04 08:13:43 +01001364 below 500 in general). If this value is not set, it will automatically be
1365 calculated based on the current file descriptors limit reported by the
1366 "ulimit -n" command, possibly reduced to a lower value if a memory limit
1367 is enforced, based on the buffer size, memory allocated to compression, SSL
1368 cache size, and use or not of SSL and the associated maxsslconn (which can
1369 also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +02001370
Willy Tarreau81c25d02011-09-07 15:17:21 +02001371maxconnrate <number>
1372 Sets the maximum per-process number of connections per second to <number>.
1373 Proxies will stop accepting connections when this limit is reached. It can be
1374 used to limit the global capacity regardless of each frontend capacity. It is
1375 important to note that this can only be used as a service protection measure,
1376 as there will not necessarily be a fair share between frontends when the
1377 limit is reached, so it's a good idea to also limit each frontend to some
1378 value close to its expected share. Also, lowering tune.maxaccept can improve
1379 fairness.
1380
William Lallemandd85f9172012-11-09 17:05:39 +01001381maxcomprate <number>
1382 Sets the maximum per-process input compression rate to <number> kilobytes
Davor Ocelice9ed2812017-12-25 17:49:28 +01001383 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +01001384 level will be decreased during the session. If the maximum is reached at the
1385 beginning of a session, the session will not compress at all. If the maximum
1386 is not reached, the compression level will be increased up to
Davor Ocelice9ed2812017-12-25 17:49:28 +01001387 tune.comp.maxlevel. A value of zero means there is no limit, this is the
William Lallemandd85f9172012-11-09 17:05:39 +01001388 default value.
1389
William Lallemand072a2bf2012-11-20 17:01:01 +01001390maxcompcpuusage <number>
1391 Sets the maximum CPU usage HAProxy can reach before stopping the compression
1392 for new requests or decreasing the compression level of current requests.
1393 It works like 'maxcomprate' but measures CPU usage instead of incoming data
1394 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
1395 case of multiple processes (nbproc > 1), each process manages its individual
1396 usage. A value of 100 disable the limit. The default value is 100. Setting
1397 a lower value will prevent the compression work from slowing the whole
1398 process down and from introducing high latencies.
1399
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001400maxpipes <number>
1401 Sets the maximum per-process number of pipes to <number>. Currently, pipes
1402 are only used by kernel-based tcp splicing. Since a pipe contains two file
1403 descriptors, the "ulimit-n" value will be increased accordingly. The default
1404 value is maxconn/4, which seems to be more than enough for most heavy usages.
1405 The splice code dynamically allocates and releases pipes, and can fall back
1406 to standard copy, so setting this value too low may only impact performance.
1407
Willy Tarreau93e7c002013-10-07 18:51:07 +02001408maxsessrate <number>
1409 Sets the maximum per-process number of sessions per second to <number>.
1410 Proxies will stop accepting connections when this limit is reached. It can be
1411 used to limit the global capacity regardless of each frontend capacity. It is
1412 important to note that this can only be used as a service protection measure,
1413 as there will not necessarily be a fair share between frontends when the
1414 limit is reached, so it's a good idea to also limit each frontend to some
1415 value close to its expected share. Also, lowering tune.maxaccept can improve
1416 fairness.
1417
Willy Tarreau403edff2012-09-06 11:58:37 +02001418maxsslconn <number>
1419 Sets the maximum per-process number of concurrent SSL connections to
1420 <number>. By default there is no SSL-specific limit, which means that the
1421 global maxconn setting will apply to all connections. Setting this limit
1422 avoids having openssl use too much memory and crash when malloc returns NULL
1423 (since it unfortunately does not reliably check for such conditions). Note
1424 that the limit applies both to incoming and outgoing connections, so one
1425 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +01001426 If this value is not set, but a memory limit is enforced, this value will be
1427 automatically computed based on the memory limit, maxconn, the buffer size,
1428 memory allocated to compression, SSL cache size, and use of SSL in either
1429 frontends, backends or both. If neither maxconn nor maxsslconn are specified
1430 when there is a memory limit, haproxy will automatically adjust these values
1431 so that 100% of the connections can be made over SSL with no risk, and will
1432 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +02001433
Willy Tarreaue43d5322013-10-07 20:01:52 +02001434maxsslrate <number>
1435 Sets the maximum per-process number of SSL sessions per second to <number>.
1436 SSL listeners will stop accepting connections when this limit is reached. It
1437 can be used to limit the global SSL CPU usage regardless of each frontend
1438 capacity. It is important to note that this can only be used as a service
1439 protection measure, as there will not necessarily be a fair share between
1440 frontends when the limit is reached, so it's a good idea to also limit each
1441 frontend to some value close to its expected share. It is also important to
1442 note that the sessions are accounted before they enter the SSL stack and not
1443 after, which also protects the stack against bad handshakes. Also, lowering
1444 tune.maxaccept can improve fairness.
1445
William Lallemand9d5f5482012-11-07 16:12:57 +01001446maxzlibmem <number>
1447 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
1448 When the maximum amount is reached, future sessions will not compress as long
1449 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +01001450 The default value is 0. The value is available in bytes on the UNIX socket
1451 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
1452 "ZlibMemUsage" in bytes.
1453
Willy Tarreau6a06a402007-07-15 20:15:28 +02001454noepoll
1455 Disables the use of the "epoll" event polling system on Linux. It is
1456 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +01001457 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001458
1459nokqueue
1460 Disables the use of the "kqueue" event polling system on BSD. It is
1461 equivalent to the command-line argument "-dk". The next polling system
1462 used will generally be "poll". See also "nopoll".
1463
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001464noevports
1465 Disables the use of the event ports event polling system on SunOS systems
1466 derived from Solaris 10 and later. It is equivalent to the command-line
1467 argument "-dv". The next polling system used will generally be "poll". See
1468 also "nopoll".
1469
Willy Tarreau6a06a402007-07-15 20:15:28 +02001470nopoll
1471 Disables the use of the "poll" event polling system. It is equivalent to the
1472 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001473 It should never be needed to disable "poll" since it's available on all
Emmanuel Hocdet0ba4f482019-04-08 16:53:32 +00001474 platforms supported by HAProxy. See also "nokqueue", "noepoll" and
1475 "noevports".
Willy Tarreau6a06a402007-07-15 20:15:28 +02001476
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001477nosplice
1478 Disables the use of kernel tcp splicing between sockets on Linux. It is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001479 equivalent to the command line argument "-dS". Data will then be copied
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001480 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001481 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +01001482 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
1483 be used. This option makes it easier to globally disable kernel splicing in
1484 case of doubt. See also "option splice-auto", "option splice-request" and
1485 "option splice-response".
1486
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001487nogetaddrinfo
1488 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
1489 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
1490
Lukas Tribusa0bcbdc2016-09-12 21:42:20 +00001491noreuseport
1492 Disables the use of SO_REUSEPORT - see socket(7). It is equivalent to the
1493 command line argument "-dR".
1494
Willy Tarreaud2d33482019-04-25 17:09:07 +02001495profiling.tasks { auto | on | off }
1496 Enables ('on') or disables ('off') per-task CPU profiling. When set to 'auto'
1497 the profiling automatically turns on a thread when it starts to suffer from
1498 an average latency of 1000 microseconds or higher as reported in the
1499 "avg_loop_us" activity field, and automatically turns off when the latency
1500 reutnrs below 990 microseconds (this value is an average over the last 1024
1501 loops so it does not vary quickly and tends to significantly smooth short
1502 spikes). It may also spontaneously trigger from time to time on overloaded
1503 systems, containers, or virtual machines, or when the system swaps (which
1504 must absolutely never happen on a load balancer).
1505
1506 CPU profiling per task can be very convenient to report where the time is
1507 spent and which requests have what effect on which other request. Enabling
1508 it will typically affect the overall's performance by less than 1%, thus it
1509 is recommended to leave it to the default 'auto' value so that it only
1510 operates when a problem is identified. This feature requires a system
Willy Tarreau75c62c22018-11-22 11:02:09 +01001511 supporting the clock_gettime(2) syscall with clock identifiers
1512 CLOCK_MONOTONIC and CLOCK_THREAD_CPUTIME_ID, otherwise the reported time will
1513 be zero. This option may be changed at run time using "set profiling" on the
1514 CLI.
1515
Willy Tarreaufe255b72007-10-14 23:09:26 +02001516spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +09001517 Sometimes it is desirable to avoid sending agent and health checks to
1518 servers at exact intervals, for instance when many logical servers are
1519 located on the same physical server. With the help of this parameter, it
1520 becomes possible to add some randomness in the check interval between 0
1521 and +/- 50%. A value between 2 and 5 seems to show good results. The
1522 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +02001523
Davor Ocelice9ed2812017-12-25 17:49:28 +01001524ssl-engine <name> [algo <comma-separated list of algorithms>]
Grant Zhang872f9c22017-01-21 01:10:18 +00001525 Sets the OpenSSL engine to <name>. List of valid values for <name> may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01001526 obtained using the command "openssl engine". This statement may be used
Grant Zhang872f9c22017-01-21 01:10:18 +00001527 multiple times, it will simply enable multiple crypto engines. Referencing an
1528 unsupported engine will prevent haproxy from starting. Note that many engines
1529 will lead to lower HTTPS performance than pure software with recent
1530 processors. The optional command "algo" sets the default algorithms an ENGINE
1531 will supply using the OPENSSL function ENGINE_set_default_string(). A value
Davor Ocelice9ed2812017-12-25 17:49:28 +01001532 of "ALL" uses the engine for all cryptographic operations. If no list of
1533 algo is specified then the value of "ALL" is used. A comma-separated list
Grant Zhang872f9c22017-01-21 01:10:18 +00001534 of different algorithms may be specified, including: RSA, DSA, DH, EC, RAND,
1535 CIPHERS, DIGESTS, PKEY, PKEY_CRYPTO, PKEY_ASN1. This is the same format that
1536 openssl configuration file uses:
1537 https://www.openssl.org/docs/man1.0.2/apps/config.html
1538
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001539ssl-mode-async
1540 Adds SSL_MODE_ASYNC mode to the SSL context. This enables asynchronous TLS
Emeric Brun3854e012017-05-17 20:42:48 +02001541 I/O operations if asynchronous capable SSL engines are used. The current
Emeric Brunb5e42a82017-06-06 12:35:14 +00001542 implementation supports a maximum of 32 engines. The Openssl ASYNC API
1543 doesn't support moving read/write buffers and is not compliant with
1544 haproxy's buffer management. So the asynchronous mode is disabled on
1545 read/write operations (it is only enabled during initial and reneg
1546 handshakes).
Grant Zhangfa6c7ee2017-01-14 01:42:15 +00001547
Willy Tarreau33cb0652014-12-23 22:52:37 +01001548tune.buffers.limit <number>
1549 Sets a hard limit on the number of buffers which may be allocated per process.
1550 The default value is zero which means unlimited. The minimum non-zero value
1551 will always be greater than "tune.buffers.reserve" and should ideally always
1552 be about twice as large. Forcing this value can be particularly useful to
1553 limit the amount of memory a process may take, while retaining a sane
Davor Ocelice9ed2812017-12-25 17:49:28 +01001554 behavior. When this limit is reached, sessions which need a buffer wait for
Willy Tarreau33cb0652014-12-23 22:52:37 +01001555 another one to be released by another session. Since buffers are dynamically
1556 allocated and released, the waiting time is very short and not perceptible
1557 provided that limits remain reasonable. In fact sometimes reducing the limit
1558 may even increase performance by increasing the CPU cache's efficiency. Tests
1559 have shown good results on average HTTP traffic with a limit to 1/10 of the
1560 expected global maxconn setting, which also significantly reduces memory
1561 usage. The memory savings come from the fact that a number of connections
1562 will not allocate 2*tune.bufsize. It is best not to touch this value unless
1563 advised to do so by an haproxy core developer.
1564
Willy Tarreau1058ae72014-12-23 22:40:40 +01001565tune.buffers.reserve <number>
1566 Sets the number of buffers which are pre-allocated and reserved for use only
1567 during memory shortage conditions resulting in failed memory allocations. The
1568 minimum value is 2 and is also the default. There is no reason a user would
1569 want to change this value, it's mostly aimed at haproxy core developers.
1570
Willy Tarreau27a674e2009-08-17 07:23:33 +02001571tune.bufsize <number>
1572 Sets the buffer size to this size (in bytes). Lower values allow more
1573 sessions to coexist in the same amount of RAM, and higher values allow some
1574 applications with very large cookies to work. The default value is 16384 and
1575 can be changed at build time. It is strongly recommended not to change this
1576 from the default value, as very low values will break some services such as
1577 statistics, and values larger than default size will increase memory usage,
1578 possibly causing the system to run out of memory. At least the global maxconn
Willy Tarreau45a66cc2017-11-24 11:28:00 +01001579 parameter should be decreased by the same factor as this one is increased. In
1580 addition, use of HTTP/2 mandates that this value must be 16384 or more. If an
1581 HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04001582 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
Willy Tarreauc77d3642018-12-12 06:19:42 +01001583 than this size, haproxy will return HTTP 502 (Bad Gateway). Note that the
1584 value set using this parameter will automatically be rounded up to the next
1585 multiple of 8 on 32-bit machines and 16 on 64-bit machines.
Willy Tarreau27a674e2009-08-17 07:23:33 +02001586
Willy Tarreau43961d52010-10-04 20:39:20 +02001587tune.chksize <number>
1588 Sets the check buffer size to this size (in bytes). Higher values may help
1589 find string or regex patterns in very large pages, though doing so may imply
1590 more memory and CPU usage. The default value is 16384 and can be changed at
1591 build time. It is not recommended to change this value, but to use better
1592 checks whenever possible.
1593
William Lallemandf3747832012-11-09 12:33:10 +01001594tune.comp.maxlevel <number>
1595 Sets the maximum compression level. The compression level affects CPU
1596 usage during compression. This value affects CPU usage during compression.
1597 Each session using compression initializes the compression algorithm with
1598 this value. The default value is 1.
1599
Willy Tarreauc299e1e2019-02-27 11:35:12 +01001600tune.fail-alloc
1601 If compiled with DEBUG_FAIL_ALLOC, gives the percentage of chances an
1602 allocation attempt fails. Must be between 0 (no failure) and 100 (no
1603 success). This is useful to debug and make sure memory failures are handled
1604 gracefully.
1605
Willy Tarreaufe20e5b2017-07-27 11:42:14 +02001606tune.h2.header-table-size <number>
1607 Sets the HTTP/2 dynamic header table size. It defaults to 4096 bytes and
1608 cannot be larger than 65536 bytes. A larger value may help certain clients
1609 send more compact requests, depending on their capabilities. This amount of
1610 memory is consumed for each HTTP/2 connection. It is recommended not to
1611 change it.
1612
Willy Tarreaue6baec02017-07-27 11:45:11 +02001613tune.h2.initial-window-size <number>
1614 Sets the HTTP/2 initial window size, which is the number of bytes the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01001615 can upload before waiting for an acknowledgment from haproxy. This setting
1616 only affects payload contents (i.e. the body of POST requests), not headers.
Willy Tarreaue6baec02017-07-27 11:45:11 +02001617 The default value is 65535, which roughly allows up to 5 Mbps of upload
1618 bandwidth per client over a network showing a 100 ms ping time, or 500 Mbps
1619 over a 1-ms local network. It can make sense to increase this value to allow
1620 faster uploads, or to reduce it to increase fairness when dealing with many
1621 clients. It doesn't affect resource usage.
1622
Willy Tarreau5242ef82017-07-27 11:47:28 +02001623tune.h2.max-concurrent-streams <number>
1624 Sets the HTTP/2 maximum number of concurrent streams per connection (ie the
1625 number of outstanding requests on a single connection). The default value is
1626 100. A larger one may slightly improve page load time for complex sites when
1627 visited over high latency networks, but increases the amount of resources a
1628 single client may allocate. A value of zero disables the limit so a single
1629 client may create as many streams as allocatable by haproxy. It is highly
1630 recommended not to change this value.
1631
Willy Tarreaua24b35c2019-02-21 13:24:36 +01001632tune.h2.max-frame-size <number>
1633 Sets the HTTP/2 maximum frame size that haproxy announces it is willing to
1634 receive to its peers. The default value is the largest between 16384 and the
1635 buffer size (tune.bufsize). In any case, haproxy will not announce support
1636 for frame sizes larger than buffers. The main purpose of this setting is to
1637 allow to limit the maximum frame size setting when using large buffers. Too
1638 large frame sizes might have performance impact or cause some peers to
1639 misbehave. It is highly recommended not to change this value.
1640
Willy Tarreau193b8c62012-11-22 00:17:38 +01001641tune.http.cookielen <number>
1642 Sets the maximum length of captured cookies. This is the maximum value that
1643 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
1644 will automatically be truncated to this one. It is important not to set too
1645 high a value because all cookie captures still allocate this size whatever
1646 their configured value (they share a same pool). This value is per request
1647 per response, so the memory allocated is twice this value per connection.
1648 When not specified, the limit is set to 63 characters. It is recommended not
1649 to change this value.
1650
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001651tune.http.logurilen <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001652 Sets the maximum length of request URI in logs. This prevents truncating long
1653 request URIs with valuable query strings in log lines. This is not related
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001654 to syslog limits. If you increase this limit, you may also increase the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001655 'log ... len yyy' parameter. Your syslog daemon may also need specific
Stéphane Cottin23e9e932017-05-18 08:58:41 +02001656 configuration directives too.
1657 The default value is 1024.
1658
Willy Tarreauac1932d2011-10-24 19:14:41 +02001659tune.http.maxhdr <number>
1660 Sets the maximum number of headers in a request. When a request comes with a
1661 number of headers greater than this value (including the first line), it is
1662 rejected with a "400 Bad Request" status code. Similarly, too large responses
1663 are blocked with "502 Bad Gateway". The default value is 101, which is enough
1664 for all usages, considering that the widely deployed Apache server uses the
1665 same limit. It can be useful to push this limit further to temporarily allow
Christopher Faulet50174f32017-06-21 16:31:35 +02001666 a buggy application to work by the time it gets fixed. The accepted range is
1667 1..32767. Keep in mind that each new header consumes 32bits of memory for
1668 each session, so don't push this limit too high.
Willy Tarreauac1932d2011-10-24 19:14:41 +02001669
Willy Tarreau7e312732014-02-12 16:35:14 +01001670tune.idletimer <timeout>
1671 Sets the duration after which haproxy will consider that an empty buffer is
1672 probably associated with an idle stream. This is used to optimally adjust
1673 some packet sizes while forwarding large and small data alternatively. The
1674 decision to use splice() or to send large buffers in SSL is modulated by this
1675 parameter. The value is in milliseconds between 0 and 65535. A value of zero
1676 means that haproxy will not try to detect idle streams. The default is 1000,
Davor Ocelice9ed2812017-12-25 17:49:28 +01001677 which seems to correctly detect end user pauses (e.g. read a page before
Willy Tarreau7e312732014-02-12 16:35:14 +01001678 clicking). There should be not reason for changing this value. Please check
1679 tune.ssl.maxrecord below.
1680
Willy Tarreau7ac908b2019-02-27 12:02:18 +01001681tune.listener.multi-queue { on | off }
1682 Enables ('on') or disables ('off') the listener's multi-queue accept which
1683 spreads the incoming traffic to all threads a "bind" line is allowed to run
1684 on instead of taking them for itself. This provides a smoother traffic
1685 distribution and scales much better, especially in environments where threads
1686 may be unevenly loaded due to external activity (network interrupts colliding
1687 with one thread for example). This option is enabled by default, but it may
1688 be forcefully disabled for troubleshooting or for situations where it is
1689 estimated that the operating system already provides a good enough
1690 distribution and connections are extremely short-lived.
1691
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001692tune.lua.forced-yield <number>
1693 This directive forces the Lua engine to execute a yield each <number> of
Tim Düsterhus4896c442016-11-29 02:15:19 +01001694 instructions executed. This permits interrupting a long script and allows the
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001695 HAProxy scheduler to process other tasks like accepting connections or
1696 forwarding traffic. The default value is 10000 instructions. If HAProxy often
Davor Ocelice9ed2812017-12-25 17:49:28 +01001697 executes some Lua code but more responsiveness is required, this value can be
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001698 lowered. If the Lua code is quite long and its result is absolutely required
1699 to process the data, the <number> can be increased.
1700
Willy Tarreau32f61e22015-03-18 17:54:59 +01001701tune.lua.maxmem
1702 Sets the maximum amount of RAM in megabytes per process usable by Lua. By
1703 default it is zero which means unlimited. It is important to set a limit to
1704 ensure that a bug in a script will not result in the system running out of
1705 memory.
1706
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001707tune.lua.session-timeout <timeout>
1708 This is the execution timeout for the Lua sessions. This is useful for
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001709 preventing infinite loops or spending too much time in Lua. This timeout
1710 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001711 not taken in account. The default timeout is 4s.
Thierry FOURNIER90da1912015-03-05 11:17:06 +01001712
1713tune.lua.task-timeout <timeout>
1714 Purpose is the same as "tune.lua.session-timeout", but this timeout is
1715 dedicated to the tasks. By default, this timeout isn't set because a task may
1716 remain alive during of the lifetime of HAProxy. For example, a task used to
1717 check servers.
1718
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001719tune.lua.service-timeout <timeout>
1720 This is the execution timeout for the Lua services. This is useful for
1721 preventing infinite loops or spending too much time in Lua. This timeout
1722 counts only the pure Lua runtime. If the Lua does a sleep, the sleep is
Davor Ocelice9ed2812017-12-25 17:49:28 +01001723 not taken in account. The default timeout is 4s.
Thierry FOURNIER7dd784b2015-10-01 14:49:33 +02001724
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001725tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +01001726 Sets the maximum number of consecutive connections a process may accept in a
1727 row before switching to other work. In single process mode, higher numbers
1728 give better performance at high connection rates. However in multi-process
1729 modes, keeping a bit of fairness between processes generally is better to
1730 increase performance. This value applies individually to each listener, so
1731 that the number of processes a listener is bound to is taken into account.
1732 This value defaults to 64. In multi-process mode, it is divided by twice
1733 the number of processes the listener is bound to. Setting this value to -1
1734 completely disables the limitation. It should normally not be needed to tweak
1735 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +01001736
1737tune.maxpollevents <number>
1738 Sets the maximum amount of events that can be processed at once in a call to
1739 the polling system. The default value is adapted to the operating system. It
1740 has been noticed that reducing it below 200 tends to slightly decrease
1741 latency at the expense of network bandwidth, and increasing it above 200
1742 tends to trade latency for slightly increased bandwidth.
1743
Willy Tarreau27a674e2009-08-17 07:23:33 +02001744tune.maxrewrite <number>
1745 Sets the reserved buffer space to this size in bytes. The reserved space is
1746 used for header rewriting or appending. The first reads on sockets will never
1747 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1748 bufsize, though that does not make much sense since there are rarely large
1749 numbers of headers to add. Setting it too high prevents processing of large
1750 requests or responses. Setting it too low prevents addition of new headers
1751 to already large requests or to POST requests. It is generally wise to set it
1752 to about 1024. It is automatically readjusted to half of bufsize if it is
1753 larger than that. This means you don't have to worry about it when changing
1754 bufsize.
1755
Willy Tarreauf3045d22015-04-29 16:24:50 +02001756tune.pattern.cache-size <number>
1757 Sets the size of the pattern lookup cache to <number> entries. This is an LRU
1758 cache which reminds previous lookups and their results. It is used by ACLs
1759 and maps on slow pattern lookups, namely the ones using the "sub", "reg",
1760 "dir", "dom", "end", "bin" match methods as well as the case-insensitive
1761 strings. It applies to pattern expressions which means that it will be able
1762 to memorize the result of a lookup among all the patterns specified on a
1763 configuration line (including all those loaded from files). It automatically
1764 invalidates entries which are updated using HTTP actions or on the CLI. The
1765 default cache size is set to 10000 entries, which limits its footprint to
1766 about 5 MB on 32-bit systems and 8 MB on 64-bit systems. There is a very low
1767 risk of collision in this cache, which is in the order of the size of the
1768 cache divided by 2^64. Typically, at 10000 requests per second with the
1769 default cache size of 10000 entries, there's 1% chance that a brute force
1770 attack could cause a single collision after 60 years, or 0.1% after 6 years.
1771 This is considered much lower than the risk of a memory corruption caused by
1772 aging components. If this is not acceptable, the cache can be disabled by
1773 setting this parameter to 0.
1774
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001775tune.pipesize <number>
1776 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1777 are the default size for the system. But sometimes when using TCP splicing,
1778 it can improve performance to increase pipe sizes, especially if it is
1779 suspected that pipes are not filled and that many calls to splice() are
1780 performed. This has an impact on the kernel's memory footprint, so this must
1781 not be changed if impacts are not understood.
1782
Olivier Houchard88698d92019-04-16 19:07:22 +02001783tune.pool-low-fd-ratio <number>
1784 This setting sets the max number of file descriptors (in percentage) used by
1785 haproxy globally against the maximum number of file descriptors haproxy can
1786 use before we stop putting connection into the idle pool for reuse. The
1787 default is 20.
1788
1789tune.pool-high-fd-ratio <number>
1790 This setting sets the max number of file descriptors (in percentage) used by
1791 haproxy globally against the maximum number of file descriptors haproxy can
1792 use before we start killing idle connections when we can't reuse a connection
1793 and we have to create a new one. The default is 25 (one quarter of the file
1794 descriptor will mean that roughly half of the maximum front connections can
1795 keep an idle connection behind, anything beyond this probably doesn't make
1796 much sense in the general case when targetting connection reuse).
1797
Willy Tarreaue803de22010-01-21 17:43:04 +01001798tune.rcvbuf.client <number>
1799tune.rcvbuf.server <number>
1800 Forces the kernel socket receive buffer size on the client or the server side
1801 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1802 and backends. It should normally never be set, and the default size (0) lets
1803 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001804 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001805 order to save kernel memory by preventing it from buffering too large amounts
1806 of received data. Lower values will significantly increase CPU usage though.
1807
Willy Tarreaub22fc302015-12-14 12:04:35 +01001808tune.recv_enough <number>
Davor Ocelice9ed2812017-12-25 17:49:28 +01001809 HAProxy uses some hints to detect that a short read indicates the end of the
Willy Tarreaub22fc302015-12-14 12:04:35 +01001810 socket buffers. One of them is that a read returns more than <recv_enough>
1811 bytes, which defaults to 10136 (7 segments of 1448 each). This default value
1812 may be changed by this setting to better deal with workloads involving lots
1813 of short messages such as telnet or SSH sessions.
1814
Olivier Houchard1599b802018-05-24 18:59:04 +02001815tune.runqueue-depth <number>
1816 Sets the maxinum amount of task that can be processed at once when running
1817 tasks. The default value is 200. Increasing it may incur latency when
1818 dealing with I/Os, making it too small can incur extra overhead.
1819
Willy Tarreaue803de22010-01-21 17:43:04 +01001820tune.sndbuf.client <number>
1821tune.sndbuf.server <number>
1822 Forces the kernel socket send buffer size on the client or the server side to
1823 the specified value in bytes. This value applies to all TCP/HTTP frontends
1824 and backends. It should normally never be set, and the default size (0) lets
1825 the kernel autotune this value depending on the amount of available memory.
Davor Ocelice9ed2812017-12-25 17:49:28 +01001826 However it can sometimes help to set it to very low values (e.g. 4096) in
Willy Tarreaue803de22010-01-21 17:43:04 +01001827 order to save kernel memory by preventing it from buffering too large amounts
1828 of received data. Lower values will significantly increase CPU usage though.
1829 Another use case is to prevent write timeouts with extremely slow clients due
1830 to the kernel waiting for a large part of the buffer to be read before
1831 notifying haproxy again.
1832
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001833tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001834 Sets the size of the global SSL session cache, in a number of blocks. A block
1835 is large enough to contain an encoded session without peer certificate.
1836 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001837 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001838 200 bytes of memory. The default value may be forced at build time, otherwise
Davor Ocelice9ed2812017-12-25 17:49:28 +01001839 defaults to 20000. When the cache is full, the most idle entries are purged
Emeric Brunaf9619d2012-11-28 18:47:52 +01001840 and reassigned. Higher values reduce the occurrence of such a purge, hence
1841 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1842 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001843 and are shared between all processes if "nbproc" is greater than 1. Setting
1844 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001845
Emeric Brun8dc60392014-05-09 13:52:00 +02001846tune.ssl.force-private-cache
Lukas Tribus27935782018-10-01 02:00:16 +02001847 This option disables SSL session cache sharing between all processes. It
Emeric Brun8dc60392014-05-09 13:52:00 +02001848 should normally not be used since it will force many renegotiations due to
1849 clients hitting a random process. But it may be required on some operating
1850 systems where none of the SSL cache synchronization method may be used. In
1851 this case, adding a first layer of hash-based load balancing before the SSL
1852 layer might limit the impact of the lack of session sharing.
1853
Emeric Brun4f65bff2012-11-16 15:11:00 +01001854tune.ssl.lifetime <timeout>
1855 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001856 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001857 does not guarantee that sessions will last that long, because if the cache is
1858 full, the longest idle sessions will be purged despite their configured
1859 lifetime. The real usefulness of this setting is to prevent sessions from
1860 being used for too long.
1861
Willy Tarreaubfd59462013-02-21 07:46:09 +01001862tune.ssl.maxrecord <number>
1863 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1864 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1865 data only once it has received a full record. With large records, it means
1866 that clients might have to download up to 16kB of data before starting to
1867 process them. Limiting the value can improve page load times on browsers
1868 located over high latency or low bandwidth networks. It is suggested to find
1869 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1870 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1871 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1872 2859 gave good results during tests. Use "strace -e trace=write" to find the
Davor Ocelice9ed2812017-12-25 17:49:28 +01001873 best value. HAProxy will automatically switch to this setting after an idle
Willy Tarreau7e312732014-02-12 16:35:14 +01001874 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001875
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001876tune.ssl.default-dh-param <number>
1877 Sets the maximum size of the Diffie-Hellman parameters used for generating
1878 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1879 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1880 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1881 this maximum value. Default value if 1024. Only 1024 or higher values are
1882 allowed. Higher values will increase the CPU load, and values greater than
1883 1024 bits are not supported by Java 7 and earlier clients. This value is not
Remi Gacogne47783ef2015-05-29 15:53:22 +02001884 used if static Diffie-Hellman parameters are supplied either directly
1885 in the certificate file or by using the ssl-dh-param-file parameter.
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001886
Christopher Faulet31af49d2015-06-09 17:29:50 +02001887tune.ssl.ssl-ctx-cache-size <number>
1888 Sets the size of the cache used to store generated certificates to <number>
1889 entries. This is a LRU cache. Because generating a SSL certificate
1890 dynamically is expensive, they are cached. The default cache size is set to
1891 1000 entries.
1892
Thierry FOURNIER5bf77322017-02-25 12:45:22 +01001893tune.ssl.capture-cipherlist-size <number>
1894 Sets the maximum size of the buffer used for capturing client-hello cipher
1895 list. If the value is 0 (default value) the capture is disabled, otherwise
1896 a buffer is allocated for each SSL/TLS connection.
1897
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001898tune.vars.global-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001899tune.vars.proc-max-size <size>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001900tune.vars.reqres-max-size <size>
1901tune.vars.sess-max-size <size>
1902tune.vars.txn-max-size <size>
Christopher Fauletff2613e2016-11-09 11:36:17 +01001903 These five tunes help to manage the maximum amount of memory used by the
1904 variables system. "global" limits the overall amount of memory available for
1905 all scopes. "proc" limits the memory for the process scope, "sess" limits the
1906 memory for the session scope, "txn" for the transaction scope, and "reqres"
1907 limits the memory for each request or response processing.
1908 Memory accounting is hierarchical, meaning more coarse grained limits include
1909 the finer grained ones: "proc" includes "sess", "sess" includes "txn", and
1910 "txn" includes "reqres".
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001911
Daniel Schneller0b547052016-03-21 20:46:57 +01001912 For example, when "tune.vars.sess-max-size" is limited to 100,
1913 "tune.vars.txn-max-size" and "tune.vars.reqres-max-size" cannot exceed
1914 100 either. If we create a variable "txn.var" that contains 100 bytes,
1915 all available space is consumed.
1916 Notice that exceeding the limits at runtime will not result in an error
1917 message, but values might be cut off or corrupted. So make sure to accurately
1918 plan for the amount of space needed to store all your variables.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02001919
William Lallemanda509e4c2012-11-07 16:54:34 +01001920tune.zlib.memlevel <number>
1921 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001922 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001923 state. A value of 1 uses minimum memory but is slow and reduces compression
Davor Ocelice9ed2812017-12-25 17:49:28 +01001924 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
William Lallemanda509e4c2012-11-07 16:54:34 +01001925 between 1 and 9. The default value is 8.
1926
1927tune.zlib.windowsize <number>
1928 Sets the window size (the size of the history buffer) as a parameter of the
1929 zlib initialization for each session. Larger values of this parameter result
Davor Ocelice9ed2812017-12-25 17:49:28 +01001930 in better compression at the expense of memory usage. Can be a value between
1931 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001932
Willy Tarreauc57f0e22009-05-10 13:12:33 +020019333.3. Debugging
1934--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001935
1936debug
1937 Enables debug mode which dumps to stdout all exchanges, and disables forking
1938 into background. It is the equivalent of the command-line argument "-d". It
1939 should never be used in a production configuration since it may prevent full
1940 system startup.
1941
1942quiet
1943 Do not display any message during startup. It is equivalent to the command-
1944 line argument "-q".
1945
Emeric Brunf099e792010-09-27 12:05:28 +02001946
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010019473.4. Userlists
1948--------------
1949It is possible to control access to frontend/backend/listen sections or to
1950http stats by allowing only authenticated and authorized users. To do this,
1951it is required to create at least one userlist and to define users.
1952
1953userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001954 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001955 used to store authentication & authorization data for independent customers.
1956
1957group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001958 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001959 attach users to this group by using a comma separated list of names
1960 proceeded by "users" keyword.
1961
Cyril Bontéf0c60612010-02-06 14:44:47 +01001962user <username> [password|insecure-password <password>]
1963 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001964 Adds user <username> to the current userlist. Both secure (encrypted) and
1965 insecure (unencrypted) passwords can be used. Encrypted passwords are
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001966 evaluated using the crypt(3) function, so depending on the system's
1967 capabilities, different algorithms are supported. For example, modern Glibc
1968 based Linux systems support MD5, SHA-256, SHA-512, and, of course, the
1969 classic DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001970
Daniel Schnellerd06f31c2017-11-06 16:51:04 +01001971 Attention: Be aware that using encrypted passwords might cause significantly
1972 increased CPU usage, depending on the number of requests, and the algorithm
1973 used. For any of the hashed variants, the password for each request must
1974 be processed through the chosen algorithm, before it can be compared to the
1975 value specified in the config file. Most current algorithms are deliberately
1976 designed to be expensive to compute to achieve resistance against brute
1977 force attacks. They do not simply salt/hash the clear text password once,
1978 but thousands of times. This can quickly become a major factor in haproxy's
1979 overall CPU consumption!
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001980
1981 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001982 userlist L1
1983 group G1 users tiger,scott
1984 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001985
Cyril Bontéf0c60612010-02-06 14:44:47 +01001986 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1987 user scott insecure-password elgato
1988 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001989
Cyril Bontéf0c60612010-02-06 14:44:47 +01001990 userlist L2
1991 group G1
1992 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001993
Cyril Bontéf0c60612010-02-06 14:44:47 +01001994 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1995 user scott insecure-password elgato groups G1,G2
1996 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001997
1998 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001999
Emeric Brunf099e792010-09-27 12:05:28 +02002000
20013.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02002002----------
Emeric Brun94900952015-06-11 18:25:54 +02002003It is possible to propagate entries of any data-types in stick-tables between
2004several haproxy instances over TCP connections in a multi-master fashion. Each
2005instance pushes its local updates and insertions to remote peers. The pushed
2006values overwrite remote ones without aggregation. Interrupted exchanges are
2007automatically detected and recovered from the last known point.
2008In addition, during a soft restart, the old process connects to the new one
2009using such a TCP connection to push all its entries before the new process
2010tries to connect to other peers. That ensures very fast replication during a
2011reload, it typically takes a fraction of a second even for large tables.
2012Note that Server IDs are used to identify servers remotely, so it is important
2013that configurations look similar or at least that the same IDs are forced on
2014each server on all participants.
Emeric Brunf099e792010-09-27 12:05:28 +02002015
2016peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002017 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02002018 which is referenced by one or more stick-tables.
2019
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002020bind [<address>]:<port_range> [, ...] [param*]
2021 Defines the binding parameters of the local peer of this "peers" section.
2022 Such lines are not supported with "peer" line in the same "peers" section.
2023
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002024disabled
2025 Disables a peers section. It disables both listening and any synchronization
2026 related to this section. This is provided to disable synchronization of stick
2027 tables without having to comment out all "peers" references.
2028
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002029default-bind [param*]
2030 Defines the binding parameters for the local peer, excepted its address.
2031
2032default-server [param*]
2033 Change default options for a server in a "peers" section.
2034
2035 Arguments:
2036 <param*> is a list of parameters for this server. The "default-server"
2037 keyword accepts an important number of options and has a complete
2038 section dedicated to it. Please refer to section 5 for more
2039 details.
2040
2041
2042 See also: "server" and section 5 about server options
2043
Willy Tarreau77e4bd12015-05-01 20:02:17 +02002044enable
2045 This re-enables a disabled peers section which was previously disabled.
2046
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002047peer <peername> <ip>:<port> [param*]
Emeric Brunf099e792010-09-27 12:05:28 +02002048 Defines a peer inside a peers section.
2049 If <peername> is set to the local peer name (by default hostname, or forced
2050 using "-L" command line option), haproxy will listen for incoming remote peer
2051 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
2052 to join the remote peer, and <peername> is used at the protocol level to
2053 identify and validate the remote peer on the server side.
2054
2055 During a soft restart, local peer <ip>:<port> is used by the old instance to
2056 connect the new one and initiate a complete replication (teaching process).
2057
2058 It is strongly recommended to have the exact same peers declaration on all
2059 peers and to only rely on the "-L" command line argument to change the local
2060 peer name. This makes it easier to maintain coherent configuration files
2061 across all peers.
2062
William Lallemandb2f07452015-05-12 14:27:13 +02002063 You may want to reference some environment variables in the address
2064 parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01002065
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002066 Note: "peer" keyword may transparently be replaced by "server" keyword (see
2067 "server" keyword explanation below).
2068
2069server <peername> [<ip>:<port>] [param*]
Michael Prokop4438c602019-05-24 10:25:45 +02002070 As previously mentioned, "peer" keyword may be replaced by "server" keyword
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002071 with a support for all "server" parameters found in 5.2 paragraph.
2072 If the underlying peer is local, <ip>:<port> parameters must not be present.
2073 These parameters must be provided on a "bind" line (see "bind" keyword
2074 of this "peers" section).
2075 Some of these parameters are irrelevant for "peers" sections.
2076
2077
Cyril Bontédc4d9032012-04-08 21:57:39 +02002078 Example:
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002079 # The old way.
Emeric Brunf099e792010-09-27 12:05:28 +02002080 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002081 peer haproxy1 192.168.0.1:1024
2082 peer haproxy2 192.168.0.2:1024
2083 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02002084
2085 backend mybackend
2086 mode tcp
2087 balance roundrobin
2088 stick-table type ip size 20k peers mypeers
2089 stick on src
2090
Willy Tarreauf7b30a92010-12-06 22:59:17 +01002091 server srv1 192.168.0.30:80
2092 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02002093
Frédéric Lécaille2f167b32019-01-11 14:13:54 +01002094 Example:
2095 peers mypeers
2096 bind 127.0.0.11:10001 ssl crt mycerts/pem
2097 default-server ssl verify none
2098 server hostA 127.0.0.10:10000
2099 server hostB #local peer
Emeric Brunf099e792010-09-27 12:05:28 +02002100
Frédéric Lécaille4f5b77c2019-03-18 14:05:58 +01002101
2102table <tablename> type {ip | integer | string [len <length>] | binary [len <length>]}
2103 size <size> [expire <expire>] [nopurge] [store <data_type>]*
2104
2105 Configure a stickiness table for the current section. This line is parsed
2106 exactly the same way as the "stick-table" keyword in others section, except
2107 for the "peers" argument which is not required here and with an aditionnal
2108 mandatory first parameter to designate the stick-table. Contrary to others
2109 sections, there may be several "table" lines in "peers" sections (see also
2110 "stick-table" keyword).
2111
2112 Also be aware of the fact that "peers" sections have their own stick-table
2113 namespaces to avoid collisions between stick-table names identical in
2114 different "peers" section. This is internally handled prepending the "peers"
2115 sections names to the name of the stick-tables followed by a '/' character.
2116 If somewhere else in the configuration file you have to refer to such
2117 stick-tables declared in "peers" sections you must use the prefixed version
2118 of the stick-table name as follows:
2119
2120 peers mypeers
2121 peer A ...
2122 peer B ...
2123 table t1 ...
2124
2125 frontend fe1
2126 tcp-request content track-sc0 src table mypeers/t1
2127
2128 This is also this prefixed version of the stick-table names which must be
2129 used to refer to stick-tables through the CLI.
2130
2131 About "peers" protocol, as only "peers" belonging to the same section may
2132 communicate with each others, there is no need to do such a distinction.
2133 Several "peers" sections may declare stick-tables with the same name.
2134 This is shorter version of the stick-table name which is sent over the network.
2135 There is only a '/' character as prefix to avoid stick-table name collisions between
2136 stick-tables declared as backends and stick-table declared in "peers" sections
2137 as follows in this weird but supported configuration:
2138
2139 peers mypeers
2140 peer A ...
2141 peer B ...
2142 table t1 type string size 10m store gpc0
2143
2144 backend t1
2145 stick-table type string size 10m store gpc0 peers mypeers
2146
2147 Here "t1" table declared in "mypeeers" section has "mypeers/t1" as global name.
2148 "t1" table declared as a backend as "t1" as global name. But at peer protocol
2149 level the former table is named "/t1", the latter is again named "t1".
2150
Simon Horman51a1cf62015-02-03 13:00:44 +090021513.6. Mailers
2152------------
2153It is possible to send email alerts when the state of servers changes.
2154If configured email alerts are sent to each mailer that is configured
2155in a mailers section. Email is sent to mailers using SMTP.
2156
Pieter Baauw386a1272015-08-16 15:26:24 +02002157mailers <mailersect>
Simon Horman51a1cf62015-02-03 13:00:44 +09002158 Creates a new mailer list with the name <mailersect>. It is an
2159 independent section which is referenced by one or more proxies.
2160
2161mailer <mailername> <ip>:<port>
2162 Defines a mailer inside a mailers section.
2163
2164 Example:
2165 mailers mymailers
2166 mailer smtp1 192.168.0.1:587
2167 mailer smtp2 192.168.0.2:587
2168
2169 backend mybackend
2170 mode tcp
2171 balance roundrobin
2172
2173 email-alert mailers mymailers
2174 email-alert from test1@horms.org
2175 email-alert to test2@horms.org
2176
2177 server srv1 192.168.0.30:80
2178 server srv2 192.168.0.31:80
2179
Pieter Baauw235fcfc2016-02-13 15:33:40 +01002180timeout mail <time>
2181 Defines the time available for a mail/connection to be made and send to
2182 the mail-server. If not defined the default value is 10 seconds. To allow
2183 for at least two SYN-ACK packets to be send during initial TCP handshake it
2184 is advised to keep this value above 4 seconds.
2185
2186 Example:
2187 mailers mymailers
2188 timeout mail 20s
2189 mailer smtp1 192.168.0.1:587
Simon Horman51a1cf62015-02-03 13:00:44 +09002190
Willy Tarreauc57f0e22009-05-10 13:12:33 +020021914. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02002192----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002193
Willy Tarreau6a06a402007-07-15 20:15:28 +02002194Proxy configuration can be located in a set of sections :
William Lallemand6e62fb62015-04-28 16:55:23 +02002195 - defaults [<name>]
Willy Tarreau6a06a402007-07-15 20:15:28 +02002196 - frontend <name>
2197 - backend <name>
2198 - listen <name>
2199
2200A "defaults" section sets default parameters for all other sections following
2201its declaration. Those default parameters are reset by the next "defaults"
2202section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01002203section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002204
2205A "frontend" section describes a set of listening sockets accepting client
2206connections.
2207
2208A "backend" section describes a set of servers to which the proxy will connect
2209to forward incoming connections.
2210
2211A "listen" section defines a complete proxy with its frontend and backend
2212parts combined in one section. It is generally useful for TCP-only traffic.
2213
Willy Tarreau0ba27502007-12-24 16:55:16 +01002214All proxy names must be formed from upper and lower case letters, digits,
2215'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
2216case-sensitive, which means that "www" and "WWW" are two different proxies.
2217
2218Historically, all proxy names could overlap, it just caused troubles in the
2219logs. Since the introduction of content switching, it is mandatory that two
2220proxies with overlapping capabilities (frontend/backend) have different names.
2221However, it is still permitted that a frontend and a backend share the same
2222name, as this configuration seems to be commonly encountered.
2223
2224Right now, two major proxy modes are supported : "tcp", also known as layer 4,
2225and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002226bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002227protocol, and can interact with it by allowing, blocking, switching, adding,
2228modifying, or removing arbitrary contents in requests or responses, based on
2229arbitrary criteria.
2230
Willy Tarreau70dffda2014-01-30 03:07:23 +01002231In HTTP mode, the processing applied to requests and responses flowing over
2232a connection depends in the combination of the frontend's HTTP options and
Christopher Faulet315b39c2018-09-21 16:26:19 +02002233the backend's. HAProxy supports 4 connection modes :
Willy Tarreau70dffda2014-01-30 03:07:23 +01002234
2235 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
2236 requests and responses are processed, and connections remain open but idle
2237 between responses and new requests.
2238
2239 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
2240 1.0 to 1.5-dev21 : only the first request and response are processed, and
2241 everything else is forwarded with no analysis at all. This mode should not
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002242 be used as it creates lots of trouble with logging and HTTP processing.
2243 And because it cannot work in HTTP/2, this option is deprecated and it is
2244 only supported on legacy HTTP frontends. In HTX, it is ignored and a
2245 warning is emitted during HAProxy startup.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002246
Willy Tarreau70dffda2014-01-30 03:07:23 +01002247 - SCL: server close ("option http-server-close") : the server-facing
2248 connection is closed after the end of the response is received, but the
2249 client-facing connection remains open.
2250
Christopher Faulet315b39c2018-09-21 16:26:19 +02002251 - CLO: close ("option httpclose"): the connection is closed after the end of
2252 the response and "Connection: close" appended in both directions.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002253
2254The effective mode that will be applied to a connection passing through a
2255frontend and a backend can be determined by both proxy modes according to the
2256following matrix, but in short, the modes are symmetric, keep-alive is the
Christopher Faulet315b39c2018-09-21 16:26:19 +02002257weakest option and close is the strongest.
Willy Tarreau70dffda2014-01-30 03:07:23 +01002258
Christopher Faulet315b39c2018-09-21 16:26:19 +02002259 Backend mode
Willy Tarreau70dffda2014-01-30 03:07:23 +01002260
Christopher Faulet315b39c2018-09-21 16:26:19 +02002261 | KAL | SCL | CLO
2262 ----+-----+-----+----
2263 KAL | KAL | SCL | CLO
2264 ----+-----+-----+----
2265 TUN | TUN | SCL | CLO
2266 Frontend ----+-----+-----+----
2267 mode SCL | SCL | SCL | CLO
2268 ----+-----+-----+----
2269 CLO | CLO | CLO | CLO
Willy Tarreau70dffda2014-01-30 03:07:23 +01002270
Willy Tarreau0ba27502007-12-24 16:55:16 +01002271
Willy Tarreau70dffda2014-01-30 03:07:23 +01002272
Willy Tarreauc57f0e22009-05-10 13:12:33 +020022734.1. Proxy keywords matrix
2274--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002275
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002276The following list of keywords is supported. Most of them may only be used in a
2277limited set of section types. Some of them are marked as "deprecated" because
2278they are inherited from an old syntax which may be confusing or functionally
2279limited, and there are new recommended keywords to replace them. Keywords
Davor Ocelice9ed2812017-12-25 17:49:28 +01002280marked with "(*)" can be optionally inverted using the "no" prefix, e.g. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002281option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02002282and must be disabled for a specific instance. Such options may also be prefixed
2283with "default" in order to restore default settings regardless of what has been
2284specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002285
Willy Tarreau6a06a402007-07-15 20:15:28 +02002286
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002287 keyword defaults frontend listen backend
2288------------------------------------+----------+----------+---------+---------
2289acl - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002290backlog X X X -
2291balance X - X X
2292bind - X X -
2293bind-process X X X X
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002294block (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002295capture cookie - X X -
2296capture request header - X X -
2297capture response header - X X -
2298clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02002299compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002300contimeout (deprecated) X - X X
2301cookie X - X X
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02002302declare capture - X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002303default-server X - X X
2304default_backend X X X -
2305description - X X X
2306disabled X X X X
2307dispatch - - X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002308email-alert from X X X X
Simon Horman64e34162015-02-06 11:11:57 +09002309email-alert level X X X X
Simon Horman51a1cf62015-02-03 13:00:44 +09002310email-alert mailers X X X X
2311email-alert myhostname X X X X
2312email-alert to X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002313enabled X X X X
2314errorfile X X X X
2315errorloc X X X X
2316errorloc302 X X X X
2317-- keyword -------------------------- defaults - frontend - listen -- backend -
2318errorloc303 X X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002319force-persist - - X X
Christopher Fauletc3fe5332016-04-07 15:30:10 +02002320filter - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002321fullconn X - X X
2322grace X X X X
2323hash-type X - X X
2324http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01002325http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02002326http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002327http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02002328http-response - X X X
Willy Tarreau30631952015-08-06 15:05:24 +02002329http-reuse X - X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02002330http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002331id - X X X
Cyril Bonté4288c5a2018-03-12 22:02:59 +01002332ignore-persist - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002333load-server-state-from-file X - X X
William Lallemand0f99e342011-10-12 17:50:54 +02002334log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01002335log-format X X X -
Dragan Dosen7ad31542015-09-28 17:16:47 +02002336log-format-sd X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01002337log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02002338max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002339maxconn X X X -
2340mode X X X X
2341monitor fail - X X -
2342monitor-net X X X -
2343monitor-uri X X X -
2344option abortonclose (*) X - X X
2345option accept-invalid-http-request (*) X X X -
2346option accept-invalid-http-response (*) X - X X
2347option allbackups (*) X - X X
2348option checkcache (*) X - X X
2349option clitcpka (*) X X X -
2350option contstats (*) X X X -
2351option dontlog-normal (*) X X X -
2352option dontlognull (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002353-- keyword -------------------------- defaults - frontend - listen -- backend -
2354option forwardfor X X X X
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02002355option http-buffer-request (*) X X X X
Willy Tarreau82649f92015-05-01 22:40:51 +02002356option http-ignore-probes (*) X X X -
Willy Tarreau16bfb022010-01-16 19:48:41 +01002357option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02002358option http-no-delay (*) X X X X
Christopher Faulet98db9762018-09-21 10:25:19 +02002359option http-pretend-keepalive (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002360option http-server-close (*) X X X X
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01002361option http-tunnel (deprecated) (*) X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002362option http-use-proxy-header (*) X X X -
Willy Tarreau68ad3a42018-10-22 11:49:15 +02002363option http-use-htx (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002364option httpchk X - X X
2365option httpclose (*) X X X X
Freddy Spierenburge88b7732019-03-25 14:35:17 +01002366option httplog X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002367option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002368option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02002369option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002370option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002371option log-health-checks (*) X - X X
2372option log-separate-errors (*) X X X -
2373option logasap (*) X X X -
2374option mysql-check X - X X
2375option nolinger (*) X X X X
2376option originalto X X X X
2377option persist (*) X - X X
Baptiste Assmann809e22a2015-10-12 20:22:55 +02002378option pgsql-check X - X X
2379option prefer-last-server (*) X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002380option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02002381option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002382option smtpchk X - X X
2383option socket-stats (*) X X X -
2384option splice-auto (*) X X X X
2385option splice-request (*) X X X X
2386option splice-response (*) X X X X
Christopher Fauletba7bc162016-11-07 21:07:38 +01002387option spop-check - - - X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002388option srvtcpka (*) X - X X
2389option ssl-hello-chk X - X X
2390-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01002391option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002392option tcp-smart-accept (*) X X X -
2393option tcp-smart-connect (*) X - X X
2394option tcpka X X X X
2395option tcplog X X X X
2396option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09002397external-check command X - X X
2398external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002399persist rdp-cookie X - X X
2400rate-limit sessions X X X -
2401redirect - X X X
2402redisp (deprecated) X - X X
2403redispatch (deprecated) X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002404reqadd (deprecated) - X X X
2405reqallow (deprecated) - X X X
2406reqdel (deprecated) - X X X
2407reqdeny (deprecated) - X X X
2408reqiallow (deprecated) - X X X
2409reqidel (deprecated) - X X X
2410reqideny (deprecated) - X X X
2411reqipass (deprecated) - X X X
2412reqirep (deprecated) - X X X
2413reqitarpit (deprecated) - X X X
2414reqpass (deprecated) - X X X
2415reqrep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002416-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreau96d51952019-05-22 20:34:35 +02002417reqtarpit (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002418retries X - X X
Olivier Houcharda254a372019-04-05 15:30:12 +02002419retry-on X - X X
Willy Tarreau96d51952019-05-22 20:34:35 +02002420rspadd (deprecated) - X X X
2421rspdel (deprecated) - X X X
2422rspdeny (deprecated) - X X X
2423rspidel (deprecated) - X X X
2424rspideny (deprecated) - X X X
2425rspirep (deprecated) - X X X
2426rsprep (deprecated) - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002427server - - X X
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02002428server-state-file-name X - X X
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02002429server-template - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002430source X - X X
2431srvtimeout (deprecated) X - X X
Baptiste Assmann5a549212015-10-12 20:30:24 +02002432stats admin - X X X
2433stats auth X X X X
2434stats enable X X X X
2435stats hide-version X X X X
2436stats http-request - X X X
2437stats realm X X X X
2438stats refresh X X X X
2439stats scope X X X X
2440stats show-desc X X X X
2441stats show-legends X X X X
2442stats show-node X X X X
2443stats uri X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002444-- keyword -------------------------- defaults - frontend - listen -- backend -
2445stick match - - X X
2446stick on - - X X
2447stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02002448stick store-response - - X X
Adam Spiers68af3c12017-04-06 16:31:39 +01002449stick-table - X X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02002450tcp-check connect - - X X
2451tcp-check expect - - X X
2452tcp-check send - - X X
2453tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02002454tcp-request connection - X X -
2455tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02002456tcp-request inspect-delay - X X X
Willy Tarreau4f614292016-10-21 17:49:36 +02002457tcp-request session - X X -
Emeric Brun0a3b67f2010-09-24 15:34:53 +02002458tcp-response content - - X X
2459tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002460timeout check X - X X
2461timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02002462timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002463timeout clitimeout (deprecated) X X X -
2464timeout connect X - X X
2465timeout contimeout (deprecated) X - X X
2466timeout http-keep-alive X X X X
2467timeout http-request X X X X
2468timeout queue X - X X
2469timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02002470timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002471timeout srvtimeout (deprecated) X - X X
2472timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02002473timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002474transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01002475unique-id-format X X X -
2476unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002477use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02002478use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01002479------------------------------------+----------+----------+---------+---------
2480 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02002481
Willy Tarreau0ba27502007-12-24 16:55:16 +01002482
Willy Tarreauc57f0e22009-05-10 13:12:33 +020024834.2. Alphabetically sorted keywords reference
2484---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01002485
2486This section provides a description of each keyword and its usage.
2487
2488
2489acl <aclname> <criterion> [flags] [operator] <value> ...
2490 Declare or complete an access list.
2491 May be used in sections : defaults | frontend | listen | backend
2492 no | yes | yes | yes
2493 Example:
2494 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2495 acl invalid_src src_port 0:1023
2496 acl local_dst hdr(host) -i localhost
2497
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002498 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002499
2500
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002501backlog <conns>
2502 Give hints to the system about the approximate listen backlog desired size
2503 May be used in sections : defaults | frontend | listen | backend
2504 yes | yes | yes | no
2505 Arguments :
2506 <conns> is the number of pending connections. Depending on the operating
2507 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02002508 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01002509
2510 In order to protect against SYN flood attacks, one solution is to increase
2511 the system's SYN backlog size. Depending on the system, sometimes it is just
2512 tunable via a system parameter, sometimes it is not adjustable at all, and
2513 sometimes the system relies on hints given by the application at the time of
2514 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
2515 to the listen() syscall. On systems which can make use of this value, it can
2516 sometimes be useful to be able to specify a different value, hence this
2517 backlog parameter.
2518
2519 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
2520 used as a hint and the system accepts up to the smallest greater power of
2521 two, and never more than some limits (usually 32768).
2522
2523 See also : "maxconn" and the target operating system's tuning guide.
2524
2525
Willy Tarreau0ba27502007-12-24 16:55:16 +01002526balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02002527balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002528 Define the load balancing algorithm to be used in a backend.
2529 May be used in sections : defaults | frontend | listen | backend
2530 yes | no | yes | yes
2531 Arguments :
2532 <algorithm> is the algorithm used to select a server when doing load
2533 balancing. This only applies when no persistence information
2534 is available, or when a connection is redispatched to another
2535 server. <algorithm> may be one of the following :
2536
2537 roundrobin Each server is used in turns, according to their weights.
2538 This is the smoothest and fairest algorithm when the server's
2539 processing time remains equally distributed. This algorithm
2540 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02002541 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08002542 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02002543 large farms, when a server becomes up after having been down
2544 for a very short time, it may sometimes take a few hundreds
2545 requests for it to be re-integrated into the farm and start
2546 receiving traffic. This is normal, though very rare. It is
2547 indicated here in case you would have the chance to observe
2548 it, so that you don't worry.
2549
2550 static-rr Each server is used in turns, according to their weights.
2551 This algorithm is as similar to roundrobin except that it is
2552 static, which means that changing a server's weight on the
2553 fly will have no effect. On the other hand, it has no design
2554 limitation on the number of servers, and when a server goes
2555 up, it is always immediately reintroduced into the farm, once
2556 the full map is recomputed. It also uses slightly less CPU to
2557 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002558
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01002559 leastconn The server with the lowest number of connections receives the
2560 connection. Round-robin is performed within groups of servers
2561 of the same load to ensure that all servers will be used. Use
2562 of this algorithm is recommended where very long sessions are
2563 expected, such as LDAP, SQL, TSE, etc... but is not very well
2564 suited for protocols using short sessions such as HTTP. This
2565 algorithm is dynamic, which means that server weights may be
2566 adjusted on the fly for slow starts for instance.
2567
Willy Tarreauf09c6602012-02-13 17:12:08 +01002568 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002569 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01002570 identifier to the highest (see server parameter "id"), which
2571 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02002572 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01002573 not make sense to use this algorithm without setting maxconn.
2574 The purpose of this algorithm is to always use the smallest
2575 number of servers so that extra servers can be powered off
2576 during non-intensive hours. This algorithm ignores the server
2577 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02002578 or IMAP than HTTP, though it can be useful there too. In
2579 order to use this algorithm efficiently, it is recommended
2580 that a cloud controller regularly checks server usage to turn
2581 them off when unused, and regularly checks backend queue to
2582 turn new servers on when the queue inflates. Alternatively,
2583 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01002584
Willy Tarreau0ba27502007-12-24 16:55:16 +01002585 source The source IP address is hashed and divided by the total
2586 weight of the running servers to designate which server will
2587 receive the request. This ensures that the same client IP
2588 address will always reach the same server as long as no
2589 server goes down or up. If the hash result changes due to the
2590 number of running servers changing, many clients will be
2591 directed to a different server. This algorithm is generally
2592 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002593 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01002594 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002595 static by default, which means that changing a server's
2596 weight on the fly will have no effect, but this can be
2597 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002598
Oskar Stolc8dc41842012-05-19 10:19:54 +01002599 uri This algorithm hashes either the left part of the URI (before
2600 the question mark) or the whole URI (if the "whole" parameter
2601 is present) and divides the hash value by the total weight of
2602 the running servers. The result designates which server will
2603 receive the request. This ensures that the same URI will
2604 always be directed to the same server as long as no server
2605 goes up or down. This is used with proxy caches and
2606 anti-virus proxies in order to maximize the cache hit rate.
2607 Note that this algorithm may only be used in an HTTP backend.
2608 This algorithm is static by default, which means that
2609 changing a server's weight on the fly will have no effect,
2610 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002611
Oskar Stolc8dc41842012-05-19 10:19:54 +01002612 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002613 "depth", both followed by a positive integer number. These
2614 options may be helpful when it is needed to balance servers
2615 based on the beginning of the URI only. The "len" parameter
2616 indicates that the algorithm should only consider that many
2617 characters at the beginning of the URI to compute the hash.
2618 Note that having "len" set to 1 rarely makes sense since most
2619 URIs start with a leading "/".
2620
2621 The "depth" parameter indicates the maximum directory depth
2622 to be used to compute the hash. One level is counted for each
2623 slash in the request. If both parameters are specified, the
2624 evaluation stops when either is reached.
2625
Willy Tarreau0ba27502007-12-24 16:55:16 +01002626 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002627 the query string of each HTTP GET request.
2628
2629 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02002630 request entity will be searched for the parameter argument,
2631 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02002632 ('?') in the URL. The message body will only start to be
2633 analyzed once either the advertised amount of data has been
2634 received or the request buffer is full. In the unlikely event
2635 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02002636 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02002637 be randomly balanced if at all. This keyword used to support
2638 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002639
2640 If the parameter is found followed by an equal sign ('=') and
2641 a value, then the value is hashed and divided by the total
2642 weight of the running servers. The result designates which
2643 server will receive the request.
2644
2645 This is used to track user identifiers in requests and ensure
2646 that a same user ID will always be sent to the same server as
2647 long as no server goes up or down. If no value is found or if
2648 the parameter is not found, then a round robin algorithm is
2649 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002650 backend. This algorithm is static by default, which means
2651 that changing a server's weight on the fly will have no
2652 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002653
Cyril Bontédc4d9032012-04-08 21:57:39 +02002654 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
2655 request. Just as with the equivalent ACL 'hdr()' function,
2656 the header name in parenthesis is not case sensitive. If the
2657 header is absent or if it does not contain any value, the
2658 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01002659
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002660 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01002661 reducing the hash algorithm to the main domain part with some
2662 specific headers such as 'Host'. For instance, in the Host
2663 value "haproxy.1wt.eu", only "1wt" will be considered.
2664
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002665 This algorithm is static by default, which means that
2666 changing a server's weight on the fly will have no effect,
2667 but this can be changed using "hash-type".
2668
Willy Tarreau21c741a2019-01-14 18:14:27 +01002669 random
2670 random(<draws>)
2671 A random number will be used as the key for the consistent
Willy Tarreau760e81d2018-05-03 07:20:40 +02002672 hashing function. This means that the servers' weights are
2673 respected, dynamic weight changes immediately take effect, as
2674 well as new server additions. Random load balancing can be
2675 useful with large farms or when servers are frequently added
Willy Tarreau21c741a2019-01-14 18:14:27 +01002676 or removed as it may avoid the hammering effect that could
2677 result from roundrobin or leastconn in this situation. The
2678 hash-balance-factor directive can be used to further improve
2679 fairness of the load balancing, especially in situations
2680 where servers show highly variable response times. When an
2681 argument <draws> is present, it must be an integer value one
2682 or greater, indicating the number of draws before selecting
2683 the least loaded of these servers. It was indeed demonstrated
2684 that picking the least loaded of two servers is enough to
2685 significantly improve the fairness of the algorithm, by
2686 always avoiding to pick the most loaded server within a farm
2687 and getting rid of any bias that could be induced by the
2688 unfair distribution of the consistent list. Higher values N
2689 will take away N-1 of the highest loaded servers at the
2690 expense of performance. With very high values, the algorithm
2691 will converge towards the leastconn's result but much slower.
2692 The default value is 2, which generally shows very good
2693 distribution and performance. This algorithm is also known as
2694 the Power of Two Random Choices and is described here :
2695 http://www.eecs.harvard.edu/~michaelm/postscripts/handbook2001.pdf
Willy Tarreau760e81d2018-05-03 07:20:40 +02002696
Emeric Brun736aa232009-06-30 17:56:00 +02002697 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02002698 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02002699 The RDP cookie <name> (or "mstshash" if omitted) will be
2700 looked up and hashed for each incoming TCP request. Just as
2701 with the equivalent ACL 'req_rdp_cookie()' function, the name
2702 is not case-sensitive. This mechanism is useful as a degraded
2703 persistence mode, as it makes it possible to always send the
2704 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002705 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02002706 used instead.
2707
2708 Note that for this to work, the frontend must ensure that an
2709 RDP cookie is already present in the request buffer. For this
2710 you must use 'tcp-request content accept' rule combined with
2711 a 'req_rdp_cookie_cnt' ACL.
2712
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002713 This algorithm is static by default, which means that
2714 changing a server's weight on the fly will have no effect,
2715 but this can be changed using "hash-type".
2716
Cyril Bontédc4d9032012-04-08 21:57:39 +02002717 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09002718
Willy Tarreau0ba27502007-12-24 16:55:16 +01002719 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02002720 algorithms. Right now, only "url_param" and "uri" support an
2721 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002722
Willy Tarreau3cd9af22009-03-15 14:06:41 +01002723 The load balancing algorithm of a backend is set to roundrobin when no other
2724 algorithm, mode nor option have been set. The algorithm may only be set once
2725 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002726
Lukas Tribus80512b12018-10-27 20:07:40 +02002727 With authentication schemes that require the same connection like NTLM, URI
2728 based alghoritms must not be used, as they would cause subsequent requests
2729 to be routed to different backend servers, breaking the invalid assumptions
2730 NTLM relies on.
2731
Willy Tarreau0ba27502007-12-24 16:55:16 +01002732 Examples :
2733 balance roundrobin
2734 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002735 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01002736 balance hdr(User-Agent)
2737 balance hdr(host)
2738 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002739
2740 Note: the following caveats and limitations on using the "check_post"
2741 extension with "url_param" must be considered :
2742
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002743 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002744 to determine if the parameters will be found in the body or entity which
2745 may contain binary data. Therefore another method may be required to
2746 restrict consideration of POST requests that have no URL parameters in
2747 the body. (see acl reqideny http_end)
2748
2749 - using a <max_wait> value larger than the request buffer size does not
2750 make sense and is useless. The buffer size is set at build time, and
2751 defaults to 16 kB.
2752
2753 - Content-Encoding is not supported, the parameter search will probably
2754 fail; and load balancing will fall back to Round Robin.
2755
2756 - Expect: 100-continue is not supported, load balancing will fall back to
2757 Round Robin.
2758
Lukas Tribus23953682017-04-28 13:24:30 +00002759 - Transfer-Encoding (RFC7230 3.3.1) is only supported in the first chunk.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002760 If the entire parameter value is not present in the first chunk, the
2761 selection of server is undefined (actually, defined by how little
2762 actually appeared in the first chunk).
2763
2764 - This feature does not support generation of a 100, 411 or 501 response.
2765
2766 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002767 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02002768 white space or control characters are found, indicating the end of what
2769 might be a URL parameter list. This is probably not a concern with SGML
2770 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002771
Willy Tarreau294d0f02015-08-10 19:40:12 +02002772 See also : "dispatch", "cookie", "transparent", "hash-type" and "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002773
2774
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002775bind [<address>]:<port_range> [, ...] [param*]
2776bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002777 Define one or several listening addresses and/or ports in a frontend.
2778 May be used in sections : defaults | frontend | listen | backend
2779 no | yes | yes | no
2780 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002781 <address> is optional and can be a host name, an IPv4 address, an IPv6
2782 address, or '*'. It designates the address the frontend will
2783 listen on. If unset, all IPv4 addresses of the system will be
2784 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01002785 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01002786 Optionally, an address family prefix may be used before the
2787 address to force the family regardless of the address format,
2788 which can be useful to specify a path to a unix socket with
2789 no slash ('/'). Currently supported prefixes are :
2790 - 'ipv4@' -> address is always IPv4
2791 - 'ipv6@' -> address is always IPv6
2792 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02002793 - 'abns@' -> address is in abstract namespace (Linux only).
2794 Note: since abstract sockets are not "rebindable", they
2795 do not cope well with multi-process mode during
2796 soft-restart, so it is better to avoid them if
2797 nbproc is greater than 1. The effect is that if the
2798 new process fails to start, only one of the old ones
2799 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01002800 - 'fd@<n>' -> use file descriptor <n> inherited from the
2801 parent. The fd must be bound and may or may not already
2802 be listening.
William Lallemand2fe7dd02018-09-11 16:51:29 +02002803 - 'sockpair@<n>'-> like fd@ but you must use the fd of a
2804 connected unix socket or of a socketpair. The bind waits
2805 to receive a FD over the unix socket and uses it as if it
2806 was the FD of an accept(). Should be used carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02002807 You may want to reference some environment variables in the
2808 address parameter, see section 2.3 about environment
2809 variables.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01002810
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002811 <port_range> is either a unique TCP port, or a port range for which the
2812 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002813 above. The port is mandatory for TCP listeners. Note that in
2814 the case of an IPv6 address, the port is always the number
2815 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002816 - a numerical port (ex: '80')
2817 - a dash-delimited ports range explicitly stating the lower
2818 and upper bounds (ex: '2000-2100') which are included in
2819 the range.
2820
2821 Particular care must be taken against port ranges, because
2822 every <address:port> couple consumes one socket (= a file
2823 descriptor), so it's easy to consume lots of descriptors
2824 with a simple range, and to run out of sockets. Also, each
2825 <address:port> couple must be used only once among all
2826 instances running on a same system. Please note that binding
2827 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04002828 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01002829 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002830
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002831 <path> is a UNIX socket path beginning with a slash ('/'). This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01002832 alternative to the TCP listening port. HAProxy will then
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002833 receive UNIX connections on the socket located at this place.
2834 The path must begin with a slash and by default is absolute.
2835 It can be relative to the prefix defined by "unix-bind" in
2836 the global section. Note that the total length of the prefix
2837 followed by the socket path cannot exceed some system limits
2838 for UNIX sockets, which commonly are set to 107 characters.
2839
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002840 <param*> is a list of parameters common to all sockets declared on the
2841 same line. These numerous parameters depend on OS and build
2842 options and have a complete section dedicated to them. Please
2843 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002844
Willy Tarreau0ba27502007-12-24 16:55:16 +01002845 It is possible to specify a list of address:port combinations delimited by
2846 commas. The frontend will then listen on all of these addresses. There is no
2847 fixed limit to the number of addresses and ports which can be listened on in
2848 a frontend, as well as there is no limit to the number of "bind" statements
2849 in a frontend.
2850
2851 Example :
2852 listen http_proxy
2853 bind :80,:443
2854 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002855 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01002856
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002857 listen http_https_proxy
2858 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02002859 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02002860
Willy Tarreau24709282013-03-10 21:32:12 +01002861 listen http_https_proxy_explicit
2862 bind ipv6@:80
2863 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
2864 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
2865
Willy Tarreaudad36a32013-03-11 01:20:04 +01002866 listen external_bind_app1
William Lallemandb2f07452015-05-12 14:27:13 +02002867 bind "fd@${FD_APP1}"
Willy Tarreaudad36a32013-03-11 01:20:04 +01002868
Willy Tarreau55dcaf62015-09-27 15:03:15 +02002869 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
2870 sun_path length is used for the address length. Some other programs
2871 such as socat use the string length only by default. Pass the option
2872 ",unix-tightsocklen=0" to any abstract socket definition in socat to
2873 make it compatible with HAProxy's.
2874
Willy Tarreauceb24bc2010-11-09 12:46:41 +01002875 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02002876 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002877
2878
Christopher Fauletff4121f2017-11-22 16:38:49 +01002879bind-process [ all | odd | even | <process_num>[-[<process_num>]] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002880 Limit visibility of an instance to a certain set of processes numbers.
2881 May be used in sections : defaults | frontend | listen | backend
2882 yes | yes | yes | yes
2883 Arguments :
2884 all All process will see this instance. This is the default. It
2885 may be used to override a default value.
2886
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002887 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002888 option may be combined with other numbers.
2889
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002890 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002891 option may be combined with other numbers. Do not use it
2892 with less than 2 processes otherwise some instances might be
2893 missing from all processes.
2894
Christopher Fauletff4121f2017-11-22 16:38:49 +01002895 process_num The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002896 whose values must all be between 1 and 32 or 64 depending on
Christopher Fauletff4121f2017-11-22 16:38:49 +01002897 the machine's word size. Ranges can be partially defined. The
2898 higher bound can be omitted. In such case, it is replaced by
2899 the corresponding maximum value. If a proxy is bound to
2900 process numbers greater than the configured global.nbproc, it
2901 will either be forced to process #1 if a single process was
Willy Tarreau102df612014-05-07 23:56:38 +02002902 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002903
2904 This keyword limits binding of certain instances to certain processes. This
2905 is useful in order not to have too many processes listening to the same
2906 ports. For instance, on a dual-core machine, it might make sense to set
2907 'nbproc 2' in the global section, then distributes the listeners among 'odd'
2908 and 'even' instances.
2909
Willy Tarreaua9db57e2013-01-18 11:29:29 +01002910 At the moment, it is not possible to reference more than 32 or 64 processes
2911 using this keyword, but this should be more than enough for most setups.
2912 Please note that 'all' really means all processes regardless of the machine's
2913 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002914
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002915 Each "bind" line may further be limited to a subset of the proxy's processes,
2916 please consult the "process" bind keyword in section 5.1.
2917
Willy Tarreaub369a042014-09-16 13:21:03 +02002918 When a frontend has no explicit "bind-process" line, it tries to bind to all
2919 the processes referenced by its "bind" lines. That means that frontends can
2920 easily adapt to their listeners' processes.
2921
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002922 If some backends are referenced by frontends bound to other processes, the
2923 backend automatically inherits the frontend's processes.
2924
2925 Example :
2926 listen app_ip1
2927 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002928 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002929
2930 listen app_ip2
2931 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002932 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002933
2934 listen management
2935 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02002936 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002937
Willy Tarreau110ecc12012-11-15 17:50:01 +01002938 listen management
2939 bind 10.0.0.4:80
2940 bind-process 1-4
2941
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02002942 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01002943
2944
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002945block { if | unless } <condition> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002946 Block a layer 7 request if/unless a condition is matched
2947 May be used in sections : defaults | frontend | listen | backend
2948 no | yes | yes | yes
2949
2950 The HTTP request will be blocked very early in the layer 7 processing
2951 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002952 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002953 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002954 conditions are met or not met. There is no fixed limit to the number of
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03002955 "block" statements per instance. To block connections at layer 4 (without
2956 sending a 403 error) see "tcp-request connection reject" and
2957 "tcp-request content reject" rules.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002958
Jarno Huuskonen8c8c3492016-12-28 18:50:29 +02002959 This form is deprecated, do not use it in any new configuration, use the new
2960 "http-request deny" instead.
2961
Willy Tarreau0ba27502007-12-24 16:55:16 +01002962 Example:
2963 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2964 acl invalid_src src_port 0:1023
2965 acl local_dst hdr(host) -i localhost
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +03002966 # block is deprecated. Use http-request deny instead:
2967 #block if invalid_src || local_dst
2968 http-request deny if invalid_src || local_dst
Willy Tarreau0ba27502007-12-24 16:55:16 +01002969
Jarno Huuskonen95b012b2017-04-06 13:59:14 +03002970 See also : section 7 about ACL usage, "http-request deny",
2971 "http-response deny", "tcp-request connection reject" and
2972 "tcp-request content reject".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002973
2974capture cookie <name> len <length>
2975 Capture and log a cookie in the request and in the response.
2976 May be used in sections : defaults | frontend | listen | backend
2977 no | yes | yes | no
2978 Arguments :
2979 <name> is the beginning of the name of the cookie to capture. In order
2980 to match the exact name, simply suffix the name with an equal
2981 sign ('='). The full name will appear in the logs, which is
2982 useful with application servers which adjust both the cookie name
Davor Ocelice9ed2812017-12-25 17:49:28 +01002983 and value (e.g. ASPSESSIONXXX).
Willy Tarreau0ba27502007-12-24 16:55:16 +01002984
2985 <length> is the maximum number of characters to report in the logs, which
2986 include the cookie name, the equal sign and the value, all in the
2987 standard "name=value" form. The string will be truncated on the
2988 right if it exceeds <length>.
2989
2990 Only the first cookie is captured. Both the "cookie" request headers and the
2991 "set-cookie" response headers are monitored. This is particularly useful to
2992 check for application bugs causing session crossing or stealing between
2993 users, because generally the user's cookies can only change on a login page.
2994
2995 When the cookie was not presented by the client, the associated log column
2996 will report "-". When a request does not cause a cookie to be assigned by the
2997 server, a "-" is reported in the response column.
2998
2999 The capture is performed in the frontend only because it is necessary that
3000 the log format does not change for a given frontend depending on the
3001 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01003002 "capture cookie" statement in a frontend. The maximum capture length is set
3003 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
3004 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003005
3006 Example:
3007 capture cookie ASPSESSION len 32
3008
3009 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003010 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003011
3012
3013capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003014 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003015 May be used in sections : defaults | frontend | listen | backend
3016 no | yes | yes | no
3017 Arguments :
3018 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003019 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003020 appear in the requests, with the first letter of each word in
3021 upper case. The header name will not appear in the logs, only the
3022 value is reported, but the position in the logs is respected.
3023
3024 <length> is the maximum number of characters to extract from the value and
3025 report in the logs. The string will be truncated on the right if
3026 it exceeds <length>.
3027
Willy Tarreau4460d032012-11-21 23:37:37 +01003028 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003029 value will be added to the logs between braces ('{}'). If multiple headers
3030 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003031 in the same order they were declared in the configuration. Non-existent
3032 headers will be logged just as an empty string. Common uses for request
3033 header captures include the "Host" field in virtual hosting environments, the
3034 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003035 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003036 environments to find where the request came from.
3037
3038 Note that when capturing headers such as "User-agent", some spaces may be
3039 logged, making the log analysis more difficult. Thus be careful about what
3040 you log if you know your log parser is not smart enough to rely on the
3041 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003042
Willy Tarreau0900abb2012-11-22 00:21:46 +01003043 There is no limit to the number of captured request headers nor to their
3044 length, though it is wise to keep them low to limit memory usage per session.
3045 In order to keep log format consistent for a same frontend, header captures
3046 can only be declared in a frontend. It is not possible to specify a capture
3047 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003048
3049 Example:
3050 capture request header Host len 15
3051 capture request header X-Forwarded-For len 15
Cyril Bontéd1b0f7c2015-10-26 22:37:39 +01003052 capture request header Referer len 15
Willy Tarreau0ba27502007-12-24 16:55:16 +01003053
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003054 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003055 about logging.
3056
3057
3058capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01003059 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003060 May be used in sections : defaults | frontend | listen | backend
3061 no | yes | yes | no
3062 Arguments :
3063 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003064 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01003065 appear in the response, with the first letter of each word in
3066 upper case. The header name will not appear in the logs, only the
3067 value is reported, but the position in the logs is respected.
3068
3069 <length> is the maximum number of characters to extract from the value and
3070 report in the logs. The string will be truncated on the right if
3071 it exceeds <length>.
3072
Willy Tarreau4460d032012-11-21 23:37:37 +01003073 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01003074 result will be added to the logs between braces ('{}') after the captured
3075 request headers. If multiple headers are captured, they will be delimited by
3076 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003077 the configuration. Non-existent headers will be logged just as an empty
3078 string. Common uses for response header captures include the "Content-length"
3079 header which indicates how many bytes are expected to be returned, the
3080 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003081
Willy Tarreau0900abb2012-11-22 00:21:46 +01003082 There is no limit to the number of captured response headers nor to their
3083 length, though it is wise to keep them low to limit memory usage per session.
3084 In order to keep log format consistent for a same frontend, header captures
3085 can only be declared in a frontend. It is not possible to specify a capture
3086 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003087
3088 Example:
3089 capture response header Content-length len 9
3090 capture response header Location len 15
3091
Willy Tarreauc57f0e22009-05-10 13:12:33 +02003092 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01003093 about logging.
3094
3095
Cyril Bontéf0c60612010-02-06 14:44:47 +01003096clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003097 Set the maximum inactivity time on the client side.
3098 May be used in sections : defaults | frontend | listen | backend
3099 yes | yes | yes | no
3100 Arguments :
3101 <timeout> is the timeout value is specified in milliseconds by default, but
3102 can be in any other unit if the number is suffixed by the unit,
3103 as explained at the top of this document.
3104
3105 The inactivity timeout applies when the client is expected to acknowledge or
3106 send data. In HTTP mode, this timeout is particularly important to consider
3107 during the first phase, when the client sends the request, and during the
3108 response while it is reading data sent by the server. The value is specified
3109 in milliseconds by default, but can be in any other unit if the number is
3110 suffixed by the unit, as specified at the top of this document. In TCP mode
3111 (and to a lesser extent, in HTTP mode), it is highly recommended that the
3112 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003113 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01003114 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +01003115 (e.g. 4 or 5 seconds).
Willy Tarreau0ba27502007-12-24 16:55:16 +01003116
3117 This parameter is specific to frontends, but can be specified once for all in
3118 "defaults" sections. This is in fact one of the easiest solutions not to
3119 forget about it. An unspecified timeout results in an infinite timeout, which
3120 is not recommended. Such a usage is accepted and works but reports a warning
3121 during startup because it may results in accumulation of expired sessions in
3122 the system if the system's timeouts are not configured either.
3123
3124 This parameter is provided for compatibility but is currently deprecated.
3125 Please use "timeout client" instead.
3126
Willy Tarreau036fae02008-01-06 13:24:40 +01003127 See also : "timeout client", "timeout http-request", "timeout server", and
3128 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003129
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003130compression algo <algorithm> ...
3131compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02003132compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02003133 Enable HTTP compression.
3134 May be used in sections : defaults | frontend | listen | backend
3135 yes | yes | yes | yes
3136 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003137 algo is followed by the list of supported compression algorithms.
3138 type is followed by the list of MIME types that will be compressed.
3139 offload makes haproxy work as a compression offloader only (see notes).
3140
3141 The currently supported algorithms are :
Willy Tarreauc91840a2015-03-28 17:00:39 +01003142 identity this is mostly for debugging, and it was useful for developing
3143 the compression feature. Identity does not apply any change on
3144 data.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003145
Willy Tarreauc91840a2015-03-28 17:00:39 +01003146 gzip applies gzip compression. This setting is only available when
Baptiste Assmannf085d632015-12-21 17:57:32 +01003147 support for zlib or libslz was built in.
Willy Tarreauc91840a2015-03-28 17:00:39 +01003148
3149 deflate same as "gzip", but with deflate algorithm and zlib format.
3150 Note that this algorithm has ambiguous support on many
3151 browsers and no support at all from recent ones. It is
3152 strongly recommended not to use it for anything else than
3153 experimentation. This setting is only available when support
Baptiste Assmannf085d632015-12-21 17:57:32 +01003154 for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003155
Willy Tarreauc91840a2015-03-28 17:00:39 +01003156 raw-deflate same as "deflate" without the zlib wrapper, and used as an
3157 alternative when the browser wants "deflate". All major
3158 browsers understand it and despite violating the standards,
3159 it is known to work better than "deflate", at least on MSIE
3160 and some versions of Safari. Do not use it in conjunction
3161 with "deflate", use either one or the other since both react
3162 to the same Accept-Encoding token. This setting is only
Baptiste Assmannf085d632015-12-21 17:57:32 +01003163 available when support for zlib or libslz was built in.
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003164
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04003165 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01003166 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003167 If backend servers support HTTP compression, these directives
3168 will be no-op: haproxy will see the compressed response and will not
3169 compress again. If backend servers do not support HTTP compression and
3170 there is Accept-Encoding header in request, haproxy will compress the
3171 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02003172
3173 The "offload" setting makes haproxy remove the Accept-Encoding header to
3174 prevent backend servers from compressing responses. It is strongly
3175 recommended not to do this because this means that all the compression work
3176 will be done on the single point where haproxy is located. However in some
3177 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04003178 with broken HTTP compression implementation which can't be turned off.
3179 In that case haproxy can be used to prevent that gateway from emitting
3180 invalid payloads. In this case, simply removing the header in the
3181 configuration does not work because it applies before the header is parsed,
3182 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02003183 then be used for such scenarios. Note: for now, the "offload" setting is
3184 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02003185
William Lallemand05097442012-11-20 12:14:28 +01003186 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003187 * the request does not advertise a supported compression algorithm in the
3188 "Accept-Encoding" header
3189 * the response message is not HTTP/1.1
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003190 * HTTP status code is not one of 200, 201, 202, or 203
Baptiste Assmann650d53d2013-01-05 15:44:44 +01003191 * response contain neither a "Content-Length" header nor a
3192 "Transfer-Encoding" whose last value is "chunked"
3193 * response contains a "Content-Type" header whose first value starts with
3194 "multipart"
3195 * the response contains the "no-transform" value in the "Cache-control"
3196 header
3197 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
3198 and later
3199 * The response contains a "Content-Encoding" header, indicating that the
3200 response is already compressed (see compression offload)
Tim Duesterhusbb48c9a2019-01-30 23:46:04 +01003201 * The response contains an invalid "ETag" header or multiple ETag headers
William Lallemand05097442012-11-20 12:14:28 +01003202
Tim Duesterhusb229f012019-01-29 16:38:56 +01003203 Note: The compression does not emit the Warning header.
William Lallemand05097442012-11-20 12:14:28 +01003204
William Lallemand82fe75c2012-10-23 10:25:10 +02003205 Examples :
3206 compression algo gzip
3207 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01003208
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003209
Cyril Bontéf0c60612010-02-06 14:44:47 +01003210contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01003211 Set the maximum time to wait for a connection attempt to a server to succeed.
3212 May be used in sections : defaults | frontend | listen | backend
3213 yes | no | yes | yes
3214 Arguments :
3215 <timeout> is the timeout value is specified in milliseconds by default, but
3216 can be in any other unit if the number is suffixed by the unit,
3217 as explained at the top of this document.
3218
3219 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003220 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01003221 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +01003222 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003223 connect timeout also presets the queue timeout to the same value if this one
3224 has not been specified. Historically, the contimeout was also used to set the
3225 tarpit timeout in a listen section, which is not possible in a pure frontend.
3226
3227 This parameter is specific to backends, but can be specified once for all in
3228 "defaults" sections. This is in fact one of the easiest solutions not to
3229 forget about it. An unspecified timeout results in an infinite timeout, which
3230 is not recommended. Such a usage is accepted and works but reports a warning
3231 during startup because it may results in accumulation of failed sessions in
3232 the system if the system's timeouts are not configured either.
3233
3234 This parameter is provided for backwards compatibility but is currently
3235 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
3236 instead.
3237
3238 See also : "timeout connect", "timeout queue", "timeout tarpit",
3239 "timeout server", "contimeout".
3240
3241
Willy Tarreau55165fe2009-05-10 12:02:55 +02003242cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02003243 [ postonly ] [ preserve ] [ httponly ] [ secure ]
3244 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Olivier Houchard4e694042017-03-14 20:01:29 +01003245 [ dynamic ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01003246 Enable cookie-based persistence in a backend.
3247 May be used in sections : defaults | frontend | listen | backend
3248 yes | no | yes | yes
3249 Arguments :
3250 <name> is the name of the cookie which will be monitored, modified or
3251 inserted in order to bring persistence. This cookie is sent to
3252 the client via a "Set-Cookie" header in the response, and is
3253 brought back by the client in a "Cookie" header in all requests.
3254 Special care should be taken to choose a name which does not
3255 conflict with any likely application cookie. Also, if the same
Davor Ocelice9ed2812017-12-25 17:49:28 +01003256 backends are subject to be used by the same clients (e.g.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003257 HTTP/HTTPS), care should be taken to use different cookie names
3258 between all backends if persistence between them is not desired.
3259
3260 rewrite This keyword indicates that the cookie will be provided by the
3261 server and that haproxy will have to modify its value to set the
3262 server's identifier in it. This mode is handy when the management
3263 of complex combinations of "Set-cookie" and "Cache-control"
3264 headers is left to the application. The application can then
3265 decide whether or not it is appropriate to emit a persistence
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003266 cookie. Since all responses should be monitored, this mode
3267 doesn't work in HTTP tunnel mode. Unless the application
Davor Ocelice9ed2812017-12-25 17:49:28 +01003268 behavior is very complex and/or broken, it is advised not to
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003269 start with this mode for new deployments. This keyword is
3270 incompatible with "insert" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003271
3272 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02003273 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003274
Willy Tarreaua79094d2010-08-31 22:54:15 +02003275 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003276 server. When used without the "preserve" option, if the server
Michael Prokop4438c602019-05-24 10:25:45 +02003277 emits a cookie with the same name, it will be removed before
Davor Ocelice9ed2812017-12-25 17:49:28 +01003278 processing. For this reason, this mode can be used to upgrade
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003279 existing configurations running in the "rewrite" mode. The cookie
3280 will only be a session cookie and will not be stored on the
3281 client's disk. By default, unless the "indirect" option is added,
3282 the server will see the cookies emitted by the client. Due to
3283 caching effects, it is generally wise to add the "nocache" or
3284 "postonly" keywords (see below). The "insert" keyword is not
3285 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003286
3287 prefix This keyword indicates that instead of relying on a dedicated
3288 cookie for the persistence, an existing one will be completed.
3289 This may be needed in some specific environments where the client
3290 does not support more than one single cookie and the application
3291 already needs it. In this case, whenever the server sets a cookie
3292 named <name>, it will be prefixed with the server's identifier
3293 and a delimiter. The prefix will be removed from all client
3294 requests so that the server still finds the cookie it emitted.
3295 Since all requests and responses are subject to being modified,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01003296 this mode doesn't work with tunnel mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02003297 not compatible with "rewrite" and "insert". Note: it is highly
3298 recommended not to use "indirect" with "prefix", otherwise server
3299 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003300
Willy Tarreaua79094d2010-08-31 22:54:15 +02003301 indirect When this option is specified, no cookie will be emitted to a
3302 client which already has a valid one for the server which has
3303 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003304 it will be removed, unless the "preserve" option is also set. In
3305 "insert" mode, this will additionally remove cookies from the
3306 requests transmitted to the server, making the persistence
3307 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02003308 Note: it is highly recommended not to use "indirect" with
3309 "prefix", otherwise server cookie updates would not be sent to
3310 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003311
3312 nocache This option is recommended in conjunction with the insert mode
3313 when there is a cache between the client and HAProxy, as it
3314 ensures that a cacheable response will be tagged non-cacheable if
3315 a cookie needs to be inserted. This is important because if all
3316 persistence cookies are added on a cacheable home page for
3317 instance, then all customers will then fetch the page from an
3318 outer cache and will all share the same persistence cookie,
3319 leading to one server receiving much more traffic than others.
3320 See also the "insert" and "postonly" options.
3321
3322 postonly This option ensures that cookie insertion will only be performed
3323 on responses to POST requests. It is an alternative to the
3324 "nocache" option, because POST responses are not cacheable, so
3325 this ensures that the persistence cookie will never get cached.
3326 Since most sites do not need any sort of persistence before the
3327 first POST which generally is a login request, this is a very
3328 efficient method to optimize caching without risking to find a
3329 persistence cookie in the cache.
3330 See also the "insert" and "nocache" options.
3331
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003332 preserve This option may only be used with "insert" and/or "indirect". It
3333 allows the server to emit the persistence cookie itself. In this
3334 case, if a cookie is found in the response, haproxy will leave it
3335 untouched. This is useful in order to end persistence after a
3336 logout request for instance. For this, the server just has to
Davor Ocelice9ed2812017-12-25 17:49:28 +01003337 emit a cookie with an invalid value (e.g. empty) or with a date in
Willy Tarreauba4c5be2010-10-23 12:46:42 +02003338 the past. By combining this mechanism with the "disable-on-404"
3339 check option, it is possible to perform a completely graceful
3340 shutdown because users will definitely leave the server after
3341 they logout.
3342
Willy Tarreau4992dd22012-05-31 21:02:17 +02003343 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
3344 when a cookie is inserted. This attribute is used so that a
3345 user agent doesn't share the cookie with non-HTTP components.
3346 Please check RFC6265 for more information on this attribute.
3347
3348 secure This option tells haproxy to add a "Secure" cookie attribute when
3349 a cookie is inserted. This attribute is used so that a user agent
3350 never emits this cookie over non-secure channels, which means
3351 that a cookie learned with this flag will be presented only over
3352 SSL/TLS connections. Please check RFC6265 for more information on
3353 this attribute.
3354
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003355 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003356 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01003357 name. If the domain begins with a dot, the browser is allowed to
3358 use it for any host ending with that name. It is also possible to
3359 specify several domain names by invoking this option multiple
3360 times. Some browsers might have small limits on the number of
3361 domains, so be careful when doing that. For the record, sending
3362 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02003363
Willy Tarreau996a92c2010-10-13 19:30:47 +02003364 maxidle This option allows inserted cookies to be ignored after some idle
3365 time. It only works with insert-mode cookies. When a cookie is
3366 sent to the client, the date this cookie was emitted is sent too.
3367 Upon further presentations of this cookie, if the date is older
3368 than the delay indicated by the parameter (in seconds), it will
3369 be ignored. Otherwise, it will be refreshed if needed when the
3370 response is sent to the client. This is particularly useful to
3371 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003372 too long on the same server (e.g. after a farm size change). When
Willy Tarreau996a92c2010-10-13 19:30:47 +02003373 this option is set and a cookie has no date, it is always
3374 accepted, but gets refreshed in the response. This maintains the
3375 ability for admins to access their sites. Cookies that have a
3376 date in the future further than 24 hours are ignored. Doing so
3377 lets admins fix timezone issues without risking kicking users off
3378 the site.
3379
3380 maxlife This option allows inserted cookies to be ignored after some life
3381 time, whether they're in use or not. It only works with insert
3382 mode cookies. When a cookie is first sent to the client, the date
3383 this cookie was emitted is sent too. Upon further presentations
3384 of this cookie, if the date is older than the delay indicated by
3385 the parameter (in seconds), it will be ignored. If the cookie in
3386 the request has no date, it is accepted and a date will be set.
3387 Cookies that have a date in the future further than 24 hours are
3388 ignored. Doing so lets admins fix timezone issues without risking
3389 kicking users off the site. Contrary to maxidle, this value is
3390 not refreshed, only the first visit date counts. Both maxidle and
3391 maxlife may be used at the time. This is particularly useful to
3392 prevent users who never close their browsers from remaining for
Davor Ocelice9ed2812017-12-25 17:49:28 +01003393 too long on the same server (e.g. after a farm size change). This
Willy Tarreau996a92c2010-10-13 19:30:47 +02003394 is stronger than the maxidle method in that it forces a
3395 redispatch after some absolute delay.
3396
Olivier Houchard4e694042017-03-14 20:01:29 +01003397 dynamic Activate dynamic cookies. When used, a session cookie is
3398 dynamically created for each server, based on the IP and port
3399 of the server, and a secret key, specified in the
3400 "dynamic-cookie-key" backend directive.
3401 The cookie will be regenerated each time the IP address change,
3402 and is only generated for IPv4/IPv6.
3403
Willy Tarreau0ba27502007-12-24 16:55:16 +01003404 There can be only one persistence cookie per HTTP backend, and it can be
3405 declared in a defaults section. The value of the cookie will be the value
3406 indicated after the "cookie" keyword in a "server" statement. If no cookie
3407 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02003408
Willy Tarreau0ba27502007-12-24 16:55:16 +01003409 Examples :
3410 cookie JSESSIONID prefix
3411 cookie SRV insert indirect nocache
3412 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02003413 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01003414
Willy Tarreau294d0f02015-08-10 19:40:12 +02003415 See also : "balance source", "capture cookie", "server" and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01003416
Willy Tarreau983e01e2010-01-11 18:42:06 +01003417
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003418declare capture [ request | response ] len <length>
3419 Declares a capture slot.
3420 May be used in sections : defaults | frontend | listen | backend
3421 no | yes | yes | no
3422 Arguments:
3423 <length> is the length allowed for the capture.
3424
3425 This declaration is only available in the frontend or listen section, but the
3426 reserved slot can be used in the backends. The "request" keyword allocates a
3427 capture slot for use in the request, and "response" allocates a capture slot
3428 for use in the response.
3429
3430 See also: "capture-req", "capture-res" (sample converters),
Baptiste Assmann5ac425c2015-10-21 23:13:46 +02003431 "capture.req.hdr", "capture.res.hdr" (sample fetches),
Thierry FOURNIERa0a1b752015-05-26 17:44:32 +02003432 "http-request capture" and "http-response capture".
3433
3434
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003435default-server [param*]
3436 Change default options for a server in a backend
3437 May be used in sections : defaults | frontend | listen | backend
3438 yes | no | yes | yes
3439 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01003440 <param*> is a list of parameters for this server. The "default-server"
3441 keyword accepts an important number of options and has a complete
3442 section dedicated to it. Please refer to section 5 for more
3443 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003444
Willy Tarreau983e01e2010-01-11 18:42:06 +01003445 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01003446 default-server inter 1000 weight 13
3447
3448 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01003449
Willy Tarreau983e01e2010-01-11 18:42:06 +01003450
Willy Tarreau0ba27502007-12-24 16:55:16 +01003451default_backend <backend>
3452 Specify the backend to use when no "use_backend" rule has been matched.
3453 May be used in sections : defaults | frontend | listen | backend
3454 yes | yes | yes | no
3455 Arguments :
3456 <backend> is the name of the backend to use.
3457
3458 When doing content-switching between frontend and backends using the
3459 "use_backend" keyword, it is often useful to indicate which backend will be
3460 used when no rule has matched. It generally is the dynamic backend which
3461 will catch all undetermined requests.
3462
Willy Tarreau0ba27502007-12-24 16:55:16 +01003463 Example :
3464
3465 use_backend dynamic if url_dyn
3466 use_backend static if url_css url_img extension_img
3467 default_backend dynamic
3468
Willy Tarreau98d04852015-05-26 12:18:29 +02003469 See also : "use_backend"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003470
Willy Tarreau0ba27502007-12-24 16:55:16 +01003471
Baptiste Assmann27f51342013-10-09 06:51:49 +02003472description <string>
3473 Describe a listen, frontend or backend.
3474 May be used in sections : defaults | frontend | listen | backend
3475 no | yes | yes | yes
3476 Arguments : string
3477
3478 Allows to add a sentence to describe the related object in the HAProxy HTML
3479 stats page. The description will be printed on the right of the object name
3480 it describes.
3481 No need to backslash spaces in the <string> arguments.
3482
3483
Willy Tarreau0ba27502007-12-24 16:55:16 +01003484disabled
3485 Disable a proxy, frontend or backend.
3486 May be used in sections : defaults | frontend | listen | backend
3487 yes | yes | yes | yes
3488 Arguments : none
3489
3490 The "disabled" keyword is used to disable an instance, mainly in order to
3491 liberate a listening port or to temporarily disable a service. The instance
3492 will still be created and its configuration will be checked, but it will be
3493 created in the "stopped" state and will appear as such in the statistics. It
3494 will not receive any traffic nor will it send any health-checks or logs. It
3495 is possible to disable many instances at once by adding the "disabled"
3496 keyword in a "defaults" section.
3497
3498 See also : "enabled"
3499
3500
Willy Tarreau5ce94572010-06-07 14:35:41 +02003501dispatch <address>:<port>
3502 Set a default server address
3503 May be used in sections : defaults | frontend | listen | backend
3504 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003505 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02003506
3507 <address> is the IPv4 address of the default server. Alternatively, a
3508 resolvable hostname is supported, but this name will be resolved
3509 during start-up.
3510
3511 <ports> is a mandatory port specification. All connections will be sent
3512 to this port, and it is not permitted to use port offsets as is
3513 possible with normal servers.
3514
Willy Tarreau787aed52011-04-15 06:45:37 +02003515 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02003516 server can take the connection. In the past it was used to forward non
3517 persistent connections to an auxiliary load balancer. Due to its simple
3518 syntax, it has also been used for simple TCP relays. It is recommended not to
3519 use it for more clarity, and to use the "server" directive instead.
3520
3521 See also : "server"
3522
Olivier Houchard4e694042017-03-14 20:01:29 +01003523
3524dynamic-cookie-key <string>
3525 Set the dynamic cookie secret key for a backend.
3526 May be used in sections : defaults | frontend | listen | backend
3527 yes | no | yes | yes
3528 Arguments : The secret key to be used.
3529
3530 When dynamic cookies are enabled (see the "dynamic" directive for cookie),
Davor Ocelice9ed2812017-12-25 17:49:28 +01003531 a dynamic cookie is created for each server (unless one is explicitly
Olivier Houchard4e694042017-03-14 20:01:29 +01003532 specified on the "server" line), using a hash of the IP address of the
3533 server, the TCP port, and the secret key.
Davor Ocelice9ed2812017-12-25 17:49:28 +01003534 That way, we can ensure session persistence across multiple load-balancers,
Olivier Houchard4e694042017-03-14 20:01:29 +01003535 even if servers are dynamically added or removed.
Willy Tarreau5ce94572010-06-07 14:35:41 +02003536
Willy Tarreau0ba27502007-12-24 16:55:16 +01003537enabled
3538 Enable a proxy, frontend or backend.
3539 May be used in sections : defaults | frontend | listen | backend
3540 yes | yes | yes | yes
3541 Arguments : none
3542
3543 The "enabled" keyword is used to explicitly enable an instance, when the
3544 defaults has been set to "disabled". This is very rarely used.
3545
3546 See also : "disabled"
3547
3548
3549errorfile <code> <file>
3550 Return a file contents instead of errors generated by HAProxy
3551 May be used in sections : defaults | frontend | listen | backend
3552 yes | yes | yes | yes
3553 Arguments :
3554 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003555 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3556 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003557
3558 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003559 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01003560 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01003561 error pages, and to use absolute paths, since files are read
3562 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003563
3564 It is important to understand that this keyword is not meant to rewrite
3565 errors returned by the server, but errors detected and returned by HAProxy.
3566 This is why the list of supported errors is limited to a small set.
3567
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003568 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3569
Willy Tarreau0ba27502007-12-24 16:55:16 +01003570 The files are returned verbatim on the TCP socket. This allows any trick such
3571 as redirections to another URL or site, as well as tricks to clean cookies,
3572 force enable or disable caching, etc... The package provides default error
3573 files returning the same contents as default errors.
3574
Willy Tarreau59140a22009-02-22 12:02:30 +01003575 The files should not exceed the configured buffer size (BUFSIZE), which
3576 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
Davor Ocelice9ed2812017-12-25 17:49:28 +01003577 not to put any reference to local contents (e.g. images) in order to avoid
Willy Tarreau59140a22009-02-22 12:02:30 +01003578 loops between the client and HAProxy when all servers are down, causing an
3579 error to be returned instead of an image. For better HTTP compliance, it is
3580 recommended that all header lines end with CR-LF and not LF alone.
3581
Willy Tarreau0ba27502007-12-24 16:55:16 +01003582 The files are read at the same time as the configuration and kept in memory.
3583 For this reason, the errors continue to be returned even when the process is
3584 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01003585 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01003586 403 status code and interrogating a blocked URL.
3587
3588 See also : "errorloc", "errorloc302", "errorloc303"
3589
Willy Tarreau59140a22009-02-22 12:02:30 +01003590 Example :
3591 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau989222a2016-01-15 10:26:26 +01003592 errorfile 408 /dev/null # work around Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01003593 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
3594 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
3595
Willy Tarreau2769aa02007-12-27 18:26:09 +01003596
3597errorloc <code> <url>
3598errorloc302 <code> <url>
3599 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3600 May be used in sections : defaults | frontend | listen | backend
3601 yes | yes | yes | yes
3602 Arguments :
3603 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003604 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3605 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003606
3607 <url> it is the exact contents of the "Location" header. It may contain
3608 either a relative URI to an error page hosted on the same site,
3609 or an absolute URI designating an error page on another site.
3610 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003611 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003612
3613 It is important to understand that this keyword is not meant to rewrite
3614 errors returned by the server, but errors detected and returned by HAProxy.
3615 This is why the list of supported errors is limited to a small set.
3616
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003617 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3618
Willy Tarreau2769aa02007-12-27 18:26:09 +01003619 Note that both keyword return the HTTP 302 status code, which tells the
3620 client to fetch the designated URL using the same HTTP method. This can be
3621 quite problematic in case of non-GET methods such as POST, because the URL
3622 sent to the client might not be allowed for something other than GET. To
Willy Tarreau989222a2016-01-15 10:26:26 +01003623 work around this problem, please use "errorloc303" which send the HTTP 303
Willy Tarreau2769aa02007-12-27 18:26:09 +01003624 status code, indicating to the client that the URL must be fetched with a GET
3625 request.
3626
3627 See also : "errorfile", "errorloc303"
3628
3629
3630errorloc303 <code> <url>
3631 Return an HTTP redirection to a URL instead of errors generated by HAProxy
3632 May be used in sections : defaults | frontend | listen | backend
3633 yes | yes | yes | yes
3634 Arguments :
3635 <code> is the HTTP status code. Currently, HAProxy is capable of
Olivier Houchard51a76d82017-10-02 16:12:07 +02003636 generating codes 200, 400, 403, 405, 408, 425, 429, 500, 502,
3637 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003638
3639 <url> it is the exact contents of the "Location" header. It may contain
3640 either a relative URI to an error page hosted on the same site,
3641 or an absolute URI designating an error page on another site.
3642 Special care should be given to relative URIs to avoid redirect
Davor Ocelice9ed2812017-12-25 17:49:28 +01003643 loops if the URI itself may generate the same error (e.g. 500).
Willy Tarreau2769aa02007-12-27 18:26:09 +01003644
3645 It is important to understand that this keyword is not meant to rewrite
3646 errors returned by the server, but errors detected and returned by HAProxy.
3647 This is why the list of supported errors is limited to a small set.
3648
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003649 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
3650
Willy Tarreau2769aa02007-12-27 18:26:09 +01003651 Note that both keyword return the HTTP 303 status code, which tells the
3652 client to fetch the designated URL using the same HTTP GET method. This
3653 solves the usual problems associated with "errorloc" and the 302 code. It is
3654 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01003655 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003656
3657 See also : "errorfile", "errorloc", "errorloc302"
3658
3659
Simon Horman51a1cf62015-02-03 13:00:44 +09003660email-alert from <emailaddr>
3661 Declare the from email address to be used in both the envelope and header
Davor Ocelice9ed2812017-12-25 17:49:28 +01003662 of email alerts. This is the address that email alerts are sent from.
Simon Horman51a1cf62015-02-03 13:00:44 +09003663 May be used in sections: defaults | frontend | listen | backend
3664 yes | yes | yes | yes
3665
3666 Arguments :
3667
3668 <emailaddr> is the from email address to use when sending email alerts
3669
3670 Also requires "email-alert mailers" and "email-alert to" to be set
3671 and if so sending email alerts is enabled for the proxy.
3672
Simon Horman64e34162015-02-06 11:11:57 +09003673 See also : "email-alert level", "email-alert mailers",
Cyril Bonté307ee1e2015-09-28 23:16:06 +02003674 "email-alert myhostname", "email-alert to", section 3.6 about
3675 mailers.
Simon Horman64e34162015-02-06 11:11:57 +09003676
3677
3678email-alert level <level>
3679 Declare the maximum log level of messages for which email alerts will be
3680 sent. This acts as a filter on the sending of email alerts.
3681 May be used in sections: defaults | frontend | listen | backend
3682 yes | yes | yes | yes
3683
3684 Arguments :
3685
3686 <level> One of the 8 syslog levels:
3687 emerg alert crit err warning notice info debug
3688 The above syslog levels are ordered from lowest to highest.
3689
3690 By default level is alert
3691
3692 Also requires "email-alert from", "email-alert mailers" and
3693 "email-alert to" to be set and if so sending email alerts is enabled
3694 for the proxy.
3695
Simon Horman1421e212015-04-30 13:10:35 +09003696 Alerts are sent when :
3697
3698 * An un-paused server is marked as down and <level> is alert or lower
3699 * A paused server is marked as down and <level> is notice or lower
3700 * A server is marked as up or enters the drain state and <level>
3701 is notice or lower
3702 * "option log-health-checks" is enabled, <level> is info or lower,
3703 and a health check status update occurs
3704
Simon Horman64e34162015-02-06 11:11:57 +09003705 See also : "email-alert from", "email-alert mailers",
3706 "email-alert myhostname", "email-alert to",
Simon Horman51a1cf62015-02-03 13:00:44 +09003707 section 3.6 about mailers.
3708
3709
3710email-alert mailers <mailersect>
3711 Declare the mailers to be used when sending email alerts
3712 May be used in sections: defaults | frontend | listen | backend
3713 yes | yes | yes | yes
3714
3715 Arguments :
3716
3717 <mailersect> is the name of the mailers section to send email alerts.
3718
3719 Also requires "email-alert from" and "email-alert to" to be set
3720 and if so sending email alerts is enabled for the proxy.
3721
Simon Horman64e34162015-02-06 11:11:57 +09003722 See also : "email-alert from", "email-alert level", "email-alert myhostname",
3723 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003724
3725
3726email-alert myhostname <hostname>
3727 Declare the to hostname address to be used when communicating with
3728 mailers.
3729 May be used in sections: defaults | frontend | listen | backend
3730 yes | yes | yes | yes
3731
3732 Arguments :
3733
Baptiste Assmann738bad92015-12-21 15:27:53 +01003734 <hostname> is the hostname to use when communicating with mailers
Simon Horman51a1cf62015-02-03 13:00:44 +09003735
3736 By default the systems hostname is used.
3737
3738 Also requires "email-alert from", "email-alert mailers" and
3739 "email-alert to" to be set and if so sending email alerts is enabled
3740 for the proxy.
3741
Simon Horman64e34162015-02-06 11:11:57 +09003742 See also : "email-alert from", "email-alert level", "email-alert mailers",
3743 "email-alert to", section 3.6 about mailers.
Simon Horman51a1cf62015-02-03 13:00:44 +09003744
3745
3746email-alert to <emailaddr>
Davor Ocelice9ed2812017-12-25 17:49:28 +01003747 Declare both the recipient address in the envelope and to address in the
Simon Horman51a1cf62015-02-03 13:00:44 +09003748 header of email alerts. This is the address that email alerts are sent to.
3749 May be used in sections: defaults | frontend | listen | backend
3750 yes | yes | yes | yes
3751
3752 Arguments :
3753
3754 <emailaddr> is the to email address to use when sending email alerts
3755
3756 Also requires "email-alert mailers" and "email-alert to" to be set
3757 and if so sending email alerts is enabled for the proxy.
3758
Simon Horman64e34162015-02-06 11:11:57 +09003759 See also : "email-alert from", "email-alert level", "email-alert mailers",
Simon Horman51a1cf62015-02-03 13:00:44 +09003760 "email-alert myhostname", section 3.6 about mailers.
3761
3762
Willy Tarreau4de91492010-01-22 19:10:05 +01003763force-persist { if | unless } <condition>
3764 Declare a condition to force persistence on down servers
3765 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01003766 no | no | yes | yes
Willy Tarreau4de91492010-01-22 19:10:05 +01003767
3768 By default, requests are not dispatched to down servers. It is possible to
3769 force this using "option persist", but it is unconditional and redispatches
3770 to a valid server if "option redispatch" is set. That leaves with very little
3771 possibilities to force some requests to reach a server which is artificially
3772 marked down for maintenance operations.
3773
3774 The "force-persist" statement allows one to declare various ACL-based
3775 conditions which, when met, will cause a request to ignore the down status of
3776 a server and still try to connect to it. That makes it possible to start a
3777 server, still replying an error to the health checks, and run a specially
3778 configured browser to test the service. Among the handy methods, one could
3779 use a specific source IP address, or a specific cookie. The cookie also has
3780 the advantage that it can easily be added/removed on the browser from a test
3781 page. Once the service is validated, it is then possible to open the service
3782 to the world by returning a valid response to health checks.
3783
3784 The forced persistence is enabled when an "if" condition is met, or unless an
3785 "unless" condition is met. The final redispatch is always disabled when this
3786 is used.
3787
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003788 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02003789 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01003790
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003791
3792filter <name> [param*]
3793 Add the filter <name> in the filter list attached to the proxy.
3794 May be used in sections : defaults | frontend | listen | backend
3795 no | yes | yes | yes
3796 Arguments :
3797 <name> is the name of the filter. Officially supported filters are
3798 referenced in section 9.
3799
Tim Düsterhus4896c442016-11-29 02:15:19 +01003800 <param*> is a list of parameters accepted by the filter <name>. The
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003801 parsing of these parameters are the responsibility of the
Tim Düsterhus4896c442016-11-29 02:15:19 +01003802 filter. Please refer to the documentation of the corresponding
3803 filter (section 9) for all details on the supported parameters.
Christopher Fauletc3fe5332016-04-07 15:30:10 +02003804
3805 Multiple occurrences of the filter line can be used for the same proxy. The
3806 same filter can be referenced many times if needed.
3807
3808 Example:
3809 listen
3810 bind *:80
3811
3812 filter trace name BEFORE-HTTP-COMP
3813 filter compression
3814 filter trace name AFTER-HTTP-COMP
3815
3816 compression algo gzip
3817 compression offload
3818
3819 server srv1 192.168.0.1:80
3820
3821 See also : section 9.
3822
Willy Tarreau4de91492010-01-22 19:10:05 +01003823
Willy Tarreau2769aa02007-12-27 18:26:09 +01003824fullconn <conns>
3825 Specify at what backend load the servers will reach their maxconn
3826 May be used in sections : defaults | frontend | listen | backend
3827 yes | no | yes | yes
3828 Arguments :
3829 <conns> is the number of connections on the backend which will make the
3830 servers use the maximal number of connections.
3831
Willy Tarreau198a7442008-01-17 12:05:32 +01003832 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01003833 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01003834 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01003835 load. The server will then always accept at least <minconn> connections,
3836 never more than <maxconn>, and the limit will be on the ramp between both
3837 values when the backend has less than <conns> concurrent connections. This
3838 makes it possible to limit the load on the servers during normal loads, but
3839 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003840 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003841
Willy Tarreaufbb78422011-06-05 15:38:35 +02003842 Since it's hard to get this value right, haproxy automatically sets it to
3843 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01003844 backend (based on "use_backend" and "default_backend" rules). That way it's
3845 safe to leave it unset. However, "use_backend" involving dynamic names are
3846 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02003847
Willy Tarreau2769aa02007-12-27 18:26:09 +01003848 Example :
3849 # The servers will accept between 100 and 1000 concurrent connections each
3850 # and the maximum of 1000 will be reached when the backend reaches 10000
3851 # connections.
3852 backend dynamic
3853 fullconn 10000
3854 server srv1 dyn1:80 minconn 100 maxconn 1000
3855 server srv2 dyn2:80 minconn 100 maxconn 1000
3856
3857 See also : "maxconn", "server"
3858
3859
3860grace <time>
3861 Maintain a proxy operational for some time after a soft stop
3862 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01003863 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01003864 Arguments :
3865 <time> is the time (by default in milliseconds) for which the instance
3866 will remain operational with the frontend sockets still listening
3867 when a soft-stop is received via the SIGUSR1 signal.
3868
3869 This may be used to ensure that the services disappear in a certain order.
3870 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003871 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01003872 needed by the equipment to detect the failure.
3873
3874 Note that currently, there is very little benefit in using this parameter,
3875 and it may in fact complicate the soft-reconfiguration process more than
3876 simplify it.
3877
Willy Tarreau0ba27502007-12-24 16:55:16 +01003878
Andrew Rodland17be45e2016-10-25 17:04:12 -04003879hash-balance-factor <factor>
3880 Specify the balancing factor for bounded-load consistent hashing
3881 May be used in sections : defaults | frontend | listen | backend
3882 yes | no | no | yes
3883 Arguments :
3884 <factor> is the control for the maximum number of concurrent requests to
3885 send to a server, expressed as a percentage of the average number
Frédéric Lécaille93d33162019-03-06 09:35:59 +01003886 of concurrent requests across all of the active servers.
Andrew Rodland17be45e2016-10-25 17:04:12 -04003887
3888 Specifying a "hash-balance-factor" for a server with "hash-type consistent"
3889 enables an algorithm that prevents any one server from getting too many
3890 requests at once, even if some hash buckets receive many more requests than
3891 others. Setting <factor> to 0 (the default) disables the feature. Otherwise,
3892 <factor> is a percentage greater than 100. For example, if <factor> is 150,
3893 then no server will be allowed to have a load more than 1.5 times the average.
3894 If server weights are used, they will be respected.
3895
3896 If the first-choice server is disqualified, the algorithm will choose another
3897 server based on the request hash, until a server with additional capacity is
3898 found. A higher <factor> allows more imbalance between the servers, while a
3899 lower <factor> means that more servers will be checked on average, affecting
3900 performance. Reasonable values are from 125 to 200.
3901
Willy Tarreau760e81d2018-05-03 07:20:40 +02003902 This setting is also used by "balance random" which internally relies on the
3903 consistent hashing mechanism.
3904
Andrew Rodland17be45e2016-10-25 17:04:12 -04003905 See also : "balance" and "hash-type".
3906
3907
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003908hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003909 Specify a method to use for mapping hashes to servers
3910 May be used in sections : defaults | frontend | listen | backend
3911 yes | no | yes | yes
3912 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04003913 <method> is the method used to select a server from the hash computed by
3914 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003915
Bhaskar98634f02013-10-29 23:30:51 -04003916 map-based the hash table is a static array containing all alive servers.
3917 The hashes will be very smooth, will consider weights, but
3918 will be static in that weight changes while a server is up
3919 will be ignored. This means that there will be no slow start.
3920 Also, since a server is selected by its position in the array,
3921 most mappings are changed when the server count changes. This
3922 means that when a server goes up or down, or when a server is
3923 added to a farm, most connections will be redistributed to
3924 different servers. This can be inconvenient with caches for
3925 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01003926
Bhaskar98634f02013-10-29 23:30:51 -04003927 consistent the hash table is a tree filled with many occurrences of each
3928 server. The hash key is looked up in the tree and the closest
3929 server is chosen. This hash is dynamic, it supports changing
3930 weights while the servers are up, so it is compatible with the
3931 slow start feature. It has the advantage that when a server
3932 goes up or down, only its associations are moved. When a
3933 server is added to the farm, only a few part of the mappings
3934 are redistributed, making it an ideal method for caches.
3935 However, due to its principle, the distribution will never be
3936 very smooth and it may sometimes be necessary to adjust a
3937 server's weight or its ID to get a more balanced distribution.
3938 In order to get the same distribution on multiple load
3939 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003940 same IDs. Note: consistent hash uses sdbm and avalanche if no
3941 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04003942
3943 <function> is the hash function to be used :
3944
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003945 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04003946 reimplementation of ndbm) database library. It was found to do
3947 well in scrambling bits, causing better distribution of the keys
3948 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003949 function with good distribution, unless the total server weight
3950 is a multiple of 64, in which case applying the avalanche
3951 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04003952
3953 djb2 this function was first proposed by Dan Bernstein many years ago
3954 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003955 function provides a better distribution than sdbm. It generally
3956 works well with text-based inputs though it can perform extremely
3957 poorly with numeric-only input or when the total server weight is
3958 a multiple of 33, unless the avalanche modifier is also used.
3959
Willy Tarreaua0f42712013-11-14 14:30:35 +01003960 wt6 this function was designed for haproxy while testing other
3961 functions in the past. It is not as smooth as the other ones, but
3962 is much less sensible to the input data set or to the number of
3963 servers. It can make sense as an alternative to sdbm+avalanche or
3964 djb2+avalanche for consistent hashing or when hashing on numeric
3965 data such as a source IP address or a visitor identifier in a URL
3966 parameter.
3967
Willy Tarreau324f07f2015-01-20 19:44:50 +01003968 crc32 this is the most common CRC32 implementation as used in Ethernet,
3969 gzip, PNG, etc. It is slower than the other ones but may provide
3970 a better distribution or less predictable results especially when
3971 used on strings.
3972
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05003973 <modifier> indicates an optional method applied after hashing the key :
3974
3975 avalanche This directive indicates that the result from the hash
3976 function above should not be used in its raw form but that
3977 a 4-byte full avalanche hash must be applied first. The
3978 purpose of this step is to mix the resulting bits from the
3979 previous hash in order to avoid any undesired effect when
3980 the input contains some limited values or when the number of
3981 servers is a multiple of one of the hash's components (64
3982 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
3983 result less predictable, but it's also not as smooth as when
3984 using the original function. Some testing might be needed
3985 with some workloads. This hash is one of the many proposed
3986 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003987
Bhaskar98634f02013-10-29 23:30:51 -04003988 The default hash type is "map-based" and is recommended for most usages. The
3989 default function is "sdbm", the selection of a function should be based on
3990 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003991
Andrew Rodland17be45e2016-10-25 17:04:12 -04003992 See also : "balance", "hash-balance-factor", "server"
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02003993
3994
Willy Tarreau0ba27502007-12-24 16:55:16 +01003995http-check disable-on-404
3996 Enable a maintenance mode upon HTTP/404 response to health-checks
3997 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01003998 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01003999 Arguments : none
4000
4001 When this option is set, a server which returns an HTTP code 404 will be
4002 excluded from further load-balancing, but will still receive persistent
4003 connections. This provides a very convenient method for Web administrators
4004 to perform a graceful shutdown of their servers. It is also important to note
4005 that a server which is detected as failed while it was in this mode will not
4006 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
4007 will immediately be reinserted into the farm. The status on the stats page
4008 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01004009 option only works in conjunction with the "httpchk" option. If this option
4010 is used with "http-check expect", then it has precedence over it so that 404
4011 responses will still be considered as soft-stop.
4012
4013 See also : "option httpchk", "http-check expect"
4014
4015
4016http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004017 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01004018 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02004019 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01004020 Arguments :
4021 <match> is a keyword indicating how to look for a specific pattern in the
4022 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004023 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01004024 exclamation mark ("!") to negate the match. Spaces are allowed
4025 between the exclamation mark and the keyword. See below for more
4026 details on the supported keywords.
4027
4028 <pattern> is the pattern to look for. It may be a string or a regular
4029 expression. If the pattern contains spaces, they must be escaped
4030 with the usual backslash ('\').
4031
4032 By default, "option httpchk" considers that response statuses 2xx and 3xx
4033 are valid, and that others are invalid. When "http-check expect" is used,
4034 it defines what is considered valid or invalid. Only one "http-check"
4035 statement is supported in a backend. If a server fails to respond or times
4036 out, the check obviously fails. The available matches are :
4037
4038 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004039 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004040 response's status code is exactly this string. If the
4041 "status" keyword is prefixed with "!", then the response
4042 will be considered invalid if the status code matches.
4043
4044 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004045 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004046 response's status code matches the expression. If the
4047 "rstatus" keyword is prefixed with "!", then the response
4048 will be considered invalid if the status code matches.
4049 This is mostly used to check for multiple codes.
4050
4051 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004052 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004053 response's body contains this exact string. If the
4054 "string" keyword is prefixed with "!", then the response
4055 will be considered invalid if the body contains this
4056 string. This can be used to look for a mandatory word at
4057 the end of a dynamic page, or to detect a failure when a
Davor Ocelice9ed2812017-12-25 17:49:28 +01004058 specific error appears on the check page (e.g. a stack
Willy Tarreaubd741542010-03-16 18:46:54 +01004059 trace).
4060
4061 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004062 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01004063 response's body matches this expression. If the "rstring"
4064 keyword is prefixed with "!", then the response will be
4065 considered invalid if the body matches the expression.
4066 This can be used to look for a mandatory word at the end
4067 of a dynamic page, or to detect a failure when a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01004068 error appears on the check page (e.g. a stack trace).
Willy Tarreaubd741542010-03-16 18:46:54 +01004069
4070 It is important to note that the responses will be limited to a certain size
4071 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
4072 Thus, too large responses may not contain the mandatory pattern when using
4073 "string" or "rstring". If a large response is absolutely required, it is
4074 possible to change the default max size by setting the global variable.
4075 However, it is worth keeping in mind that parsing very large responses can
4076 waste some CPU cycles, especially when regular expressions are used, and that
4077 it is always better to focus the checks on smaller resources.
4078
Cyril Bonté32602d22015-01-30 00:07:07 +01004079 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
4080 will automatically append a "Connection: close" header, meaning that this
4081 header should not be present in the request provided by "option httpchk".
4082
Willy Tarreaubd741542010-03-16 18:46:54 +01004083 Last, if "http-check expect" is combined with "http-check disable-on-404",
4084 then this last one has precedence when the server responds with 404.
4085
4086 Examples :
4087 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004088 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01004089
4090 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004091 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01004092
4093 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01004094 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01004095
4096 # check that we have a correct hexadecimal tag before /html
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03004097 http-check expect rstring <!--tag:[0-9a-f]*--></html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01004098
Willy Tarreaubd741542010-03-16 18:46:54 +01004099 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01004100
4101
Willy Tarreauef781042010-01-27 11:53:01 +01004102http-check send-state
4103 Enable emission of a state header with HTTP health checks
4104 May be used in sections : defaults | frontend | listen | backend
4105 yes | no | yes | yes
4106 Arguments : none
4107
4108 When this option is set, haproxy will systematically send a special header
4109 "X-Haproxy-Server-State" with a list of parameters indicating to each server
4110 how they are seen by haproxy. This can be used for instance when a server is
4111 manipulated without access to haproxy and the operator needs to know whether
4112 haproxy still sees it up or not, or if the server is the last one in a farm.
4113
4114 The header is composed of fields delimited by semi-colons, the first of which
4115 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
4116 checks on the total number before transition, just as appears in the stats
4117 interface. Next headers are in the form "<variable>=<value>", indicating in
4118 no specific order some values available in the stats interface :
Joseph Lynch514061c2015-01-15 17:52:59 -08004119 - a variable "address", containing the address of the backend server.
4120 This corresponds to the <address> field in the server declaration. For
4121 unix domain sockets, it will read "unix".
4122
4123 - a variable "port", containing the port of the backend server. This
4124 corresponds to the <port> field in the server declaration. For unix
4125 domain sockets, it will read "unix".
4126
Willy Tarreauef781042010-01-27 11:53:01 +01004127 - a variable "name", containing the name of the backend followed by a slash
4128 ("/") then the name of the server. This can be used when a server is
4129 checked in multiple backends.
4130
4131 - a variable "node" containing the name of the haproxy node, as set in the
4132 global "node" variable, otherwise the system's hostname if unspecified.
4133
4134 - a variable "weight" indicating the weight of the server, a slash ("/")
4135 and the total weight of the farm (just counting usable servers). This
4136 helps to know if other servers are available to handle the load when this
4137 one fails.
4138
4139 - a variable "scur" indicating the current number of concurrent connections
4140 on the server, followed by a slash ("/") then the total number of
4141 connections on all servers of the same backend.
4142
4143 - a variable "qcur" indicating the current number of requests in the
4144 server's queue.
4145
4146 Example of a header received by the application server :
4147 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
4148 scur=13/22; qcur=0
4149
4150 See also : "option httpchk", "http-check disable-on-404"
4151
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004152
4153http-request <action> [options...] [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004154 Access control for Layer 7 requests
4155
4156 May be used in sections: defaults | frontend | listen | backend
4157 no | yes | yes | yes
4158
Willy Tarreau20b0de52012-12-24 15:45:22 +01004159 The http-request statement defines a set of rules which apply to layer 7
4160 processing. The rules are evaluated in their declaration order when they are
4161 met in a frontend, listen or backend section. Any rule may optionally be
4162 followed by an ACL-based condition, in which case it will only be evaluated
4163 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004164
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004165 The first keyword is the rule's action. The supported actions are described
4166 below.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004167
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004168 There is no limit to the number of http-request statements per instance.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004169
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004170 It is important to know that http-request rules are processed very early in
4171 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
4172 or "reqadd" rules. That way, headers added by "add-header"/"set-header" are
4173 visible by almost all further ACL rules.
Willy Tarreau53275e82017-11-24 07:52:01 +01004174
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004175 Using "reqadd"/"reqdel"/"reqrep" to manipulate request headers is discouraged
4176 in newer versions (>= 1.5). But if you need to use regular expression to
4177 delete headers, you can still use "reqdel". Also please use
4178 "http-request deny/allow/tarpit" instead of "reqdeny"/"reqpass"/"reqtarpit".
Willy Tarreauccbcc372012-12-27 12:37:57 +01004179
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004180 Example:
4181 acl nagios src 192.168.129.3
4182 acl local_net src 192.168.0.0/16
4183 acl auth_ok http_auth(L1)
Willy Tarreau20b0de52012-12-24 15:45:22 +01004184
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004185 http-request allow if nagios
4186 http-request allow if local_net auth_ok
4187 http-request auth realm Gimme if local_net auth_ok
4188 http-request deny
Willy Tarreau81499eb2012-12-27 12:19:02 +01004189
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004190 Example:
4191 acl key req.hdr(X-Add-Acl-Key) -m found
4192 acl add path /addacl
4193 acl del path /delacl
Willy Tarreau20b0de52012-12-24 15:45:22 +01004194
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004195 acl myhost hdr(Host) -f myhost.lst
Willy Tarreau20b0de52012-12-24 15:45:22 +01004196
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004197 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
4198 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004199
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004200 Example:
4201 acl value req.hdr(X-Value) -m found
4202 acl setmap path /setmap
4203 acl delmap path /delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004204
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004205 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004206
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004207 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
4208 http-request del-map(map.lst) %[src] if delmap
Sasha Pachev218f0642014-06-16 12:05:59 -06004209
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004210 See also : "stats http-request", section 3.4 about userlists and section 7
4211 about ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004212
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004213http-request add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004214
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004215 This is used to add a new entry into an ACL. The ACL must be loaded from a
4216 file (even a dummy empty file). The file name of the ACL to be updated is
4217 passed between parentheses. It takes one argument: <key fmt>, which follows
4218 log-format rules, to collect content of the new entry. It performs a lookup
4219 in the ACL before insertion, to avoid duplicated (or more) values. This
4220 lookup is done by a linear search and can be expensive with large lists!
4221 It is the equivalent of the "add acl" command from the stats socket, but can
4222 be triggered by an HTTP request.
Sasha Pachev218f0642014-06-16 12:05:59 -06004223
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004224http-request add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004225
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004226 This appends an HTTP header field whose name is specified in <name> and
4227 whose value is defined by <fmt> which follows the log-format rules (see
4228 Custom Log Format in section 8.2.4). This is particularly useful to pass
4229 connection-specific information to the server (e.g. the client's SSL
4230 certificate), or to combine several headers into one. This rule is not
4231 final, so it is possible to add other similar rules. Note that header
4232 addition is performed immediately, so one rule might reuse the resulting
4233 header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004234
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004235http-request allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004236
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004237 This stops the evaluation of the rules and lets the request pass the check.
4238 No further "http-request" rules are evaluated.
Sasha Pachev218f0642014-06-16 12:05:59 -06004239
Sasha Pachev218f0642014-06-16 12:05:59 -06004240
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004241http-request auth [realm <realm>] [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004242
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004243 This stops the evaluation of the rules and immediately responds with an
4244 HTTP 401 or 407 error code to invite the user to present a valid user name
4245 and password. No further "http-request" rules are evaluated. An optional
4246 "realm" parameter is supported, it sets the authentication realm that is
4247 returned with the response (typically the application's name).
Sasha Pachev218f0642014-06-16 12:05:59 -06004248
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004249 Example:
4250 acl auth_ok http_auth_group(L1) G1
4251 http-request auth unless auth_ok
Sasha Pachev218f0642014-06-16 12:05:59 -06004252
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004253http-request cache-use <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004254
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004255 See section 10.2 about cache setup.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004256
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004257http-request capture <sample> [ len <length> | id <id> ]
4258 [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004259
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004260 This captures sample expression <sample> from the request buffer, and
4261 converts it to a string of at most <len> characters. The resulting string is
4262 stored into the next request "capture" slot, so it will possibly appear next
4263 to some captured HTTP headers. It will then automatically appear in the logs,
4264 and it will be possible to extract it using sample fetch rules to feed it
4265 into headers or anything. The length should be limited given that this size
4266 will be allocated for each capture during the whole session life.
4267 Please check section 7.3 (Fetching samples) and "capture request header" for
4268 more information.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004269
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004270 If the keyword "id" is used instead of "len", the action tries to store the
4271 captured string in a previously declared capture slot. This is useful to run
4272 captures in backends. The slot id can be declared by a previous directive
4273 "http-request capture" or with the "declare capture" keyword. If the slot
4274 <id> doesn't exist, then HAProxy fails parsing the configuration to prevent
4275 unexpected behavior at run time.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004276
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004277http-request del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004278
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004279 This is used to delete an entry from an ACL. The ACL must be loaded from a
4280 file (even a dummy empty file). The file name of the ACL to be updated is
4281 passed between parentheses. It takes one argument: <key fmt>, which follows
4282 log-format rules, to collect content of the entry to delete.
4283 It is the equivalent of the "del acl" command from the stats socket, but can
4284 be triggered by an HTTP request.
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01004285
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004286http-request del-header <name> [ { if | unless } <condition> ]
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004287
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004288 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004289
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004290http-request del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004291
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004292 This is used to delete an entry from a MAP. The MAP must be loaded from a
4293 file (even a dummy empty file). The file name of the MAP to be updated is
4294 passed between parentheses. It takes one argument: <key fmt>, which follows
4295 log-format rules, to collect content of the entry to delete.
4296 It takes one argument: "file name" It is the equivalent of the "del map"
4297 command from the stats socket, but can be triggered by an HTTP request.
Willy Tarreau51347ed2013-06-11 19:34:13 +02004298
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004299http-request deny [deny_status <status>] [ { if | unless } <condition> ]
Patrick Hemmer268a7072018-05-11 12:52:31 -04004300
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004301 This stops the evaluation of the rules and immediately rejects the request
4302 and emits an HTTP 403 error, or optionally the status code specified as an
4303 argument to "deny_status". The list of permitted status codes is limited to
4304 those that can be overridden by the "errorfile" directive.
4305 No further "http-request" rules are evaluated.
Patrick Hemmer268a7072018-05-11 12:52:31 -04004306
Olivier Houchard602bf7d2019-05-10 13:59:15 +02004307http-request disable-l7-retry [ { if | unless } <condition> ]
4308 This disables any attempt to retry the request if it fails for any other
4309 reason than a connection failure. This can be useful for example to make
4310 sure POST requests aren't retried on failure.
4311
Baptiste Assmann333939c2019-01-21 08:34:50 +01004312http-request do-resolve(<var>,<resolvers>,[ipv4,ipv6]) <expr> :
4313
4314 This action performs a DNS resolution of the output of <expr> and stores
4315 the result in the variable <var>. It uses the DNS resolvers section
4316 pointed by <resolvers>.
4317 It is possible to choose a resolution preference using the optional
4318 arguments 'ipv4' or 'ipv6'.
4319 When performing the DNS resolution, the client side connection is on
4320 pause waiting till the end of the resolution.
4321 If an IP address can be found, it is stored into <var>. If any kind of
4322 error occurs, then <var> is not set.
4323 One can use this action to discover a server IP address at run time and
4324 based on information found in the request (IE a Host header).
4325 If this action is used to find the server's IP address (using the
4326 "set-dst" action), then the server IP address in the backend must be set
4327 to 0.0.0.0.
4328
4329 Example:
4330 resolvers mydns
4331 nameserver local 127.0.0.53:53
4332 nameserver google 8.8.8.8:53
4333 timeout retry 1s
4334 hold valid 10s
4335 hold nx 3s
4336 hold other 3s
4337 hold obsolete 0s
4338 accepted_payload_size 8192
4339
4340 frontend fe
4341 bind 10.42.0.1:80
4342 http-request do-resolve(txn.myip,mydns,ipv4) hdr(Host),lower
4343 http-request capture var(txn.myip) len 40
4344
4345 # return 503 when the variable is not set,
4346 # which mean DNS resolution error
4347 use_backend b_503 unless { var(txn.myip) -m found }
4348
4349 default_backend be
4350
4351 backend b_503
4352 # dummy backend used to return 503.
4353 # one can use the errorfile directive to send a nice
4354 # 503 error page to end users
4355
4356 backend be
4357 # rule to prevent HAProxy from reconnecting to services
4358 # on the local network (forged DNS name used to scan the network)
4359 http-request deny if { var(txn.myip) -m ip 127.0.0.0/8 10.0.0.0/8 }
4360 http-request set-dst var(txn.myip)
4361 server clear 0.0.0.0:0
4362
4363 NOTE: Don't forget to set the "protection" rules to ensure HAProxy won't
4364 be used to scan the network or worst won't loop over itself...
4365
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004366http-request early-hint <name> <fmt> [ { if | unless } <condition> ]
4367
4368 This is used to build an HTTP 103 Early Hints response prior to any other one.
4369 This appends an HTTP header field to this response whose name is specified in
4370 <name> and whose value is defined by <fmt> which follows the log-format rules
4371 (see Custom Log Format in section 8.2.4). This is particularly useful to pass
Frédéric Lécaille3aac1062018-11-13 09:42:13 +01004372 to the client some Link headers to preload resources required to render the
4373 HTML documents.
Frédéric Lécaille06f5b642018-11-12 11:01:10 +01004374
4375 See RFC 8297 for more information.
4376
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004377http-request redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004378
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004379 This performs an HTTP redirection based on a redirect rule. This is exactly
4380 the same as the "redirect" statement except that it inserts a redirect rule
4381 which can be processed in the middle of other "http-request" rules and that
4382 these rules use the "log-format" strings. See the "redirect" keyword for the
4383 rule's syntax.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004384
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004385http-request reject [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004386
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004387 This stops the evaluation of the rules and immediately closes the connection
4388 without sending any response. It acts similarly to the
4389 "tcp-request content reject" rules. It can be useful to force an immediate
4390 connection closure on HTTP/2 connections.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004391
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004392http-request replace-header <name> <match-regex> <replace-fmt>
4393 [ { if | unless } <condition> ]
Willy Tarreaua9083d02015-05-08 15:27:59 +02004394
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004395 This matches the regular expression in all occurrences of header field
4396 <name> according to <match-regex>, and replaces them with the
4397 <replace-fmt> argument. Format characters are allowed in replace-fmt and
4398 work like in <fmt> arguments in "http-request add-header". The match is
4399 only case-sensitive. It is important to understand that this action only
4400 considers whole header lines, regardless of the number of values they may
4401 contain. This usage is suited to headers naturally containing commas in
4402 their value, such as If-Modified-Since and so on.
Thierry FOURNIER82bf70d2015-05-26 17:58:29 +02004403
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004404 Example:
4405 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
William Lallemand86d0df02017-11-24 21:36:45 +01004406
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004407 # applied to:
4408 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
Willy Tarreau09448f72014-06-25 18:12:15 +02004409
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004410 # outputs:
4411 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
Willy Tarreau09448f72014-06-25 18:12:15 +02004412
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004413 # assuming the backend IP is 192.168.1.20
Willy Tarreau09448f72014-06-25 18:12:15 +02004414
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004415http-request replace-value <name> <match-regex> <replace-fmt>
4416 [ { if | unless } <condition> ]
Willy Tarreau09448f72014-06-25 18:12:15 +02004417
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004418 This works like "replace-header" except that it matches the regex against
4419 every comma-delimited value of the header field <name> instead of the
4420 entire header. This is suited for all headers which are allowed to carry
4421 more than one value. An example could be the Accept header.
Willy Tarreau09448f72014-06-25 18:12:15 +02004422
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004423 Example:
4424 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004425
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004426 # applied to:
4427 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004428
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004429 # outputs:
4430 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004431
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004432http-request sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4433http-request sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004434
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004435 This actions increments the GPC0 or GPC1 counter according with the sticky
4436 counter designated by <sc-id>. If an error occurs, this action silently fails
4437 and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004438
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004439http-request sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004440
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004441 This action sets the GPT0 tag according to the sticky counter designated by
4442 <sc-id> and the value of <int>. The expected result is a boolean. If an error
4443 occurs, this action silently fails and the actions evaluation continues.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004444
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004445http-request set-dst <expr> [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004446
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004447 This is used to set the destination IP address to the value of specified
4448 expression. Useful when a proxy in front of HAProxy rewrites destination IP,
4449 but provides the correct IP in a HTTP header; or you want to mask the IP for
4450 privacy. If you want to connect to the new address/port, use '0.0.0.0:0' as a
4451 server address in the backend.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004452
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004453 Arguments:
4454 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4455 by some converters.
Christopher Faulet85d79c92016-11-09 16:54:56 +01004456
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004457 Example:
4458 http-request set-dst hdr(x-dst)
4459 http-request set-dst dst,ipmask(24)
Christopher Faulet85d79c92016-11-09 16:54:56 +01004460
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004461 When possible, set-dst preserves the original destination port as long as the
4462 address family allows it, otherwise the destination port is set to 0.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004463
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004464http-request set-dst-port <expr> [ { if | unless } <condition> ]
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004465
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004466 This is used to set the destination port address to the value of specified
4467 expression. If you want to connect to the new address/port, use '0.0.0.0:0'
4468 as a server address in the backend.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004469
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004470 Arguments:
4471 <expr> Is a standard HAProxy expression formed by a sample-fetch
4472 followed by some converters.
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004473
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004474 Example:
4475 http-request set-dst-port hdr(x-port)
4476 http-request set-dst-port int(4000)
Adis Nezirovic2fbcafc2015-07-06 15:44:30 +02004477
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004478 When possible, set-dst-port preserves the original destination address as
4479 long as the address family supports a port, otherwise it forces the
4480 destination address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02004481
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004482http-request set-header <name> <fmt> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004483
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004484 This does the same as "http-request add-header" except that the header name
4485 is first removed if it existed. This is useful when passing security
4486 information to the server, where the header must not be manipulated by
4487 external users. Note that the new value is computed before the removal so it
4488 is possible to concatenate a value to an existing header.
William Lallemand44be6402016-05-25 01:51:35 +02004489
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004490 Example:
4491 http-request set-header X-Haproxy-Current-Date %T
4492 http-request set-header X-SSL %[ssl_fc]
4493 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id,hex]
4494 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
4495 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
4496 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
4497 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
4498 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
4499 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
William Lallemand44be6402016-05-25 01:51:35 +02004500
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004501http-request set-log-level <level> [ { if | unless } <condition> ]
William Lallemand44be6402016-05-25 01:51:35 +02004502
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004503 This is used to change the log level of the current request when a certain
4504 condition is met. Valid levels are the 8 syslog levels (see the "log"
4505 keyword) plus the special level "silent" which disables logging for this
4506 request. This rule is not final so the last matching rule wins. This rule
4507 can be useful to disable health checks coming from another equipment.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004508
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004509http-request set-map(<file-name>) <key fmt> <value fmt>
4510 [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004511
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004512 This is used to add a new entry into a MAP. The MAP must be loaded from a
4513 file (even a dummy empty file). The file name of the MAP to be updated is
4514 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
4515 log-format rules, used to collect MAP key, and <value fmt>, which follows
4516 log-format rules, used to collect content for the new entry.
4517 It performs a lookup in the MAP before insertion, to avoid duplicated (or
4518 more) values. This lookup is done by a linear search and can be expensive
4519 with large lists! It is the equivalent of the "set map" command from the
4520 stats socket, but can be triggered by an HTTP request.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004521
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004522http-request set-mark <mark> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004523
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004524 This is used to set the Netfilter MARK on all packets sent to the client to
4525 the value passed in <mark> on platforms which support it. This value is an
4526 unsigned 32 bit value which can be matched by netfilter and by the routing
4527 table. It can be expressed both in decimal or hexadecimal format (prefixed by
4528 "0x"). This can be useful to force certain packets to take a different route
4529 (for example a cheaper network path for bulk downloads). This works on Linux
4530 kernels 2.6.32 and above and requires admin privileges.
Willy Tarreau00005ce2016-10-21 15:07:45 +02004531
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004532http-request set-method <fmt> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004533
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004534 This rewrites the request method with the result of the evaluation of format
4535 string <fmt>. There should be very few valid reasons for having to do so as
4536 this is more likely to break something than to fix it.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004537
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004538http-request set-nice <nice> [ { if | unless } <condition> ]
William Lallemand13e9b0c2016-05-25 02:34:07 +02004539
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004540 This sets the "nice" factor of the current request being processed. It only
4541 has effect against the other requests being processed at the same time.
4542 The default value is 0, unless altered by the "nice" setting on the "bind"
4543 line. The accepted range is -1024..1024. The higher the value, the nicest
4544 the request will be. Lower values will make the request more important than
4545 other ones. This can be useful to improve the speed of some requests, or
4546 lower the priority of non-important requests. Using this setting without
4547 prior experimentation can cause some major slowdown.
William Lallemand13e9b0c2016-05-25 02:34:07 +02004548
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004549http-request set-path <fmt> [ { if | unless } <condition> ]
Willy Tarreau00005ce2016-10-21 15:07:45 +02004550
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004551 This rewrites the request path with the result of the evaluation of format
4552 string <fmt>. The query string, if any, is left intact. If a scheme and
4553 authority is found before the path, they are left intact as well. If the
4554 request doesn't have a path ("*"), this one is replaced with the format.
4555 This can be used to prepend a directory component in front of a path for
4556 example. See also "http-request set-query" and "http-request set-uri".
Willy Tarreau2d392c22015-08-24 01:43:45 +02004557
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004558 Example :
4559 # prepend the host name before the path
4560 http-request set-path /%[hdr(host)]%[path]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004561
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004562http-request set-priority-class <expr> [ { if | unless } <condition> ]
Olivier Houchardccaa7de2017-10-02 11:51:03 +02004563
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004564 This is used to set the queue priority class of the current request.
4565 The value must be a sample expression which converts to an integer in the
4566 range -2047..2047. Results outside this range will be truncated.
4567 The priority class determines the order in which queued requests are
4568 processed. Lower values have higher priority.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004569
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004570http-request set-priority-offset <expr> [ { if | unless } <condition> ]
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004571
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004572 This is used to set the queue priority timestamp offset of the current
4573 request. The value must be a sample expression which converts to an integer
4574 in the range -524287..524287. Results outside this range will be truncated.
4575 When a request is queued, it is ordered first by the priority class, then by
4576 the current timestamp adjusted by the given offset in milliseconds. Lower
4577 values have higher priority.
4578 Note that the resulting timestamp is is only tracked with enough precision
4579 for 524,287ms (8m44s287ms). If the request is queued long enough to where the
4580 adjusted timestamp exceeds this value, it will be misidentified as highest
4581 priority. Thus it is important to set "timeout queue" to a value, where when
4582 combined with the offset, does not exceed this limit.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004583
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004584http-request set-query <fmt> [ { if | unless } <condition> ]
Willy Tarreau20b0de52012-12-24 15:45:22 +01004585
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004586 This rewrites the request's query string which appears after the first
4587 question mark ("?") with the result of the evaluation of format string <fmt>.
4588 The part prior to the question mark is left intact. If the request doesn't
4589 contain a question mark and the new value is not empty, then one is added at
4590 the end of the URI, followed by the new value. If a question mark was
4591 present, it will never be removed even if the value is empty. This can be
4592 used to add or remove parameters from the query string.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08004593
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004594 See also "http-request set-query" and "http-request set-uri".
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004595
4596 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004597 # replace "%3D" with "=" in the query string
4598 http-request set-query %[query,regsub(%3D,=,g)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004599
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004600http-request set-src <expr> [ { if | unless } <condition> ]
4601 This is used to set the source IP address to the value of specified
4602 expression. Useful when a proxy in front of HAProxy rewrites source IP, but
4603 provides the correct IP in a HTTP header; or you want to mask source IP for
4604 privacy.
4605
4606 Arguments :
4607 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4608 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004609
Cyril Bonté78caf842010-03-10 22:41:43 +01004610 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004611 http-request set-src hdr(x-forwarded-for)
4612 http-request set-src src,ipmask(24)
4613
4614 When possible, set-src preserves the original source port as long as the
4615 address family allows it, otherwise the source port is set to 0.
4616
4617http-request set-src-port <expr> [ { if | unless } <condition> ]
4618
4619 This is used to set the source port address to the value of specified
4620 expression.
4621
4622 Arguments:
4623 <expr> Is a standard HAProxy expression formed by a sample-fetch followed
4624 by some converters.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01004625
Willy Tarreau20b0de52012-12-24 15:45:22 +01004626 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004627 http-request set-src-port hdr(x-port)
4628 http-request set-src-port int(4000)
4629
4630 When possible, set-src-port preserves the original source address as long as
4631 the address family supports a port, otherwise it forces the source address to
4632 IPv4 "0.0.0.0" before rewriting the port.
4633
4634http-request set-tos <tos> [ { if | unless } <condition> ]
4635
4636 This is used to set the TOS or DSCP field value of packets sent to the client
4637 to the value passed in <tos> on platforms which support this. This value
4638 represents the whole 8 bits of the IP TOS field, and can be expressed both in
4639 decimal or hexadecimal format (prefixed by "0x"). Note that only the 6 higher
4640 bits are used in DSCP or TOS, and the two lower bits are always 0. This can
4641 be used to adjust some routing behavior on border routers based on some
4642 information from the request.
4643
4644 See RFC 2474, 2597, 3260 and 4594 for more information.
4645
4646http-request set-uri <fmt> [ { if | unless } <condition> ]
4647
4648 This rewrites the request URI with the result of the evaluation of format
4649 string <fmt>. The scheme, authority, path and query string are all replaced
4650 at once. This can be used to rewrite hosts in front of proxies, or to
4651 perform complex modifications to the URI such as moving parts between the
4652 path and the query string.
4653 See also "http-request set-path" and "http-request set-query".
4654
4655http-request set-var(<var-name>) <expr> [ { if | unless } <condition> ]
4656
4657 This is used to set the contents of a variable. The variable is declared
4658 inline.
4659
4660 Arguments:
4661 <var-name> The name of the variable starts with an indication about its
4662 scope. The scopes allowed are:
4663 "proc" : the variable is shared with the whole process
4664 "sess" : the variable is shared with the whole session
4665 "txn" : the variable is shared with the transaction
4666 (request and response)
4667 "req" : the variable is shared only during request
4668 processing
4669 "res" : the variable is shared only during response
4670 processing
4671 This prefix is followed by a name. The separator is a '.'.
4672 The name may only contain characters 'a-z', 'A-Z', '0-9'
4673 and '_'.
4674
4675 <expr> Is a standard HAProxy expression formed by a sample-fetch
4676 followed by some converters.
Willy Tarreau20b0de52012-12-24 15:45:22 +01004677
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004678 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004679 http-request set-var(req.my_var) req.fhdr(user-agent),lower
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004680
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004681http-request send-spoe-group <engine-name> <group-name>
4682 [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004683
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004684 This action is used to trigger sending of a group of SPOE messages. To do so,
4685 the SPOE engine used to send messages must be defined, as well as the SPOE
4686 group to send. Of course, the SPOE engine must refer to an existing SPOE
4687 filter. If not engine name is provided on the SPOE filter line, the SPOE
4688 agent name must be used.
4689
4690 Arguments:
4691 <engine-name> The SPOE engine name.
4692
4693 <group-name> The SPOE group name as specified in the engine
4694 configuration.
4695
4696http-request silent-drop [ { if | unless } <condition> ]
4697
4698 This stops the evaluation of the rules and makes the client-facing connection
4699 suddenly disappear using a system-dependent way that tries to prevent the
4700 client from being notified. The effect it then that the client still sees an
4701 established connection while there's none on HAProxy. The purpose is to
4702 achieve a comparable effect to "tarpit" except that it doesn't use any local
4703 resource at all on the machine running HAProxy. It can resist much higher
4704 loads than "tarpit", and slow down stronger attackers. It is important to
4705 understand the impact of using this mechanism. All stateful equipment placed
4706 between the client and HAProxy (firewalls, proxies, load balancers) will also
4707 keep the established connection for a long time and may suffer from this
4708 action.
4709 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
4710 option is used to block the emission of a TCP reset. On other systems, the
4711 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
4712 router, though it's still delivered to local networks. Do not use it unless
4713 you fully understand how it works.
4714
4715http-request tarpit [deny_status <status>] [ { if | unless } <condition> ]
4716
4717 This stops the evaluation of the rules and immediately blocks the request
4718 without responding for a delay specified by "timeout tarpit" or
4719 "timeout connect" if the former is not set. After that delay, if the client
4720 is still connected, an HTTP error 500 (or optionally the status code
4721 specified as an argument to "deny_status") is returned so that the client
4722 does not suspect it has been tarpitted. Logs will report the flags "PT".
4723 The goal of the tarpit rule is to slow down robots during an attack when
4724 they're limited on the number of concurrent requests. It can be very
4725 efficient against very dumb robots, and will significantly reduce the load
4726 on firewalls compared to a "deny" rule. But when facing "correctly"
4727 developed robots, it can make things worse by forcing haproxy and the front
4728 firewall to support insane number of concurrent connections.
4729 See also the "silent-drop" action.
4730
4731http-request track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
4732http-request track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
4733http-request track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
4734
4735 This enables tracking of sticky counters from current request. These rules do
4736 not stop evaluation and do not change default action. The number of counters
4737 that may be simultaneously tracked by the same connection is set in
4738 MAX_SESS_STKCTR at build time (reported in haproxy -vv) which defaults to 3,
4739 so the track-sc number is between 0 and (MAX_SESS_STCKTR-1). The first
4740 "track-sc0" rule executed enables tracking of the counters of the specified
4741 table as the first set. The first "track-sc1" rule executed enables tracking
4742 of the counters of the specified table as the second set. The first
4743 "track-sc2" rule executed enables tracking of the counters of the specified
4744 table as the third set. It is a recommended practice to use the first set of
4745 counters for the per-frontend counters and the second set for the per-backend
4746 ones. But this is just a guideline, all may be used everywhere.
4747
4748 Arguments :
4749 <key> is mandatory, and is a sample expression rule as described in
4750 section 7.3. It describes what elements of the incoming request or
4751 connection will be analyzed, extracted, combined, and used to
4752 select which table entry to update the counters.
4753
4754 <table> is an optional table to be used instead of the default one, which
4755 is the stick-table declared in the current proxy. All the counters
4756 for the matches and updates for the key will then be performed in
4757 that table until the session ends.
4758
4759 Once a "track-sc*" rule is executed, the key is looked up in the table and if
4760 it is not found, an entry is allocated for it. Then a pointer to that entry
4761 is kept during all the session's life, and this entry's counters are updated
4762 as often as possible, every time the session's counters are updated, and also
4763 systematically when the session ends. Counters are only updated for events
4764 that happen after the tracking has been started. As an exception, connection
4765 counters and request counters are systematically updated so that they reflect
4766 useful information.
4767
4768 If the entry tracks concurrent connection counters, one connection is counted
4769 for as long as the entry is tracked, and the entry will not expire during
4770 that time. Tracking counters also provides a performance advantage over just
4771 checking the keys, because only one table lookup is performed for all ACL
4772 checks that make use of it.
4773
4774http-request unset-var(<var-name>) [ { if | unless } <condition> ]
4775
4776 This is used to unset a variable. See above for details about <var-name>.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004777
4778 Example:
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004779 http-request unset-var(req.my_var)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004780
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004781http-request wait-for-handshake [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004782
Cyril Bontéc6ad23b2018-10-17 00:14:50 +02004783 This will delay the processing of the request until the SSL handshake
4784 happened. This is mostly useful to delay processing early data until we're
4785 sure they are valid.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004786
Willy Tarreauef781042010-01-27 11:53:01 +01004787
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004788http-response <action> <options...> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004789 Access control for Layer 7 responses
4790
4791 May be used in sections: defaults | frontend | listen | backend
4792 no | yes | yes | yes
4793
4794 The http-response statement defines a set of rules which apply to layer 7
4795 processing. The rules are evaluated in their declaration order when they are
4796 met in a frontend, listen or backend section. Any rule may optionally be
4797 followed by an ACL-based condition, in which case it will only be evaluated
4798 if the condition is true. Since these rules apply on responses, the backend
4799 rules are applied first, followed by the frontend's rules.
4800
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004801 The first keyword is the rule's action. The supported actions are described
4802 below.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004803
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004804 There is no limit to the number of http-response statements per instance.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004805
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004806 It is important to know that http-response rules are processed very early in
4807 the HTTP processing, before "rspdel" or "rsprep" or "rspadd" rules. That way,
4808 headers added by "add-header"/"set-header" are visible by almost all further
4809 ACL rules.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004810
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004811 Using "rspadd"/"rspdel"/"rsprep" to manipulate request headers is discouraged
4812 in newer versions (>= 1.5). But if you need to use regular expression to
4813 delete headers, you can still use "rspdel". Also please use
4814 "http-response deny" instead of "rspdeny".
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004815
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004816 Example:
4817 acl key_acl res.hdr(X-Acl-Key) -m found
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02004818
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004819 acl myhost hdr(Host) -f myhost.lst
Sasha Pachev218f0642014-06-16 12:05:59 -06004820
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004821 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
4822 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
Sasha Pachev218f0642014-06-16 12:05:59 -06004823
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004824 Example:
4825 acl value res.hdr(X-Value) -m found
Sasha Pachev218f0642014-06-16 12:05:59 -06004826
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004827 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
Sasha Pachev218f0642014-06-16 12:05:59 -06004828
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004829 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
4830 http-response del-map(map.lst) %[src] if ! value
Sasha Pachev218f0642014-06-16 12:05:59 -06004831
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004832 See also : "http-request", section 3.4 about userlists and section 7 about
4833 ACL usage.
Sasha Pachev218f0642014-06-16 12:05:59 -06004834
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004835http-response add-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004836
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004837 This is used to add a new entry into an ACL. The ACL must be loaded from a
4838 file (even a dummy empty file). The file name of the ACL to be updated is
4839 passed between parentheses. It takes one argument: <key fmt>, which follows
4840 log-format rules, to collect content of the new entry. It performs a lookup
4841 in the ACL before insertion, to avoid duplicated (or more) values.
4842 This lookup is done by a linear search and can be expensive with large lists!
4843 It is the equivalent of the "add acl" command from the stats socket, but can
4844 be triggered by an HTTP response.
Sasha Pachev218f0642014-06-16 12:05:59 -06004845
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004846http-response add-header <name> <fmt> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004847
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004848 This appends an HTTP header field whose name is specified in <name> and whose
4849 value is defined by <fmt> which follows the log-format rules (see Custom Log
4850 Format in section 8.2.4). This may be used to send a cookie to a client for
4851 example, or to pass some internal information.
4852 This rule is not final, so it is possible to add other similar rules.
4853 Note that header addition is performed immediately, so one rule might reuse
4854 the resulting header from a previous rule.
Sasha Pachev218f0642014-06-16 12:05:59 -06004855
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004856http-response allow [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004857
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004858 This stops the evaluation of the rules and lets the response pass the check.
4859 No further "http-response" rules are evaluated for the current section.
Sasha Pachev218f0642014-06-16 12:05:59 -06004860
Jarno Huuskonen251a6b72019-01-04 14:05:02 +02004861http-response cache-store <name> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004862
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004863 See section 10.2 about cache setup.
Sasha Pachev218f0642014-06-16 12:05:59 -06004864
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004865http-response capture <sample> id <id> [ { if | unless } <condition> ]
Sasha Pachev218f0642014-06-16 12:05:59 -06004866
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004867 This captures sample expression <sample> from the response buffer, and
4868 converts it to a string. The resulting string is stored into the next request
4869 "capture" slot, so it will possibly appear next to some captured HTTP
4870 headers. It will then automatically appear in the logs, and it will be
4871 possible to extract it using sample fetch rules to feed it into headers or
4872 anything. Please check section 7.3 (Fetching samples) and
4873 "capture response header" for more information.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004874
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004875 The keyword "id" is the id of the capture slot which is used for storing the
4876 string. The capture slot must be defined in an associated frontend.
4877 This is useful to run captures in backends. The slot id can be declared by a
4878 previous directive "http-response capture" or with the "declare capture"
4879 keyword.
4880 If the slot <id> doesn't exist, then HAProxy fails parsing the configuration
4881 to prevent unexpected behavior at run time.
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004882
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004883http-response del-acl(<file-name>) <key fmt> [ { if | unless } <condition> ]
Thierry FOURNIER35d70ef2015-08-26 16:21:56 +02004884
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004885 This is used to delete an entry from an ACL. The ACL must be loaded from a
4886 file (even a dummy empty file). The file name of the ACL to be updated is
4887 passed between parentheses. It takes one argument: <key fmt>, which follows
4888 log-format rules, to collect content of the entry to delete.
4889 It is the equivalent of the "del acl" command from the stats socket, but can
4890 be triggered by an HTTP response.
Willy Tarreauf4c43c12013-06-11 17:01:13 +02004891
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004892http-response del-header <name> [ { if | unless } <condition> ]
Willy Tarreau9a355ec2013-06-11 17:45:46 +02004893
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004894 This removes all HTTP header fields whose name is specified in <name>.
Willy Tarreau42cf39e2013-06-11 18:51:32 +02004895
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004896http-response del-map(<file-name>) <key fmt> [ { if | unless } <condition> ]
Willy Tarreau51347ed2013-06-11 19:34:13 +02004897
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004898 This is used to delete an entry from a MAP. The MAP must be loaded from a
4899 file (even a dummy empty file). The file name of the MAP to be updated is
4900 passed between parentheses. It takes one argument: <key fmt>, which follows
4901 log-format rules, to collect content of the entry to delete.
4902 It takes one argument: "file name" It is the equivalent of the "del map"
4903 command from the stats socket, but can be triggered by an HTTP response.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004904
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004905http-response deny [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004906
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004907 This stops the evaluation of the rules and immediately rejects the response
4908 and emits an HTTP 502 error. No further "http-response" rules are evaluated.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004909
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004910http-response redirect <rule> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02004911
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004912 This performs an HTTP redirection based on a redirect rule.
4913 This supports a format string similarly to "http-request redirect" rules,
4914 with the exception that only the "location" type of redirect is possible on
4915 the response. See the "redirect" keyword for the rule's syntax. When a
4916 redirect rule is applied during a response, connections to the server are
4917 closed so that no data can be forwarded from the server to the client.
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004918
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004919http-response replace-header <name> <regex-match> <replace-fmt>
4920 [ { if | unless } <condition> ]
Thierry FOURNIERe80fada2015-05-26 18:06:31 +02004921
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004922 This matches the regular expression in all occurrences of header field <name>
4923 according to <match-regex>, and replaces them with the <replace-fmt> argument.
4924 Format characters are allowed in replace-fmt and work like in <fmt> arguments
4925 in "add-header". The match is only case-sensitive. It is important to
4926 understand that this action only considers whole header lines, regardless of
4927 the number of values they may contain. This usage is suited to headers
4928 naturally containing commas in their value, such as Set-Cookie, Expires and
4929 so on.
William Lallemand86d0df02017-11-24 21:36:45 +01004930
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004931 Example:
4932 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
Willy Tarreau51d861a2015-05-22 17:30:48 +02004933
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004934 # applied to:
4935 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004936
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004937 # outputs:
4938 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 +02004939
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004940 # assuming the backend IP is 192.168.1.20.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004941
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004942http-response replace-value <name> <regex-match> <replace-fmt>
4943 [ { if | unless } <condition> ]
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004944
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004945 This works like "replace-header" except that it matches the regex against
4946 every comma-delimited value of the header field <name> instead of the entire
4947 header. This is suited for all headers which are allowed to carry more than
4948 one value. An example could be the Accept header.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +02004949
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004950 Example:
4951 http-response replace-value Cache-control ^public$ private
Christopher Faulet85d79c92016-11-09 16:54:56 +01004952
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004953 # applied to:
4954 Cache-Control: max-age=3600, public
Christopher Faulet85d79c92016-11-09 16:54:56 +01004955
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004956 # outputs:
4957 Cache-Control: max-age=3600, private
Christopher Faulet85d79c92016-11-09 16:54:56 +01004958
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004959http-response sc-inc-gpc0(<sc-id>) [ { if | unless } <condition> ]
4960http-response sc-inc-gpc1(<sc-id>) [ { if | unless } <condition> ]
Ruoshan Huange4edc6b2016-07-14 15:07:45 +08004961
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004962 This action increments the GPC0 or GPC1 counter according with the sticky
4963 counter designated by <sc-id>. If an error occurs, this action silently fails
4964 and the actions evaluation continues.
Thierry FOURNIER236657b2015-08-19 08:25:14 +02004965
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004966http-response sc-set-gpt0(<sc-id>) <int> [ { if | unless } <condition> ]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02004967
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004968 This action sets the GPT0 tag according to the sticky counter designated by
4969 <sc-id> and the value of <int>. The expected result is a boolean. If an error
4970 occurs, this action silently fails and the actions evaluation continues.
Frédéric Lécaille6778b272018-01-29 15:22:53 +01004971
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004972http-response send-spoe-group [ { if | unless } <condition> ]
Willy Tarreau2d392c22015-08-24 01:43:45 +02004973
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004974 This action is used to trigger sending of a group of SPOE messages. To do so,
4975 the SPOE engine used to send messages must be defined, as well as the SPOE
4976 group to send. Of course, the SPOE engine must refer to an existing SPOE
4977 filter. If not engine name is provided on the SPOE filter line, the SPOE
4978 agent name must be used.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004979
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004980 Arguments:
4981 <engine-name> The SPOE engine name.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004982
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004983 <group-name> The SPOE group name as specified in the engine
4984 configuration.
Christopher Faulet76c09ef2017-09-21 11:03:52 +02004985
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004986http-response set-header <name> <fmt> [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004987
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004988 This does the same as "add-header" except that the header name is first
4989 removed if it existed. This is useful when passing security information to
4990 the server, where the header must not be manipulated by external users.
Willy Tarreaue365c0b2013-06-11 16:06:12 +02004991
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02004992http-response set-log-level <level> [ { if | unless } <condition> ]
4993
4994 This is used to change the log level of the current request when a certain
4995 condition is met. Valid levels are the 8 syslog levels (see the "log"
4996 keyword) plus the special level "silent" which disables logging for this
4997 request. This rule is not final so the last matching rule wins. This rule can
4998 be useful to disable health checks coming from another equipment.
4999
5000http-response set-map(<file-name>) <key fmt> <value fmt>
5001
5002 This is used to add a new entry into a MAP. The MAP must be loaded from a
5003 file (even a dummy empty file). The file name of the MAP to be updated is
5004 passed between parentheses. It takes 2 arguments: <key fmt>, which follows
5005 log-format rules, used to collect MAP key, and <value fmt>, which follows
5006 log-format rules, used to collect content for the new entry. It performs a
5007 lookup in the MAP before insertion, to avoid duplicated (or more) values.
5008 This lookup is done by a linear search and can be expensive with large lists!
5009 It is the equivalent of the "set map" command from the stats socket, but can
5010 be triggered by an HTTP response.
5011
5012http-response set-mark <mark> [ { if | unless } <condition> ]
5013
5014 This is used to set the Netfilter MARK on all packets sent to the client to
5015 the value passed in <mark> on platforms which support it. This value is an
5016 unsigned 32 bit value which can be matched by netfilter and by the routing
5017 table. It can be expressed both in decimal or hexadecimal format (prefixed
5018 by "0x"). This can be useful to force certain packets to take a different
5019 route (for example a cheaper network path for bulk downloads). This works on
5020 Linux kernels 2.6.32 and above and requires admin privileges.
5021
5022http-response set-nice <nice> [ { if | unless } <condition> ]
5023
5024 This sets the "nice" factor of the current request being processed.
5025 It only has effect against the other requests being processed at the same
5026 time. The default value is 0, unless altered by the "nice" setting on the
5027 "bind" line. The accepted range is -1024..1024. The higher the value, the
5028 nicest the request will be. Lower values will make the request more important
5029 than other ones. This can be useful to improve the speed of some requests, or
5030 lower the priority of non-important requests. Using this setting without
5031 prior experimentation can cause some major slowdown.
5032
5033http-response set-status <status> [reason <str>]
5034 [ { if | unless } <condition> ]
5035
5036 This replaces the response status code with <status> which must be an integer
5037 between 100 and 999. Optionally, a custom reason text can be provided defined
5038 by <str>, or the default reason for the specified code will be used as a
5039 fallback.
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08005040
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005041 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005042 # return "431 Request Header Fields Too Large"
5043 http-response set-status 431
5044 # return "503 Slow Down", custom reason
5045 http-response set-status 503 reason "Slow Down".
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005046
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005047http-response set-tos <tos> [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005048
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005049 This is used to set the TOS or DSCP field value of packets sent to the client
5050 to the value passed in <tos> on platforms which support this.
5051 This value represents the whole 8 bits of the IP TOS field, and can be
5052 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note that
5053 only the 6 higher bits are used in DSCP or TOS, and the two lower bits are
5054 always 0. This can be used to adjust some routing behavior on border routers
5055 based on some information from the request.
5056
5057 See RFC 2474, 2597, 3260 and 4594 for more information.
5058
5059http-response set-var(<var-name>) <expr> [ { if | unless } <condition> ]
5060
5061 This is used to set the contents of a variable. The variable is declared
5062 inline.
5063
5064 Arguments:
5065 <var-name> The name of the variable starts with an indication about its
5066 scope. The scopes allowed are:
5067 "proc" : the variable is shared with the whole process
5068 "sess" : the variable is shared with the whole session
5069 "txn" : the variable is shared with the transaction
5070 (request and response)
5071 "req" : the variable is shared only during request
5072 processing
5073 "res" : the variable is shared only during response
5074 processing
5075 This prefix is followed by a name. The separator is a '.'.
5076 The name may only contain characters 'a-z', 'A-Z', '0-9', '.'
5077 and '_'.
5078
5079 <expr> Is a standard HAProxy expression formed by a sample-fetch
5080 followed by some converters.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005081
5082 Example:
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005083 http-response set-var(sess.last_redir) res.hdr(location)
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005084
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005085http-response silent-drop [ { if | unless } <condition> ]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005086
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005087 This stops the evaluation of the rules and makes the client-facing connection
5088 suddenly disappear using a system-dependent way that tries to prevent the
5089 client from being notified. The effect it then that the client still sees an
5090 established connection while there's none on HAProxy. The purpose is to
5091 achieve a comparable effect to "tarpit" except that it doesn't use any local
5092 resource at all on the machine running HAProxy. It can resist much higher
5093 loads than "tarpit", and slow down stronger attackers. It is important to
5094 understand the impact of using this mechanism. All stateful equipment placed
5095 between the client and HAProxy (firewalls, proxies, load balancers) will also
5096 keep the established connection for a long time and may suffer from this
5097 action.
5098 On modern Linux systems running with enough privileges, the TCP_REPAIR socket
5099 option is used to block the emission of a TCP reset. On other systems, the
5100 socket's TTL is reduced to 1 so that the TCP reset doesn't pass the first
5101 router, though it's still delivered to local networks. Do not use it unless
5102 you fully understand how it works.
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02005103
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005104http-response track-sc0 <key> [table <table>] [ { if | unless } <condition> ]
5105http-response track-sc1 <key> [table <table>] [ { if | unless } <condition> ]
5106http-response track-sc2 <key> [table <table>] [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02005107
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02005108 This enables tracking of sticky counters from current response. Please refer
5109 to "http-request track-sc" for a complete description. The only difference
5110 from "http-request track-sc" is the <key> sample expression can only make use
5111 of samples in response (e.g. res.*, status etc.) and samples below Layer 6
5112 (e.g. SSL-related samples, see section 7.3.4). If the sample is not
5113 supported, haproxy will fail and warn while parsing the config.
5114
5115http-response unset-var(<var-name>) [ { if | unless } <condition> ]
5116
5117 This is used to unset a variable. See "http-response set-var" for details
5118 about <var-name>.
5119
5120 Example:
5121 http-response unset-var(sess.last_redir)
5122
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02005123
Willy Tarreau30631952015-08-06 15:05:24 +02005124http-reuse { never | safe | aggressive | always }
5125 Declare how idle HTTP connections may be shared between requests
5126
5127 May be used in sections: defaults | frontend | listen | backend
5128 yes | no | yes | yes
5129
5130 By default, a connection established between haproxy and the backend server
Olivier Houchard86006a52018-12-14 19:37:49 +01005131 which is considered safe for reuse is moved back to the server's idle
5132 connections pool so that any other request can make use of it. This is the
5133 "safe" strategy below.
Willy Tarreau30631952015-08-06 15:05:24 +02005134
5135 The argument indicates the desired connection reuse strategy :
5136
Olivier Houchard86006a52018-12-14 19:37:49 +01005137 - "never" : idle connections are never shared between sessions. This mode
5138 may be enforced to cancel a different strategy inherited from
5139 a defaults section or for troubleshooting. For example, if an
5140 old bogus application considers that multiple requests over
5141 the same connection come from the same client and it is not
5142 possible to fix the application, it may be desirable to
5143 disable connection sharing in a single backend. An example of
5144 such an application could be an old haproxy using cookie
5145 insertion in tunnel mode and not checking any request past the
5146 first one.
Willy Tarreau30631952015-08-06 15:05:24 +02005147
Olivier Houchard86006a52018-12-14 19:37:49 +01005148 - "safe" : this is the default and the recommended strategy. The first
5149 request of a session is always sent over its own connection,
5150 and only subsequent requests may be dispatched over other
5151 existing connections. This ensures that in case the server
5152 closes the connection when the request is being sent, the
5153 browser can decide to silently retry it. Since it is exactly
5154 equivalent to regular keep-alive, there should be no side
5155 effects.
Willy Tarreau30631952015-08-06 15:05:24 +02005156
5157 - "aggressive" : this mode may be useful in webservices environments where
5158 all servers are not necessarily known and where it would be
5159 appreciable to deliver most first requests over existing
5160 connections. In this case, first requests are only delivered
5161 over existing connections that have been reused at least once,
5162 proving that the server correctly supports connection reuse.
5163 It should only be used when it's sure that the client can
5164 retry a failed request once in a while and where the benefit
Michael Prokop4438c602019-05-24 10:25:45 +02005165 of aggressive connection reuse significantly outweighs the
Willy Tarreau30631952015-08-06 15:05:24 +02005166 downsides of rare connection failures.
5167
5168 - "always" : this mode is only recommended when the path to the server is
5169 known for never breaking existing connections quickly after
5170 releasing them. It allows the first request of a session to be
5171 sent to an existing connection. This can provide a significant
5172 performance increase over the "safe" strategy when the backend
5173 is a cache farm, since such components tend to show a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005174 consistent behavior and will benefit from the connection
Willy Tarreau30631952015-08-06 15:05:24 +02005175 sharing. It is recommended that the "http-keep-alive" timeout
5176 remains low in this mode so that no dead connections remain
5177 usable. In most cases, this will lead to the same performance
5178 gains as "aggressive" but with more risks. It should only be
5179 used when it improves the situation over "aggressive".
5180
5181 When http connection sharing is enabled, a great care is taken to respect the
Davor Ocelice9ed2812017-12-25 17:49:28 +01005182 connection properties and compatibility. Specifically :
5183 - connections made with "usesrc" followed by a client-dependent value
5184 ("client", "clientip", "hdr_ip") are marked private and never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005185
5186 - connections sent to a server with a TLS SNI extension are marked private
Davor Ocelice9ed2812017-12-25 17:49:28 +01005187 and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005188
Lukas Tribusfd9b68c2018-10-27 20:06:59 +02005189 - connections with certain bogus authentication schemes (relying on the
5190 connection) like NTLM are detected, marked private and are never shared;
Willy Tarreau30631952015-08-06 15:05:24 +02005191
5192 No connection pool is involved, once a session dies, the last idle connection
5193 it was attached to is deleted at the same time. This ensures that connections
5194 may not last after all sessions are closed.
5195
5196 Note: connection reuse improves the accuracy of the "server maxconn" setting,
5197 because almost no new connection will be established while idle connections
5198 remain available. This is particularly true with the "always" strategy.
5199
5200 See also : "option http-keep-alive", "server maxconn"
5201
5202
Mark Lamourinec2247f02012-01-04 13:02:01 -05005203http-send-name-header [<header>]
5204 Add the server name to a request. Use the header string given by <header>
5205
5206 May be used in sections: defaults | frontend | listen | backend
5207 yes | no | yes | yes
5208
5209 Arguments :
5210
5211 <header> The header string to use to send the server name
5212
5213 The "http-send-name-header" statement causes the name of the target
Davor Ocelice9ed2812017-12-25 17:49:28 +01005214 server to be added to the headers of an HTTP request. The name
Mark Lamourinec2247f02012-01-04 13:02:01 -05005215 is added with the header string proved.
5216
5217 See also : "server"
5218
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005219id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02005220 Set a persistent ID to a proxy.
5221 May be used in sections : defaults | frontend | listen | backend
5222 no | yes | yes | yes
5223 Arguments : none
5224
5225 Set a persistent ID for the proxy. This ID must be unique and positive.
5226 An unused ID will automatically be assigned if unset. The first assigned
5227 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01005228
5229
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005230ignore-persist { if | unless } <condition>
5231 Declare a condition to ignore persistence
5232 May be used in sections: defaults | frontend | listen | backend
Cyril Bonté4288c5a2018-03-12 22:02:59 +01005233 no | no | yes | yes
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005234
5235 By default, when cookie persistence is enabled, every requests containing
5236 the cookie are unconditionally persistent (assuming the target server is up
5237 and running).
5238
5239 The "ignore-persist" statement allows one to declare various ACL-based
5240 conditions which, when met, will cause a request to ignore persistence.
5241 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005242 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005243 persistence for a specific User-Agent (for example, some web crawler bots).
5244
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005245 The persistence is ignored when an "if" condition is met, or unless an
5246 "unless" condition is met.
5247
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03005248 Example:
5249 acl url_static path_beg /static /images /img /css
5250 acl url_static path_end .gif .png .jpg .css .js
5251 ignore-persist if url_static
5252
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005253 See also : "force-persist", "cookie", and section 7 about ACL usage.
5254
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005255load-server-state-from-file { global | local | none }
5256 Allow seamless reload of HAProxy
5257 May be used in sections: defaults | frontend | listen | backend
5258 yes | no | yes | yes
5259
5260 This directive points HAProxy to a file where server state from previous
5261 running process has been saved. That way, when starting up, before handling
5262 traffic, the new process can apply old states to servers exactly has if no
Davor Ocelice9ed2812017-12-25 17:49:28 +01005263 reload occurred. The purpose of the "load-server-state-from-file" directive is
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005264 to tell haproxy which file to use. For now, only 2 arguments to either prevent
5265 loading state or load states from a file containing all backends and servers.
5266 The state file can be generated by running the command "show servers state"
5267 over the stats socket and redirect output.
5268
Davor Ocelice9ed2812017-12-25 17:49:28 +01005269 The format of the file is versioned and is very specific. To understand it,
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005270 please read the documentation of the "show servers state" command (chapter
Willy Tarreau1af20c72017-06-23 16:01:14 +02005271 9.3 of Management Guide).
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005272
5273 Arguments:
5274 global load the content of the file pointed by the global directive
5275 named "server-state-file".
5276
5277 local load the content of the file pointed by the directive
5278 "server-state-file-name" if set. If not set, then the backend
5279 name is used as a file name.
5280
5281 none don't load any stat for this backend
5282
5283 Notes:
Willy Tarreaue5a60682016-11-09 14:54:53 +01005284 - server's IP address is preserved across reloads by default, but the
5285 order can be changed thanks to the server's "init-addr" setting. This
5286 means that an IP address change performed on the CLI at run time will
Davor Ocelice9ed2812017-12-25 17:49:28 +01005287 be preserved, and that any change to the local resolver (e.g. /etc/hosts)
Willy Tarreaue5a60682016-11-09 14:54:53 +01005288 will possibly not have any effect if the state file is in use.
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005289
5290 - server's weight is applied from previous running process unless it has
5291 has changed between previous and new configuration files.
5292
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005293 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005294
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005295 global
5296 stats socket /tmp/socket
5297 server-state-file /tmp/server_state
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005298
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005299 defaults
5300 load-server-state-from-file global
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005301
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005302 backend bk
5303 server s1 127.0.0.1:22 check weight 11
5304 server s2 127.0.0.1:22 check weight 12
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005305
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005306
5307 Then one can run :
5308
5309 socat /tmp/socket - <<< "show servers state" > /tmp/server_state
5310
5311 Content of the file /tmp/server_state would be like this:
5312
5313 1
5314 # <field names skipped for the doc example>
5315 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5316 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5317
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005318 Example: Minimal configuration
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005319
5320 global
5321 stats socket /tmp/socket
5322 server-state-base /etc/haproxy/states
5323
5324 defaults
5325 load-server-state-from-file local
5326
5327 backend bk
5328 server s1 127.0.0.1:22 check weight 11
5329 server s2 127.0.0.1:22 check weight 12
5330
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02005331
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02005332 Then one can run :
5333
5334 socat /tmp/socket - <<< "show servers state bk" > /etc/haproxy/states/bk
5335
5336 Content of the file /etc/haproxy/states/bk would be like this:
5337
5338 1
5339 # <field names skipped for the doc example>
5340 1 bk 1 s1 127.0.0.1 2 0 11 11 4 6 3 4 6 0 0
5341 1 bk 2 s2 127.0.0.1 2 0 12 12 4 6 3 4 6 0 0
5342
5343 See also: "server-state-file", "server-state-file-name", and
5344 "show servers state"
5345
Cyril Bonté0d4bf012010-04-25 23:21:46 +02005346
Willy Tarreau2769aa02007-12-27 18:26:09 +01005347log global
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005348log <address> [len <length>] [format <format>] [sample <ranges>:<smp_size>]
5349 <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02005350no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01005351 Enable per-instance logging of events and traffic.
5352 May be used in sections : defaults | frontend | listen | backend
5353 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02005354
5355 Prefix :
5356 no should be used when the logger list must be flushed. For example,
5357 if you don't want to inherit from the default logger list. This
5358 prefix does not allow arguments.
5359
Willy Tarreau2769aa02007-12-27 18:26:09 +01005360 Arguments :
5361 global should be used when the instance's logging parameters are the
5362 same as the global ones. This is the most common usage. "global"
5363 replaces <address>, <facility> and <level> with those of the log
5364 entries found in the "global" section. Only one "log global"
5365 statement may be used per instance, and this form takes no other
5366 parameter.
5367
5368 <address> indicates where to send the logs. It takes the same format as
5369 for the "global" section's logs, and can be one of :
5370
5371 - An IPv4 address optionally followed by a colon (':') and a UDP
5372 port. If no port is specified, 514 is used by default (the
5373 standard syslog port).
5374
David du Colombier24bb5f52011-03-17 10:40:23 +01005375 - An IPv6 address followed by a colon (':') and optionally a UDP
5376 port. If no port is specified, 514 is used by default (the
5377 standard syslog port).
5378
Willy Tarreau2769aa02007-12-27 18:26:09 +01005379 - A filesystem path to a UNIX domain socket, keeping in mind
5380 considerations for chroot (be sure the path is accessible
5381 inside the chroot) and uid/gid (be sure the path is
Davor Ocelice9ed2812017-12-25 17:49:28 +01005382 appropriately writable).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005383
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005384 - A file descriptor number in the form "fd@<number>", which may
5385 point to a pipe, terminal, or socket. In this case unbuffered
5386 logs are used and one writev() call per log is performed. This
5387 is a bit expensive but acceptable for most workloads. Messages
5388 sent this way will not be truncated but may be dropped, in
5389 which case the DroppedLogs counter will be incremented. The
5390 writev() call is atomic even on pipes for messages up to
5391 PIPE_BUF size, which POSIX recommends to be at least 512 and
5392 which is 4096 bytes on most modern operating systems. Any
5393 larger message may be interleaved with messages from other
5394 processes. Exceptionally for debugging purposes the file
5395 descriptor may also be directed to a file, but doing so will
5396 significantly slow haproxy down as non-blocking calls will be
5397 ignored. Also there will be no way to purge nor rotate this
5398 file without restarting the process. Note that the configured
5399 syslog format is preserved, so the output is suitable for use
Willy Tarreauc1b06452018-11-12 11:57:56 +01005400 with a TCP syslog server. See also the "short" and "raw"
5401 formats below.
Willy Tarreau5a32ecc2018-11-12 07:34:59 +01005402
5403 - "stdout" / "stderr", which are respectively aliases for "fd@1"
5404 and "fd@2", see above.
5405
5406 You may want to reference some environment variables in the
5407 address parameter, see section 2.3 about environment variables.
Willy Tarreaudad36a32013-03-11 01:20:04 +01005408
Willy Tarreau18324f52014-06-27 18:10:07 +02005409 <length> is an optional maximum line length. Log lines larger than this
5410 value will be truncated before being sent. The reason is that
5411 syslog servers act differently on log line length. All servers
5412 support the default value of 1024, but some servers simply drop
5413 larger lines while others do log them. If a server supports long
5414 lines, it may make sense to set this value here in order to avoid
5415 truncating long lines. Similarly, if a server drops long lines,
5416 it is preferable to truncate them before sending them. Accepted
5417 values are 80 to 65535 inclusive. The default value of 1024 is
5418 generally fine for all standard usages. Some specific cases of
Davor Ocelice9ed2812017-12-25 17:49:28 +01005419 long captures or JSON-formatted logs may require larger values.
Willy Tarreau18324f52014-06-27 18:10:07 +02005420
Frédéric Lécailled690dfa2019-04-25 10:52:17 +02005421 <ranges> A list of comma-separated ranges to identify the logs to sample.
5422 This is used to balance the load of the logs to send to the log
5423 server. The limits of the ranges cannot be null. They are numbered
5424 from 1. The size or period (in number of logs) of the sample must
5425 be set with <sample_size> parameter.
5426
5427 <sample_size>
5428 The size of the sample in number of logs to consider when balancing
5429 their logging loads. It is used to balance the load of the logs to
5430 send to the syslog server. This size must be greater or equal to the
5431 maximum of the high limits of the ranges.
5432 (see also <ranges> parameter).
5433
Willy Tarreauadb345d2018-11-12 07:56:13 +01005434 <format> is the log format used when generating syslog messages. It may be
5435 one of the following :
5436
5437 rfc3164 The RFC3164 syslog message format. This is the default.
5438 (https://tools.ietf.org/html/rfc3164)
5439
5440 rfc5424 The RFC5424 syslog message format.
5441 (https://tools.ietf.org/html/rfc5424)
5442
Willy Tarreaue8746a02018-11-12 08:45:00 +01005443 short A message containing only a level between angle brackets such as
5444 '<3>', followed by the text. The PID, date, time, process name
5445 and system name are omitted. This is designed to be used with a
5446 local log server. This format is compatible with what the
5447 systemd logger consumes.
5448
Willy Tarreauc1b06452018-11-12 11:57:56 +01005449 raw A message containing only the text. The level, PID, date, time,
5450 process name and system name are omitted. This is designed to
5451 be used in containers or during development, where the severity
5452 only depends on the file descriptor used (stdout/stderr).
5453
Willy Tarreau2769aa02007-12-27 18:26:09 +01005454 <facility> must be one of the 24 standard syslog facilities :
5455
Willy Tarreaue8746a02018-11-12 08:45:00 +01005456 kern user mail daemon auth syslog lpr news
5457 uucp cron auth2 ftp ntp audit alert cron2
5458 local0 local1 local2 local3 local4 local5 local6 local7
5459
Willy Tarreauc1b06452018-11-12 11:57:56 +01005460 Note that the facility is ignored for the "short" and "raw"
5461 formats, but still required as a positional field. It is
5462 recommended to use "daemon" in this case to make it clear that
5463 it's only supposed to be used locally.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005464
5465 <level> is optional and can be specified to filter outgoing messages. By
5466 default, all messages are sent. If a level is specified, only
5467 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005468 will be sent. An optional minimum level can be specified. If it
5469 is set, logs emitted with a more severe level than this one will
5470 be capped to this level. This is used to avoid sending "emerg"
5471 messages on all terminals on some default syslog configurations.
5472 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005473
5474 emerg alert crit err warning notice info debug
5475
William Lallemand0f99e342011-10-12 17:50:54 +02005476 It is important to keep in mind that it is the frontend which decides what to
5477 log from a connection, and that in case of content switching, the log entries
5478 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01005479
5480 However, backend log declaration define how and where servers status changes
5481 will be logged. Level "notice" will be used to indicate a server going up,
5482 "warning" will be used for termination signals and definitive service
5483 termination, and "alert" will be used for when a server goes down.
5484
5485 Note : According to RFC3164, messages are truncated to 1024 bytes before
5486 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005487
5488 Example :
5489 log global
Willy Tarreauc1b06452018-11-12 11:57:56 +01005490 log stdout format short daemon # send log to systemd
5491 log stdout format raw daemon # send everything to stdout
5492 log stderr format raw daemon notice # send important events to stderr
Willy Tarreauf7edefa2009-05-10 17:20:05 +02005493 log 127.0.0.1:514 local0 notice # only send important events
5494 log 127.0.0.1:514 local0 notice notice # same but limit output level
William Lallemandb2f07452015-05-12 14:27:13 +02005495 log "${LOCAL_SYSLOG}:514" local0 notice # send to local server
Willy Tarreaudad36a32013-03-11 01:20:04 +01005496
Willy Tarreau2769aa02007-12-27 18:26:09 +01005497
William Lallemand48940402012-01-30 16:47:22 +01005498log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005499 Specifies the log format string to use for traffic logs
5500 May be used in sections: defaults | frontend | listen | backend
5501 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01005502
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01005503 This directive specifies the log format string that will be used for all logs
5504 resulting from traffic passing through the frontend using this line. If the
5505 directive is used in a defaults section, all subsequent frontends will use
5506 the same log format. Please see section 8.2.4 which covers the log format
5507 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01005508
Guillaume de Lafond29f45602017-03-31 19:52:15 +02005509 "log-format" directive overrides previous "option tcplog", "log-format" and
5510 "option httplog" directives.
5511
Dragan Dosen7ad31542015-09-28 17:16:47 +02005512log-format-sd <string>
5513 Specifies the RFC5424 structured-data log format string
5514 May be used in sections: defaults | frontend | listen | backend
5515 yes | yes | yes | no
5516
5517 This directive specifies the RFC5424 structured-data log format string that
5518 will be used for all logs resulting from traffic passing through the frontend
5519 using this line. If the directive is used in a defaults section, all
5520 subsequent frontends will use the same log format. Please see section 8.2.4
5521 which covers the log format string in depth.
5522
5523 See https://tools.ietf.org/html/rfc5424#section-6.3 for more information
5524 about the RFC5424 structured-data part.
5525
5526 Note : This log format string will be used only for loggers that have set
5527 log format to "rfc5424".
5528
5529 Example :
5530 log-format-sd [exampleSDID@1234\ bytes=\"%B\"\ status=\"%ST\"]
5531
5532
Willy Tarreau094af4e2015-01-07 15:03:42 +01005533log-tag <string>
5534 Specifies the log tag to use for all outgoing logs
5535 May be used in sections: defaults | frontend | listen | backend
5536 yes | yes | yes | yes
5537
5538 Sets the tag field in the syslog header to this string. It defaults to the
5539 log-tag set in the global section, otherwise the program name as launched
5540 from the command line, which usually is "haproxy". Sometimes it can be useful
5541 to differentiate between multiple processes running on the same host, or to
5542 differentiate customer instances running in the same process. In the backend,
5543 logs about servers up/down will use this tag. As a hint, it can be convenient
5544 to set a log-tag related to a hosted customer in a defaults section then put
5545 all the frontends and backends for that customer, then start another customer
5546 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005547
Willy Tarreauc35362a2014-04-25 13:58:37 +02005548max-keep-alive-queue <value>
5549 Set the maximum server queue size for maintaining keep-alive connections
5550 May be used in sections: defaults | frontend | listen | backend
5551 yes | no | yes | yes
5552
5553 HTTP keep-alive tries to reuse the same server connection whenever possible,
5554 but sometimes it can be counter-productive, for example if a server has a lot
5555 of connections while other ones are idle. This is especially true for static
5556 servers.
5557
5558 The purpose of this setting is to set a threshold on the number of queued
5559 connections at which haproxy stops trying to reuse the same server and prefers
5560 to find another one. The default value, -1, means there is no limit. A value
5561 of zero means that keep-alive requests will never be queued. For very close
5562 servers which can be reached with a low latency and which are not sensible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01005563 breaking keep-alive, a low value is recommended (e.g. local static server can
Willy Tarreauc35362a2014-04-25 13:58:37 +02005564 use a value of 10 or less). For remote servers suffering from a high latency,
5565 higher values might be needed to cover for the latency and/or the cost of
5566 picking a different server.
5567
5568 Note that this has no impact on responses which are maintained to the same
5569 server consecutively to a 401 response. They will still go to the same server
5570 even if they have to be queued.
5571
5572 See also : "option http-server-close", "option prefer-last-server", server
5573 "maxconn" and cookie persistence.
5574
Olivier Houcharda4d4fdf2018-12-14 19:27:06 +01005575max-session-srv-conns <nb>
5576 Set the maximum number of outgoing connections we can keep idling for a given
5577 client session. The default is 5 (it precisely equals MAX_SRV_LIST which is
5578 defined at build time).
Willy Tarreauc35362a2014-04-25 13:58:37 +02005579
Willy Tarreau2769aa02007-12-27 18:26:09 +01005580maxconn <conns>
5581 Fix the maximum number of concurrent connections on a frontend
5582 May be used in sections : defaults | frontend | listen | backend
5583 yes | yes | yes | no
5584 Arguments :
5585 <conns> is the maximum number of concurrent connections the frontend will
5586 accept to serve. Excess connections will be queued by the system
5587 in the socket's listen queue and will be served once a connection
5588 closes.
5589
5590 If the system supports it, it can be useful on big sites to raise this limit
5591 very high so that haproxy manages connection queues, instead of leaving the
5592 clients with unanswered connection attempts. This value should not exceed the
5593 global maxconn. Also, keep in mind that a connection contains two buffers
Baptiste Assmann79fb45d2016-03-06 23:34:31 +01005594 of tune.bufsize (16kB by default) each, as well as some other data resulting
5595 in about 33 kB of RAM being consumed per established connection. That means
5596 that a medium system equipped with 1GB of RAM can withstand around
5597 20000-25000 concurrent connections if properly tuned.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005598
5599 Also, when <conns> is set to large values, it is possible that the servers
5600 are not sized to accept such loads, and for this reason it is generally wise
5601 to assign them some reasonable connection limits.
5602
Willy Tarreauc8d5b952019-02-27 17:25:52 +01005603 When this value is set to zero, which is the default, the global "maxconn"
5604 value is used.
Vincent Bernat6341be52012-06-27 17:18:30 +02005605
Willy Tarreau2769aa02007-12-27 18:26:09 +01005606 See also : "server", global section's "maxconn", "fullconn"
5607
5608
5609mode { tcp|http|health }
5610 Set the running mode or protocol of the instance
5611 May be used in sections : defaults | frontend | listen | backend
5612 yes | yes | yes | yes
5613 Arguments :
5614 tcp The instance will work in pure TCP mode. A full-duplex connection
5615 will be established between clients and servers, and no layer 7
5616 examination will be performed. This is the default mode. It
5617 should be used for SSL, SSH, SMTP, ...
5618
5619 http The instance will work in HTTP mode. The client request will be
5620 analyzed in depth before connecting to any server. Any request
5621 which is not RFC-compliant will be rejected. Layer 7 filtering,
5622 processing and switching will be possible. This is the mode which
5623 brings HAProxy most of its value.
5624
5625 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02005626 to incoming connections and close the connection. Alternatively,
5627 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
5628 instead. Nothing will be logged in either case. This mode is used
5629 to reply to external components health checks. This mode is
5630 deprecated and should not be used anymore as it is possible to do
5631 the same and even better by combining TCP or HTTP modes with the
5632 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005633
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005634 When doing content switching, it is mandatory that the frontend and the
5635 backend are in the same mode (generally HTTP), otherwise the configuration
5636 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005637
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005638 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01005639 defaults http_instances
5640 mode http
5641
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005642 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005643
Willy Tarreau0ba27502007-12-24 16:55:16 +01005644
Cyril Bontéf0c60612010-02-06 14:44:47 +01005645monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01005646 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005647 May be used in sections : defaults | frontend | listen | backend
5648 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01005649 Arguments :
5650 if <cond> the monitor request will fail if the condition is satisfied,
5651 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005652 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01005653 are met, for instance a low number of servers both in a
5654 backend and its backup.
5655
5656 unless <cond> the monitor request will succeed only if the condition is
5657 satisfied, and will fail otherwise. Such a condition may be
5658 based on a test on the presence of a minimum number of active
5659 servers in a list of backends.
5660
5661 This statement adds a condition which can force the response to a monitor
5662 request to report a failure. By default, when an external component queries
5663 the URI dedicated to monitoring, a 200 response is returned. When one of the
5664 conditions above is met, haproxy will return 503 instead of 200. This is
5665 very useful to report a site failure to an external component which may base
5666 routing advertisements between multiple sites on the availability reported by
5667 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005668 criterion. Note that "monitor fail" only works in HTTP mode. Both status
5669 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01005670
5671 Example:
5672 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01005673 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01005674 acl site_dead nbsrv(dynamic) lt 2
5675 acl site_dead nbsrv(static) lt 2
5676 monitor-uri /site_alive
5677 monitor fail if site_dead
5678
Willy Tarreauae94d4d2011-05-11 16:28:49 +02005679 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01005680
5681
5682monitor-net <source>
5683 Declare a source network which is limited to monitor requests
5684 May be used in sections : defaults | frontend | listen | backend
5685 yes | yes | yes | no
5686 Arguments :
5687 <source> is the source IPv4 address or network which will only be able to
5688 get monitor responses to any request. It can be either an IPv4
5689 address, a host name, or an address followed by a slash ('/')
5690 followed by a mask.
5691
5692 In TCP mode, any connection coming from a source matching <source> will cause
5693 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005694 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01005695 forwarding the connection to a remote server.
5696
5697 In HTTP mode, a connection coming from a source matching <source> will be
5698 accepted, the following response will be sent without waiting for a request,
5699 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
5700 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02005701 running without forwarding the request to a backend server. Note that this
5702 response is sent in raw format, without any transformation. This is important
5703 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01005704
Willy Tarreau82569f92012-09-27 23:48:56 +02005705 Monitor requests are processed very early, just after tcp-request connection
5706 ACLs which are the only ones able to block them. These connections are short
5707 lived and never wait for any data from the client. They cannot be logged, and
5708 it is the intended purpose. They are only used to report HAProxy's health to
5709 an upper component, nothing more. Please note that "monitor fail" rules do
5710 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01005711
Willy Tarreau95cd2832010-03-04 23:36:33 +01005712 Last, please note that only one "monitor-net" statement can be specified in
5713 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02005714
Willy Tarreau2769aa02007-12-27 18:26:09 +01005715 Example :
5716 # addresses .252 and .253 are just probing us.
5717 frontend www
5718 monitor-net 192.168.0.252/31
5719
5720 See also : "monitor fail", "monitor-uri"
5721
5722
5723monitor-uri <uri>
5724 Intercept a URI used by external components' monitor requests
5725 May be used in sections : defaults | frontend | listen | backend
5726 yes | yes | yes | no
5727 Arguments :
5728 <uri> is the exact URI which we want to intercept to return HAProxy's
5729 health status instead of forwarding the request.
5730
5731 When an HTTP request referencing <uri> will be received on a frontend,
5732 HAProxy will not forward it nor log it, but instead will return either
5733 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
5734 conditions defined with "monitor fail". This is normally enough for any
5735 front-end HTTP probe to detect that the service is UP and running without
5736 forwarding the request to a backend server. Note that the HTTP method, the
5737 version and all headers are ignored, but the request must at least be valid
5738 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
5739
Willy Tarreau721d8e02017-12-01 18:25:08 +01005740 Monitor requests are processed very early, just after the request is parsed
5741 and even before any "http-request" or "block" rulesets. The only rulesets
5742 applied before are the tcp-request ones. They cannot be logged either, and it
5743 is the intended purpose. They are only used to report HAProxy's health to an
5744 upper component, nothing more. However, it is possible to add any number of
5745 conditions using "monitor fail" and ACLs so that the result can be adjusted
5746 to whatever check can be imagined (most often the number of available servers
5747 in a backend).
Willy Tarreau2769aa02007-12-27 18:26:09 +01005748
5749 Example :
5750 # Use /haproxy_test to report haproxy's status
5751 frontend www
5752 mode http
5753 monitor-uri /haproxy_test
5754
5755 See also : "monitor fail", "monitor-net"
5756
Willy Tarreau0ba27502007-12-24 16:55:16 +01005757
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005758option abortonclose
5759no option abortonclose
5760 Enable or disable early dropping of aborted requests pending in queues.
5761 May be used in sections : defaults | frontend | listen | backend
5762 yes | no | yes | yes
5763 Arguments : none
5764
5765 In presence of very high loads, the servers will take some time to respond.
5766 The per-instance connection queue will inflate, and the response time will
5767 increase respective to the size of the queue times the average per-session
5768 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01005769 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005770 the queue, and slowing down other users, and the servers as well, because the
5771 request will eventually be served, then aborted at the first error
5772 encountered while delivering the response.
5773
5774 As there is no way to distinguish between a full STOP and a simple output
5775 close on the client side, HTTP agents should be conservative and consider
5776 that the client might only have closed its output channel while waiting for
5777 the response. However, this introduces risks of congestion when lots of users
5778 do the same, and is completely useless nowadays because probably no client at
5779 all will close the session while waiting for the response. Some HTTP agents
Davor Ocelice9ed2812017-12-25 17:49:28 +01005780 support this behavior (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005781 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01005782 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005783 of being the single component to break rare but valid traffic is extremely
5784 low, which adds to the temptation to be able to abort a session early while
5785 still not served and not pollute the servers.
5786
Davor Ocelice9ed2812017-12-25 17:49:28 +01005787 In HAProxy, the user can choose the desired behavior using the option
5788 "abortonclose". By default (without the option) the behavior is HTTP
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005789 compliant and aborted requests will be served. But when the option is
5790 specified, a session with an incoming channel closed will be aborted while
5791 it is still possible, either pending in the queue for a connection slot, or
5792 during the connection establishment if the server has not yet acknowledged
5793 the connection request. This considerably reduces the queue size and the load
5794 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01005795 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005796
5797 If this option has been enabled in a "defaults" section, it can be disabled
5798 in a specific instance by prepending the "no" keyword before it.
5799
5800 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
5801
5802
Willy Tarreau4076a152009-04-02 15:18:36 +02005803option accept-invalid-http-request
5804no option accept-invalid-http-request
5805 Enable or disable relaxing of HTTP request parsing
5806 May be used in sections : defaults | frontend | listen | backend
5807 yes | yes | yes | no
5808 Arguments : none
5809
Willy Tarreau91852eb2015-05-01 13:26:00 +02005810 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005811 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005812 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005813 forbidden characters are essentially used to build attacks exploiting server
5814 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5815 server will emit invalid header names for whatever reason (configuration,
5816 implementation) and the issue will not be immediately fixed. In such a case,
5817 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01005818 even if that does not make sense, by specifying this option. Similarly, the
5819 list of characters allowed to appear in a URI is well defined by RFC3986, and
5820 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
5821 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
Davor Ocelice9ed2812017-12-25 17:49:28 +01005822 not allowed at all. HAProxy always blocks a number of them (0..32, 127). The
Willy Tarreau91852eb2015-05-01 13:26:00 +02005823 remaining ones are blocked by default unless this option is enabled. This
Willy Tarreau13317662015-05-01 13:47:08 +02005824 option also relaxes the test on the HTTP version, it allows HTTP/0.9 requests
5825 to pass through (no version specified) and multiple digits for both the major
5826 and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005827
5828 This option should never be enabled by default as it hides application bugs
5829 and open security breaches. It should only be deployed after a problem has
5830 been confirmed.
5831
5832 When this option is enabled, erroneous header names will still be accepted in
5833 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01005834 analysis using the "show errors" request on the UNIX stats socket. Similarly,
5835 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02005836 also helps confirming that the issue has been solved.
5837
5838 If this option has been enabled in a "defaults" section, it can be disabled
5839 in a specific instance by prepending the "no" keyword before it.
5840
5841 See also : "option accept-invalid-http-response" and "show errors" on the
5842 stats socket.
5843
5844
5845option accept-invalid-http-response
5846no option accept-invalid-http-response
5847 Enable or disable relaxing of HTTP response parsing
5848 May be used in sections : defaults | frontend | listen | backend
5849 yes | no | yes | yes
5850 Arguments : none
5851
Willy Tarreau91852eb2015-05-01 13:26:00 +02005852 By default, HAProxy complies with RFC7230 in terms of message parsing. This
Willy Tarreau4076a152009-04-02 15:18:36 +02005853 means that invalid characters in header names are not permitted and cause an
Davor Ocelice9ed2812017-12-25 17:49:28 +01005854 error to be returned to the client. This is the desired behavior as such
Willy Tarreau4076a152009-04-02 15:18:36 +02005855 forbidden characters are essentially used to build attacks exploiting server
5856 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
5857 server will emit invalid header names for whatever reason (configuration,
5858 implementation) and the issue will not be immediately fixed. In such a case,
5859 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau91852eb2015-05-01 13:26:00 +02005860 even if that does not make sense, by specifying this option. This option also
5861 relaxes the test on the HTTP version format, it allows multiple digits for
5862 both the major and the minor version.
Willy Tarreau4076a152009-04-02 15:18:36 +02005863
5864 This option should never be enabled by default as it hides application bugs
5865 and open security breaches. It should only be deployed after a problem has
5866 been confirmed.
5867
5868 When this option is enabled, erroneous header names will still be accepted in
5869 responses, but the complete response will be captured in order to permit
5870 later analysis using the "show errors" request on the UNIX stats socket.
5871 Doing this also helps confirming that the issue has been solved.
5872
5873 If this option has been enabled in a "defaults" section, it can be disabled
5874 in a specific instance by prepending the "no" keyword before it.
5875
5876 See also : "option accept-invalid-http-request" and "show errors" on the
5877 stats socket.
5878
5879
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005880option allbackups
5881no option allbackups
5882 Use either all backup servers at a time or only the first one
5883 May be used in sections : defaults | frontend | listen | backend
5884 yes | no | yes | yes
5885 Arguments : none
5886
5887 By default, the first operational backup server gets all traffic when normal
5888 servers are all down. Sometimes, it may be preferred to use multiple backups
5889 at once, because one will not be enough. When "option allbackups" is enabled,
5890 the load balancing will be performed among all backup servers when all normal
5891 ones are unavailable. The same load balancing algorithm will be used and the
5892 servers' weights will be respected. Thus, there will not be any priority
5893 order between the backup servers anymore.
5894
5895 This option is mostly used with static server farms dedicated to return a
5896 "sorry" page when an application is completely offline.
5897
5898 If this option has been enabled in a "defaults" section, it can be disabled
5899 in a specific instance by prepending the "no" keyword before it.
5900
5901
5902option checkcache
5903no option checkcache
Godbach7056a352013-12-11 20:01:07 +08005904 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005905 May be used in sections : defaults | frontend | listen | backend
5906 yes | no | yes | yes
5907 Arguments : none
5908
5909 Some high-level frameworks set application cookies everywhere and do not
5910 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005911 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005912 high risk of session crossing or stealing between users traversing the same
5913 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02005914 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005915
5916 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005917 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01005918 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005919 response to check if there's a risk of caching a cookie on a client-side
5920 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01005921 to the client are :
Davor Ocelice9ed2812017-12-25 17:49:28 +01005922 - all those without "Set-Cookie" header;
Willy Tarreauc55ddce2017-12-21 11:41:38 +01005923 - all those with a return code other than 200, 203, 204, 206, 300, 301,
5924 404, 405, 410, 414, 501, provided that the server has not set a
Davor Ocelice9ed2812017-12-25 17:49:28 +01005925 "Cache-control: public" header field;
Willy Tarreau24ea0bc2017-12-21 11:32:55 +01005926 - all those that result from a request using a method other than GET, HEAD,
5927 OPTIONS, TRACE, provided that the server has not set a 'Cache-Control:
Davor Ocelice9ed2812017-12-25 17:49:28 +01005928 public' header field;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005929 - those with a 'Pragma: no-cache' header
5930 - those with a 'Cache-control: private' header
5931 - those with a 'Cache-control: no-store' header
5932 - those with a 'Cache-control: max-age=0' header
5933 - those with a 'Cache-control: s-maxage=0' header
5934 - those with a 'Cache-control: no-cache' header
5935 - those with a 'Cache-control: no-cache="set-cookie"' header
5936 - those with a 'Cache-control: no-cache="set-cookie,' header
5937 (allowing other fields after set-cookie)
5938
5939 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01005940 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005941 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005942 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005943 that admins are informed that there's something to be fixed.
5944
5945 Due to the high impact on the application, the application should be tested
5946 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005947 good practice to always activate it during tests, even if it is not used in
Davor Ocelice9ed2812017-12-25 17:49:28 +01005948 production, as it will report potentially dangerous application behaviors.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005949
5950 If this option has been enabled in a "defaults" section, it can be disabled
5951 in a specific instance by prepending the "no" keyword before it.
5952
5953
5954option clitcpka
5955no option clitcpka
5956 Enable or disable the sending of TCP keepalive packets on the client side
5957 May be used in sections : defaults | frontend | listen | backend
5958 yes | yes | yes | no
5959 Arguments : none
5960
5961 When there is a firewall or any session-aware component between a client and
5962 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01005963 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005964 components decides to expire a session which has remained idle for too long.
5965
5966 Enabling socket-level TCP keep-alives makes the system regularly send packets
5967 to the other end of the connection, leaving it active. The delay between
5968 keep-alive probes is controlled by the system only and depends both on the
5969 operating system and its tuning parameters.
5970
5971 It is important to understand that keep-alive packets are neither emitted nor
5972 received at the application level. It is only the network stacks which sees
5973 them. For this reason, even if one side of the proxy already uses keep-alives
5974 to maintain its connection alive, those keep-alive packets will not be
5975 forwarded to the other side of the proxy.
5976
5977 Please note that this has nothing to do with HTTP keep-alive.
5978
5979 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
5980 client side of a connection, which should help when session expirations are
5981 noticed between HAProxy and a client.
5982
5983 If this option has been enabled in a "defaults" section, it can be disabled
5984 in a specific instance by prepending the "no" keyword before it.
5985
5986 See also : "option srvtcpka", "option tcpka"
5987
5988
Willy Tarreau0ba27502007-12-24 16:55:16 +01005989option contstats
5990 Enable continuous traffic statistics updates
5991 May be used in sections : defaults | frontend | listen | backend
5992 yes | yes | yes | no
5993 Arguments : none
5994
5995 By default, counters used for statistics calculation are incremented
5996 only when a session finishes. It works quite well when serving small
5997 objects, but with big ones (for example large images or archives) or
5998 with A/V streaming, a graph generated from haproxy counters looks like
Willy Tarreaudef0d222016-11-08 22:03:00 +01005999 a hedgehog. With this option enabled counters get incremented frequently
6000 along the session, typically every 5 seconds, which is often enough to
6001 produce clean graphs. Recounting touches a hotpath directly so it is not
6002 not enabled by default, as it can cause a lot of wakeups for very large
6003 session counts and cause a small performance drop.
Willy Tarreau0ba27502007-12-24 16:55:16 +01006004
6005
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006006option dontlog-normal
6007no option dontlog-normal
6008 Enable or disable logging of normal, successful connections
6009 May be used in sections : defaults | frontend | listen | backend
6010 yes | yes | yes | no
6011 Arguments : none
6012
6013 There are large sites dealing with several thousand connections per second
6014 and for which logging is a major pain. Some of them are even forced to turn
6015 logs off and cannot debug production issues. Setting this option ensures that
6016 normal connections, those which experience no error, no timeout, no retry nor
6017 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
6018 mode, the response status code is checked and return codes 5xx will still be
6019 logged.
6020
6021 It is strongly discouraged to use this option as most of the time, the key to
6022 complex issues is in the normal logs which will not be logged here. If you
6023 need to separate logs, see the "log-separate-errors" option instead.
6024
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006025 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006026 logging.
6027
6028
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006029option dontlognull
6030no option dontlognull
6031 Enable or disable logging of null connections
6032 May be used in sections : defaults | frontend | listen | backend
6033 yes | yes | yes | no
6034 Arguments : none
6035
6036 In certain environments, there are components which will regularly connect to
6037 various systems to ensure that they are still alive. It can be the case from
6038 another load balancer as well as from monitoring systems. By default, even a
6039 simple port probe or scan will produce a log. If those connections pollute
6040 the logs too much, it is possible to enable option "dontlognull" to indicate
6041 that a connection on which no data has been transferred will not be logged,
Willy Tarreau0f228a02015-05-01 15:37:53 +02006042 which typically corresponds to those probes. Note that errors will still be
6043 returned to the client and accounted for in the stats. If this is not what is
6044 desired, option http-ignore-probes can be used instead.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006045
6046 It is generally recommended not to use this option in uncontrolled
Davor Ocelice9ed2812017-12-25 17:49:28 +01006047 environments (e.g. internet), otherwise scans and other malicious activities
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006048 would not be logged.
6049
6050 If this option has been enabled in a "defaults" section, it can be disabled
6051 in a specific instance by prepending the "no" keyword before it.
6052
Willy Tarreau0f228a02015-05-01 15:37:53 +02006053 See also : "log", "http-ignore-probes", "monitor-net", "monitor-uri", and
6054 section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006055
Willy Tarreaubf1f8162007-12-28 17:42:56 +01006056
Willy Tarreau87cf5142011-08-19 22:57:24 +02006057option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006058 Enable insertion of the X-Forwarded-For header to requests sent to servers
6059 May be used in sections : defaults | frontend | listen | backend
6060 yes | yes | yes | yes
6061 Arguments :
6062 <network> is an optional argument used to disable this option for sources
6063 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02006064 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01006065 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006066
6067 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
6068 their client address. This is sometimes annoying when the client's IP address
6069 is expected in server logs. To solve this problem, the well-known HTTP header
6070 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
6071 This header contains a value representing the client's IP address. Since this
6072 header is always appended at the end of the existing header list, the server
6073 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02006074 the server's manual to find how to enable use of this standard header. Note
6075 that only the last occurrence of the header must be used, since it is really
6076 possible that the client has already brought one.
6077
Willy Tarreaud72758d2010-01-12 10:42:19 +01006078 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02006079 the default "X-Forwarded-For". This can be useful where you might already
Davor Ocelice9ed2812017-12-25 17:49:28 +01006080 have a "X-Forwarded-For" header from a different application (e.g. stunnel),
Willy Tarreaud72758d2010-01-12 10:42:19 +01006081 and you need preserve it. Also if your backend server doesn't use the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006082 "X-Forwarded-For" header and requires different one (e.g. Zeus Web Servers
Ross Westaf72a1d2008-08-03 10:51:45 +02006083 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01006084
6085 Sometimes, a same HAProxy instance may be shared between a direct client
6086 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6087 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6088 header for a known source address or network by adding the "except" keyword
6089 followed by the network address. In this case, any source IP matching the
6090 network will not cause an addition of this header. Most common uses are with
6091 private networks or 127.0.0.1.
6092
Willy Tarreau87cf5142011-08-19 22:57:24 +02006093 Alternatively, the keyword "if-none" states that the header will only be
6094 added if it is not present. This should only be used in perfectly trusted
6095 environment, as this might cause a security issue if headers reaching haproxy
6096 are under the control of the end-user.
6097
Willy Tarreauc27debf2008-01-06 08:57:02 +01006098 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02006099 least one of them uses it, the header will be added. Note that the backend's
6100 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02006101 both are defined. In the case of the "if-none" argument, if at least one of
6102 the frontend or the backend does not specify it, it wants the addition to be
6103 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006104
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +02006105 Example :
Willy Tarreauc27debf2008-01-06 08:57:02 +01006106 # Public HTTP address also used by stunnel on the same machine
6107 frontend www
6108 mode http
6109 option forwardfor except 127.0.0.1 # stunnel already adds the header
6110
Ross Westaf72a1d2008-08-03 10:51:45 +02006111 # Those servers want the IP Address in X-Client
6112 backend www
6113 mode http
6114 option forwardfor header X-Client
6115
Willy Tarreau87cf5142011-08-19 22:57:24 +02006116 See also : "option httpclose", "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006117 "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01006118
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006119
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006120option http-buffer-request
6121no option http-buffer-request
6122 Enable or disable waiting for whole HTTP request body before proceeding
6123 May be used in sections : defaults | frontend | listen | backend
6124 yes | yes | yes | yes
6125 Arguments : none
6126
6127 It is sometimes desirable to wait for the body of an HTTP request before
6128 taking a decision. This is what is being done by "balance url_param" for
6129 example. The first use case is to buffer requests from slow clients before
6130 connecting to the server. Another use case consists in taking the routing
6131 decision based on the request body's contents. This option placed in a
6132 frontend or backend forces the HTTP processing to wait until either the whole
6133 body is received, or the request buffer is full, or the first chunk is
6134 complete in case of chunked encoding. It can have undesired side effects with
Tim Düsterhus4896c442016-11-29 02:15:19 +01006135 some applications abusing HTTP by expecting unbuffered transmissions between
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006136 the frontend and the backend, so this should definitely not be used by
6137 default.
6138
Baptiste Assmanneccdf432015-10-28 13:49:01 +01006139 See also : "option http-no-delay", "timeout http-request"
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006140
6141
Willy Tarreau0f228a02015-05-01 15:37:53 +02006142option http-ignore-probes
6143no option http-ignore-probes
6144 Enable or disable logging of null connections and request timeouts
6145 May be used in sections : defaults | frontend | listen | backend
6146 yes | yes | yes | no
6147 Arguments : none
6148
6149 Recently some browsers started to implement a "pre-connect" feature
6150 consisting in speculatively connecting to some recently visited web sites
6151 just in case the user would like to visit them. This results in many
6152 connections being established to web sites, which end up in 408 Request
6153 Timeout if the timeout strikes first, or 400 Bad Request when the browser
6154 decides to close them first. These ones pollute the log and feed the error
6155 counters. There was already "option dontlognull" but it's insufficient in
6156 this case. Instead, this option does the following things :
6157 - prevent any 400/408 message from being sent to the client if nothing
Davor Ocelice9ed2812017-12-25 17:49:28 +01006158 was received over a connection before it was closed;
6159 - prevent any log from being emitted in this situation;
Willy Tarreau0f228a02015-05-01 15:37:53 +02006160 - prevent any error counter from being incremented
6161
6162 That way the empty connection is silently ignored. Note that it is better
6163 not to use this unless it is clear that it is needed, because it will hide
6164 real problems. The most common reason for not receiving a request and seeing
6165 a 408 is due to an MTU inconsistency between the client and an intermediary
6166 element such as a VPN, which blocks too large packets. These issues are
6167 generally seen with POST requests as well as GET with large cookies. The logs
6168 are often the only way to detect them.
6169
6170 If this option has been enabled in a "defaults" section, it can be disabled
6171 in a specific instance by prepending the "no" keyword before it.
6172
6173 See also : "log", "dontlognull", "errorfile", and section 8 about logging.
6174
6175
Willy Tarreau16bfb022010-01-16 19:48:41 +01006176option http-keep-alive
6177no option http-keep-alive
6178 Enable or disable HTTP keep-alive from client to server
6179 May be used in sections : defaults | frontend | listen | backend
6180 yes | yes | yes | yes
6181 Arguments : none
6182
Willy Tarreau70dffda2014-01-30 03:07:23 +01006183 By default HAProxy operates in keep-alive mode with regards to persistent
6184 connections: for each connection it processes each request and response, and
Christopher Faulet315b39c2018-09-21 16:26:19 +02006185 leaves the connection idle on both sides between the end of a response and
6186 the start of a new request. This mode may be changed by several options such
6187 as "option http-server-close", "option httpclose" or "option http-tunnel".
6188 This option allows to set back the keep-alive mode, which can be useful when
6189 another mode was used in a defaults section.
Willy Tarreau70dffda2014-01-30 03:07:23 +01006190
6191 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
6192 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01006193 network) and the fastest session reuse on the server side at the expense
6194 of maintaining idle connections to the servers. In general, it is possible
6195 with this option to achieve approximately twice the request rate that the
6196 "http-server-close" option achieves on small objects. There are mainly two
6197 situations where this option may be useful :
6198
6199 - when the server is non-HTTP compliant and authenticates the connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01006200 instead of requests (e.g. NTLM authentication)
Willy Tarreau16bfb022010-01-16 19:48:41 +01006201
6202 - when the cost of establishing the connection to the server is significant
6203 compared to the cost of retrieving the associated object from the server.
6204
6205 This last case can happen when the server is a fast static server of cache.
6206 In this case, the server will need to be properly tuned to support high enough
6207 connection counts because connections will last until the client sends another
6208 request.
6209
6210 If the client request has to go to another backend or another server due to
6211 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01006212 immediately be closed and a new one re-opened. Option "prefer-last-server" is
6213 available to try optimize server selection so that if the server currently
6214 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01006215
Willy Tarreau16bfb022010-01-16 19:48:41 +01006216 At the moment, logs will not indicate whether requests came from the same
6217 session or not. The accept date reported in the logs corresponds to the end
6218 of the previous request, and the request time corresponds to the time spent
6219 waiting for a new request. The keep-alive request time is still bound to the
6220 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6221 not set.
6222
Cyril Bonté653dcd62014-02-20 00:13:15 +01006223 This option disables and replaces any previous "option httpclose", "option
Christopher Faulet315b39c2018-09-21 16:26:19 +02006224 http-server-close" or "option http-tunnel". When backend and frontend options
6225 differ, all of these 4 options have precedence over "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006226
Christopher Faulet315b39c2018-09-21 16:26:19 +02006227 See also : "option httpclose",, "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01006228 "option prefer-last-server", "option http-pretend-keepalive",
Frédéric Lécaille93d33162019-03-06 09:35:59 +01006229 and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01006230
6231
Willy Tarreau96e31212011-05-30 18:10:30 +02006232option http-no-delay
6233no option http-no-delay
6234 Instruct the system to favor low interactive delays over performance in HTTP
6235 May be used in sections : defaults | frontend | listen | backend
6236 yes | yes | yes | yes
6237 Arguments : none
6238
6239 In HTTP, each payload is unidirectional and has no notion of interactivity.
6240 Any agent is expected to queue data somewhat for a reasonably low delay.
6241 There are some very rare server-to-server applications that abuse the HTTP
6242 protocol and expect the payload phase to be highly interactive, with many
6243 interleaved data chunks in both directions within a single request. This is
6244 absolutely not supported by the HTTP specification and will not work across
6245 most proxies or servers. When such applications attempt to do this through
6246 haproxy, it works but they will experience high delays due to the network
6247 optimizations which favor performance by instructing the system to wait for
6248 enough data to be available in order to only send full packets. Typical
6249 delays are around 200 ms per round trip. Note that this only happens with
6250 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
6251 affected.
6252
6253 When "option http-no-delay" is present in either the frontend or the backend
6254 used by a connection, all such optimizations will be disabled in order to
6255 make the exchanges as fast as possible. Of course this offers no guarantee on
6256 the functionality, as it may break at any other place. But if it works via
6257 HAProxy, it will work as fast as possible. This option should never be used
6258 by default, and should never be used at all unless such a buggy application
6259 is discovered. The impact of using this option is an increase of bandwidth
6260 usage and CPU usage, which may significantly lower performance in high
6261 latency environments.
6262
Willy Tarreau9fbe18e2015-05-01 22:42:08 +02006263 See also : "option http-buffer-request"
6264
Willy Tarreau96e31212011-05-30 18:10:30 +02006265
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006266option http-pretend-keepalive
6267no option http-pretend-keepalive
6268 Define whether haproxy will announce keepalive to the server or not
6269 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet98db9762018-09-21 10:25:19 +02006270 yes | no | yes | yes
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006271 Arguments : none
6272
Christopher Faulet315b39c2018-09-21 16:26:19 +02006273 When running with "option http-server-close" or "option httpclose", haproxy
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006274 adds a "Connection: close" header to the request forwarded to the server.
6275 Unfortunately, when some servers see this header, they automatically refrain
6276 from using the chunked encoding for responses of unknown length, while this
6277 is totally unrelated. The immediate effect is that this prevents haproxy from
6278 maintaining the client connection alive. A second effect is that a client or
6279 a cache could receive an incomplete response without being aware of it, and
6280 consider the response complete.
6281
6282 By setting "option http-pretend-keepalive", haproxy will make the server
6283 believe it will keep the connection alive. The server will then not fall back
6284 to the abnormal undesired above. When haproxy gets the whole response, it
6285 will close the connection with the server just as it would do with the
Christopher Faulet315b39c2018-09-21 16:26:19 +02006286 "option httpclose". That way the client gets a normal response and the
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006287 connection is correctly closed on the server side.
6288
6289 It is recommended not to enable this option by default, because most servers
6290 will more efficiently close the connection themselves after the last packet,
6291 and release its buffers slightly earlier. Also, the added packet on the
6292 network could slightly reduce the overall peak performance. However it is
6293 worth noting that when this option is enabled, haproxy will have slightly
6294 less work to do. So if haproxy is the bottleneck on the whole architecture,
6295 enabling this option might save a few CPU cycles.
6296
Christopher Faulet98db9762018-09-21 10:25:19 +02006297 This option may be set in backend and listen sections. Using it in a frontend
6298 section will be ignored and a warning will be reported during startup. It is
6299 a backend related option, so there is no real reason to set it on a
6300 frontend. This option may be combined with "option httpclose", which will
6301 cause keepalive to be announced to the server and close to be announced to
6302 the client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006303
6304 If this option has been enabled in a "defaults" section, it can be disabled
6305 in a specific instance by prepending the "no" keyword before it.
6306
Christopher Faulet315b39c2018-09-21 16:26:19 +02006307 See also : "option httpclose", "option http-server-close", and
Willy Tarreau16bfb022010-01-16 19:48:41 +01006308 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02006309
Willy Tarreauc27debf2008-01-06 08:57:02 +01006310
Willy Tarreaub608feb2010-01-02 22:47:18 +01006311option http-server-close
6312no option http-server-close
6313 Enable or disable HTTP connection closing on the server side
6314 May be used in sections : defaults | frontend | listen | backend
6315 yes | yes | yes | yes
6316 Arguments : none
6317
Willy Tarreau70dffda2014-01-30 03:07:23 +01006318 By default HAProxy operates in keep-alive mode with regards to persistent
6319 connections: for each connection it processes each request and response, and
6320 leaves the connection idle on both sides between the end of a response and
6321 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006322 as "option http-server-close", "option httpclose" or "option http-tunnel".
6323 Setting "option http-server-close" enables HTTP connection-close mode on the
6324 server side while keeping the ability to support HTTP keep-alive and
6325 pipelining on the client side. This provides the lowest latency on the client
6326 side (slow network) and the fastest session reuse on the server side to save
6327 server resources, similarly to "option httpclose". It also permits
6328 non-keepalive capable servers to be served in keep-alive mode to the clients
6329 if they conform to the requirements of RFC7230. Please note that some servers
6330 do not always conform to those requirements when they see "Connection: close"
6331 in the request. The effect will be that keep-alive will never be used. A
6332 workaround consists in enabling "option http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006333
6334 At the moment, logs will not indicate whether requests came from the same
6335 session or not. The accept date reported in the logs corresponds to the end
6336 of the previous request, and the request time corresponds to the time spent
6337 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01006338 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
6339 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006340
6341 This option may be set both in a frontend and in a backend. It is enabled if
6342 at least one of the frontend or backend holding a connection has it enabled.
Christopher Faulet315b39c2018-09-21 16:26:19 +02006343 It disables and replaces any previous "option httpclose", "option http-tunnel"
6344 or "option http-keep-alive". Please check section 4 ("Proxies") to see how
6345 this option combines with others when frontend and backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01006346
6347 If this option has been enabled in a "defaults" section, it can be disabled
6348 in a specific instance by prepending the "no" keyword before it.
6349
Christopher Faulet315b39c2018-09-21 16:26:19 +02006350 See also : "option httpclose", "option http-pretend-keepalive",
6351 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01006352
6353
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006354option http-tunnel (deprecated)
6355no option http-tunnel (deprecated)
6356 Disable or enable HTTP connection processing after first transaction.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006357 May be used in sections : defaults | frontend | listen | backend
Christopher Faulet4212a302018-09-21 10:42:19 +02006358 yes | yes | yes | no
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006359 Arguments : none
6360
Christopher Faulet6c9bbb22019-03-26 21:37:23 +01006361 Warning : Because it cannot work in HTTP/2, this option is deprecated and it
6362 is only supported on legacy HTTP frontends. In HTX, it is ignored and a
6363 warning is emitted during HAProxy startup.
6364
Willy Tarreau70dffda2014-01-30 03:07:23 +01006365 By default HAProxy operates in keep-alive mode with regards to persistent
6366 connections: for each connection it processes each request and response, and
6367 leaves the connection idle on both sides between the end of a response and
6368 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006369 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006370
6371 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006372 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01006373 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
6374 is normally not needed anymore unless in very specific cases such as when
6375 using an in-house protocol that looks like HTTP but is not compatible, or
6376 just to log one request per client in order to reduce log size. Note that
6377 everything which works at the HTTP level, including header parsing/addition,
6378 cookie processing or content switching will only work for the first request
6379 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006380
Christopher Faulet4212a302018-09-21 10:42:19 +02006381 This option may be set on frontend and listen sections. Using it on a backend
6382 section will be ignored and a warning will be reported during the startup. It
6383 is a frontend related option, so there is no real reason to set it on a
6384 backend.
6385
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006386 If this option has been enabled in a "defaults" section, it can be disabled
6387 in a specific instance by prepending the "no" keyword before it.
6388
Christopher Faulet315b39c2018-09-21 16:26:19 +02006389 See also : "option httpclose", "option http-server-close",
6390 "option http-keep-alive", and "1.1. The HTTP transaction model".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01006391
6392
Willy Tarreau88d349d2010-01-25 12:15:43 +01006393option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01006394no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01006395 Make use of non-standard Proxy-Connection header instead of Connection
6396 May be used in sections : defaults | frontend | listen | backend
6397 yes | yes | yes | no
6398 Arguments : none
6399
Lukas Tribus23953682017-04-28 13:24:30 +00006400 While RFC7230 explicitly states that HTTP/1.1 agents must use the
Willy Tarreau88d349d2010-01-25 12:15:43 +01006401 Connection header to indicate their wish of persistent or non-persistent
6402 connections, both browsers and proxies ignore this header for proxied
6403 connections and make use of the undocumented, non-standard Proxy-Connection
6404 header instead. The issue begins when trying to put a load balancer between
6405 browsers and such proxies, because there will be a difference between what
6406 haproxy understands and what the client and the proxy agree on.
6407
6408 By setting this option in a frontend, haproxy can automatically switch to use
6409 that non-standard header if it sees proxied requests. A proxied request is
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006410 defined here as one where the URI begins with neither a '/' nor a '*'. This
6411 is incompatible with the HTTP tunnel mode. Note that this option can only be
6412 specified in a frontend and will affect the request along its whole life.
Willy Tarreau88d349d2010-01-25 12:15:43 +01006413
Willy Tarreau844a7e72010-01-31 21:46:18 +01006414 Also, when this option is set, a request which requires authentication will
6415 automatically switch to use proxy authentication headers if it is itself a
6416 proxied request. That makes it possible to check or enforce authentication in
6417 front of an existing proxy.
6418
Willy Tarreau88d349d2010-01-25 12:15:43 +01006419 This option should normally never be used, except in front of a proxy.
6420
Christopher Faulet315b39c2018-09-21 16:26:19 +02006421 See also : "option httpclose", and "option http-server-close".
Willy Tarreau88d349d2010-01-25 12:15:43 +01006422
6423
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006424option http-use-htx
6425no option http-use-htx
6426 Switch to the new HTX internal representation for HTTP protocol elements
6427 May be used in sections : defaults | frontend | listen | backend
6428 yes | yes | yes | yes
6429 Arguments : none
6430
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006431 Historically, the HTTP protocol is processed as-is. Inserting, deleting, or
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006432 modifying a header field requires to rewrite the affected part in the buffer
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006433 and to move the buffer's tail accordingly. This mode is known as the legacy
6434 HTTP mode. Since this principle has deep roots in haproxy, the HTTP/2
6435 protocol is converted to HTTP/1.1 before being processed this way. It also
6436 results in the inability to establish HTTP/2 connections to servers because
6437 of the loss of HTTP/2 semantics in the HTTP/1 representation.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006438
6439 HTX is the name of a totally new native internal representation for the HTTP
6440 protocol, that is agnostic to the version and aims at preserving semantics
6441 all along the chain. It relies on a fast parsing, tokenizing and indexing of
6442 the protocol elements so that no more memory moves are necessary and that
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006443 most elements are directly accessed. It supports using either HTTP/1 or
6444 HTTP/2 on any side regardless of the other side's version. It also supports
6445 upgrades from TCP to HTTP and implicit ones from HTTP/1 to HTTP/2 (matching
6446 the HTTP/2 preface).
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006447
Christopher Faulet1d2b5862019-04-12 16:10:51 +02006448 This option indicates that HTX needs to be used. Since the version 2.0-dev3,
6449 the HTX is the default mode. To switch back on the legacy HTTP mode, the
6450 option must be explicitly disabled using the "no" prefix. For prior versions,
6451 the feature has incomplete functional coverage, so it is not enabled by
6452 default.
Willy Tarreau68ad3a42018-10-22 11:49:15 +02006453
6454 See also : "mode http"
6455
6456
Willy Tarreaud63335a2010-02-26 12:56:52 +01006457option httpchk
6458option httpchk <uri>
6459option httpchk <method> <uri>
6460option httpchk <method> <uri> <version>
6461 Enable HTTP protocol to check on the servers health
6462 May be used in sections : defaults | frontend | listen | backend
6463 yes | no | yes | yes
6464 Arguments :
6465 <method> is the optional HTTP method used with the requests. When not set,
6466 the "OPTIONS" method is used, as it generally requires low server
6467 processing and is easy to filter out from the logs. Any method
6468 may be used, though it is not recommended to invent non-standard
6469 ones.
6470
6471 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
6472 which is accessible by default on almost any server, but may be
6473 changed to any other URI. Query strings are permitted.
6474
6475 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
6476 but some servers might behave incorrectly in HTTP 1.0, so turning
6477 it to HTTP/1.1 may sometimes help. Note that the Host field is
6478 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
6479 after "\r\n" following the version string.
6480
6481 By default, server health checks only consist in trying to establish a TCP
6482 connection. When "option httpchk" is specified, a complete HTTP request is
6483 sent once the TCP connection is established, and responses 2xx and 3xx are
6484 considered valid, while all other ones indicate a server failure, including
6485 the lack of any response.
6486
6487 The port and interval are specified in the server configuration.
6488
6489 This option does not necessarily require an HTTP backend, it also works with
6490 plain TCP backends. This is particularly useful to check simple scripts bound
6491 to some dedicated ports using the inetd daemon.
6492
6493 Examples :
6494 # Relay HTTPS traffic to Apache instance and check service availability
6495 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
6496 backend https_relay
6497 mode tcp
6498 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
6499 server apache1 192.168.1.1:443 check port 80
6500
Simon Hormanafc47ee2013-11-25 10:46:35 +09006501 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
6502 "option pgsql-check", "http-check" and the "check", "port" and
6503 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006504
6505
Willy Tarreauc27debf2008-01-06 08:57:02 +01006506option httpclose
6507no option httpclose
Christopher Faulet315b39c2018-09-21 16:26:19 +02006508 Enable or disable HTTP connection closing
Willy Tarreauc27debf2008-01-06 08:57:02 +01006509 May be used in sections : defaults | frontend | listen | backend
6510 yes | yes | yes | yes
6511 Arguments : none
6512
Willy Tarreau70dffda2014-01-30 03:07:23 +01006513 By default HAProxy operates in keep-alive mode with regards to persistent
6514 connections: for each connection it processes each request and response, and
6515 leaves the connection idle on both sides between the end of a response and
6516 the start of a new request. This mode may be changed by several options such
Christopher Faulet315b39c2018-09-21 16:26:19 +02006517 as "option http-server-close", "option httpclose" or "option http-tunnel".
Willy Tarreau70dffda2014-01-30 03:07:23 +01006518
Christopher Faulet315b39c2018-09-21 16:26:19 +02006519 If "option httpclose" is set, HAProxy will close connections with the server
6520 and the client as soon as the request and the response are received. It will
6521 alos check if a "Connection: close" header is already set in each direction,
6522 and will add one if missing. Any "Connection" header different from "close"
6523 will also be removed.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006524
Christopher Faulet315b39c2018-09-21 16:26:19 +02006525 This option may also be combined with "option http-pretend-keepalive", which
6526 will disable sending of the "Connection: close" header, but will still cause
6527 the connection to be closed once the whole response is received.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006528
6529 This option may be set both in a frontend and in a backend. It is enabled if
6530 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01006531 It disables and replaces any previous "option http-server-close",
Christopher Faulet315b39c2018-09-21 16:26:19 +02006532 "option http-keep-alive" or "option http-tunnel". Please check section 4
6533 ("Proxies") to see how this option combines with others when frontend and
6534 backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006535
6536 If this option has been enabled in a "defaults" section, it can be disabled
6537 in a specific instance by prepending the "no" keyword before it.
6538
Christopher Faulet315b39c2018-09-21 16:26:19 +02006539 See also : "option http-server-close" and "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01006540
6541
Emeric Brun3a058f32009-06-30 18:26:00 +02006542option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01006543 Enable logging of HTTP request, session state and timers
6544 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01006545 yes | yes | yes | no
Emeric Brun3a058f32009-06-30 18:26:00 +02006546 Arguments :
6547 clf if the "clf" argument is added, then the output format will be
6548 the CLF format instead of HAProxy's default HTTP format. You can
6549 use this when you need to feed HAProxy's logs through a specific
Davor Ocelice9ed2812017-12-25 17:49:28 +01006550 log analyzer which only support the CLF format and which is not
Emeric Brun3a058f32009-06-30 18:26:00 +02006551 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006552
6553 By default, the log output format is very poor, as it only contains the
6554 source and destination addresses, and the instance name. By specifying
6555 "option httplog", each log line turns into a much richer format including,
6556 but not limited to, the HTTP request, the connection timers, the session
6557 status, the connections numbers, the captured headers and cookies, the
6558 frontend, backend and server name, and of course the source address and
6559 ports.
6560
PiBa-NLbd556bf2014-12-11 21:31:54 +01006561 Specifying only "option httplog" will automatically clear the 'clf' mode
6562 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02006563
Guillaume de Lafond29f45602017-03-31 19:52:15 +02006564 "option httplog" overrides any previous "log-format" directive.
6565
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006566 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01006567
Willy Tarreau55165fe2009-05-10 12:02:55 +02006568
6569option http_proxy
6570no option http_proxy
6571 Enable or disable plain HTTP proxy mode
6572 May be used in sections : defaults | frontend | listen | backend
6573 yes | yes | yes | yes
6574 Arguments : none
6575
6576 It sometimes happens that people need a pure HTTP proxy which understands
6577 basic proxy requests without caching nor any fancy feature. In this case,
6578 it may be worth setting up an HAProxy instance with the "option http_proxy"
6579 set. In this mode, no server is declared, and the connection is forwarded to
6580 the IP address and port found in the URL after the "http://" scheme.
6581
6582 No host address resolution is performed, so this only works when pure IP
6583 addresses are passed. Since this option's usage perimeter is rather limited,
Lukas Tribusf01a9cd2016-02-03 18:09:37 +01006584 it will probably be used only by experts who know they need exactly it. This
6585 is incompatible with the HTTP tunnel mode.
Willy Tarreau55165fe2009-05-10 12:02:55 +02006586
6587 If this option has been enabled in a "defaults" section, it can be disabled
6588 in a specific instance by prepending the "no" keyword before it.
6589
6590 Example :
6591 # this backend understands HTTP proxy requests and forwards them directly.
6592 backend direct_forward
6593 option httpclose
6594 option http_proxy
6595
6596 See also : "option httpclose"
6597
Willy Tarreau211ad242009-10-03 21:45:07 +02006598
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006599option independent-streams
6600no option independent-streams
6601 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006602 May be used in sections : defaults | frontend | listen | backend
6603 yes | yes | yes | yes
6604 Arguments : none
6605
6606 By default, when data is sent over a socket, both the write timeout and the
6607 read timeout for that socket are refreshed, because we consider that there is
6608 activity on that socket, and we have no other means of guessing if we should
6609 receive data or not.
6610
Davor Ocelice9ed2812017-12-25 17:49:28 +01006611 While this default behavior is desirable for almost all applications, there
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006612 exists a situation where it is desirable to disable it, and only refresh the
6613 read timeout if there are incoming data. This happens on sessions with large
6614 timeouts and low amounts of exchanged data such as telnet session. If the
6615 server suddenly disappears, the output data accumulates in the system's
6616 socket buffers, both timeouts are correctly refreshed, and there is no way
6617 to know the server does not receive them, so we don't timeout. However, when
6618 the underlying protocol always echoes sent data, it would be enough by itself
6619 to detect the issue using the read timeout. Note that this problem does not
6620 happen with more verbose protocols because data won't accumulate long in the
6621 socket buffers.
6622
6623 When this option is set on the frontend, it will disable read timeout updates
6624 on data sent to the client. There probably is little use of this case. When
6625 the option is set on the backend, it will disable read timeout updates on
6626 data sent to the server. Doing so will typically break large HTTP posts from
6627 slow lines, so use it with caution.
6628
Lukas Tribus745f15e2018-11-08 12:41:42 +01006629 Note: older versions used to call this setting "option independant-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04006630 with a spelling mistake. This spelling is still supported but
6631 deprecated.
6632
Willy Tarreauce887fd2012-05-12 12:50:00 +02006633 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02006634
6635
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02006636option ldap-check
6637 Use LDAPv3 health checks for server testing
6638 May be used in sections : defaults | frontend | listen | backend
6639 yes | no | yes | yes
6640 Arguments : none
6641
6642 It is possible to test that the server correctly talks LDAPv3 instead of just
6643 testing that it accepts the TCP connection. When this option is set, an
6644 LDAPv3 anonymous simple bind message is sent to the server, and the response
6645 is analyzed to find an LDAPv3 bind response message.
6646
6647 The server is considered valid only when the LDAP response contains success
6648 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
6649
6650 Logging of bind requests is server dependent see your documentation how to
6651 configure it.
6652
6653 Example :
6654 option ldap-check
6655
6656 See also : "option httpchk"
6657
6658
Simon Horman98637e52014-06-20 12:30:16 +09006659option external-check
6660 Use external processes for server health checks
6661 May be used in sections : defaults | frontend | listen | backend
6662 yes | no | yes | yes
6663
6664 It is possible to test the health of a server using an external command.
6665 This is achieved by running the executable set using "external-check
6666 command".
6667
6668 Requires the "external-check" global to be set.
6669
6670 See also : "external-check", "external-check command", "external-check path"
6671
6672
Willy Tarreau211ad242009-10-03 21:45:07 +02006673option log-health-checks
6674no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02006675 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02006676 May be used in sections : defaults | frontend | listen | backend
6677 yes | no | yes | yes
6678 Arguments : none
6679
Willy Tarreaubef1b322014-05-13 21:01:39 +02006680 By default, failed health check are logged if server is UP and successful
6681 health checks are logged if server is DOWN, so the amount of additional
6682 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02006683
Willy Tarreaubef1b322014-05-13 21:01:39 +02006684 When this option is enabled, any change of the health check status or to
6685 the server's health will be logged, so that it becomes possible to know
6686 that a server was failing occasional checks before crashing, or exactly when
6687 it failed to respond a valid HTTP status, then when the port started to
6688 reject connections, then when the server stopped responding at all.
6689
Davor Ocelice9ed2812017-12-25 17:49:28 +01006690 Note that status changes not caused by health checks (e.g. enable/disable on
Willy Tarreaubef1b322014-05-13 21:01:39 +02006691 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02006692
Willy Tarreaubef1b322014-05-13 21:01:39 +02006693 See also: "option httpchk", "option ldap-check", "option mysql-check",
6694 "option pgsql-check", "option redis-check", "option smtpchk",
6695 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02006696
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006697
6698option log-separate-errors
6699no option log-separate-errors
6700 Change log level for non-completely successful connections
6701 May be used in sections : defaults | frontend | listen | backend
6702 yes | yes | yes | no
6703 Arguments : none
6704
6705 Sometimes looking for errors in logs is not easy. This option makes haproxy
6706 raise the level of logs containing potentially interesting information such
6707 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
6708 level changes from "info" to "err". This makes it possible to log them
6709 separately to a different file with most syslog daemons. Be careful not to
6710 remove them from the original file, otherwise you would lose ordering which
6711 provides very important information.
6712
6713 Using this option, large sites dealing with several thousand connections per
6714 second may log normal traffic to a rotating buffer and only archive smaller
6715 error logs.
6716
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006717 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02006718 logging.
6719
Willy Tarreauc27debf2008-01-06 08:57:02 +01006720
6721option logasap
6722no option logasap
6723 Enable or disable early logging of HTTP requests
6724 May be used in sections : defaults | frontend | listen | backend
6725 yes | yes | yes | no
6726 Arguments : none
6727
6728 By default, HTTP requests are logged upon termination so that the total
6729 transfer time and the number of bytes appear in the logs. When large objects
6730 are being transferred, it may take a while before the request appears in the
6731 logs. Using "option logasap", the request gets logged as soon as the server
6732 sends the complete headers. The only missing information in the logs will be
6733 the total number of bytes which will indicate everything except the amount
6734 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006735 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01006736 "Content-Length" response header so that the logs at least indicate how many
6737 bytes are expected to be transferred.
6738
Willy Tarreaucc6c8912009-02-22 10:53:55 +01006739 Examples :
6740 listen http_proxy 0.0.0.0:80
6741 mode http
6742 option httplog
6743 option logasap
6744 log 192.168.2.200 local3
6745
6746 >>> Feb 6 12:14:14 localhost \
6747 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
6748 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
6749 "GET /image.iso HTTP/1.0"
6750
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006751 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01006752 logging.
6753
6754
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006755option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006756 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006757 May be used in sections : defaults | frontend | listen | backend
6758 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006759 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02006760 <username> This is the username which will be used when connecting to MySQL
6761 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02006762 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006763
6764 If you specify a username, the check consists of sending two MySQL packet,
6765 one Client Authentication packet, and one QUIT packet, to correctly close
Davor Ocelice9ed2812017-12-25 17:49:28 +01006766 MySQL session. We then parse the MySQL Handshake Initialization packet and/or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006767 Error packet. It is a basic but useful test which does not produce error nor
6768 aborted connect on the server. However, it requires adding an authorization
6769 in the MySQL table, like this :
6770
6771 USE mysql;
6772 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
6773 FLUSH PRIVILEGES;
6774
6775 If you don't specify a username (it is deprecated and not recommended), the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006776 check only consists in parsing the Mysql Handshake Initialization packet or
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02006777 Error packet, we don't send anything in this mode. It was reported that it
6778 can generate lockout if check is too frequent and/or if there is not enough
6779 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
6780 value as if a connection is established successfully within fewer than MySQL
6781 "max_connect_errors" attempts after a previous connection was interrupted,
6782 the error count for the host is cleared to zero. If HAProxy's server get
6783 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
6784
6785 Remember that this does not check database presence nor database consistency.
6786 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006787
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02006788 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006789
6790 Most often, an incoming MySQL server needs to see the client's IP address for
6791 various purposes, including IP privilege matching and connection logging.
6792 When possible, it is often wise to masquerade the client's IP address when
6793 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02006794 which requires the transparent proxy feature to be compiled in, and the MySQL
6795 server to route the client via the machine hosting haproxy.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01006796
6797 See also: "option httpchk"
6798
6799
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006800option nolinger
6801no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006802 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006803 May be used in sections: defaults | frontend | listen | backend
6804 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006805 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006806
Davor Ocelice9ed2812017-12-25 17:49:28 +01006807 When clients or servers abort connections in a dirty way (e.g. they are
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006808 physically disconnected), the session timeouts triggers and the session is
6809 closed. But it will remain in FIN_WAIT1 state for some time in the system,
6810 using some resources and possibly limiting the ability to establish newer
6811 connections.
6812
6813 When this happens, it is possible to activate "option nolinger" which forces
6814 the system to immediately remove any socket's pending data on close. Thus,
6815 the session is instantly purged from the system's tables. This usually has
6816 side effects such as increased number of TCP resets due to old retransmits
6817 getting immediately rejected. Some firewalls may sometimes complain about
6818 this too.
6819
6820 For this reason, it is not recommended to use this option when not absolutely
6821 needed. You know that you need it when you have thousands of FIN_WAIT1
6822 sessions on your system (TIME_WAIT ones do not count).
6823
6824 This option may be used both on frontends and backends, depending on the side
6825 where it is required. Use it on the frontend for clients, and on the backend
6826 for servers.
6827
6828 If this option has been enabled in a "defaults" section, it can be disabled
6829 in a specific instance by prepending the "no" keyword before it.
6830
6831
Willy Tarreau55165fe2009-05-10 12:02:55 +02006832option originalto [ except <network> ] [ header <name> ]
6833 Enable insertion of the X-Original-To header to requests sent to servers
6834 May be used in sections : defaults | frontend | listen | backend
6835 yes | yes | yes | yes
6836 Arguments :
6837 <network> is an optional argument used to disable this option for sources
6838 matching <network>
6839 <name> an optional argument to specify a different "X-Original-To"
6840 header name.
6841
6842 Since HAProxy can work in transparent mode, every request from a client can
6843 be redirected to the proxy and HAProxy itself can proxy every request to a
6844 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
6845 be lost. This is annoying when you want access rules based on destination ip
6846 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
6847 added by HAProxy to all requests sent to the server. This header contains a
6848 value representing the original destination IP address. Since this must be
6849 configured to always use the last occurrence of this header only. Note that
6850 only the last occurrence of the header must be used, since it is really
6851 possible that the client has already brought one.
6852
6853 The keyword "header" may be used to supply a different header name to replace
6854 the default "X-Original-To". This can be useful where you might already
6855 have a "X-Original-To" header from a different application, and you need
6856 preserve it. Also if your backend server doesn't use the "X-Original-To"
6857 header and requires different one.
6858
6859 Sometimes, a same HAProxy instance may be shared between a direct client
6860 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
6861 used to decrypt HTTPS traffic). It is possible to disable the addition of the
6862 header for a known source address or network by adding the "except" keyword
6863 followed by the network address. In this case, any source IP matching the
6864 network will not cause an addition of this header. Most common uses are with
6865 private networks or 127.0.0.1.
6866
6867 This option may be specified either in the frontend or in the backend. If at
6868 least one of them uses it, the header will be added. Note that the backend's
6869 setting of the header subargument takes precedence over the frontend's if
6870 both are defined.
6871
Willy Tarreau55165fe2009-05-10 12:02:55 +02006872 Examples :
6873 # Original Destination address
6874 frontend www
6875 mode http
6876 option originalto except 127.0.0.1
6877
6878 # Those servers want the IP Address in X-Client-Dst
6879 backend www
6880 mode http
6881 option originalto header X-Client-Dst
6882
Christopher Faulet315b39c2018-09-21 16:26:19 +02006883 See also : "option httpclose", "option http-server-close".
Willy Tarreau55165fe2009-05-10 12:02:55 +02006884
6885
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006886option persist
6887no option persist
6888 Enable or disable forced persistence on down servers
6889 May be used in sections: defaults | frontend | listen | backend
6890 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006891 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006892
6893 When an HTTP request reaches a backend with a cookie which references a dead
6894 server, by default it is redispatched to another server. It is possible to
6895 force the request to be sent to the dead server first using "option persist"
6896 if absolutely needed. A common use case is when servers are under extreme
6897 load and spend their time flapping. In this case, the users would still be
6898 directed to the server they opened the session on, in the hope they would be
6899 correctly served. It is recommended to use "option redispatch" in conjunction
6900 with this option so that in the event it would not be possible to connect to
6901 the server at all (server definitely dead), the client would finally be
6902 redirected to another valid server.
6903
6904 If this option has been enabled in a "defaults" section, it can be disabled
6905 in a specific instance by prepending the "no" keyword before it.
6906
Willy Tarreau4de91492010-01-22 19:10:05 +01006907 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006908
6909
Willy Tarreau0c122822013-12-15 18:49:01 +01006910option pgsql-check [ user <username> ]
6911 Use PostgreSQL health checks for server testing
6912 May be used in sections : defaults | frontend | listen | backend
6913 yes | no | yes | yes
6914 Arguments :
6915 <username> This is the username which will be used when connecting to
6916 PostgreSQL server.
6917
6918 The check sends a PostgreSQL StartupMessage and waits for either
6919 Authentication request or ErrorResponse message. It is a basic but useful
6920 test which does not produce error nor aborted connect on the server.
6921 This check is identical with the "mysql-check".
6922
6923 See also: "option httpchk"
6924
6925
Willy Tarreau9420b122013-12-15 18:58:25 +01006926option prefer-last-server
6927no option prefer-last-server
6928 Allow multiple load balanced requests to remain on the same server
6929 May be used in sections: defaults | frontend | listen | backend
6930 yes | no | yes | yes
6931 Arguments : none
6932
6933 When the load balancing algorithm in use is not deterministic, and a previous
6934 request was sent to a server to which haproxy still holds a connection, it is
6935 sometimes desirable that subsequent requests on a same session go to the same
6936 server as much as possible. Note that this is different from persistence, as
6937 we only indicate a preference which haproxy tries to apply without any form
6938 of warranty. The real use is for keep-alive connections sent to servers. When
6939 this option is used, haproxy will try to reuse the same connection that is
6940 attached to the server instead of rebalancing to another server, causing a
6941 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01006942 not make much sense to use this in combination with hashing algorithms. Note,
6943 haproxy already automatically tries to stick to a server which sends a 401 or
Lukas Tribus80512b12018-10-27 20:07:40 +02006944 to a proxy which sends a 407 (authentication required), when the load
6945 balancing algorithm is not deterministic. This is mandatory for use with the
6946 broken NTLM authentication challenge, and significantly helps in
Willy Tarreau068621e2013-12-23 15:11:25 +01006947 troubleshooting some faulty applications. Option prefer-last-server might be
6948 desirable in these environments as well, to avoid redistributing the traffic
6949 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01006950
6951 If this option has been enabled in a "defaults" section, it can be disabled
6952 in a specific instance by prepending the "no" keyword before it.
6953
6954 See also: "option http-keep-alive"
6955
6956
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006957option redispatch
Joseph Lynch726ab712015-05-11 23:25:34 -07006958option redispatch <interval>
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006959no option redispatch
6960 Enable or disable session redistribution in case of connection failure
6961 May be used in sections: defaults | frontend | listen | backend
6962 yes | no | yes | yes
Joseph Lynch726ab712015-05-11 23:25:34 -07006963 Arguments :
6964 <interval> The optional integer value that controls how often redispatches
6965 occur when retrying connections. Positive value P indicates a
6966 redispatch is desired on every Pth retry, and negative value
Davor Ocelice9ed2812017-12-25 17:49:28 +01006967 N indicate a redispatch is desired on the Nth retry prior to the
Joseph Lynch726ab712015-05-11 23:25:34 -07006968 last retry. For example, the default of -1 preserves the
Davor Ocelice9ed2812017-12-25 17:49:28 +01006969 historical behavior of redispatching on the last retry, a
Joseph Lynch726ab712015-05-11 23:25:34 -07006970 positive value of 1 would indicate a redispatch on every retry,
6971 and a positive value of 3 would indicate a redispatch on every
6972 third retry. You can disable redispatches with a value of 0.
6973
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006974
6975 In HTTP mode, if a server designated by a cookie is down, clients may
6976 definitely stick to it because they cannot flush the cookie, so they will not
6977 be able to access the service anymore.
6978
Willy Tarreau59884a62019-01-02 14:48:31 +01006979 Specifying "option redispatch" will allow the proxy to break cookie or
6980 consistent hash based persistence and redistribute them to a working server.
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006981
Joseph Lynch726ab712015-05-11 23:25:34 -07006982 It also allows to retry connections to another server in case of multiple
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006983 connection failures. Of course, it requires having "retries" set to a nonzero
6984 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01006985
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006986 This form is the preferred form, which replaces both the "redispatch" and
6987 "redisp" keywords.
6988
6989 If this option has been enabled in a "defaults" section, it can be disabled
6990 in a specific instance by prepending the "no" keyword before it.
6991
Willy Tarreau4de91492010-01-22 19:10:05 +01006992 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006993
Willy Tarreaua453bdd2008-01-08 19:50:52 +01006994
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02006995option redis-check
6996 Use redis health checks for server testing
6997 May be used in sections : defaults | frontend | listen | backend
6998 yes | no | yes | yes
6999 Arguments : none
7000
7001 It is possible to test that the server correctly talks REDIS protocol instead
7002 of just testing that it accepts the TCP connection. When this option is set,
7003 a PING redis command is sent to the server, and the response is analyzed to
7004 find the "+PONG" response message.
7005
7006 Example :
7007 option redis-check
7008
Jarno Huuskonene5ae7022017-04-03 14:36:21 +03007009 See also : "option httpchk", "option tcp-check", "tcp-check expect"
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02007010
7011
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007012option smtpchk
7013option smtpchk <hello> <domain>
7014 Use SMTP health checks for server testing
7015 May be used in sections : defaults | frontend | listen | backend
7016 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01007017 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007018 <hello> is an optional argument. It is the "hello" command to use. It can
Lukas Tribus27935782018-10-01 02:00:16 +02007019 be either "HELO" (for SMTP) or "EHLO" (for ESMTP). All other
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007020 values will be turned into the default command ("HELO").
7021
7022 <domain> is the domain name to present to the server. It may only be
7023 specified (and is mandatory) if the hello command has been
7024 specified. By default, "localhost" is used.
7025
7026 When "option smtpchk" is set, the health checks will consist in TCP
7027 connections followed by an SMTP command. By default, this command is
7028 "HELO localhost". The server's return code is analyzed and only return codes
7029 starting with a "2" will be considered as valid. All other responses,
7030 including a lack of response will constitute an error and will indicate a
7031 dead server.
7032
7033 This test is meant to be used with SMTP servers or relays. Depending on the
7034 request, it is possible that some servers do not log each connection attempt,
Davor Ocelice9ed2812017-12-25 17:49:28 +01007035 so you may want to experiment to improve the behavior. Using telnet on port
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007036 25 is often easier than adjusting the configuration.
7037
7038 Most often, an incoming SMTP server needs to see the client's IP address for
7039 various purposes, including spam filtering, anti-spoofing and logging. When
7040 possible, it is often wise to masquerade the client's IP address when
7041 connecting to the server using the "usesrc" argument of the "source" keyword,
Willy Tarreau29fbe512015-08-20 19:35:14 +02007042 which requires the transparent proxy feature to be compiled in.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007043
7044 Example :
7045 option smtpchk HELO mydomain.org
7046
7047 See also : "option httpchk", "source"
7048
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007049
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02007050option socket-stats
7051no option socket-stats
7052
7053 Enable or disable collecting & providing separate statistics for each socket.
7054 May be used in sections : defaults | frontend | listen | backend
7055 yes | yes | yes | no
7056
7057 Arguments : none
7058
7059
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007060option splice-auto
7061no option splice-auto
7062 Enable or disable automatic kernel acceleration on sockets in both directions
7063 May be used in sections : defaults | frontend | listen | backend
7064 yes | yes | yes | yes
7065 Arguments : none
7066
7067 When this option is enabled either on a frontend or on a backend, haproxy
7068 will automatically evaluate the opportunity to use kernel tcp splicing to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007069 forward data between the client and the server, in either direction. HAProxy
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007070 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007071 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007072 are not much aggressive in order to limit excessive use of splicing. This
7073 option requires splicing to be enabled at compile time, and may be globally
7074 disabled with the global option "nosplice". Since splice uses pipes, using it
7075 requires that there are enough spare pipes.
7076
7077 Important note: kernel-based TCP splicing is a Linux-specific feature which
7078 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
7079 transfer data between sockets without copying these data to user-space, thus
7080 providing noticeable performance gains and CPU cycles savings. Since many
7081 early implementations are buggy, corrupt data and/or are inefficient, this
7082 feature is not enabled by default, and it should be used with extreme care.
7083 While it is not possible to detect the correctness of an implementation,
7084 2.6.29 is the first version offering a properly working implementation. In
7085 case of doubt, splicing may be globally disabled using the global "nosplice"
7086 keyword.
7087
7088 Example :
7089 option splice-auto
7090
7091 If this option has been enabled in a "defaults" section, it can be disabled
7092 in a specific instance by prepending the "no" keyword before it.
7093
7094 See also : "option splice-request", "option splice-response", and global
7095 options "nosplice" and "maxpipes"
7096
7097
7098option splice-request
7099no option splice-request
7100 Enable or disable automatic kernel acceleration on sockets for requests
7101 May be used in sections : defaults | frontend | listen | backend
7102 yes | yes | yes | yes
7103 Arguments : none
7104
7105 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007106 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007107 the client to the server. It might still use the recv/send scheme if there
7108 are no spare pipes left. This option requires splicing to be enabled at
7109 compile time, and may be globally disabled with the global option "nosplice".
7110 Since splice uses pipes, using it requires that there are enough spare pipes.
7111
7112 Important note: see "option splice-auto" for usage limitations.
7113
7114 Example :
7115 option splice-request
7116
7117 If this option has been enabled in a "defaults" section, it can be disabled
7118 in a specific instance by prepending the "no" keyword before it.
7119
7120 See also : "option splice-auto", "option splice-response", and global options
7121 "nosplice" and "maxpipes"
7122
7123
7124option splice-response
7125no option splice-response
7126 Enable or disable automatic kernel acceleration on sockets for responses
7127 May be used in sections : defaults | frontend | listen | backend
7128 yes | yes | yes | yes
7129 Arguments : none
7130
7131 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007132 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01007133 the server to the client. It might still use the recv/send scheme if there
7134 are no spare pipes left. This option requires splicing to be enabled at
7135 compile time, and may be globally disabled with the global option "nosplice".
7136 Since splice uses pipes, using it requires that there are enough spare pipes.
7137
7138 Important note: see "option splice-auto" for usage limitations.
7139
7140 Example :
7141 option splice-response
7142
7143 If this option has been enabled in a "defaults" section, it can be disabled
7144 in a specific instance by prepending the "no" keyword before it.
7145
7146 See also : "option splice-auto", "option splice-request", and global options
7147 "nosplice" and "maxpipes"
7148
7149
Christopher Fauletba7bc162016-11-07 21:07:38 +01007150option spop-check
7151 Use SPOP health checks for server testing
7152 May be used in sections : defaults | frontend | listen | backend
7153 no | no | no | yes
7154 Arguments : none
7155
7156 It is possible to test that the server correctly talks SPOP protocol instead
7157 of just testing that it accepts the TCP connection. When this option is set,
7158 a HELLO handshake is performed between HAProxy and the server, and the
7159 response is analyzed to check no error is reported.
7160
7161 Example :
7162 option spop-check
7163
7164 See also : "option httpchk"
7165
7166
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007167option srvtcpka
7168no option srvtcpka
7169 Enable or disable the sending of TCP keepalive packets on the server side
7170 May be used in sections : defaults | frontend | listen | backend
7171 yes | no | yes | yes
7172 Arguments : none
7173
7174 When there is a firewall or any session-aware component between a client and
7175 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007176 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007177 components decides to expire a session which has remained idle for too long.
7178
7179 Enabling socket-level TCP keep-alives makes the system regularly send packets
7180 to the other end of the connection, leaving it active. The delay between
7181 keep-alive probes is controlled by the system only and depends both on the
7182 operating system and its tuning parameters.
7183
7184 It is important to understand that keep-alive packets are neither emitted nor
7185 received at the application level. It is only the network stacks which sees
7186 them. For this reason, even if one side of the proxy already uses keep-alives
7187 to maintain its connection alive, those keep-alive packets will not be
7188 forwarded to the other side of the proxy.
7189
7190 Please note that this has nothing to do with HTTP keep-alive.
7191
7192 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
7193 server side of a connection, which should help when session expirations are
7194 noticed between HAProxy and a server.
7195
7196 If this option has been enabled in a "defaults" section, it can be disabled
7197 in a specific instance by prepending the "no" keyword before it.
7198
7199 See also : "option clitcpka", "option tcpka"
7200
7201
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007202option ssl-hello-chk
7203 Use SSLv3 client hello health checks for server testing
7204 May be used in sections : defaults | frontend | listen | backend
7205 yes | no | yes | yes
7206 Arguments : none
7207
7208 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
7209 possible to test that the server correctly talks SSL instead of just testing
7210 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
7211 SSLv3 client hello messages are sent once the connection is established to
7212 the server, and the response is analyzed to find an SSL server hello message.
7213 The server is considered valid only when the response contains this server
7214 hello message.
7215
7216 All servers tested till there correctly reply to SSLv3 client hello messages,
7217 and most servers tested do not even log the requests containing only hello
7218 messages, which is appreciable.
7219
Willy Tarreau763a95b2012-10-04 23:15:39 +02007220 Note that this check works even when SSL support was not built into haproxy
7221 because it forges the SSL message. When SSL support is available, it is best
7222 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007223
Willy Tarreau763a95b2012-10-04 23:15:39 +02007224 See also: "option httpchk", "check-ssl"
7225
Willy Tarreaua453bdd2008-01-08 19:50:52 +01007226
Willy Tarreaued179852013-12-16 01:07:00 +01007227option tcp-check
7228 Perform health checks using tcp-check send/expect sequences
7229 May be used in sections: defaults | frontend | listen | backend
7230 yes | no | yes | yes
7231
7232 This health check method is intended to be combined with "tcp-check" command
7233 lists in order to support send/expect types of health check sequences.
7234
7235 TCP checks currently support 4 modes of operations :
7236 - no "tcp-check" directive : the health check only consists in a connection
7237 attempt, which remains the default mode.
7238
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007239 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01007240 used to send a string along with a connection opening. With some
7241 protocols, it helps sending a "QUIT" message for example that prevents
7242 the server from logging a connection error for each health check. The
7243 check result will still be based on the ability to open the connection
7244 only.
7245
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007246 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01007247 The connection is opened and haproxy waits for the server to present some
7248 contents which must validate some rules. The check result will be based
7249 on the matching between the contents and the rules. This is suited for
7250 POP, IMAP, SMTP, FTP, SSH, TELNET.
7251
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007252 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Davor Ocelice9ed2812017-12-25 17:49:28 +01007253 used to test a hello-type protocol. HAProxy sends a message, the server
7254 responds and its response is analyzed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007255 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01007256 suited for protocols which require a binding or a request/response model.
7257 LDAP, MySQL, Redis and SSL are example of such protocols, though they
7258 already all have their dedicated checks with a deeper understanding of
7259 the respective protocols.
7260 In this mode, many questions may be sent and many answers may be
Davor Ocelice9ed2812017-12-25 17:49:28 +01007261 analyzed.
Willy Tarreaued179852013-12-16 01:07:00 +01007262
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007263 A fifth mode can be used to insert comments in different steps of the
7264 script.
7265
7266 For each tcp-check rule you create, you can add a "comment" directive,
7267 followed by a string. This string will be reported in the log and stderr
7268 in debug mode. It is useful to make user-friendly error reporting.
7269 The "comment" is of course optional.
7270
7271
Willy Tarreaued179852013-12-16 01:07:00 +01007272 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007273 # perform a POP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007274 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007275 tcp-check expect string +OK\ POP3\ ready comment POP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007276
Davor Ocelice9ed2812017-12-25 17:49:28 +01007277 # perform an IMAP check (analyze only server's banner)
Willy Tarreaued179852013-12-16 01:07:00 +01007278 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007279 tcp-check expect string *\ OK\ IMAP4\ ready comment IMAP\ protocol
Willy Tarreaued179852013-12-16 01:07:00 +01007280
7281 # look for the redis master server after ensuring it speaks well
7282 # redis protocol, then it exits properly.
Davor Ocelice9ed2812017-12-25 17:49:28 +01007283 # (send a command then analyze the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01007284 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007285 tcp-check comment PING\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007286 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02007287 tcp-check expect string +PONG
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007288 tcp-check comment role\ check
Willy Tarreaued179852013-12-16 01:07:00 +01007289 tcp-check send info\ replication\r\n
7290 tcp-check expect string role:master
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007291 tcp-check comment QUIT\ phase
Willy Tarreaued179852013-12-16 01:07:00 +01007292 tcp-check send QUIT\r\n
7293 tcp-check expect string +OK
7294
Davor Ocelice9ed2812017-12-25 17:49:28 +01007295 forge a HTTP request, then analyze the response
Willy Tarreaued179852013-12-16 01:07:00 +01007296 (send many headers before analyzing)
7297 option tcp-check
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007298 tcp-check comment forge\ and\ send\ HTTP\ request
Willy Tarreaued179852013-12-16 01:07:00 +01007299 tcp-check send HEAD\ /\ HTTP/1.1\r\n
7300 tcp-check send Host:\ www.mydomain.com\r\n
7301 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
7302 tcp-check send \r\n
Baptiste Assmannd60a9e52015-04-25 16:27:23 +02007303 tcp-check expect rstring HTTP/1\..\ (2..|3..) comment check\ HTTP\ response
Willy Tarreaued179852013-12-16 01:07:00 +01007304
7305
7306 See also : "tcp-check expect", "tcp-check send"
7307
7308
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007309option tcp-smart-accept
7310no option tcp-smart-accept
7311 Enable or disable the saving of one ACK packet during the accept sequence
7312 May be used in sections : defaults | frontend | listen | backend
7313 yes | yes | yes | no
7314 Arguments : none
7315
7316 When an HTTP connection request comes in, the system acknowledges it on
7317 behalf of HAProxy, then the client immediately sends its request, and the
7318 system acknowledges it too while it is notifying HAProxy about the new
7319 connection. HAProxy then reads the request and responds. This means that we
7320 have one TCP ACK sent by the system for nothing, because the request could
7321 very well be acknowledged by HAProxy when it sends its response.
7322
7323 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
7324 sending this useless ACK on platforms which support it (currently at least
7325 Linux). It must not cause any problem, because the system will send it anyway
7326 after 40 ms if the response takes more time than expected to come.
7327
7328 During complex network debugging sessions, it may be desirable to disable
7329 this optimization because delayed ACKs can make troubleshooting more complex
7330 when trying to identify where packets are delayed. It is then possible to
Davor Ocelice9ed2812017-12-25 17:49:28 +01007331 fall back to normal behavior by specifying "no option tcp-smart-accept".
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007332
7333 It is also possible to force it for non-HTTP proxies by simply specifying
7334 "option tcp-smart-accept". For instance, it can make sense with some services
7335 such as SMTP where the server speaks first.
7336
7337 It is recommended to avoid forcing this option in a defaults section. In case
7338 of doubt, consider setting it back to automatic values by prepending the
7339 "default" keyword before it, or disabling it using the "no" keyword.
7340
Willy Tarreaud88edf22009-06-14 15:48:17 +02007341 See also : "option tcp-smart-connect"
7342
7343
7344option tcp-smart-connect
7345no option tcp-smart-connect
7346 Enable or disable the saving of one ACK packet during the connect sequence
7347 May be used in sections : defaults | frontend | listen | backend
7348 yes | no | yes | yes
7349 Arguments : none
7350
7351 On certain systems (at least Linux), HAProxy can ask the kernel not to
7352 immediately send an empty ACK upon a connection request, but to directly
7353 send the buffer request instead. This saves one packet on the network and
7354 thus boosts performance. It can also be useful for some servers, because they
7355 immediately get the request along with the incoming connection.
7356
7357 This feature is enabled when "option tcp-smart-connect" is set in a backend.
7358 It is not enabled by default because it makes network troubleshooting more
7359 complex.
7360
7361 It only makes sense to enable it with protocols where the client speaks first
7362 such as HTTP. In other situations, if there is no data to send in place of
7363 the ACK, a normal ACK is sent.
7364
7365 If this option has been enabled in a "defaults" section, it can be disabled
7366 in a specific instance by prepending the "no" keyword before it.
7367
7368 See also : "option tcp-smart-accept"
7369
Willy Tarreau9ea05a72009-06-14 12:07:01 +02007370
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007371option tcpka
7372 Enable or disable the sending of TCP keepalive packets on both sides
7373 May be used in sections : defaults | frontend | listen | backend
7374 yes | yes | yes | yes
7375 Arguments : none
7376
7377 When there is a firewall or any session-aware component between a client and
7378 a server, and when the protocol involves very long sessions with long idle
Davor Ocelice9ed2812017-12-25 17:49:28 +01007379 periods (e.g. remote desktops), there is a risk that one of the intermediate
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007380 components decides to expire a session which has remained idle for too long.
7381
7382 Enabling socket-level TCP keep-alives makes the system regularly send packets
7383 to the other end of the connection, leaving it active. The delay between
7384 keep-alive probes is controlled by the system only and depends both on the
7385 operating system and its tuning parameters.
7386
7387 It is important to understand that keep-alive packets are neither emitted nor
7388 received at the application level. It is only the network stacks which sees
7389 them. For this reason, even if one side of the proxy already uses keep-alives
7390 to maintain its connection alive, those keep-alive packets will not be
7391 forwarded to the other side of the proxy.
7392
7393 Please note that this has nothing to do with HTTP keep-alive.
7394
7395 Using option "tcpka" enables the emission of TCP keep-alive probes on both
7396 the client and server sides of a connection. Note that this is meaningful
7397 only in "defaults" or "listen" sections. If this option is used in a
7398 frontend, only the client side will get keep-alives, and if this option is
7399 used in a backend, only the server side will get keep-alives. For this
7400 reason, it is strongly recommended to explicitly use "option clitcpka" and
7401 "option srvtcpka" when the configuration is split between frontends and
7402 backends.
7403
7404 See also : "option clitcpka", "option srvtcpka"
7405
Willy Tarreau844e3c52008-01-11 16:28:18 +01007406
7407option tcplog
7408 Enable advanced logging of TCP connections with session state and timers
7409 May be used in sections : defaults | frontend | listen | backend
Tim Duesterhus9ad9f352018-02-05 20:52:27 +01007410 yes | yes | yes | no
Willy Tarreau844e3c52008-01-11 16:28:18 +01007411 Arguments : none
7412
7413 By default, the log output format is very poor, as it only contains the
7414 source and destination addresses, and the instance name. By specifying
7415 "option tcplog", each log line turns into a much richer format including, but
7416 not limited to, the connection timers, the session status, the connections
7417 numbers, the frontend, backend and server name, and of course the source
7418 address and ports. This option is useful for pure TCP proxies in order to
7419 find which of the client or server disconnects or times out. For normal HTTP
7420 proxies, it's better to use "option httplog" which is even more complete.
7421
Guillaume de Lafond29f45602017-03-31 19:52:15 +02007422 "option tcplog" overrides any previous "log-format" directive.
7423
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007424 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007425
7426
Willy Tarreau844e3c52008-01-11 16:28:18 +01007427option transparent
7428no option transparent
7429 Enable client-side transparent proxying
7430 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01007431 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01007432 Arguments : none
7433
7434 This option was introduced in order to provide layer 7 persistence to layer 3
7435 load balancers. The idea is to use the OS's ability to redirect an incoming
7436 connection for a remote address to a local process (here HAProxy), and let
7437 this process know what address was initially requested. When this option is
7438 used, sessions without cookies will be forwarded to the original destination
7439 IP address of the incoming request (which should match that of another
7440 equipment), while requests with cookies will still be forwarded to the
7441 appropriate server.
7442
7443 Note that contrary to a common belief, this option does NOT make HAProxy
7444 present the client's IP to the server when establishing the connection.
7445
Willy Tarreaua1146052011-03-01 09:51:54 +01007446 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007447 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01007448
Willy Tarreaubf1f8162007-12-28 17:42:56 +01007449
Simon Horman98637e52014-06-20 12:30:16 +09007450external-check command <command>
7451 Executable to run when performing an external-check
7452 May be used in sections : defaults | frontend | listen | backend
7453 yes | no | yes | yes
7454
7455 Arguments :
7456 <command> is the external command to run
7457
Simon Horman98637e52014-06-20 12:30:16 +09007458 The arguments passed to the to the command are:
7459
Cyril Bonté777be862014-12-02 21:21:35 +01007460 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09007461
Cyril Bonté777be862014-12-02 21:21:35 +01007462 The <proxy_address> and <proxy_port> are derived from the first listener
7463 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
7464 listener the proxy_address will be the path of the socket and the
7465 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
7466 possible to determine a listener, and both <proxy_address> and <proxy_port>
7467 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09007468
Cyril Bonté72cda2a2014-12-27 22:28:39 +01007469 Some values are also provided through environment variables.
7470
7471 Environment variables :
7472 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
7473 applicable, for example in a "backend" section).
7474
7475 HAPROXY_PROXY_ID The backend id.
7476
7477 HAPROXY_PROXY_NAME The backend name.
7478
7479 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
7480 applicable, for example in a "backend" section or
7481 for a UNIX socket).
7482
7483 HAPROXY_SERVER_ADDR The server address.
7484
7485 HAPROXY_SERVER_CURCONN The current number of connections on the server.
7486
7487 HAPROXY_SERVER_ID The server id.
7488
7489 HAPROXY_SERVER_MAXCONN The server max connections.
7490
7491 HAPROXY_SERVER_NAME The server name.
7492
7493 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
7494 socket).
7495
7496 PATH The PATH environment variable used when executing
7497 the command may be set using "external-check path".
7498
Simon Horman98637e52014-06-20 12:30:16 +09007499 If the command executed and exits with a zero status then the check is
7500 considered to have passed, otherwise the check is considered to have
7501 failed.
7502
7503 Example :
7504 external-check command /bin/true
7505
7506 See also : "external-check", "option external-check", "external-check path"
7507
7508
7509external-check path <path>
7510 The value of the PATH environment variable used when running an external-check
7511 May be used in sections : defaults | frontend | listen | backend
7512 yes | no | yes | yes
7513
7514 Arguments :
7515 <path> is the path used when executing external command to run
7516
7517 The default path is "".
7518
7519 Example :
7520 external-check path "/usr/bin:/bin"
7521
7522 See also : "external-check", "option external-check",
7523 "external-check command"
7524
7525
Emeric Brun647caf12009-06-30 17:57:00 +02007526persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02007527persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02007528 Enable RDP cookie-based persistence
7529 May be used in sections : defaults | frontend | listen | backend
7530 yes | no | yes | yes
7531 Arguments :
7532 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02007533 default cookie name "msts" will be used. There currently is no
7534 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02007535
7536 This statement enables persistence based on an RDP cookie. The RDP cookie
7537 contains all information required to find the server in the list of known
Davor Ocelice9ed2812017-12-25 17:49:28 +01007538 servers. So when this option is set in the backend, the request is analyzed
Emeric Brun647caf12009-06-30 17:57:00 +02007539 and if an RDP cookie is found, it is decoded. If it matches a known server
7540 which is still UP (or if "option persist" is set), then the connection is
7541 forwarded to this server.
7542
7543 Note that this only makes sense in a TCP backend, but for this to work, the
7544 frontend must have waited long enough to ensure that an RDP cookie is present
7545 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007546 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02007547 a single "listen" section.
7548
Willy Tarreau61e28f22010-05-16 22:31:05 +02007549 Also, it is important to understand that the terminal server will emit this
7550 RDP cookie only if it is configured for "token redirection mode", which means
7551 that the "IP address redirection" option is disabled.
7552
Emeric Brun647caf12009-06-30 17:57:00 +02007553 Example :
7554 listen tse-farm
7555 bind :3389
7556 # wait up to 5s for an RDP cookie in the request
7557 tcp-request inspect-delay 5s
7558 tcp-request content accept if RDP_COOKIE
7559 # apply RDP cookie persistence
7560 persist rdp-cookie
7561 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02007562 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02007563 balance rdp-cookie
7564 server srv1 1.1.1.1:3389
7565 server srv2 1.1.1.2:3389
7566
Simon Hormanab814e02011-06-24 14:50:20 +09007567 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
7568 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02007569
7570
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007571rate-limit sessions <rate>
7572 Set a limit on the number of new sessions accepted per second on a frontend
7573 May be used in sections : defaults | frontend | listen | backend
7574 yes | yes | yes | no
7575 Arguments :
7576 <rate> The <rate> parameter is an integer designating the maximum number
7577 of new sessions per second to accept on the frontend.
7578
7579 When the frontend reaches the specified number of new sessions per second, it
7580 stops accepting new connections until the rate drops below the limit again.
7581 During this time, the pending sessions will be kept in the socket's backlog
7582 (in system buffers) and haproxy will not even be aware that sessions are
7583 pending. When applying very low limit on a highly loaded service, it may make
7584 sense to increase the socket's backlog using the "backlog" keyword.
7585
7586 This feature is particularly efficient at blocking connection-based attacks
7587 or service abuse on fragile servers. Since the session rate is measured every
7588 millisecond, it is extremely accurate. Also, the limit applies immediately,
7589 no delay is needed at all to detect the threshold.
7590
7591 Example : limit the connection rate on SMTP to 10 per second max
7592 listen smtp
7593 mode tcp
7594 bind :25
7595 rate-limit sessions 10
Panagiotis Panagiotopoulos7282d8e2016-02-11 16:37:15 +02007596 server smtp1 127.0.0.1:1025
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007597
Willy Tarreaua17c2d92011-07-25 08:16:20 +02007598 Note : when the maximum rate is reached, the frontend's status is not changed
7599 but its sockets appear as "WAITING" in the statistics if the
7600 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01007601
7602 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
7603
7604
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007605redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
7606redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
7607redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007608 Return an HTTP redirection if/unless a condition is matched
7609 May be used in sections : defaults | frontend | listen | backend
7610 no | yes | yes | yes
7611
7612 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01007613 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007614
Willy Tarreau0140f252008-11-19 21:07:09 +01007615 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007616 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007617 the HTTP "Location" header. When used in an "http-request" rule,
7618 <loc> value follows the log-format rules and can include some
7619 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007620
7621 <pfx> With "redirect prefix", the "Location" header is built from the
7622 concatenation of <pfx> and the complete URI path, including the
7623 query string, unless the "drop-query" option is specified (see
7624 below). As a special case, if <pfx> equals exactly "/", then
7625 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007626 redirect to the same URL (for instance, to insert a cookie). When
7627 used in an "http-request" rule, <pfx> value follows the log-format
7628 rules and can include some dynamic values (see Custom Log Format
7629 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007630
7631 <sch> With "redirect scheme", then the "Location" header is built by
7632 concatenating <sch> with "://" then the first occurrence of the
7633 "Host" header, and then the URI path, including the query string
7634 unless the "drop-query" option is specified (see below). If no
7635 path is found or if the path is "*", then "/" is used instead. If
7636 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007637 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007638 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007639 HTTPS. When used in an "http-request" rule, <sch> value follows
7640 the log-format rules and can include some dynamic values (see
7641 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01007642
7643 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007644 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
7645 with 302 used by default if no code is specified. 301 means
7646 "Moved permanently", and a browser may cache the Location. 302
Baptiste Assmannea849c02015-08-03 11:42:50 +02007647 means "Moved temporarily" and means that the browser should not
Willy Tarreaub67fdc42013-03-29 19:28:11 +01007648 cache the redirection. 303 is equivalent to 302 except that the
7649 browser will fetch the location with a GET method. 307 is just
7650 like 302 but makes it clear that the same method must be reused.
7651 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01007652
7653 <option> There are several options which can be specified to adjust the
Davor Ocelice9ed2812017-12-25 17:49:28 +01007654 expected behavior of a redirection :
Willy Tarreau0140f252008-11-19 21:07:09 +01007655
7656 - "drop-query"
7657 When this keyword is used in a prefix-based redirection, then the
7658 location will be set without any possible query-string, which is useful
7659 for directing users to a non-secure page for instance. It has no effect
7660 with a location-type redirect.
7661
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007662 - "append-slash"
7663 This keyword may be used in conjunction with "drop-query" to redirect
7664 users who use a URL not ending with a '/' to the same one with the '/'.
7665 It can be useful to ensure that search engines will only see one URL.
7666 For this, a return code 301 is preferred.
7667
Willy Tarreau0140f252008-11-19 21:07:09 +01007668 - "set-cookie NAME[=value]"
7669 A "Set-Cookie" header will be added with NAME (and optionally "=value")
7670 to the response. This is sometimes used to indicate that a user has
7671 been seen, for instance to protect against some types of DoS. No other
7672 cookie option is added, so the cookie will be a session cookie. Note
7673 that for a browser, a sole cookie name without an equal sign is
7674 different from a cookie with an equal sign.
7675
7676 - "clear-cookie NAME[=]"
7677 A "Set-Cookie" header will be added with NAME (and optionally "="), but
7678 with the "Max-Age" attribute set to zero. This will tell the browser to
7679 delete this cookie. It is useful for instance on logout pages. It is
7680 important to note that clearing the cookie "NAME" will not remove a
7681 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
7682 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007683
7684 Example: move the login URL only to HTTPS.
7685 acl clear dst_port 80
7686 acl secure dst_port 8080
7687 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01007688 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01007689 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01007690 acl cookie_set hdr_sub(cookie) SEEN=1
7691
7692 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01007693 redirect prefix https://mysite.com if login_page !secure
7694 redirect prefix http://mysite.com drop-query if login_page !uid_given
7695 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01007696 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007697
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01007698 Example: send redirects for request for articles without a '/'.
7699 acl missing_slash path_reg ^/article/[^/]*$
7700 redirect code 301 prefix / drop-query append-slash if missing_slash
7701
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007702 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01007703 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02007704
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007705 Example: append 'www.' prefix in front of all hosts not having it
Coen Rosdorff596659b2016-04-11 11:33:49 +02007706 http-request redirect code 301 location \
7707 http://www.%[hdr(host)]%[capture.req.uri] \
7708 unless { hdr_beg(host) -i www }
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01007709
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007710 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02007711
7712
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007713redisp (deprecated)
7714redispatch (deprecated)
7715 Enable or disable session redistribution in case of connection failure
7716 May be used in sections: defaults | frontend | listen | backend
7717 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007718 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007719
7720 In HTTP mode, if a server designated by a cookie is down, clients may
7721 definitely stick to it because they cannot flush the cookie, so they will not
7722 be able to access the service anymore.
7723
7724 Specifying "redispatch" will allow the proxy to break their persistence and
7725 redistribute them to a working server.
7726
7727 It also allows to retry last connection to another server in case of multiple
7728 connection failures. Of course, it requires having "retries" set to a nonzero
7729 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01007730
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01007731 This form is deprecated, do not use it in any new configuration, use the new
7732 "option redispatch" instead.
7733
7734 See also : "option redispatch"
7735
Willy Tarreaueabeafa2008-01-16 16:17:06 +01007736
Willy Tarreau96d51952019-05-22 20:34:35 +02007737reqadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007738 Add a header at the end of the HTTP request
7739 May be used in sections : defaults | frontend | listen | backend
7740 no | yes | yes | yes
7741 Arguments :
7742 <string> is the complete line to be added. Any space or known delimiter
7743 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007744 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007745
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007746 <cond> is an optional matching condition built from ACLs. It makes it
7747 possible to ignore this rule when other conditions are not met.
7748
Willy Tarreau303c0352008-01-17 19:01:39 +01007749 A new line consisting in <string> followed by a line feed will be added after
7750 the last header of an HTTP request.
7751
7752 Header transformations only apply to traffic which passes through HAProxy,
7753 and not to traffic generated by HAProxy, such as health-checks or error
7754 responses.
7755
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01007756 Example : add "X-Proto: SSL" to requests coming via port 81
7757 acl is-ssl dst_port 81
7758 reqadd X-Proto:\ SSL if is-ssl
7759
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007760 See also: "rspadd", "http-request", section 6 about HTTP header manipulation,
7761 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007762
7763
Willy Tarreau96d51952019-05-22 20:34:35 +02007764reqallow <search> [{if | unless} <cond>] (deprecated)
7765reqiallow <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007766 Definitely allow an HTTP request if a line matches a regular expression
7767 May be used in sections : defaults | frontend | listen | backend
7768 no | yes | yes | yes
7769 Arguments :
7770 <search> is the regular expression applied to HTTP headers and to the
7771 request line. This is an extended regular expression. Parenthesis
7772 grouping is supported and no preliminary backslash is required.
7773 Any space or known delimiter must be escaped using a backslash
7774 ('\'). The pattern applies to a full line at a time. The
7775 "reqallow" keyword strictly matches case while "reqiallow"
7776 ignores case.
7777
Willy Tarreau5321c422010-01-28 20:35:13 +01007778 <cond> is an optional matching condition built from ACLs. It makes it
7779 possible to ignore this rule when other conditions are not met.
7780
Willy Tarreau303c0352008-01-17 19:01:39 +01007781 A request containing any line which matches extended regular expression
7782 <search> will mark the request as allowed, even if any later test would
7783 result in a deny. The test applies both to the request line and to request
7784 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007785 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007786
7787 It is easier, faster and more powerful to use ACLs to write access policies.
7788 Reqdeny, reqallow and reqpass should be avoided in new designs.
7789
7790 Example :
7791 # allow www.* but refuse *.local
7792 reqiallow ^Host:\ www\.
7793 reqideny ^Host:\ .*\.local
7794
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007795 See also: "reqdeny", "block", "http-request", section 6 about HTTP header
7796 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007797
7798
Willy Tarreau96d51952019-05-22 20:34:35 +02007799reqdel <search> [{if | unless} <cond>] (deprecated)
7800reqidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007801 Delete all headers matching a regular expression in an HTTP request
7802 May be used in sections : defaults | frontend | listen | backend
7803 no | yes | yes | yes
7804 Arguments :
7805 <search> is the regular expression applied to HTTP headers and to the
7806 request line. This is an extended regular expression. Parenthesis
7807 grouping is supported and no preliminary backslash is required.
7808 Any space or known delimiter must be escaped using a backslash
7809 ('\'). The pattern applies to a full line at a time. The "reqdel"
7810 keyword strictly matches case while "reqidel" ignores case.
7811
Willy Tarreau5321c422010-01-28 20:35:13 +01007812 <cond> is an optional matching condition built from ACLs. It makes it
7813 possible to ignore this rule when other conditions are not met.
7814
Willy Tarreau303c0352008-01-17 19:01:39 +01007815 Any header line matching extended regular expression <search> in the request
7816 will be completely deleted. Most common use of this is to remove unwanted
7817 and/or dangerous headers or cookies from a request before passing it to the
7818 next servers.
7819
7820 Header transformations only apply to traffic which passes through HAProxy,
7821 and not to traffic generated by HAProxy, such as health-checks or error
7822 responses. Keep in mind that header names are not case-sensitive.
7823
7824 Example :
7825 # remove X-Forwarded-For header and SERVER cookie
7826 reqidel ^X-Forwarded-For:.*
7827 reqidel ^Cookie:.*SERVER=
7828
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007829 See also: "reqadd", "reqrep", "rspdel", "http-request", section 6 about
7830 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007831
7832
Willy Tarreau96d51952019-05-22 20:34:35 +02007833reqdeny <search> [{if | unless} <cond>] (deprecated)
7834reqideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007835 Deny an HTTP request if a line matches a regular expression
7836 May be used in sections : defaults | frontend | listen | backend
7837 no | yes | yes | yes
7838 Arguments :
7839 <search> is the regular expression applied to HTTP headers and to the
7840 request line. This is an extended regular expression. Parenthesis
7841 grouping is supported and no preliminary backslash is required.
7842 Any space or known delimiter must be escaped using a backslash
7843 ('\'). The pattern applies to a full line at a time. The
7844 "reqdeny" keyword strictly matches case while "reqideny" ignores
7845 case.
7846
Willy Tarreau5321c422010-01-28 20:35:13 +01007847 <cond> is an optional matching condition built from ACLs. It makes it
7848 possible to ignore this rule when other conditions are not met.
7849
Willy Tarreau303c0352008-01-17 19:01:39 +01007850 A request containing any line which matches extended regular expression
7851 <search> will mark the request as denied, even if any later test would
7852 result in an allow. The test applies both to the request line and to request
7853 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01007854 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01007855
Willy Tarreauced27012008-01-17 20:35:34 +01007856 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007857 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01007858 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01007859
Willy Tarreau303c0352008-01-17 19:01:39 +01007860 It is easier, faster and more powerful to use ACLs to write access policies.
7861 Reqdeny, reqallow and reqpass should be avoided in new designs.
7862
7863 Example :
7864 # refuse *.local, then allow www.*
7865 reqideny ^Host:\ .*\.local
7866 reqiallow ^Host:\ www\.
7867
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007868 See also: "reqallow", "rspdeny", "block", "http-request", section 6 about
7869 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007870
7871
Willy Tarreau96d51952019-05-22 20:34:35 +02007872reqpass <search> [{if | unless} <cond>] (deprecated)
7873reqipass <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007874 Ignore any HTTP request line matching a regular expression in next rules
7875 May be used in sections : defaults | frontend | listen | backend
7876 no | yes | yes | yes
7877 Arguments :
7878 <search> is the regular expression applied to HTTP headers and to the
7879 request line. This is an extended regular expression. Parenthesis
7880 grouping is supported and no preliminary backslash is required.
7881 Any space or known delimiter must be escaped using a backslash
7882 ('\'). The pattern applies to a full line at a time. The
7883 "reqpass" keyword strictly matches case while "reqipass" ignores
7884 case.
7885
Willy Tarreau5321c422010-01-28 20:35:13 +01007886 <cond> is an optional matching condition built from ACLs. It makes it
7887 possible to ignore this rule when other conditions are not met.
7888
Willy Tarreau303c0352008-01-17 19:01:39 +01007889 A request containing any line which matches extended regular expression
7890 <search> will skip next rules, without assigning any deny or allow verdict.
7891 The test applies both to the request line and to request headers. Keep in
7892 mind that URLs in request line are case-sensitive while header names are not.
7893
7894 It is easier, faster and more powerful to use ACLs to write access policies.
7895 Reqdeny, reqallow and reqpass should be avoided in new designs.
7896
7897 Example :
7898 # refuse *.local, then allow www.*, but ignore "www.private.local"
7899 reqipass ^Host:\ www.private\.local
7900 reqideny ^Host:\ .*\.local
7901 reqiallow ^Host:\ www\.
7902
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007903 See also: "reqallow", "reqdeny", "block", "http-request", section 6 about
7904 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007905
7906
Willy Tarreau96d51952019-05-22 20:34:35 +02007907reqrep <search> <string> [{if | unless} <cond>] (deprecated)
7908reqirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007909 Replace a regular expression with a string in an HTTP request line
7910 May be used in sections : defaults | frontend | listen | backend
7911 no | yes | yes | yes
7912 Arguments :
7913 <search> is the regular expression applied to HTTP headers and to the
7914 request line. This is an extended regular expression. Parenthesis
7915 grouping is supported and no preliminary backslash is required.
7916 Any space or known delimiter must be escaped using a backslash
7917 ('\'). The pattern applies to a full line at a time. The "reqrep"
7918 keyword strictly matches case while "reqirep" ignores case.
7919
7920 <string> is the complete line to be added. Any space or known delimiter
7921 must be escaped using a backslash ('\'). References to matched
7922 pattern groups are possible using the common \N form, with N
7923 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007924 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01007925
Willy Tarreau5321c422010-01-28 20:35:13 +01007926 <cond> is an optional matching condition built from ACLs. It makes it
7927 possible to ignore this rule when other conditions are not met.
7928
Willy Tarreau303c0352008-01-17 19:01:39 +01007929 Any line matching extended regular expression <search> in the request (both
7930 the request line and header lines) will be completely replaced with <string>.
7931 Most common use of this is to rewrite URLs or domain names in "Host" headers.
7932
7933 Header transformations only apply to traffic which passes through HAProxy,
7934 and not to traffic generated by HAProxy, such as health-checks or error
7935 responses. Note that for increased readability, it is suggested to add enough
7936 spaces between the request and the response. Keep in mind that URLs in
7937 request line are case-sensitive while header names are not.
7938
7939 Example :
7940 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04007941 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01007942 # replace "www.mydomain.com" with "www" in the host name.
7943 reqirep ^Host:\ www.mydomain.com Host:\ www
7944
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007945 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", "http-request",
7946 section 6 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007947
7948
Willy Tarreau96d51952019-05-22 20:34:35 +02007949reqtarpit <search> [{if | unless} <cond>] (deprecated)
7950reqitarpit <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01007951 Tarpit an HTTP request containing a line matching a regular expression
7952 May be used in sections : defaults | frontend | listen | backend
7953 no | yes | yes | yes
7954 Arguments :
7955 <search> is the regular expression applied to HTTP headers and to the
7956 request line. This is an extended regular expression. Parenthesis
7957 grouping is supported and no preliminary backslash is required.
7958 Any space or known delimiter must be escaped using a backslash
7959 ('\'). The pattern applies to a full line at a time. The
7960 "reqtarpit" keyword strictly matches case while "reqitarpit"
7961 ignores case.
7962
Willy Tarreau5321c422010-01-28 20:35:13 +01007963 <cond> is an optional matching condition built from ACLs. It makes it
7964 possible to ignore this rule when other conditions are not met.
7965
Willy Tarreau303c0352008-01-17 19:01:39 +01007966 A request containing any line which matches extended regular expression
7967 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01007968 be kept open for a pre-defined time, then will return an HTTP error 500 so
7969 that the attacker does not suspect it has been tarpitted. The status 500 will
7970 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01007971 delay is defined by "timeout tarpit", or "timeout connect" if the former is
7972 not set.
7973
7974 The goal of the tarpit is to slow down robots attacking servers with
7975 identifiable requests. Many robots limit their outgoing number of connections
7976 and stay connected waiting for a reply which can take several minutes to
7977 come. Depending on the environment and attack, it may be particularly
7978 efficient at reducing the load on the network and firewalls.
7979
Willy Tarreau5321c422010-01-28 20:35:13 +01007980 Examples :
Davor Ocelice9ed2812017-12-25 17:49:28 +01007981 # ignore user-agents reporting any flavor of "Mozilla" or "MSIE", but
Willy Tarreau303c0352008-01-17 19:01:39 +01007982 # block all others.
7983 reqipass ^User-Agent:\.*(Mozilla|MSIE)
7984 reqitarpit ^User-Agent:
7985
Willy Tarreau5321c422010-01-28 20:35:13 +01007986 # block bad guys
7987 acl badguys src 10.1.0.3 172.16.13.20/28
7988 reqitarpit . if badguys
7989
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08007990 See also: "reqallow", "reqdeny", "reqpass", "http-request", section 6
7991 about HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01007992
7993
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02007994retries <value>
7995 Set the number of retries to perform on a server after a connection failure
7996 May be used in sections: defaults | frontend | listen | backend
7997 yes | no | yes | yes
7998 Arguments :
7999 <value> is the number of times a connection attempt should be retried on
8000 a server when a connection either is refused or times out. The
8001 default value is 3.
8002
8003 It is important to understand that this value applies to the number of
8004 connection attempts, not full requests. When a connection has effectively
8005 been established to a server, there will be no more retry.
8006
8007 In order to avoid immediate reconnections to a server which is restarting,
Joseph Lynch726ab712015-05-11 23:25:34 -07008008 a turn-around timer of min("timeout connect", one second) is applied before
8009 a retry occurs.
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02008010
8011 When "option redispatch" is set, the last retry may be performed on another
8012 server even if a cookie references a different server.
8013
8014 See also : "option redispatch"
8015
8016
Olivier Houcharda254a372019-04-05 15:30:12 +02008017retry-on [list of keywords]
8018 Specify when to attempt to automatically retry a failed request
8019 May be used in sections: defaults | frontend | listen | backend
8020 yes | no | yes | yes
8021 Arguments :
8022 <keywords> is a list of keywords or HTTP status codes, each representing a
8023 type of failure event on which an attempt to retry the request
8024 is desired. Please read the notes at the bottom before changing
8025 this setting. The following keywords are supported :
8026
8027 none never retry
8028
8029 conn-failure retry when the connection or the SSL handshake failed
8030 and the request could not be sent. This is the default.
8031
8032 empty-response retry when the server connection was closed after part
8033 of the request was sent, and nothing was received from
8034 the server. This type of failure may be caused by the
8035 request timeout on the server side, poor network
8036 condition, or a server crash or restart while
8037 processing the request.
8038
Olivier Houcharde3249a92019-05-03 23:01:47 +02008039 junk-response retry when the server returned something not looking
8040 like a complete HTTP response. This includes partial
8041 responses headers as well as non-HTTP contents. It
8042 usually is a bad idea to retry on such events, which
8043 may be caused a configuration issue (wrong server port)
8044 or by the request being harmful to the server (buffer
8045 overflow attack for example).
8046
Olivier Houcharda254a372019-04-05 15:30:12 +02008047 response-timeout the server timeout stroke while waiting for the server
8048 to respond to the request. This may be caused by poor
8049 network condition, the reuse of an idle connection
8050 which has expired on the path, or by the request being
8051 extremely expensive to process. It generally is a bad
8052 idea to retry on such events on servers dealing with
8053 heavy database processing (full scans, etc) as it may
8054 amplify denial of service attacks.
8055
Olivier Houchard865d8392019-05-03 22:46:27 +02008056 0rtt-rejected retry requests which were sent over early data and were
8057 rejected by the server. These requests are generally
8058 considered to be safe to retry.
8059
Olivier Houcharda254a372019-04-05 15:30:12 +02008060 <status> any HTTP status code among "404" (Not Found), "408"
8061 (Request Timeout), "425" (Too Early), "500" (Server
8062 Error), "501" (Not Implemented), "502" (Bad Gateway),
8063 "503" (Service Unavailable), "504" (Gateway Timeout).
8064
Olivier Houchardddf0e032019-05-10 18:05:40 +02008065 all-retryable-errors
8066 retry request for any error that are considered
8067 retryable. This currently activates "conn-failure",
8068 "empty-response", "junk-response", "response-timeout",
8069 "0rtt-rejected", "500", "502", "503", and "504".
8070
Olivier Houcharda254a372019-04-05 15:30:12 +02008071 Using this directive replaces any previous settings with the new ones; it is
8072 not cumulative.
8073
8074 Please note that using anything other than "none" and "conn-failure" requires
8075 to allocate a buffer and copy the whole request into it, so it has memory and
8076 performance impacts. Requests not fitting in a single buffer will never be
8077 retried (see the global tune.bufsize setting).
8078
8079 You have to make sure the application has a replay protection mechanism built
8080 in such as a unique transaction IDs passed in requests, or that replaying the
8081 same request has no consequence, or it is very dangerous to use any retry-on
8082 value beside "conn-failure" and "none". Static file servers and caches are
8083 generally considered safe against any type of retry. Using a status code can
8084 be useful to quickly leave a server showing an abnormal behavior (out of
8085 memory, file system issues, etc), but in this case it may be a good idea to
8086 immediately redispatch the connection to another server (please see "option
8087 redispatch" for this). Last, it is important to understand that most causes
8088 of failures are the requests themselves and that retrying a request causing a
8089 server to misbehave will often make the situation even worse for this server,
8090 or for the whole service in case of redispatch.
8091
8092 Unless you know exactly how the application deals with replayed requests, you
8093 should not use this directive.
8094
8095 The default is "conn-failure".
8096
8097 See also: "retries", "option redispatch", "tune.bufsize"
8098
Willy Tarreau96d51952019-05-22 20:34:35 +02008099rspadd <string> [{if | unless} <cond>] (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008100 Add a header at the end of the HTTP response
8101 May be used in sections : defaults | frontend | listen | backend
8102 no | yes | yes | yes
8103 Arguments :
8104 <string> is the complete line to be added. Any space or known delimiter
8105 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008106 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008107
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008108 <cond> is an optional matching condition built from ACLs. It makes it
8109 possible to ignore this rule when other conditions are not met.
8110
Willy Tarreau303c0352008-01-17 19:01:39 +01008111 A new line consisting in <string> followed by a line feed will be added after
8112 the last header of an HTTP response.
8113
8114 Header transformations only apply to traffic which passes through HAProxy,
8115 and not to traffic generated by HAProxy, such as health-checks or error
8116 responses.
8117
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008118 See also: "rspdel" "reqadd", "http-response", section 6 about HTTP header
8119 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008120
8121
Willy Tarreau96d51952019-05-22 20:34:35 +02008122rspdel <search> [{if | unless} <cond>] (deprecated)
8123rspidel <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008124 Delete all headers matching a regular expression in an HTTP response
8125 May be used in sections : defaults | frontend | listen | backend
8126 no | yes | yes | yes
8127 Arguments :
8128 <search> is the regular expression applied to HTTP headers and to the
8129 response line. This is an extended regular expression, so
8130 parenthesis grouping is supported and no preliminary backslash
8131 is required. Any space or known delimiter must be escaped using
8132 a backslash ('\'). The pattern applies to a full line at a time.
8133 The "rspdel" keyword strictly matches case while "rspidel"
8134 ignores case.
8135
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008136 <cond> is an optional matching condition built from ACLs. It makes it
8137 possible to ignore this rule when other conditions are not met.
8138
Willy Tarreau303c0352008-01-17 19:01:39 +01008139 Any header line matching extended regular expression <search> in the response
8140 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008141 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01008142 client.
8143
8144 Header transformations only apply to traffic which passes through HAProxy,
8145 and not to traffic generated by HAProxy, such as health-checks or error
8146 responses. Keep in mind that header names are not case-sensitive.
8147
8148 Example :
8149 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02008150 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01008151
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008152 See also: "rspadd", "rsprep", "reqdel", "http-response", section 6 about
8153 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008154
8155
Willy Tarreau96d51952019-05-22 20:34:35 +02008156rspdeny <search> [{if | unless} <cond>] (deprecated)
8157rspideny <search> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008158 Block an HTTP response if a line matches a regular expression
8159 May be used in sections : defaults | frontend | listen | backend
8160 no | yes | yes | yes
8161 Arguments :
8162 <search> is the regular expression applied to HTTP headers and to the
8163 response line. This is an extended regular expression, so
8164 parenthesis grouping is supported and no preliminary backslash
8165 is required. Any space or known delimiter must be escaped using
8166 a backslash ('\'). The pattern applies to a full line at a time.
8167 The "rspdeny" keyword strictly matches case while "rspideny"
8168 ignores case.
8169
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008170 <cond> is an optional matching condition built from ACLs. It makes it
8171 possible to ignore this rule when other conditions are not met.
8172
Willy Tarreau303c0352008-01-17 19:01:39 +01008173 A response containing any line which matches extended regular expression
8174 <search> will mark the request as denied. The test applies both to the
8175 response line and to response headers. Keep in mind that header names are not
8176 case-sensitive.
8177
8178 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01008179 block the response before it reaches the client. If a response is denied, it
8180 will be replaced with an HTTP 502 error so that the client never retrieves
8181 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01008182
8183 It is easier, faster and more powerful to use ACLs to write access policies.
8184 Rspdeny should be avoided in new designs.
8185
8186 Example :
8187 # Ensure that no content type matching ms-word will leak
8188 rspideny ^Content-type:\.*/ms-word
8189
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008190 See also: "reqdeny", "acl", "block", "http-response", section 6 about
8191 HTTP header manipulation and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008192
8193
Willy Tarreau96d51952019-05-22 20:34:35 +02008194rsprep <search> <string> [{if | unless} <cond>] (deprecated)
8195rspirep <search> <string> [{if | unless} <cond>] (ignore case) (deprecated)
Willy Tarreau303c0352008-01-17 19:01:39 +01008196 Replace a regular expression with a string in an HTTP response line
8197 May be used in sections : defaults | frontend | listen | backend
8198 no | yes | yes | yes
8199 Arguments :
8200 <search> is the regular expression applied to HTTP headers and to the
8201 response line. This is an extended regular expression, so
8202 parenthesis grouping is supported and no preliminary backslash
8203 is required. Any space or known delimiter must be escaped using
8204 a backslash ('\'). The pattern applies to a full line at a time.
8205 The "rsprep" keyword strictly matches case while "rspirep"
8206 ignores case.
8207
8208 <string> is the complete line to be added. Any space or known delimiter
8209 must be escaped using a backslash ('\'). References to matched
8210 pattern groups are possible using the common \N form, with N
8211 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008212 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01008213
Willy Tarreaufdb563c2010-01-31 15:43:27 +01008214 <cond> is an optional matching condition built from ACLs. It makes it
8215 possible to ignore this rule when other conditions are not met.
8216
Willy Tarreau303c0352008-01-17 19:01:39 +01008217 Any line matching extended regular expression <search> in the response (both
8218 the response line and header lines) will be completely replaced with
8219 <string>. Most common use of this is to rewrite Location headers.
8220
8221 Header transformations only apply to traffic which passes through HAProxy,
8222 and not to traffic generated by HAProxy, such as health-checks or error
8223 responses. Note that for increased readability, it is suggested to add enough
8224 spaces between the request and the response. Keep in mind that header names
8225 are not case-sensitive.
8226
8227 Example :
8228 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
8229 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
8230
Ruoshan Huangeb5a3632015-12-08 21:00:23 +08008231 See also: "rspadd", "rspdel", "reqrep", "http-response", section 6 about
8232 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01008233
8234
David du Colombier486df472011-03-17 10:40:26 +01008235server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008236 Declare a server in a backend
8237 May be used in sections : defaults | frontend | listen | backend
8238 no | no | yes | yes
8239 Arguments :
8240 <name> is the internal name assigned to this server. This name will
Davor Ocelice9ed2812017-12-25 17:49:28 +01008241 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05008242 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008243
David du Colombier486df472011-03-17 10:40:26 +01008244 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
8245 resolvable hostname is supported, but this name will be resolved
8246 during start-up. Address "0.0.0.0" or "*" has a special meaning.
8247 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02008248 address as the one from the client connection. This is useful in
8249 transparent proxy architectures where the client's connection is
8250 intercepted and haproxy must forward to the original destination
8251 address. This is more or less what the "transparent" keyword does
8252 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01008253 to report statistics. Optionally, an address family prefix may be
8254 used before the address to force the family regardless of the
8255 address format, which can be useful to specify a path to a unix
8256 socket with no slash ('/'). Currently supported prefixes are :
8257 - 'ipv4@' -> address is always IPv4
8258 - 'ipv6@' -> address is always IPv6
8259 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008260 - 'abns@' -> address is in abstract namespace (Linux only)
William Lallemand2fe7dd02018-09-11 16:51:29 +02008261 - 'sockpair@' -> address is the FD of a connected unix
8262 socket or of a socketpair. During a connection, the
8263 backend creates a pair of connected sockets, and passes
8264 one of them over the FD. The bind part will use the
8265 received socket as the client FD. Should be used
8266 carefully.
William Lallemandb2f07452015-05-12 14:27:13 +02008267 You may want to reference some environment variables in the
8268 address parameter, see section 2.3 about environment
Willy Tarreau6a031d12016-11-07 19:42:35 +01008269 variables. The "init-addr" setting can be used to modify the way
8270 IP addresses should be resolved upon startup.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008271
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008272 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008273 be sent to this port. If unset, the same port the client
8274 connected to will be used. The port may also be prefixed by a "+"
8275 or a "-". In this case, the server's port will be determined by
8276 adding this value to the client's port.
8277
8278 <param*> is a list of parameters for this server. The "server" keywords
8279 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008280 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008281
8282 Examples :
8283 server first 10.1.1.1:1080 cookie first check inter 1000
8284 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01008285 server transp ipv4@
William Lallemandb2f07452015-05-12 14:27:13 +02008286 server backup "${SRV_BACKUP}:1080" backup
8287 server www1_dc1 "${LAN_DC1}.101:80"
8288 server www1_dc2 "${LAN_DC2}.101:80"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008289
Willy Tarreau55dcaf62015-09-27 15:03:15 +02008290 Note: regarding Linux's abstract namespace sockets, HAProxy uses the whole
8291 sun_path length is used for the address length. Some other programs
8292 such as socat use the string length only by default. Pass the option
8293 ",unix-tightsocklen=0" to any abstract socket definition in socat to
8294 make it compatible with HAProxy's.
8295
Mark Lamourinec2247f02012-01-04 13:02:01 -05008296 See also: "default-server", "http-send-name-header" and section 5 about
8297 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008298
Baptiste Assmann01c6cc32015-08-23 11:45:29 +02008299server-state-file-name [<file>]
8300 Set the server state file to read, load and apply to servers available in
8301 this backend. It only applies when the directive "load-server-state-from-file"
8302 is set to "local". When <file> is not provided or if this directive is not
8303 set, then backend name is used. If <file> starts with a slash '/', then it is
8304 considered as an absolute path. Otherwise, <file> is concatenated to the
8305 global directive "server-state-file-base".
8306
8307 Example: the minimal configuration below would make HAProxy look for the
8308 state server file '/etc/haproxy/states/bk':
8309
8310 global
8311 server-state-file-base /etc/haproxy/states
8312
8313 backend bk
8314 load-server-state-from-file
8315
8316 See also: "server-state-file-base", "load-server-state-from-file", and
8317 "show servers state"
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008318
Frédéric Lécaillecb4502e2017-04-20 13:36:25 +02008319server-template <prefix> <num | range> <fqdn>[:<port>] [params*]
8320 Set a template to initialize servers with shared parameters.
8321 The names of these servers are built from <prefix> and <num | range> parameters.
8322 May be used in sections : defaults | frontend | listen | backend
8323 no | no | yes | yes
8324
8325 Arguments:
8326 <prefix> A prefix for the server names to be built.
8327
8328 <num | range>
8329 If <num> is provided, this template initializes <num> servers
8330 with 1 up to <num> as server name suffixes. A range of numbers
8331 <num_low>-<num_high> may also be used to use <num_low> up to
8332 <num_high> as server name suffixes.
8333
8334 <fqdn> A FQDN for all the servers this template initializes.
8335
8336 <port> Same meaning as "server" <port> argument (see "server" keyword).
8337
8338 <params*>
8339 Remaining server parameters among all those supported by "server"
8340 keyword.
8341
8342 Examples:
8343 # Initializes 3 servers with srv1, srv2 and srv3 as names,
8344 # google.com as FQDN, and health-check enabled.
8345 server-template srv 1-3 google.com:80 check
8346
8347 # or
8348 server-template srv 3 google.com:80 check
8349
8350 # would be equivalent to:
8351 server srv1 google.com:80 check
8352 server srv2 google.com:80 check
8353 server srv3 google.com:80 check
8354
8355
8356
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008357source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02008358source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008359source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008360 Set the source address for outgoing connections
8361 May be used in sections : defaults | frontend | listen | backend
8362 yes | no | yes | yes
8363 Arguments :
8364 <addr> is the IPv4 address HAProxy will bind to before connecting to a
8365 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01008366
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008367 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01008368 the most appropriate address to reach its destination. Optionally
8369 an address family prefix may be used before the address to force
8370 the family regardless of the address format, which can be useful
8371 to specify a path to a unix socket with no slash ('/'). Currently
8372 supported prefixes are :
8373 - 'ipv4@' -> address is always IPv4
8374 - 'ipv6@' -> address is always IPv6
8375 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02008376 - 'abns@' -> address is in abstract namespace (Linux only)
Cyril Bonté307ee1e2015-09-28 23:16:06 +02008377 You may want to reference some environment variables in the
8378 address parameter, see section 2.3 about environment variables.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008379
8380 <port> is an optional port. It is normally not needed but may be useful
8381 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02008382 the system will select a free port. Note that port ranges are not
8383 supported in the backend. If you want to force port ranges, you
8384 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008385
8386 <addr2> is the IP address to present to the server when connections are
8387 forwarded in full transparent proxy mode. This is currently only
8388 supported on some patched Linux kernels. When this address is
8389 specified, clients connecting to the server will be presented
8390 with this address, while health checks will still use the address
8391 <addr>.
8392
8393 <port2> is the optional port to present to the server when connections
8394 are forwarded in full transparent proxy mode (see <addr2> above).
8395 The default value of zero means the system will select a free
8396 port.
8397
Willy Tarreaubce70882009-09-07 11:51:47 +02008398 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
8399 This is the name of a comma-separated header list which can
8400 contain multiple IP addresses. By default, the last occurrence is
8401 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01008402 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02008403 by previous proxy, typically Stunnel. In order to use another
8404 occurrence from the last one, please see the <occ> parameter
8405 below. When the header (or occurrence) is not found, no binding
8406 is performed so that the proxy's default IP address is used. Also
8407 keep in mind that the header name is case insensitive, as for any
8408 HTTP header.
8409
8410 <occ> is the occurrence number of a value to be used in a multi-value
8411 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008412 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02008413 address. Positive values indicate a position from the first
8414 occurrence, 1 being the first one. Negative values indicate
8415 positions relative to the last one, -1 being the last one. This
8416 is helpful for situations where an X-Forwarded-For header is set
8417 at the entry point of an infrastructure and must be used several
8418 proxy layers away. When this value is not specified, -1 is
8419 assumed. Passing a zero here disables the feature.
8420
Willy Tarreaud53f96b2009-02-04 18:46:54 +01008421 <name> is an optional interface name to which to bind to for outgoing
8422 traffic. On systems supporting this features (currently, only
8423 Linux), this allows one to bind all traffic to the server to
8424 this interface even if it is not the one the system would select
8425 based on routing tables. This should be used with extreme care.
8426 Note that using this option requires root privileges.
8427
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008428 The "source" keyword is useful in complex environments where a specific
8429 address only is allowed to connect to the servers. It may be needed when a
8430 private address must be used through a public gateway for instance, and it is
8431 known that the system cannot determine the adequate source address by itself.
8432
8433 An extension which is available on certain patched Linux kernels may be used
8434 through the "usesrc" optional keyword. It makes it possible to connect to the
8435 servers with an IP address which does not belong to the system itself. This
8436 is called "full transparent proxy mode". For this to work, the destination
8437 servers have to route their traffic back to this address through the machine
8438 running HAProxy, and IP forwarding must generally be enabled on this machine.
8439
8440 In this "full transparent proxy" mode, it is possible to force a specific IP
8441 address to be presented to the servers. This is not much used in fact. A more
8442 common use is to tell HAProxy to present the client's IP address. For this,
8443 there are two methods :
8444
8445 - present the client's IP and port addresses. This is the most transparent
8446 mode, but it can cause problems when IP connection tracking is enabled on
8447 the machine, because a same connection may be seen twice with different
8448 states. However, this solution presents the huge advantage of not
8449 limiting the system to the 64k outgoing address+port couples, because all
8450 of the client ranges may be used.
8451
8452 - present only the client's IP address and select a spare port. This
8453 solution is still quite elegant but slightly less transparent (downstream
8454 firewalls logs will not match upstream's). It also presents the downside
8455 of limiting the number of concurrent connections to the usual 64k ports.
8456 However, since the upstream and downstream ports are different, local IP
8457 connection tracking on the machine will not be upset by the reuse of the
8458 same session.
8459
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008460 This option sets the default source for all servers in the backend. It may
8461 also be specified in a "defaults" section. Finer source address specification
8462 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008463 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008464
Baptiste Assmann91bd3372015-07-17 21:59:42 +02008465 In order to work, "usesrc" requires root privileges.
8466
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008467 Examples :
8468 backend private
8469 # Connect to the servers using our 192.168.1.200 source address
8470 source 192.168.1.200
8471
8472 backend transparent_ssl1
8473 # Connect to the SSL farm from the client's source address
8474 source 192.168.1.200 usesrc clientip
8475
8476 backend transparent_ssl2
8477 # Connect to the SSL farm from the client's source address and port
8478 # not recommended if IP conntrack is present on the local machine.
8479 source 192.168.1.200 usesrc client
8480
8481 backend transparent_ssl3
8482 # Connect to the SSL farm from the client's source address. It
8483 # is more conntrack-friendly.
8484 source 192.168.1.200 usesrc clientip
8485
8486 backend transparent_smtp
8487 # Connect to the SMTP farm from the client's source address/port
8488 # with Tproxy version 4.
8489 source 0.0.0.0 usesrc clientip
8490
Willy Tarreaubce70882009-09-07 11:51:47 +02008491 backend transparent_http
8492 # Connect to the servers using the client's IP as seen by previous
8493 # proxy.
8494 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
8495
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008496 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008497 the Linux kernel on www.balabit.com, the "bind" keyword.
8498
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01008499
Willy Tarreau844e3c52008-01-11 16:28:18 +01008500srvtimeout <timeout> (deprecated)
8501 Set the maximum inactivity time on the server side.
8502 May be used in sections : defaults | frontend | listen | backend
8503 yes | no | yes | yes
8504 Arguments :
8505 <timeout> is the timeout value specified in milliseconds by default, but
8506 can be in any other unit if the number is suffixed by the unit,
8507 as explained at the top of this document.
8508
8509 The inactivity timeout applies when the server is expected to acknowledge or
8510 send data. In HTTP mode, this timeout is particularly important to consider
8511 during the first phase of the server's response, when it has to send the
8512 headers, as it directly represents the server's processing time for the
8513 request. To find out what value to put there, it's often good to start with
8514 what would be considered as unacceptable response times, then check the logs
8515 to observe the response time distribution, and adjust the value accordingly.
8516
8517 The value is specified in milliseconds by default, but can be in any other
8518 unit if the number is suffixed by the unit, as specified at the top of this
8519 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
8520 recommended that the client timeout remains equal to the server timeout in
8521 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008522 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01008523 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +01008524 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01008525
8526 This parameter is specific to backends, but can be specified once for all in
8527 "defaults" sections. This is in fact one of the easiest solutions not to
8528 forget about it. An unspecified timeout results in an infinite timeout, which
8529 is not recommended. Such a usage is accepted and works but reports a warning
8530 during startup because it may results in accumulation of expired sessions in
8531 the system if the system's timeouts are not configured either.
8532
8533 This parameter is provided for compatibility but is currently deprecated.
8534 Please use "timeout server" instead.
8535
Willy Tarreauce887fd2012-05-12 12:50:00 +02008536 See also : "timeout server", "timeout tunnel", "timeout client" and
8537 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008538
8539
Cyril Bonté66c327d2010-10-12 00:14:37 +02008540stats admin { if | unless } <cond>
8541 Enable statistics admin level if/unless a condition is matched
8542 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008543 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02008544
8545 This statement enables the statistics admin level if/unless a condition is
8546 matched.
8547
8548 The admin level allows to enable/disable servers from the web interface. By
8549 default, statistics page is read-only for security reasons.
8550
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008551 Note : Consider not using this feature in multi-process mode (nbproc > 1)
8552 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01008553 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008554
Cyril Bonté23b39d92011-02-10 22:54:44 +01008555 Currently, the POST request is limited to the buffer size minus the reserved
8556 buffer space, which means that if the list of servers is too long, the
8557 request won't be processed. It is recommended to alter few servers at a
8558 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008559
8560 Example :
8561 # statistics admin level only for localhost
8562 backend stats_localhost
8563 stats enable
8564 stats admin if LOCALHOST
8565
8566 Example :
8567 # statistics admin level always enabled because of the authentication
8568 backend stats_auth
8569 stats enable
8570 stats auth admin:AdMiN123
8571 stats admin if TRUE
8572
8573 Example :
8574 # statistics admin level depends on the authenticated user
8575 userlist stats-auth
8576 group admin users admin
8577 user admin insecure-password AdMiN123
8578 group readonly users haproxy
8579 user haproxy insecure-password haproxy
8580
8581 backend stats_auth
8582 stats enable
8583 acl AUTH http_auth(stats-auth)
8584 acl AUTH_ADMIN http_auth_group(stats-auth) admin
8585 stats http-request auth unless AUTH
8586 stats admin if AUTH_ADMIN
8587
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01008588 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
8589 "bind-process", section 3.4 about userlists and section 7 about
8590 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02008591
8592
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008593stats auth <user>:<passwd>
8594 Enable statistics with authentication and grant access to an account
8595 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008596 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008597 Arguments :
8598 <user> is a user name to grant access to
8599
8600 <passwd> is the cleartext password associated to this user
8601
8602 This statement enables statistics with default settings, and restricts access
8603 to declared users only. It may be repeated as many times as necessary to
8604 allow as many users as desired. When a user tries to access the statistics
8605 without a valid account, a "401 Forbidden" response will be returned so that
8606 the browser asks the user to provide a valid user and password. The real
8607 which will be returned to the browser is configurable using "stats realm".
8608
8609 Since the authentication method is HTTP Basic Authentication, the passwords
8610 circulate in cleartext on the network. Thus, it was decided that the
8611 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02008612 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008613
8614 It is also possible to reduce the scope of the proxies which appear in the
8615 report using "stats scope".
8616
8617 Though this statement alone is enough to enable statistics reporting, it is
8618 recommended to set all other settings in order to avoid relying on default
8619 unobvious parameters.
8620
8621 Example :
8622 # public access (limited to this backend only)
8623 backend public_www
8624 server srv1 192.168.0.1:80
8625 stats enable
8626 stats hide-version
8627 stats scope .
8628 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008629 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008630 stats auth admin1:AdMiN123
8631 stats auth admin2:AdMiN321
8632
8633 # internal monitoring access (unlimited)
8634 backend private_monitoring
8635 stats enable
8636 stats uri /admin?stats
8637 stats refresh 5s
8638
8639 See also : "stats enable", "stats realm", "stats scope", "stats uri"
8640
8641
8642stats enable
8643 Enable statistics reporting with default settings
8644 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008645 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008646 Arguments : none
8647
8648 This statement enables statistics reporting with default settings defined
8649 at build time. Unless stated otherwise, these settings are used :
8650 - stats uri : /haproxy?stats
8651 - stats realm : "HAProxy Statistics"
8652 - stats auth : no authentication
8653 - stats scope : no restriction
8654
8655 Though this statement alone is enough to enable statistics reporting, it is
8656 recommended to set all other settings in order to avoid relying on default
8657 unobvious parameters.
8658
8659 Example :
8660 # public access (limited to this backend only)
8661 backend public_www
8662 server srv1 192.168.0.1:80
8663 stats enable
8664 stats hide-version
8665 stats scope .
8666 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008667 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008668 stats auth admin1:AdMiN123
8669 stats auth admin2:AdMiN321
8670
8671 # internal monitoring access (unlimited)
8672 backend private_monitoring
8673 stats enable
8674 stats uri /admin?stats
8675 stats refresh 5s
8676
8677 See also : "stats auth", "stats realm", "stats uri"
8678
8679
Willy Tarreaud63335a2010-02-26 12:56:52 +01008680stats hide-version
8681 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008682 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008683 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008684 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008685
Willy Tarreaud63335a2010-02-26 12:56:52 +01008686 By default, the stats page reports some useful status information along with
8687 the statistics. Among them is HAProxy's version. However, it is generally
8688 considered dangerous to report precise version to anyone, as it can help them
8689 target known weaknesses with specific attacks. The "stats hide-version"
8690 statement removes the version from the statistics report. This is recommended
8691 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008692
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008693 Though this statement alone is enough to enable statistics reporting, it is
8694 recommended to set all other settings in order to avoid relying on default
8695 unobvious parameters.
8696
Willy Tarreaud63335a2010-02-26 12:56:52 +01008697 Example :
8698 # public access (limited to this backend only)
8699 backend public_www
8700 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02008701 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008702 stats hide-version
8703 stats scope .
8704 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008705 stats realm HAProxy\ Statistics
Willy Tarreaud63335a2010-02-26 12:56:52 +01008706 stats auth admin1:AdMiN123
8707 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008708
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008709 # internal monitoring access (unlimited)
8710 backend private_monitoring
8711 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01008712 stats uri /admin?stats
8713 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01008714
Willy Tarreaud63335a2010-02-26 12:56:52 +01008715 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02008716
Willy Tarreau983e01e2010-01-11 18:42:06 +01008717
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02008718stats http-request { allow | deny | auth [realm <realm>] }
8719 [ { if | unless } <condition> ]
8720 Access control for statistics
8721
8722 May be used in sections: defaults | frontend | listen | backend
8723 no | no | yes | yes
8724
8725 As "http-request", these set of options allow to fine control access to
8726 statistics. Each option may be followed by if/unless and acl.
8727 First option with matched condition (or option without condition) is final.
8728 For "deny" a 403 error will be returned, for "allow" normal processing is
8729 performed, for "auth" a 401/407 error code is returned so the client
8730 should be asked to enter a username and password.
8731
8732 There is no fixed limit to the number of http-request statements per
8733 instance.
8734
8735 See also : "http-request", section 3.4 about userlists and section 7
8736 about ACL usage.
8737
8738
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008739stats realm <realm>
8740 Enable statistics and set authentication realm
8741 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008742 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008743 Arguments :
8744 <realm> is the name of the HTTP Basic Authentication realm reported to
8745 the browser. The browser uses it to display it in the pop-up
8746 inviting the user to enter a valid username and password.
8747
8748 The realm is read as a single word, so any spaces in it should be escaped
8749 using a backslash ('\').
8750
8751 This statement is useful only in conjunction with "stats auth" since it is
8752 only related to authentication.
8753
8754 Though this statement alone is enough to enable statistics reporting, it is
8755 recommended to set all other settings in order to avoid relying on default
8756 unobvious parameters.
8757
8758 Example :
8759 # public access (limited to this backend only)
8760 backend public_www
8761 server srv1 192.168.0.1:80
8762 stats enable
8763 stats hide-version
8764 stats scope .
8765 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008766 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008767 stats auth admin1:AdMiN123
8768 stats auth admin2:AdMiN321
8769
8770 # internal monitoring access (unlimited)
8771 backend private_monitoring
8772 stats enable
8773 stats uri /admin?stats
8774 stats refresh 5s
8775
8776 See also : "stats auth", "stats enable", "stats uri"
8777
8778
8779stats refresh <delay>
8780 Enable statistics with automatic refresh
8781 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008782 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008783 Arguments :
8784 <delay> is the suggested refresh delay, specified in seconds, which will
8785 be returned to the browser consulting the report page. While the
8786 browser is free to apply any delay, it will generally respect it
8787 and refresh the page this every seconds. The refresh interval may
8788 be specified in any other non-default time unit, by suffixing the
8789 unit after the value, as explained at the top of this document.
8790
8791 This statement is useful on monitoring displays with a permanent page
8792 reporting the load balancer's activity. When set, the HTML report page will
8793 include a link "refresh"/"stop refresh" so that the user can select whether
8794 he wants automatic refresh of the page or not.
8795
8796 Though this statement alone is enough to enable statistics reporting, it is
8797 recommended to set all other settings in order to avoid relying on default
8798 unobvious parameters.
8799
8800 Example :
8801 # public access (limited to this backend only)
8802 backend public_www
8803 server srv1 192.168.0.1:80
8804 stats enable
8805 stats hide-version
8806 stats scope .
8807 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008808 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008809 stats auth admin1:AdMiN123
8810 stats auth admin2:AdMiN321
8811
8812 # internal monitoring access (unlimited)
8813 backend private_monitoring
8814 stats enable
8815 stats uri /admin?stats
8816 stats refresh 5s
8817
8818 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8819
8820
8821stats scope { <name> | "." }
8822 Enable statistics and limit access scope
8823 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008824 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008825 Arguments :
8826 <name> is the name of a listen, frontend or backend section to be
8827 reported. The special name "." (a single dot) designates the
8828 section in which the statement appears.
8829
8830 When this statement is specified, only the sections enumerated with this
8831 statement will appear in the report. All other ones will be hidden. This
8832 statement may appear as many times as needed if multiple sections need to be
8833 reported. Please note that the name checking is performed as simple string
8834 comparisons, and that it is never checked that a give section name really
8835 exists.
8836
8837 Though this statement alone is enough to enable statistics reporting, it is
8838 recommended to set all other settings in order to avoid relying on default
8839 unobvious parameters.
8840
8841 Example :
8842 # public access (limited to this backend only)
8843 backend public_www
8844 server srv1 192.168.0.1:80
8845 stats enable
8846 stats hide-version
8847 stats scope .
8848 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008849 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008850 stats auth admin1:AdMiN123
8851 stats auth admin2:AdMiN321
8852
8853 # internal monitoring access (unlimited)
8854 backend private_monitoring
8855 stats enable
8856 stats uri /admin?stats
8857 stats refresh 5s
8858
8859 See also : "stats auth", "stats enable", "stats realm", "stats uri"
8860
Willy Tarreaud63335a2010-02-26 12:56:52 +01008861
Willy Tarreauc9705a12010-07-27 20:05:50 +02008862stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01008863 Enable reporting of a description on the statistics page.
8864 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008865 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008866
Willy Tarreauc9705a12010-07-27 20:05:50 +02008867 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01008868 description from global section is automatically used instead.
8869
8870 This statement is useful for users that offer shared services to their
8871 customers, where node or description should be different for each customer.
8872
8873 Though this statement alone is enough to enable statistics reporting, it is
8874 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008875 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008876
8877 Example :
8878 # internal monitoring access (unlimited)
8879 backend private_monitoring
8880 stats enable
8881 stats show-desc Master node for Europe, Asia, Africa
8882 stats uri /admin?stats
8883 stats refresh 5s
8884
8885 See also: "show-node", "stats enable", "stats uri" and "description" in
8886 global section.
8887
8888
8889stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02008890 Enable reporting additional information on the statistics page
8891 May be used in sections : defaults | frontend | listen | backend
8892 yes | yes | yes | yes
8893 Arguments : none
8894
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008895 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01008896 - cap: capabilities (proxy)
8897 - mode: one of tcp, http or health (proxy)
8898 - id: SNMP ID (proxy, socket, server)
8899 - IP (socket, server)
8900 - cookie (backend, server)
8901
8902 Though this statement alone is enough to enable statistics reporting, it is
8903 recommended to set all other settings in order to avoid relying on default
Davor Ocelice9ed2812017-12-25 17:49:28 +01008904 unobvious parameters. Default behavior is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008905
8906 See also: "stats enable", "stats uri".
8907
8908
8909stats show-node [ <name> ]
8910 Enable reporting of a host name on the statistics page.
8911 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008912 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01008913 Arguments:
8914 <name> is an optional name to be reported. If unspecified, the
8915 node name from global section is automatically used instead.
8916
8917 This statement is useful for users that offer shared services to their
8918 customers, where node or description might be different on a stats page
Davor Ocelice9ed2812017-12-25 17:49:28 +01008919 provided for each customer. Default behavior is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01008920
8921 Though this statement alone is enough to enable statistics reporting, it is
8922 recommended to set all other settings in order to avoid relying on default
8923 unobvious parameters.
8924
8925 Example:
8926 # internal monitoring access (unlimited)
8927 backend private_monitoring
8928 stats enable
8929 stats show-node Europe-1
8930 stats uri /admin?stats
8931 stats refresh 5s
8932
8933 See also: "show-desc", "stats enable", "stats uri", and "node" in global
8934 section.
8935
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008936
8937stats uri <prefix>
8938 Enable statistics and define the URI prefix to access them
8939 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02008940 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008941 Arguments :
8942 <prefix> is the prefix of any URI which will be redirected to stats. This
8943 prefix may contain a question mark ('?') to indicate part of a
8944 query string.
8945
8946 The statistics URI is intercepted on the relayed traffic, so it appears as a
8947 page within the normal application. It is strongly advised to ensure that the
8948 selected URI will never appear in the application, otherwise it will never be
8949 possible to reach it in the application.
8950
8951 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008952 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008953 It is generally a good idea to include a question mark in the URI so that
8954 intermediate proxies refrain from caching the results. Also, since any string
8955 beginning with the prefix will be accepted as a stats request, the question
8956 mark helps ensuring that no valid URI will begin with the same words.
8957
8958 It is sometimes very convenient to use "/" as the URI prefix, and put that
8959 statement in a "listen" instance of its own. That makes it easy to dedicate
8960 an address or a port to statistics only.
8961
8962 Though this statement alone is enough to enable statistics reporting, it is
8963 recommended to set all other settings in order to avoid relying on default
8964 unobvious parameters.
8965
8966 Example :
8967 # public access (limited to this backend only)
8968 backend public_www
8969 server srv1 192.168.0.1:80
8970 stats enable
8971 stats hide-version
8972 stats scope .
8973 stats uri /admin?stats
Davor Ocelice9ed2812017-12-25 17:49:28 +01008974 stats realm HAProxy\ Statistics
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008975 stats auth admin1:AdMiN123
8976 stats auth admin2:AdMiN321
8977
8978 # internal monitoring access (unlimited)
8979 backend private_monitoring
8980 stats enable
8981 stats uri /admin?stats
8982 stats refresh 5s
8983
8984 See also : "stats auth", "stats enable", "stats realm"
8985
8986
Willy Tarreaud63335a2010-02-26 12:56:52 +01008987stick match <pattern> [table <table>] [{if | unless} <cond>]
8988 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01008989 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01008990 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008991
8992 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02008993 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008994 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01008995 will be analyzed in the hope to find a matching entry in a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01008996 stickiness table. This rule is mandatory.
8997
8998 <table> is an optional stickiness table name. If unspecified, the same
8999 backend's table is used. A stickiness table is declared using
9000 the "stick-table" statement.
9001
9002 <cond> is an optional matching condition. It makes it possible to match
9003 on a certain criterion only when other conditions are met (or
9004 not met). For instance, it could be used to match on a source IP
9005 address except when a request passes through a known proxy, in
9006 which case we'd match on a header containing that IP address.
9007
9008 Some protocols or applications require complex stickiness rules and cannot
9009 always simply rely on cookies nor hashing. The "stick match" statement
9010 describes a rule to extract the stickiness criterion from an incoming request
9011 or connection. See section 7 for a complete list of possible patterns and
9012 transformation rules.
9013
9014 The table has to be declared using the "stick-table" statement. It must be of
9015 a type compatible with the pattern. By default it is the one which is present
9016 in the same backend. It is possible to share a table with other backends by
9017 referencing it using the "table" keyword. If another table is referenced,
9018 the server's ID inside the backends are used. By default, all server IDs
9019 start at 1 in each backend, so the server ordering is enough. But in case of
9020 doubt, it is highly recommended to force server IDs using their "id" setting.
9021
9022 It is possible to restrict the conditions where a "stick match" statement
9023 will apply, using "if" or "unless" followed by a condition. See section 7 for
9024 ACL based conditions.
9025
9026 There is no limit on the number of "stick match" statements. The first that
9027 applies and matches will cause the request to be directed to the same server
9028 as was used for the request which created the entry. That way, multiple
9029 matches can be used as fallbacks.
9030
9031 The stick rules are checked after the persistence cookies, so they will not
9032 affect stickiness if a cookie has already been used to select a server. That
9033 way, it becomes very easy to insert cookies and match on IP addresses in
9034 order to maintain stickiness between HTTP and HTTPS.
9035
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009036 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9037 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009038 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009039
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009040 Example :
9041 # forward SMTP users to the same server they just used for POP in the
9042 # last 30 minutes
9043 backend pop
9044 mode tcp
9045 balance roundrobin
9046 stick store-request src
9047 stick-table type ip size 200k expire 30m
9048 server s1 192.168.1.1:110
9049 server s2 192.168.1.1:110
9050
9051 backend smtp
9052 mode tcp
9053 balance roundrobin
9054 stick match src table pop
9055 server s1 192.168.1.1:25
9056 server s2 192.168.1.1:25
9057
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009058 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009059 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009060
9061
9062stick on <pattern> [table <table>] [{if | unless} <condition>]
9063 Define a request pattern to associate a user to a server
9064 May be used in sections : defaults | frontend | listen | backend
9065 no | no | yes | yes
9066
9067 Note : This form is exactly equivalent to "stick match" followed by
9068 "stick store-request", all with the same arguments. Please refer
9069 to both keywords for details. It is only provided as a convenience
9070 for writing more maintainable configurations.
9071
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009072 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9073 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009074 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009075
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009076 Examples :
9077 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01009078 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009079
9080 # ...is strictly equivalent to this one :
9081 stick match src table pop if !localhost
9082 stick store-request src table pop if !localhost
9083
9084
9085 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
9086 # well as HTTP without cookie. Share the same table between both accesses.
9087 backend http
9088 mode http
9089 balance roundrobin
9090 stick on src table https
9091 cookie SRV insert indirect nocache
9092 server s1 192.168.1.1:80 cookie s1
9093 server s2 192.168.1.1:80 cookie s2
9094
9095 backend https
9096 mode tcp
9097 balance roundrobin
9098 stick-table type ip size 200k expire 30m
9099 stick on src
9100 server s1 192.168.1.1:443
9101 server s2 192.168.1.1:443
9102
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009103 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009104
9105
9106stick store-request <pattern> [table <table>] [{if | unless} <condition>]
9107 Define a request pattern used to create an entry in a stickiness table
9108 May be used in sections : defaults | frontend | listen | backend
9109 no | no | yes | yes
9110
9111 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009112 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009113 describes what elements of the incoming request or connection
Davor Ocelice9ed2812017-12-25 17:49:28 +01009114 will be analyzed, extracted and stored in the table once a
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009115 server is selected.
9116
9117 <table> is an optional stickiness table name. If unspecified, the same
9118 backend's table is used. A stickiness table is declared using
9119 the "stick-table" statement.
9120
9121 <cond> is an optional storage condition. It makes it possible to store
9122 certain criteria only when some conditions are met (or not met).
9123 For instance, it could be used to store the source IP address
9124 except when the request passes through a known proxy, in which
9125 case we'd store a converted form of a header containing that IP
9126 address.
9127
9128 Some protocols or applications require complex stickiness rules and cannot
9129 always simply rely on cookies nor hashing. The "stick store-request" statement
9130 describes a rule to decide what to extract from the request and when to do
9131 it, in order to store it into a stickiness table for further requests to
9132 match it using the "stick match" statement. Obviously the extracted part must
9133 make sense and have a chance to be matched in a further request. Storing a
9134 client's IP address for instance often makes sense. Storing an ID found in a
9135 URL parameter also makes sense. Storing a source port will almost never make
9136 any sense because it will be randomly matched. See section 7 for a complete
9137 list of possible patterns and transformation rules.
9138
9139 The table has to be declared using the "stick-table" statement. It must be of
9140 a type compatible with the pattern. By default it is the one which is present
9141 in the same backend. It is possible to share a table with other backends by
9142 referencing it using the "table" keyword. If another table is referenced,
9143 the server's ID inside the backends are used. By default, all server IDs
9144 start at 1 in each backend, so the server ordering is enough. But in case of
9145 doubt, it is highly recommended to force server IDs using their "id" setting.
9146
9147 It is possible to restrict the conditions where a "stick store-request"
9148 statement will apply, using "if" or "unless" followed by a condition. This
9149 condition will be evaluated while parsing the request, so any criteria can be
9150 used. See section 7 for ACL based conditions.
9151
9152 There is no limit on the number of "stick store-request" statements, but
9153 there is a limit of 8 simultaneous stores per request or response. This
9154 makes it possible to store up to 8 criteria, all extracted from either the
9155 request or the response, regardless of the number of rules. Only the 8 first
9156 ones which match will be kept. Using this, it is possible to feed multiple
9157 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009158 another protocol or access method. Using multiple store-request rules with
9159 the same table is possible and may be used to find the best criterion to rely
9160 on, by arranging the rules by decreasing preference order. Only the first
9161 extracted criterion for a given table will be stored. All subsequent store-
9162 request rules referencing the same table will be skipped and their ACLs will
9163 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009164
9165 The "store-request" rules are evaluated once the server connection has been
9166 established, so that the table will contain the real server that processed
9167 the request.
9168
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009169 Note : Consider not using this feature in multi-process mode (nbproc > 1)
9170 unless you know what you do : memory is not shared between the
Davor Ocelice9ed2812017-12-25 17:49:28 +01009171 processes, which can result in random behaviors.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009172
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009173 Example :
9174 # forward SMTP users to the same server they just used for POP in the
9175 # last 30 minutes
9176 backend pop
9177 mode tcp
9178 balance roundrobin
9179 stick store-request src
9180 stick-table type ip size 200k expire 30m
9181 server s1 192.168.1.1:110
9182 server s2 192.168.1.1:110
9183
9184 backend smtp
9185 mode tcp
9186 balance roundrobin
9187 stick match src table pop
9188 server s1 192.168.1.1:25
9189 server s2 192.168.1.1:25
9190
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009191 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02009192 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009193
9194
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009195stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02009196 size <size> [expire <expire>] [nopurge] [peers <peersect>]
9197 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08009198 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009199 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009200 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009201
9202 Arguments :
9203 ip a table declared with "type ip" will only store IPv4 addresses.
9204 This form is very compact (about 50 bytes per entry) and allows
9205 very fast entry lookup and stores with almost no overhead. This
9206 is mainly used to store client source IP addresses.
9207
David du Colombier9a6d3c92011-03-17 10:40:24 +01009208 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
9209 This form is very compact (about 60 bytes per entry) and allows
9210 very fast entry lookup and stores with almost no overhead. This
9211 is mainly used to store client source IP addresses.
9212
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009213 integer a table declared with "type integer" will store 32bit integers
9214 which can represent a client identifier found in a request for
9215 instance.
9216
9217 string a table declared with "type string" will store substrings of up
9218 to <len> characters. If the string provided by the pattern
9219 extractor is larger than <len>, it will be truncated before
9220 being stored. During matching, at most <len> characters will be
9221 compared between the string in the table and the extracted
9222 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009223 to 32 characters.
9224
9225 binary a table declared with "type binary" will store binary blocks
9226 of <len> bytes. If the block provided by the pattern
9227 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02009228 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009229 is shorter than <len>, it will be padded by 0. When not
9230 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009231
9232 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02009233 "string" type table (See type "string" above). Or the number
9234 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009235 changing this parameter as memory usage will proportionally
9236 increase.
9237
9238 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01009239 value directly impacts memory usage. Count approximately
9240 50 bytes per entry, plus the size of a string if any. The size
9241 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009242
9243 [nopurge] indicates that we refuse to purge older entries when the table
9244 is full. When not specified and the table is full when haproxy
9245 wants to store an entry in it, it will flush a few of the oldest
9246 entries in order to release some space for the new ones. This is
Davor Ocelice9ed2812017-12-25 17:49:28 +01009247 most often the desired behavior. In some specific cases, it
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009248 be desirable to refuse new entries instead of purging the older
9249 ones. That may be the case when the amount of data to store is
9250 far above the hardware limits and we prefer not to offer access
9251 to new clients than to reject the ones already connected. When
9252 using this parameter, be sure to properly set the "expire"
9253 parameter (see below).
9254
Emeric Brunf099e792010-09-27 12:05:28 +02009255 <peersect> is the name of the peers section to use for replication. Entries
9256 which associate keys to server IDs are kept synchronized with
9257 the remote peers declared in this section. All entries are also
9258 automatically learned from the local peer (old process) during a
9259 soft restart.
9260
Willy Tarreau1abc6732015-05-01 19:21:02 +02009261 NOTE : each peers section may be referenced only by tables
9262 belonging to the same unique process.
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01009263
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009264 <expire> defines the maximum duration of an entry in the table since it
9265 was last created, refreshed or matched. The expiration delay is
9266 defined using the standard time format, similarly as the various
9267 timeouts. The maximum duration is slightly above 24 days. See
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009268 section 2.4 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02009269 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009270 be removed once full. Be sure not to use the "nopurge" parameter
9271 if not expiration delay is specified.
9272
Willy Tarreau08d5f982010-06-06 13:34:54 +02009273 <data_type> is used to store additional information in the stick-table. This
9274 may be used by ACLs in order to control various criteria related
9275 to the activity of the client matching the stick-table. For each
9276 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02009277 that the additional data can fit. Several data types may be
9278 stored with an entry. Multiple data types may be specified after
9279 the "store" keyword, as a comma-separated list. Alternatively,
9280 it is possible to repeat the "store" keyword followed by one or
9281 several data types. Except for the "server_id" type which is
9282 automatically detected and enabled, all data types must be
9283 explicitly declared to be stored. If an ACL references a data
9284 type which is not stored, the ACL will simply not match. Some
9285 data types require an argument which must be passed just after
9286 the type between parenthesis. See below for the supported data
9287 types and their arguments.
9288
9289 The data types that can be stored with an entry are the following :
9290 - server_id : this is an integer which holds the numeric ID of the server a
9291 request was assigned to. It is used by the "stick match", "stick store",
9292 and "stick on" rules. It is automatically enabled when referenced.
9293
9294 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
9295 integer which may be used for anything. Most of the time it will be used
9296 to put a special tag on some entries, for instance to note that a
Davor Ocelice9ed2812017-12-25 17:49:28 +01009297 specific behavior was detected and must be known for future matches.
Willy Tarreauc9705a12010-07-27 20:05:50 +02009298
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009299 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
9300 over a period. It is a positive 32-bit integer integer which may be used
9301 for anything. Just like <gpc0>, it counts events, but instead of keeping
Davor Ocelice9ed2812017-12-25 17:49:28 +01009302 a cumulative number, it maintains the rate at which the counter is
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009303 incremented. Most of the time it will be used to measure the frequency of
Davor Ocelice9ed2812017-12-25 17:49:28 +01009304 occurrence of certain events (e.g. requests to a specific URL).
Willy Tarreauba2ffd12013-05-29 15:54:14 +02009305
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009306 - gpc1 : second General Purpose Counter. It is a positive 32-bit integer
9307 integer which may be used for anything. Most of the time it will be used
9308 to put a special tag on some entries, for instance to note that a
9309 specific behavior was detected and must be known for future matches.
9310
9311 - gpc1_rate(<period>) : increment rate of the second General Purpose Counter
9312 over a period. It is a positive 32-bit integer integer which may be used
9313 for anything. Just like <gpc1>, it counts events, but instead of keeping
9314 a cumulative number, it maintains the rate at which the counter is
9315 incremented. Most of the time it will be used to measure the frequency of
9316 occurrence of certain events (e.g. requests to a specific URL).
9317
Willy Tarreauc9705a12010-07-27 20:05:50 +02009318 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
9319 the absolute number of connections received from clients which matched
9320 this entry. It does not mean the connections were accepted, just that
9321 they were received.
9322
9323 - conn_cur : Current Connections. It is a positive 32-bit integer which
9324 stores the concurrent connection counts for the entry. It is incremented
9325 once an incoming connection matches the entry, and decremented once the
9326 connection leaves. That way it is possible to know at any time the exact
9327 number of concurrent connections for an entry.
9328
9329 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9330 integer parameter <period> which indicates in milliseconds the length
9331 of the period over which the average is measured. It reports the average
9332 incoming connection rate over that period, in connections per period. The
9333 result is an integer which can be matched using ACLs.
9334
9335 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
9336 the absolute number of sessions received from clients which matched this
9337 entry. A session is a connection that was accepted by the layer 4 rules.
9338
9339 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9340 integer parameter <period> which indicates in milliseconds the length
9341 of the period over which the average is measured. It reports the average
9342 incoming session rate over that period, in sessions per period. The
9343 result is an integer which can be matched using ACLs.
9344
9345 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
9346 counts the absolute number of HTTP requests received from clients which
9347 matched this entry. It does not matter whether they are valid requests or
9348 not. Note that this is different from sessions when keep-alive is used on
9349 the client side.
9350
9351 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9352 integer parameter <period> which indicates in milliseconds the length
9353 of the period over which the average is measured. It reports the average
9354 HTTP request rate over that period, in requests per period. The result is
9355 an integer which can be matched using ACLs. It does not matter whether
9356 they are valid requests or not. Note that this is different from sessions
9357 when keep-alive is used on the client side.
9358
9359 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
9360 counts the absolute number of HTTP requests errors induced by clients
9361 which matched this entry. Errors are counted on invalid and truncated
9362 requests, as well as on denied or tarpitted requests, and on failed
9363 authentications. If the server responds with 4xx, then the request is
9364 also counted as an error since it's an error triggered by the client
Davor Ocelice9ed2812017-12-25 17:49:28 +01009365 (e.g. vulnerability scan).
Willy Tarreauc9705a12010-07-27 20:05:50 +02009366
9367 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9368 integer parameter <period> which indicates in milliseconds the length
9369 of the period over which the average is measured. It reports the average
9370 HTTP request error rate over that period, in requests per period (see
9371 http_err_cnt above for what is accounted as an error). The result is an
9372 integer which can be matched using ACLs.
9373
9374 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009375 integer which counts the cumulative number of bytes received from clients
Willy Tarreauc9705a12010-07-27 20:05:50 +02009376 which matched this entry. Headers are included in the count. This may be
9377 used to limit abuse of upload features on photo or video servers.
9378
9379 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
9380 integer parameter <period> which indicates in milliseconds the length
9381 of the period over which the average is measured. It reports the average
9382 incoming bytes rate over that period, in bytes per period. It may be used
9383 to detect users which upload too much and too fast. Warning: with large
9384 uploads, it is possible that the amount of uploaded data will be counted
9385 once upon termination, thus causing spikes in the average transfer speed
9386 instead of having a smooth one. This may partially be smoothed with
9387 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
9388 recommended for better fairness.
9389
9390 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
Davor Ocelice9ed2812017-12-25 17:49:28 +01009391 integer which counts the cumulative number of bytes sent to clients which
Willy Tarreauc9705a12010-07-27 20:05:50 +02009392 matched this entry. Headers are included in the count. This may be used
9393 to limit abuse of bots sucking the whole site.
9394
9395 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
9396 an integer parameter <period> which indicates in milliseconds the length
9397 of the period over which the average is measured. It reports the average
9398 outgoing bytes rate over that period, in bytes per period. It may be used
9399 to detect users which download too much and too fast. Warning: with large
9400 transfers, it is possible that the amount of transferred data will be
9401 counted once upon termination, thus causing spikes in the average
9402 transfer speed instead of having a smooth one. This may partially be
9403 smoothed with "option contstats" though this is not perfect yet. Use of
9404 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02009405
Willy Tarreauc00cdc22010-06-06 16:48:26 +02009406 There is only one stick-table per proxy. At the moment of writing this doc,
9407 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009408 to be required, simply create a dummy backend with a stick-table in it and
9409 reference it.
9410
9411 It is important to understand that stickiness based on learning information
9412 has some limitations, including the fact that all learned associations are
Baptiste Assmann123ff042016-03-06 23:29:28 +01009413 lost upon restart unless peers are properly configured to transfer such
9414 information upon restart (recommended). In general it can be good as a
9415 complement but not always as an exclusive stickiness.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009416
Willy Tarreauc9705a12010-07-27 20:05:50 +02009417 Last, memory requirements may be important when storing many data types.
9418 Indeed, storing all indicators above at once in each entry requires 116 bytes
9419 per entry, or 116 MB for a 1-million entries table. This is definitely not
9420 something that can be ignored.
9421
9422 Example:
9423 # Keep track of counters of up to 1 million IP addresses over 5 minutes
9424 # and store a general purpose counter and the average connection rate
9425 # computed over a sliding window of 30 seconds.
9426 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
9427
Jarno Huuskonene0ee0be2017-07-04 10:35:12 +03009428 See also : "stick match", "stick on", "stick store-request", section 2.4
David du Colombiera13d1b92011-03-17 10:40:22 +01009429 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01009430
9431
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009432stick store-response <pattern> [table <table>] [{if | unless} <condition>]
Baptiste Assmann2f2d2ec2016-03-06 23:27:24 +01009433 Define a response pattern used to create an entry in a stickiness table
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009434 May be used in sections : defaults | frontend | listen | backend
9435 no | no | yes | yes
9436
9437 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009438 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009439 describes what elements of the response or connection will
Davor Ocelice9ed2812017-12-25 17:49:28 +01009440 be analyzed, extracted and stored in the table once a
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009441 server is selected.
9442
9443 <table> is an optional stickiness table name. If unspecified, the same
9444 backend's table is used. A stickiness table is declared using
9445 the "stick-table" statement.
9446
9447 <cond> is an optional storage condition. It makes it possible to store
9448 certain criteria only when some conditions are met (or not met).
9449 For instance, it could be used to store the SSL session ID only
9450 when the response is a SSL server hello.
9451
9452 Some protocols or applications require complex stickiness rules and cannot
9453 always simply rely on cookies nor hashing. The "stick store-response"
9454 statement describes a rule to decide what to extract from the response and
9455 when to do it, in order to store it into a stickiness table for further
9456 requests to match it using the "stick match" statement. Obviously the
9457 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009458 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009459 See section 7 for a complete list of possible patterns and transformation
9460 rules.
9461
9462 The table has to be declared using the "stick-table" statement. It must be of
9463 a type compatible with the pattern. By default it is the one which is present
9464 in the same backend. It is possible to share a table with other backends by
9465 referencing it using the "table" keyword. If another table is referenced,
9466 the server's ID inside the backends are used. By default, all server IDs
9467 start at 1 in each backend, so the server ordering is enough. But in case of
9468 doubt, it is highly recommended to force server IDs using their "id" setting.
9469
9470 It is possible to restrict the conditions where a "stick store-response"
9471 statement will apply, using "if" or "unless" followed by a condition. This
9472 condition will be evaluated while parsing the response, so any criteria can
9473 be used. See section 7 for ACL based conditions.
9474
9475 There is no limit on the number of "stick store-response" statements, but
9476 there is a limit of 8 simultaneous stores per request or response. This
9477 makes it possible to store up to 8 criteria, all extracted from either the
9478 request or the response, regardless of the number of rules. Only the 8 first
9479 ones which match will be kept. Using this, it is possible to feed multiple
9480 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01009481 another protocol or access method. Using multiple store-response rules with
9482 the same table is possible and may be used to find the best criterion to rely
9483 on, by arranging the rules by decreasing preference order. Only the first
9484 extracted criterion for a given table will be stored. All subsequent store-
9485 response rules referencing the same table will be skipped and their ACLs will
9486 not be evaluated. However, even if a store-request rule references a table, a
9487 store-response rule may also use the same table. This means that each table
9488 may learn exactly one element from the request and one element from the
9489 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009490
9491 The table will contain the real server that processed the request.
9492
9493 Example :
9494 # Learn SSL session ID from both request and response and create affinity.
9495 backend https
9496 mode tcp
9497 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02009498 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009499 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009500
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009501 acl clienthello req_ssl_hello_type 1
9502 acl serverhello rep_ssl_hello_type 2
9503
9504 # use tcp content accepts to detects ssl client and server hello.
9505 tcp-request inspect-delay 5s
9506 tcp-request content accept if clienthello
9507
9508 # no timeout on response inspect delay by default.
9509 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02009510
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009511 # SSL session ID (SSLID) may be present on a client or server hello.
9512 # Its length is coded on 1 byte at offset 43 and its value starts
9513 # at offset 44.
9514
9515 # Match and learn on request if client hello.
9516 stick on payload_lv(43,1) if clienthello
9517
9518 # Learn on response if server hello.
9519 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02009520
Emeric Brun6a1cefa2010-09-24 18:15:17 +02009521 server s1 192.168.1.1:443
9522 server s2 192.168.1.1:443
9523
9524 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
9525 extraction.
9526
9527
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009528tcp-check connect [params*]
9529 Opens a new connection
9530 May be used in sections: defaults | frontend | listen | backend
9531 no | no | yes | yes
9532
9533 When an application lies on more than a single TCP port or when HAProxy
9534 load-balance many services in a single backend, it makes sense to probe all
9535 the services individually before considering a server as operational.
9536
9537 When there are no TCP port configured on the server line neither server port
9538 directive, then the 'tcp-check connect port <port>' must be the first step
9539 of the sequence.
9540
9541 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
9542 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
9543 do.
9544
9545 Parameters :
9546 They are optional and can be used to describe how HAProxy should open and
9547 use the TCP connection.
9548
9549 port if not set, check port or server port is used.
9550 It tells HAProxy where to open the connection to.
9551 <port> must be a valid TCP port source integer, from 1 to 65535.
9552
9553 send-proxy send a PROXY protocol string
9554
9555 ssl opens a ciphered connection
9556
9557 Examples:
9558 # check HTTP and HTTPs services on a server.
9559 # first open port 80 thanks to server line port directive, then
9560 # tcp-check opens port 443, ciphered and run a request on it:
9561 option tcp-check
9562 tcp-check connect
9563 tcp-check send GET\ /\ HTTP/1.0\r\n
9564 tcp-check send Host:\ haproxy.1wt.eu\r\n
9565 tcp-check send \r\n
9566 tcp-check expect rstring (2..|3..)
9567 tcp-check connect port 443 ssl
9568 tcp-check send GET\ /\ HTTP/1.0\r\n
9569 tcp-check send Host:\ haproxy.1wt.eu\r\n
9570 tcp-check send \r\n
9571 tcp-check expect rstring (2..|3..)
9572 server www 10.0.0.1 check port 80
9573
9574 # check both POP and IMAP from a single server:
9575 option tcp-check
9576 tcp-check connect port 110
9577 tcp-check expect string +OK\ POP3\ ready
9578 tcp-check connect port 143
9579 tcp-check expect string *\ OK\ IMAP4\ ready
9580 server mail 10.0.0.1 check
9581
9582 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
9583
9584
9585tcp-check expect [!] <match> <pattern>
Davor Ocelice9ed2812017-12-25 17:49:28 +01009586 Specify data to be collected and analyzed during a generic health check
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009587 May be used in sections: defaults | frontend | listen | backend
9588 no | no | yes | yes
9589
9590 Arguments :
9591 <match> is a keyword indicating how to look for a specific pattern in the
9592 response. The keyword may be one of "string", "rstring" or
9593 binary.
9594 The keyword may be preceded by an exclamation mark ("!") to negate
9595 the match. Spaces are allowed between the exclamation mark and the
9596 keyword. See below for more details on the supported keywords.
9597
9598 <pattern> is the pattern to look for. It may be a string or a regular
9599 expression. If the pattern contains spaces, they must be escaped
9600 with the usual backslash ('\').
9601 If the match is set to binary, then the pattern must be passed as
Davor Ocelice9ed2812017-12-25 17:49:28 +01009602 a series of hexadecimal digits in an even number. Each sequence of
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009603 two digits will represent a byte. The hexadecimal digits may be
9604 used upper or lower case.
9605
9606
9607 The available matches are intentionally similar to their http-check cousins :
9608
9609 string <string> : test the exact string matches in the response buffer.
9610 A health check response will be considered valid if the
9611 response's buffer contains this exact string. If the
9612 "string" keyword is prefixed with "!", then the response
9613 will be considered invalid if the body contains this
9614 string. This can be used to look for a mandatory pattern
9615 in a protocol response, or to detect a failure when a
9616 specific error appears in a protocol banner.
9617
9618 rstring <regex> : test a regular expression on the response buffer.
9619 A health check response will be considered valid if the
9620 response's buffer matches this expression. If the
9621 "rstring" keyword is prefixed with "!", then the response
9622 will be considered invalid if the body matches the
9623 expression.
9624
9625 binary <hexstring> : test the exact string in its hexadecimal form matches
9626 in the response buffer. A health check response will
9627 be considered valid if the response's buffer contains
9628 this exact hexadecimal string.
9629 Purpose is to match data on binary protocols.
9630
9631 It is important to note that the responses will be limited to a certain size
9632 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
9633 Thus, too large responses may not contain the mandatory pattern when using
9634 "string", "rstring" or binary. If a large response is absolutely required, it
9635 is possible to change the default max size by setting the global variable.
9636 However, it is worth keeping in mind that parsing very large responses can
9637 waste some CPU cycles, especially when regular expressions are used, and that
9638 it is always better to focus the checks on smaller resources. Also, in its
9639 current state, the check will not find any string nor regex past a null
9640 character in the response. Similarly it is not possible to request matching
9641 the null character.
9642
9643 Examples :
9644 # perform a POP check
9645 option tcp-check
9646 tcp-check expect string +OK\ POP3\ ready
9647
9648 # perform an IMAP check
9649 option tcp-check
9650 tcp-check expect string *\ OK\ IMAP4\ ready
9651
9652 # look for the redis master server
9653 option tcp-check
9654 tcp-check send PING\r\n
Baptiste Assmanna3322992015-08-04 10:12:18 +02009655 tcp-check expect string +PONG
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009656 tcp-check send info\ replication\r\n
9657 tcp-check expect string role:master
9658 tcp-check send QUIT\r\n
9659 tcp-check expect string +OK
9660
9661
9662 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
9663 "tcp-check send-binary", "http-check expect", tune.chksize
9664
9665
9666tcp-check send <data>
9667 Specify a string to be sent as a question during a generic health check
9668 May be used in sections: defaults | frontend | listen | backend
9669 no | no | yes | yes
9670
9671 <data> : the data to be sent as a question during a generic health check
9672 session. For now, <data> must be a string.
9673
9674 Examples :
9675 # look for the redis master server
9676 option tcp-check
9677 tcp-check send info\ replication\r\n
9678 tcp-check expect string role:master
9679
9680 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9681 "tcp-check send-binary", tune.chksize
9682
9683
Davor Ocelice9ed2812017-12-25 17:49:28 +01009684tcp-check send-binary <hexstring>
9685 Specify a hex digits string to be sent as a binary question during a raw
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009686 tcp health check
9687 May be used in sections: defaults | frontend | listen | backend
9688 no | no | yes | yes
9689
9690 <data> : the data to be sent as a question during a generic health check
9691 session. For now, <data> must be a string.
Davor Ocelice9ed2812017-12-25 17:49:28 +01009692 <hexstring> : test the exact string in its hexadecimal form matches in the
Willy Tarreau938c7fe2014-04-25 14:21:39 +02009693 response buffer. A health check response will be considered
9694 valid if the response's buffer contains this exact
9695 hexadecimal string.
9696 Purpose is to send binary data to ask on binary protocols.
9697
9698 Examples :
9699 # redis check in binary
9700 option tcp-check
9701 tcp-check send-binary 50494e470d0a # PING\r\n
9702 tcp-check expect binary 2b504F4e47 # +PONG
9703
9704
9705 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
9706 "tcp-check send", tune.chksize
9707
9708
Willy Tarreaue9656522010-08-17 15:40:09 +02009709tcp-request connection <action> [{if | unless} <condition>]
9710 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02009711 May be used in sections : defaults | frontend | listen | backend
9712 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02009713 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009714 <action> defines the action to perform if the condition applies. See
9715 below.
Willy Tarreau1a687942010-05-23 22:40:30 +02009716
Willy Tarreaue9656522010-08-17 15:40:09 +02009717 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009718
9719 Immediately after acceptance of a new incoming connection, it is possible to
9720 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02009721 or dropped or have its counters tracked. Those conditions cannot make use of
9722 any data contents because the connection has not been read from yet, and the
9723 buffers are not yet allocated. This is used to selectively and very quickly
9724 accept or drop connections from various sources with a very low overhead. If
9725 some contents need to be inspected in order to take the decision, the
9726 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009727
Willy Tarreaue9656522010-08-17 15:40:09 +02009728 The "tcp-request connection" rules are evaluated in their exact declaration
9729 order. If no rule matches or if there is no rule, the default action is to
9730 accept the incoming connection. There is no specific limit to the number of
9731 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009732
Willy Tarreaua9083d02015-05-08 15:27:59 +02009733 Four types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02009734 - accept :
9735 accepts the connection if the condition is true (when used with "if")
9736 or false (when used with "unless"). The first such rule executed ends
9737 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009738
Willy Tarreaue9656522010-08-17 15:40:09 +02009739 - reject :
9740 rejects the connection if the condition is true (when used with "if")
9741 or false (when used with "unless"). The first such rule executed ends
9742 the rules evaluation. Rejected connections do not even become a
9743 session, which is why they are accounted separately for in the stats,
9744 as "denied connections". They are not considered for the session
9745 rate-limit and are not logged either. The reason is that these rules
9746 should only be used to filter extremely high connection rates such as
9747 the ones encountered during a massive DDoS attack. Under these extreme
9748 conditions, the simple action of logging each event would make the
9749 system collapse and would considerably lower the filtering capacity. If
9750 logging is absolutely desired, then "tcp-request content" rules should
Willy Tarreau4f614292016-10-21 17:49:36 +02009751 be used instead, as "tcp-request session" rules will not log either.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009752
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009753 - expect-proxy layer4 :
9754 configures the client-facing connection to receive a PROXY protocol
9755 header before any byte is read from the socket. This is equivalent to
9756 having the "accept-proxy" keyword on the "bind" line, except that using
9757 the TCP rule allows the PROXY protocol to be accepted only for certain
9758 IP address ranges using an ACL. This is convenient when multiple layers
9759 of load balancers are passed through by traffic coming from public
9760 hosts.
9761
Bertrand Jacquin90759682016-06-06 15:35:39 +01009762 - expect-netscaler-cip layer4 :
9763 configures the client-facing connection to receive a NetScaler Client
9764 IP insertion protocol header before any byte is read from the socket.
9765 This is equivalent to having the "accept-netscaler-cip" keyword on the
9766 "bind" line, except that using the TCP rule allows the PROXY protocol
9767 to be accepted only for certain IP address ranges using an ACL. This
9768 is convenient when multiple layers of load balancers are passed
9769 through by traffic coming from public hosts.
9770
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009771 - capture <sample> len <length> :
9772 This only applies to "tcp-request content" rules. It captures sample
9773 expression <sample> from the request buffer, and converts it to a
9774 string of at most <len> characters. The resulting string is stored into
9775 the next request "capture" slot, so it will possibly appear next to
9776 some captured HTTP headers. It will then automatically appear in the
9777 logs, and it will be possible to extract it using sample fetch rules to
9778 feed it into headers or anything. The length should be limited given
9779 that this size will be allocated for each capture during the whole
Willy Tarreaua9083d02015-05-08 15:27:59 +02009780 session life. Please check section 7.3 (Fetching samples) and "capture
9781 request header" for more information.
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009782
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009783 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02009784 enables tracking of sticky counters from current connection. These
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +02009785 rules do not stop evaluation and do not change default action. The
9786 number of counters that may be simultaneously tracked by the same
9787 connection is set in MAX_SESS_STKCTR at build time (reported in
9788 haproxy -vv) whichs defaults to 3, so the track-sc number is between 0
9789 and (MAX_SESS_STCKTR-1). The first "track-sc0" rule executed enables
9790 tracking of the counters of the specified table as the first set. The
9791 first "track-sc1" rule executed enables tracking of the counters of the
9792 specified table as the second set. The first "track-sc2" rule executed
9793 enables tracking of the counters of the specified table as the third
9794 set. It is a recommended practice to use the first set of counters for
9795 the per-frontend counters and the second set for the per-backend ones.
9796 But this is just a guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009797
Willy Tarreaue9656522010-08-17 15:40:09 +02009798 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02009799 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02009800 in section 7.3. It describes what elements of the incoming
Davor Ocelice9ed2812017-12-25 17:49:28 +01009801 request or connection will be analyzed, extracted, combined,
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009802 and used to select which table entry to update the counters.
9803 Note that "tcp-request connection" cannot use content-based
9804 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009805
Willy Tarreaue9656522010-08-17 15:40:09 +02009806 <table> is an optional table to be used instead of the default one,
9807 which is the stick-table declared in the current proxy. All
9808 the counters for the matches and updates for the key will
9809 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009810
Willy Tarreaue9656522010-08-17 15:40:09 +02009811 Once a "track-sc*" rule is executed, the key is looked up in the table
9812 and if it is not found, an entry is allocated for it. Then a pointer to
9813 that entry is kept during all the session's life, and this entry's
9814 counters are updated as often as possible, every time the session's
9815 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01009816 Counters are only updated for events that happen after the tracking has
9817 been started. For example, connection counters will not be updated when
9818 tracking layer 7 information, since the connection event happens before
9819 layer7 information is extracted.
9820
Willy Tarreaue9656522010-08-17 15:40:09 +02009821 If the entry tracks concurrent connection counters, one connection is
9822 counted for as long as the entry is tracked, and the entry will not
9823 expire during that time. Tracking counters also provides a performance
9824 advantage over just checking the keys, because only one table lookup is
9825 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009826
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009827 - sc-inc-gpc0(<sc-id>):
9828 The "sc-inc-gpc0" increments the GPC0 counter according to the sticky
9829 counter designated by <sc-id>. If an error occurs, this action silently
9830 fails and the actions evaluation continues.
9831
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009832 - sc-inc-gpc1(<sc-id>):
9833 The "sc-inc-gpc1" increments the GPC1 counter according to the sticky
9834 counter designated by <sc-id>. If an error occurs, this action silently
9835 fails and the actions evaluation continues.
9836
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009837 - sc-set-gpt0(<sc-id>) <int>:
9838 This action sets the GPT0 tag according to the sticky counter designated
9839 by <sc-id> and the value of <int>. The expected result is a boolean. If
9840 an error occurs, this action silently fails and the actions evaluation
9841 continues.
9842
William Lallemand2e785f22016-05-25 01:48:42 +02009843 - set-src <expr> :
9844 Is used to set the source IP address to the value of specified
9845 expression. Useful if you want to mask source IP for privacy.
9846 If you want to provide an IP from a HTTP header use "http-request
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009847 set-src".
William Lallemand2e785f22016-05-25 01:48:42 +02009848
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009849 Arguments:
9850 <expr> Is a standard HAProxy expression formed by a sample-fetch
9851 followed by some converters.
William Lallemand2e785f22016-05-25 01:48:42 +02009852
9853 Example:
William Lallemand2e785f22016-05-25 01:48:42 +02009854 tcp-request connection set-src src,ipmask(24)
9855
Willy Tarreau0c630532016-10-21 17:52:58 +02009856 When possible, set-src preserves the original source port as long as the
9857 address family allows it, otherwise the source port is set to 0.
William Lallemand2e785f22016-05-25 01:48:42 +02009858
William Lallemand44be6402016-05-25 01:51:35 +02009859 - set-src-port <expr> :
9860 Is used to set the source port address to the value of specified
9861 expression.
9862
Cyril Bonté6c81d5f2018-10-17 00:14:51 +02009863 Arguments:
9864 <expr> Is a standard HAProxy expression formed by a sample-fetch
9865 followed by some converters.
William Lallemand44be6402016-05-25 01:51:35 +02009866
9867 Example:
William Lallemand44be6402016-05-25 01:51:35 +02009868 tcp-request connection set-src-port int(4000)
9869
Willy Tarreau0c630532016-10-21 17:52:58 +02009870 When possible, set-src-port preserves the original source address as long
9871 as the address family supports a port, otherwise it forces the source
9872 address to IPv4 "0.0.0.0" before rewriting the port.
William Lallemand44be6402016-05-25 01:51:35 +02009873
William Lallemand13e9b0c2016-05-25 02:34:07 +02009874 - set-dst <expr> :
9875 Is used to set the destination IP address to the value of specified
9876 expression. Useful if you want to mask IP for privacy in log.
9877 If you want to provide an IP from a HTTP header use "http-request
9878 set-dst". If you want to connect to the new address/port, use
9879 '0.0.0.0:0' as a server address in the backend.
9880
9881 <expr> Is a standard HAProxy expression formed by a sample-fetch
9882 followed by some converters.
9883
9884 Example:
9885
9886 tcp-request connection set-dst dst,ipmask(24)
9887 tcp-request connection set-dst ipv4(10.0.0.1)
9888
Willy Tarreau0c630532016-10-21 17:52:58 +02009889 When possible, set-dst preserves the original destination port as long as
9890 the address family allows it, otherwise the destination port is set to 0.
9891
William Lallemand13e9b0c2016-05-25 02:34:07 +02009892 - set-dst-port <expr> :
9893 Is used to set the destination port address to the value of specified
9894 expression. If you want to connect to the new address/port, use
9895 '0.0.0.0:0' as a server address in the backend.
9896
9897
9898 <expr> Is a standard HAProxy expression formed by a sample-fetch
9899 followed by some converters.
9900
9901 Example:
9902
9903 tcp-request connection set-dst-port int(4000)
9904
Willy Tarreau0c630532016-10-21 17:52:58 +02009905 When possible, set-dst-port preserves the original destination address as
9906 long as the address family supports a port, otherwise it forces the
9907 destination address to IPv4 "0.0.0.0" before rewriting the port.
9908
Willy Tarreau2d392c22015-08-24 01:43:45 +02009909 - "silent-drop" :
9910 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +01009911 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +02009912 to prevent the client from being notified. The effect it then that the
9913 client still sees an established connection while there's none on
9914 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
9915 except that it doesn't use any local resource at all on the machine
9916 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +01009917 slow down stronger attackers. It is important to understand the impact
9918 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009919 client and HAProxy (firewalls, proxies, load balancers) will also keep
9920 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +01009921 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +02009922 TCP_REPAIR socket option is used to block the emission of a TCP
9923 reset. On other systems, the socket's TTL is reduced to 1 so that the
9924 TCP reset doesn't pass the first router, though it's still delivered to
9925 local networks. Do not use it unless you fully understand how it works.
9926
Willy Tarreaue9656522010-08-17 15:40:09 +02009927 Note that the "if/unless" condition is optional. If no condition is set on
9928 the action, it is simply performed unconditionally. That can be useful for
9929 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009930
Willy Tarreaue9656522010-08-17 15:40:09 +02009931 Example: accept all connections from white-listed hosts, reject too fast
9932 connection without counting them, and track accepted connections.
9933 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009934
Willy Tarreaue9656522010-08-17 15:40:09 +02009935 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009936 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009937 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009938
Willy Tarreaue9656522010-08-17 15:40:09 +02009939 Example: accept all connections from white-listed hosts, count all other
9940 connections and reject too fast ones. This results in abusive ones
9941 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009942
Willy Tarreaue9656522010-08-17 15:40:09 +02009943 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009944 tcp-request connection track-sc0 src
9945 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009946
Willy Tarreau4f0d9192013-06-11 20:40:55 +02009947 Example: enable the PROXY protocol for traffic coming from all known proxies.
9948
9949 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
9950
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009951 See section 7 about ACL usage.
9952
Willy Tarreau4f614292016-10-21 17:49:36 +02009953 See also : "tcp-request session", "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02009954
9955
Willy Tarreaue9656522010-08-17 15:40:09 +02009956tcp-request content <action> [{if | unless} <condition>]
9957 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02009958 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02009959 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02009960 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +02009961 <action> defines the action to perform if the condition applies. See
9962 below.
Willy Tarreau62644772008-07-16 18:36:06 +02009963
Willy Tarreaue9656522010-08-17 15:40:09 +02009964 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02009965
Davor Ocelice9ed2812017-12-25 17:49:28 +01009966 A request's contents can be analyzed at an early stage of request processing
Willy Tarreaue9656522010-08-17 15:40:09 +02009967 called "TCP content inspection". During this stage, ACL-based rules are
9968 evaluated every time the request contents are updated, until either an
9969 "accept" or a "reject" rule matches, or the TCP request inspection delay
9970 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02009971
Willy Tarreaue9656522010-08-17 15:40:09 +02009972 The first difference between these rules and "tcp-request connection" rules
9973 is that "tcp-request content" rules can make use of contents to take a
9974 decision. Most often, these decisions will consider a protocol recognition or
9975 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01009976 both frontends and backends. In case of HTTP keep-alive with the client, all
9977 tcp-request content rules are evaluated again, so haproxy keeps a record of
9978 what sticky counters were assigned by a "tcp-request connection" versus a
9979 "tcp-request content" rule, and flushes all the content-related ones after
9980 processing an HTTP request, so that they may be evaluated again by the rules
9981 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009982 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01009983 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02009984
Willy Tarreaue9656522010-08-17 15:40:09 +02009985 Content-based rules are evaluated in their exact declaration order. If no
9986 rule matches or if there is no rule, the default action is to accept the
9987 contents. There is no specific limit to the number of rules which may be
9988 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02009989
Thierry FOURNIER236657b2015-08-19 08:25:14 +02009990 Several types of actions are supported :
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009991 - accept : the request is accepted
Baptiste Assmann333939c2019-01-21 08:34:50 +01009992 - do-resolve: perform a DNS resolution
Willy Tarreau18bf01e2014-06-13 16:18:52 +02009993 - reject : the request is rejected and the connection is closed
9994 - capture : the specified sample expression is captured
Patrick Hemmer268a7072018-05-11 12:52:31 -04009995 - set-priority-class <expr> | set-priority-offset <expr>
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02009996 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +02009997 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +01009998 - sc-inc-gpc1(<sc-id>)
Thierry Fournierb9125672016-03-29 19:34:37 +02009999 - sc-set-gpt0(<sc-id>) <int>
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010000 - set-dst <expr>
10001 - set-dst-port <expr>
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010002 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010003 - unset-var(<var-name>)
Willy Tarreau2d392c22015-08-24 01:43:45 +020010004 - silent-drop
Davor Ocelice9ed2812017-12-25 17:49:28 +010010005 - send-spoe-group <engine-name> <group-name>
Willy Tarreau62644772008-07-16 18:36:06 +020010006
Willy Tarreaue9656522010-08-17 15:40:09 +020010007 They have the same meaning as their counter-parts in "tcp-request connection"
10008 so please refer to that section for a complete description.
Baptiste Assmann333939c2019-01-21 08:34:50 +010010009 For "do-resolve" action, please check the "http-request do-resolve"
10010 configuration section.
Willy Tarreau62644772008-07-16 18:36:06 +020010011
Willy Tarreauf3338342014-01-28 21:40:28 +010010012 While there is nothing mandatory about it, it is recommended to use the
10013 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
10014 content" rules in the frontend, and track-sc2 for "tcp-request content"
10015 rules in the backend, because that makes the configuration more readable
10016 and easier to troubleshoot, but this is just a guideline and all counters
10017 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +020010018
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010019 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +020010020 the action, it is simply performed unconditionally. That can be useful for
10021 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +020010022
Willy Tarreaue9656522010-08-17 15:40:09 +020010023 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +020010024 rules, since HTTP-specific ACL matches are able to preliminarily parse the
10025 contents of a buffer before extracting the required data. If the buffered
10026 contents do not parse as a valid HTTP message, then the ACL does not match.
10027 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +010010028 processing, so there is no risk of parsing something differently. In an HTTP
10029 backend connected to from an HTTP frontend, it is guaranteed that HTTP
10030 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +020010031
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010032 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010033 are present when the rule is processed. The rule processing engine is able to
10034 wait until the inspect delay expires when the data to be tracked is not yet
10035 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010036
Baptiste Assmanne1afd4f2019-04-18 16:21:13 +020010037 The "set-dst" and "set-dst-port" are used to set respectively the destination
10038 IP and port. More information on how to use it at "http-request set-dst".
10039
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010040 The "set-var" is used to set the content of a variable. The variable is
Willy Tarreau4f614292016-10-21 17:49:36 +020010041 declared inline. For "tcp-request session" rules, only session-level
10042 variables can be used, without any layer7 contents.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010043
Daniel Schneller0b547052016-03-21 20:46:57 +010010044 <var-name> The name of the variable starts with an indication about
10045 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010046 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010047 "sess" : the variable is shared with the whole session
10048 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010049 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010050 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010051 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010052 "res" : the variable is shared only during response
10053 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010054 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010055 The name may only contain characters 'a-z', 'A-Z', '0-9',
10056 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010057
10058 <expr> Is a standard HAProxy expression formed by a sample-fetch
10059 followed by some converters.
10060
Christopher Faulet85d79c92016-11-09 16:54:56 +010010061 The "unset-var" is used to unset a variable. See above for details about
10062 <var-name>.
10063
Patrick Hemmer268a7072018-05-11 12:52:31 -040010064 The "set-priority-class" is used to set the queue priority class of the
10065 current request. The value must be a sample expression which converts to an
10066 integer in the range -2047..2047. Results outside this range will be
10067 truncated. The priority class determines the order in which queued requests
10068 are processed. Lower values have higher priority.
10069
10070 The "set-priority-offset" is used to set the queue priority timestamp offset
10071 of the current request. The value must be a sample expression which converts
10072 to an integer in the range -524287..524287. Results outside this range will be
10073 truncated. When a request is queued, it is ordered first by the priority
10074 class, then by the current timestamp adjusted by the given offset in
10075 milliseconds. Lower values have higher priority.
10076 Note that the resulting timestamp is is only tracked with enough precision for
10077 524,287ms (8m44s287ms). If the request is queued long enough to where the
10078 adjusted timestamp exceeds this value, it will be misidentified as highest
10079 priority. Thus it is important to set "timeout queue" to a value, where when
10080 combined with the offset, does not exceed this limit.
10081
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010082 The "send-spoe-group" is used to trigger sending of a group of SPOE
10083 messages. To do so, the SPOE engine used to send messages must be defined, as
10084 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10085 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10086 the SPOE agent name must be used.
10087
10088 <engine-name> The SPOE engine name.
10089
10090 <group-name> The SPOE group name as specified in the engine configuration.
10091
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010092 Example:
10093
10094 tcp-request content set-var(sess.my_var) src
Christopher Faulet85d79c92016-11-09 16:54:56 +010010095 tcp-request content unset-var(sess.my_var2)
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010096
Willy Tarreau62644772008-07-16 18:36:06 +020010097 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +020010098 # Accept HTTP requests containing a Host header saying "example.com"
10099 # and reject everything else.
10100 acl is_host_com hdr(Host) -i example.com
10101 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +020010102 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +020010103 tcp-request content reject
10104
10105 Example:
Willy Tarreau62644772008-07-16 18:36:06 +020010106 # reject SMTP connection if client speaks first
10107 tcp-request inspect-delay 30s
10108 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010109 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +020010110
10111 # Forward HTTPS connection only if client speaks
10112 tcp-request inspect-delay 30s
10113 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +020010114 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +020010115 tcp-request content reject
10116
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010117 Example:
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010118 # Track the last IP(stick-table type string) from X-Forwarded-For
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010119 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010120 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010121 # Or track the last IP(stick-table type ip|ipv6) from X-Forwarded-For
10122 tcp-request content track-sc0 req.hdr_ip(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010123
10124 Example:
10125 # track request counts per "base" (concatenation of Host+URL)
10126 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +020010127 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +010010128
Willy Tarreaue9656522010-08-17 15:40:09 +020010129 Example: track per-frontend and per-backend counters, block abusers at the
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010130 frontend when the backend detects abuse(and marks gpc0).
Willy Tarreaue9656522010-08-17 15:40:09 +020010131
10132 frontend http
Davor Ocelice9ed2812017-12-25 17:49:28 +010010133 # Use General Purpose Counter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +020010134 # protecting all our sites
10135 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010136 tcp-request connection track-sc0 src
10137 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +020010138 ...
10139 use_backend http_dynamic if { path_end .php }
10140
10141 backend http_dynamic
10142 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010143 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +020010144 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010145 acl click_too_fast sc1_http_req_rate gt 10
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010146 acl mark_as_abuser sc0_inc_gpc0(http) gt 0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010147 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +020010148 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +020010149
Willy Tarreauc57f0e22009-05-10 13:12:33 +020010150 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +020010151
Jarno Huuskonen95b012b2017-04-06 13:59:14 +030010152 See also : "tcp-request connection", "tcp-request session",
10153 "tcp-request inspect-delay", and "http-request".
Willy Tarreau62644772008-07-16 18:36:06 +020010154
10155
10156tcp-request inspect-delay <timeout>
10157 Set the maximum allowed time to wait for data during content inspection
10158 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +020010159 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +020010160 Arguments :
10161 <timeout> is the timeout value specified in milliseconds by default, but
10162 can be in any other unit if the number is suffixed by the unit,
10163 as explained at the top of this document.
10164
10165 People using haproxy primarily as a TCP relay are often worried about the
10166 risk of passing any type of protocol to a server without any analysis. In
10167 order to be able to analyze the request contents, we must first withhold
10168 the data then analyze them. This statement simply enables withholding of
10169 data for at most the specified amount of time.
10170
Willy Tarreaufb356202010-08-03 14:02:05 +020010171 TCP content inspection applies very early when a connection reaches a
10172 frontend, then very early when the connection is forwarded to a backend. This
10173 means that a connection may experience a first delay in the frontend and a
10174 second delay in the backend if both have tcp-request rules.
10175
Willy Tarreau62644772008-07-16 18:36:06 +020010176 Note that when performing content inspection, haproxy will evaluate the whole
10177 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010178 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +020010179 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +010010180 contents are definitive. If no delay is set, haproxy will not wait at all
10181 and will immediately apply a verdict based on the available information.
10182 Obviously this is unlikely to be very useful and might even be racy, so such
10183 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +020010184
10185 As soon as a rule matches, the request is released and continues as usual. If
10186 the timeout is reached and no rule matches, the default policy will be to let
10187 it pass through unaffected.
10188
10189 For most protocols, it is enough to set it to a few seconds, as most clients
10190 send the full request immediately upon connection. Add 3 or more seconds to
10191 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +010010192 to use large values, for instance to ensure that the client never talks
Davor Ocelice9ed2812017-12-25 17:49:28 +010010193 before the server (e.g. SMTP), or to wait for a client to talk before passing
10194 data to the server (e.g. SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +020010195 least the inspection delay, otherwise it will expire first. If the client
10196 closes the connection or if the buffer is full, the delay immediately expires
10197 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +020010198
Willy Tarreau55165fe2009-05-10 12:02:55 +020010199 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +020010200 "timeout client".
10201
10202
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010203tcp-response content <action> [{if | unless} <condition>]
10204 Perform an action on a session response depending on a layer 4-7 condition
10205 May be used in sections : defaults | frontend | listen | backend
10206 no | no | yes | yes
10207 Arguments :
Willy Tarreauc870bfd2015-09-28 18:47:38 +020010208 <action> defines the action to perform if the condition applies. See
10209 below.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010210
10211 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
10212
Davor Ocelice9ed2812017-12-25 17:49:28 +010010213 Response contents can be analyzed at an early stage of response processing
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010214 called "TCP content inspection". During this stage, ACL-based rules are
10215 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010216 "accept", "close" or a "reject" rule matches, or a TCP response inspection
10217 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010218
10219 Most often, these decisions will consider a protocol recognition or validity.
10220
10221 Content-based rules are evaluated in their exact declaration order. If no
10222 rule matches or if there is no rule, the default action is to accept the
10223 contents. There is no specific limit to the number of rules which may be
10224 inserted.
10225
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010226 Several types of actions are supported :
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010227 - accept :
10228 accepts the response if the condition is true (when used with "if")
10229 or false (when used with "unless"). The first such rule executed ends
10230 the rules evaluation.
10231
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010232 - close :
10233 immediately closes the connection with the server if the condition is
10234 true (when used with "if"), or false (when used with "unless"). The
10235 first such rule executed ends the rules evaluation. The main purpose of
10236 this action is to force a connection to be finished between a client
10237 and a server after an exchange when the application protocol expects
10238 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010239 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +020010240 protocols.
10241
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010242 - reject :
10243 rejects the response if the condition is true (when used with "if")
10244 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010245 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010246
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010247 - set-var(<var-name>) <expr>
10248 Sets a variable.
10249
Christopher Faulet85d79c92016-11-09 16:54:56 +010010250 - unset-var(<var-name>)
10251 Unsets a variable.
10252
Thierry FOURNIERe0627bd2015-08-04 08:20:33 +020010253 - sc-inc-gpc0(<sc-id>):
10254 This action increments the GPC0 counter according to the sticky
10255 counter designated by <sc-id>. If an error occurs, this action fails
10256 silently and the actions evaluation continues.
10257
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010258 - sc-inc-gpc1(<sc-id>):
10259 This action increments the GPC1 counter according to the sticky
10260 counter designated by <sc-id>. If an error occurs, this action fails
10261 silently and the actions evaluation continues.
10262
Thierry FOURNIER236657b2015-08-19 08:25:14 +020010263 - sc-set-gpt0(<sc-id>) <int> :
10264 This action sets the GPT0 tag according to the sticky counter designated
10265 by <sc-id> and the value of <int>. The expected result is a boolean. If
10266 an error occurs, this action silently fails and the actions evaluation
10267 continues.
10268
Willy Tarreau2d392c22015-08-24 01:43:45 +020010269 - "silent-drop" :
10270 This stops the evaluation of the rules and makes the client-facing
Davor Ocelice9ed2812017-12-25 17:49:28 +010010271 connection suddenly disappear using a system-dependent way that tries
Willy Tarreau2d392c22015-08-24 01:43:45 +020010272 to prevent the client from being notified. The effect it then that the
10273 client still sees an established connection while there's none on
10274 HAProxy. The purpose is to achieve a comparable effect to "tarpit"
10275 except that it doesn't use any local resource at all on the machine
10276 running HAProxy. It can resist much higher loads than "tarpit", and
Davor Ocelice9ed2812017-12-25 17:49:28 +010010277 slow down stronger attackers. It is important to understand the impact
10278 of using this mechanism. All stateful equipment placed between the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010279 client and HAProxy (firewalls, proxies, load balancers) will also keep
10280 the established connection for a long time and may suffer from this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010281 action. On modern Linux systems running with enough privileges, the
Willy Tarreau2d392c22015-08-24 01:43:45 +020010282 TCP_REPAIR socket option is used to block the emission of a TCP
10283 reset. On other systems, the socket's TTL is reduced to 1 so that the
10284 TCP reset doesn't pass the first router, though it's still delivered to
10285 local networks. Do not use it unless you fully understand how it works.
10286
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010287 - send-spoe-group <engine-name> <group-name>
10288 Send a group of SPOE messages.
10289
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010290 Note that the "if/unless" condition is optional. If no condition is set on
10291 the action, it is simply performed unconditionally. That can be useful for
10292 for changing the default action to a reject.
10293
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010294 It is perfectly possible to match layer 7 contents with "tcp-response
10295 content" rules, but then it is important to ensure that a full response has
10296 been buffered, otherwise no contents will match. In order to achieve this,
10297 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010298 period.
10299
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010300 The "set-var" is used to set the content of a variable. The variable is
10301 declared inline.
10302
Daniel Schneller0b547052016-03-21 20:46:57 +010010303 <var-name> The name of the variable starts with an indication about
10304 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010010305 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010010306 "sess" : the variable is shared with the whole session
10307 "txn" : the variable is shared with the transaction
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010308 (request and response)
Daniel Schneller0b547052016-03-21 20:46:57 +010010309 "req" : the variable is shared only during request
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010310 processing
Daniel Schneller0b547052016-03-21 20:46:57 +010010311 "res" : the variable is shared only during response
10312 processing
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010313 This prefix is followed by a name. The separator is a '.'.
Christopher Fauletb71557a2016-10-31 10:49:03 +010010314 The name may only contain characters 'a-z', 'A-Z', '0-9',
10315 '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020010316
10317 <expr> Is a standard HAProxy expression formed by a sample-fetch
10318 followed by some converters.
10319
10320 Example:
10321
10322 tcp-request content set-var(sess.my_var) src
10323
Christopher Faulet85d79c92016-11-09 16:54:56 +010010324 The "unset-var" is used to unset a variable. See above for details about
10325 <var-name>.
10326
10327 Example:
10328
10329 tcp-request content unset-var(sess.my_var)
10330
Christopher Faulet76c09ef2017-09-21 11:03:52 +020010331 The "send-spoe-group" is used to trigger sending of a group of SPOE
10332 messages. To do so, the SPOE engine used to send messages must be defined, as
10333 well as the SPOE group to send. Of course, the SPOE engine must refer to an
10334 existing SPOE filter. If not engine name is provided on the SPOE filter line,
10335 the SPOE agent name must be used.
10336
10337 <engine-name> The SPOE engine name.
10338
10339 <group-name> The SPOE group name as specified in the engine configuration.
10340
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010341 See section 7 about ACL usage.
10342
10343 See also : "tcp-request content", "tcp-response inspect-delay"
10344
10345
Willy Tarreau4f614292016-10-21 17:49:36 +020010346tcp-request session <action> [{if | unless} <condition>]
10347 Perform an action on a validated session depending on a layer 5 condition
10348 May be used in sections : defaults | frontend | listen | backend
10349 no | yes | yes | no
10350 Arguments :
10351 <action> defines the action to perform if the condition applies. See
10352 below.
10353
10354 <condition> is a standard layer5-only ACL-based condition (see section 7).
10355
Davor Ocelice9ed2812017-12-25 17:49:28 +010010356 Once a session is validated, (i.e. after all handshakes have been completed),
Willy Tarreau4f614292016-10-21 17:49:36 +020010357 it is possible to evaluate some conditions to decide whether this session
10358 must be accepted or dropped or have its counters tracked. Those conditions
10359 cannot make use of any data contents because no buffers are allocated yet and
10360 the processing cannot wait at this stage. The main use case it to copy some
10361 early information into variables (since variables are accessible in the
10362 session), or to keep track of some information collected after the handshake,
10363 such as SSL-level elements (SNI, ciphers, client cert's CN) or information
Davor Ocelice9ed2812017-12-25 17:49:28 +010010364 from the PROXY protocol header (e.g. track a source forwarded this way). The
Willy Tarreau4f614292016-10-21 17:49:36 +020010365 extracted information can thus be copied to a variable or tracked using
10366 "track-sc" rules. Of course it is also possible to decide to accept/reject as
10367 with other rulesets. Most operations performed here could also be performed
10368 in "tcp-request content" rules, except that in HTTP these rules are evaluated
10369 for each new request, and that might not always be acceptable. For example a
10370 rule might increment a counter on each evaluation. It would also be possible
10371 that a country is resolved by geolocation from the source IP address,
10372 assigned to a session-wide variable, then the source address rewritten from
10373 an HTTP header for all requests. If some contents need to be inspected in
10374 order to take the decision, the "tcp-request content" statements must be used
10375 instead.
10376
10377 The "tcp-request session" rules are evaluated in their exact declaration
10378 order. If no rule matches or if there is no rule, the default action is to
10379 accept the incoming session. There is no specific limit to the number of
10380 rules which may be inserted.
10381
10382 Several types of actions are supported :
10383 - accept : the request is accepted
10384 - reject : the request is rejected and the connection is closed
10385 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
10386 - sc-inc-gpc0(<sc-id>)
Frédéric Lécaille6778b272018-01-29 15:22:53 +010010387 - sc-inc-gpc1(<sc-id>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010388 - sc-set-gpt0(<sc-id>) <int>
10389 - set-var(<var-name>) <expr>
Christopher Faulet85d79c92016-11-09 16:54:56 +010010390 - unset-var(<var-name>)
Willy Tarreau4f614292016-10-21 17:49:36 +020010391 - silent-drop
10392
10393 These actions have the same meaning as their respective counter-parts in
10394 "tcp-request connection" and "tcp-request content", so please refer to these
10395 sections for a complete description.
10396
10397 Note that the "if/unless" condition is optional. If no condition is set on
10398 the action, it is simply performed unconditionally. That can be useful for
10399 "track-sc*" actions as well as for changing the default action to a reject.
10400
10401 Example: track the original source address by default, or the one advertised
10402 in the PROXY protocol header for connection coming from the local
10403 proxies. The first connection-level rule enables receipt of the
10404 PROXY protocol for these ones, the second rule tracks whatever
10405 address we decide to keep after optional decoding.
10406
10407 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
10408 tcp-request session track-sc0 src
10409
10410 Example: accept all sessions from white-listed hosts, reject too fast
10411 sessions without counting them, and track accepted sessions.
10412 This results in session rate being capped from abusive sources.
10413
10414 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10415 tcp-request session reject if { src_sess_rate gt 10 }
10416 tcp-request session track-sc0 src
10417
10418 Example: accept all sessions from white-listed hosts, count all other
10419 sessions and reject too fast ones. This results in abusive ones
10420 being blocked as long as they don't slow down.
10421
10422 tcp-request session accept if { src -f /etc/haproxy/whitelist.lst }
10423 tcp-request session track-sc0 src
10424 tcp-request session reject if { sc0_sess_rate gt 10 }
10425
10426 See section 7 about ACL usage.
10427
10428 See also : "tcp-request connection", "tcp-request content", "stick-table"
10429
10430
Emeric Brun0a3b67f2010-09-24 15:34:53 +020010431tcp-response inspect-delay <timeout>
10432 Set the maximum allowed time to wait for a response during content inspection
10433 May be used in sections : defaults | frontend | listen | backend
10434 no | no | yes | yes
10435 Arguments :
10436 <timeout> is the timeout value specified in milliseconds by default, but
10437 can be in any other unit if the number is suffixed by the unit,
10438 as explained at the top of this document.
10439
10440 See also : "tcp-response content", "tcp-request inspect-delay".
10441
10442
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010443timeout check <timeout>
10444 Set additional check timeout, but only after a connection has been already
10445 established.
10446
10447 May be used in sections: defaults | frontend | listen | backend
10448 yes | no | yes | yes
10449 Arguments:
10450 <timeout> is the timeout value specified in milliseconds by default, but
10451 can be in any other unit if the number is suffixed by the unit,
10452 as explained at the top of this document.
10453
10454 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
10455 for check and "timeout check" as an additional read timeout. The "min" is
Davor Ocelice9ed2812017-12-25 17:49:28 +010010456 used so that people running with *very* long "timeout connect" (e.g. those
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010457 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +010010458 (Please also note that there is no valid reason to have such long connect
10459 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
10460 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010461
10462 If "timeout check" is not set haproxy uses "inter" for complete check
10463 timeout (connect + read) exactly like all <1.3.15 version.
10464
10465 In most cases check request is much simpler and faster to handle than normal
10466 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +010010467 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010468
10469 This parameter is specific to backends, but can be specified once for all in
10470 "defaults" sections. This is in fact one of the easiest solutions not to
10471 forget about it.
10472
Willy Tarreau41a340d2008-01-22 12:25:31 +010010473 See also: "timeout connect", "timeout queue", "timeout server",
10474 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010475
10476
Willy Tarreau0ba27502007-12-24 16:55:16 +010010477timeout client <timeout>
10478timeout clitimeout <timeout> (deprecated)
10479 Set the maximum inactivity time on the client side.
10480 May be used in sections : defaults | frontend | listen | backend
10481 yes | yes | yes | no
10482 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010483 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010484 can be in any other unit if the number is suffixed by the unit,
10485 as explained at the top of this document.
10486
10487 The inactivity timeout applies when the client is expected to acknowledge or
10488 send data. In HTTP mode, this timeout is particularly important to consider
10489 during the first phase, when the client sends the request, and during the
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010490 response while it is reading data sent by the server. That said, for the
10491 first phase, it is preferable to set the "timeout http-request" to better
10492 protect HAProxy from Slowloris like attacks. The value is specified in
10493 milliseconds by default, but can be in any other unit if the number is
Willy Tarreau0ba27502007-12-24 16:55:16 +010010494 suffixed by the unit, as specified at the top of this document. In TCP mode
10495 (and to a lesser extent, in HTTP mode), it is highly recommended that the
10496 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010497 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +010010498 losses by specifying timeouts that are slightly above multiples of 3 seconds
Davor Ocelice9ed2812017-12-25 17:49:28 +010010499 (e.g. 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
10500 sessions (e.g. WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +020010501 which overrides "timeout client" and "timeout server" for tunnels, as well as
10502 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010503
10504 This parameter is specific to frontends, but can be specified once for all in
10505 "defaults" sections. This is in fact one of the easiest solutions not to
10506 forget about it. An unspecified timeout results in an infinite timeout, which
10507 is not recommended. Such a usage is accepted and works but reports a warning
10508 during startup because it may results in accumulation of expired sessions in
10509 the system if the system's timeouts are not configured either.
10510
Willy Tarreau95c4e142017-11-26 12:18:55 +010010511 This also applies to HTTP/2 connections, which will be closed with GOAWAY.
Lukas Tribus75df9d72017-11-24 19:05:12 +010010512
Willy Tarreau0ba27502007-12-24 16:55:16 +010010513 This parameter replaces the old, deprecated "clitimeout". It is recommended
10514 to use it to write new configurations. The form "timeout clitimeout" is
10515 provided only by backwards compatibility but its use is strongly discouraged.
10516
Baptiste Assmann2e1941e2016-03-06 23:24:12 +010010517 See also : "clitimeout", "timeout server", "timeout tunnel",
10518 "timeout http-request".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010519
10520
Willy Tarreau05cdd962014-05-10 14:30:07 +020010521timeout client-fin <timeout>
10522 Set the inactivity timeout on the client side for half-closed connections.
10523 May be used in sections : defaults | frontend | listen | backend
10524 yes | yes | yes | no
10525 Arguments :
10526 <timeout> is the timeout value specified in milliseconds by default, but
10527 can be in any other unit if the number is suffixed by the unit,
10528 as explained at the top of this document.
10529
10530 The inactivity timeout applies when the client is expected to acknowledge or
10531 send data while one direction is already shut down. This timeout is different
10532 from "timeout client" in that it only applies to connections which are closed
10533 in one direction. This is particularly useful to avoid keeping connections in
10534 FIN_WAIT state for too long when clients do not disconnect cleanly. This
10535 problem is particularly common long connections such as RDP or WebSocket.
10536 Note that this timeout can override "timeout tunnel" when a connection shuts
Willy Tarreau599391a2017-11-24 10:16:00 +010010537 down in one direction. It is applied to idle HTTP/2 connections once a GOAWAY
10538 frame was sent, often indicating an expectation that the connection quickly
10539 ends.
Willy Tarreau05cdd962014-05-10 14:30:07 +020010540
10541 This parameter is specific to frontends, but can be specified once for all in
10542 "defaults" sections. By default it is not set, so half-closed connections
10543 will use the other timeouts (timeout.client or timeout.tunnel).
10544
10545 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
10546
10547
Willy Tarreau0ba27502007-12-24 16:55:16 +010010548timeout connect <timeout>
10549timeout contimeout <timeout> (deprecated)
10550 Set the maximum time to wait for a connection attempt to a server to succeed.
10551 May be used in sections : defaults | frontend | listen | backend
10552 yes | no | yes | yes
10553 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010554 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +010010555 can be in any other unit if the number is suffixed by the unit,
10556 as explained at the top of this document.
10557
10558 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010559 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +010010560 cover one or several TCP packet losses by specifying timeouts that are
Davor Ocelice9ed2812017-12-25 17:49:28 +010010561 slightly above multiples of 3 seconds (e.g. 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +010010562 connect timeout also presets both queue and tarpit timeouts to the same value
10563 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010564
10565 This parameter is specific to backends, but can be specified once for all in
10566 "defaults" sections. This is in fact one of the easiest solutions not to
10567 forget about it. An unspecified timeout results in an infinite timeout, which
10568 is not recommended. Such a usage is accepted and works but reports a warning
10569 during startup because it may results in accumulation of failed sessions in
10570 the system if the system's timeouts are not configured either.
10571
10572 This parameter replaces the old, deprecated "contimeout". It is recommended
10573 to use it to write new configurations. The form "timeout contimeout" is
10574 provided only by backwards compatibility but its use is strongly discouraged.
10575
Willy Tarreau41a340d2008-01-22 12:25:31 +010010576 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
10577 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +010010578
10579
Willy Tarreaub16a5742010-01-10 14:46:16 +010010580timeout http-keep-alive <timeout>
10581 Set the maximum allowed time to wait for a new HTTP request to appear
10582 May be used in sections : defaults | frontend | listen | backend
10583 yes | yes | yes | yes
10584 Arguments :
10585 <timeout> is the timeout value specified in milliseconds by default, but
10586 can be in any other unit if the number is suffixed by the unit,
10587 as explained at the top of this document.
10588
10589 By default, the time to wait for a new request in case of keep-alive is set
10590 by "timeout http-request". However this is not always convenient because some
10591 people want very short keep-alive timeouts in order to release connections
10592 faster, and others prefer to have larger ones but still have short timeouts
10593 once the request has started to present itself.
10594
10595 The "http-keep-alive" timeout covers these needs. It will define how long to
10596 wait for a new HTTP request to start coming after a response was sent. Once
10597 the first byte of request has been seen, the "http-request" timeout is used
10598 to wait for the complete request to come. Note that empty lines prior to a
10599 new request do not refresh the timeout and are not counted as a new request.
10600
10601 There is also another difference between the two timeouts : when a connection
10602 expires during timeout http-keep-alive, no error is returned, the connection
10603 just closes. If the connection expires in "http-request" while waiting for a
10604 connection to complete, a HTTP 408 error is returned.
10605
10606 In general it is optimal to set this value to a few tens to hundreds of
10607 milliseconds, to allow users to fetch all objects of a page at once but
Davor Ocelice9ed2812017-12-25 17:49:28 +010010608 without waiting for further clicks. Also, if set to a very small value (e.g.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010609 1 millisecond) it will probably only accept pipelined requests but not the
10610 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +020010611 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +010010612
10613 If this parameter is not set, the "http-request" timeout applies, and if both
10614 are not set, "timeout client" still applies at the lower level. It should be
10615 set in the frontend to take effect, unless the frontend is in TCP mode, in
10616 which case the HTTP backend's timeout will be used.
10617
Willy Tarreau95c4e142017-11-26 12:18:55 +010010618 When using HTTP/2 "timeout client" is applied instead. This is so we can keep
10619 using short keep-alive timeouts in HTTP/1.1 while using longer ones in HTTP/2
Lukas Tribus75df9d72017-11-24 19:05:12 +010010620 (where we only have one connection per client and a connection setup).
10621
Willy Tarreaub16a5742010-01-10 14:46:16 +010010622 See also : "timeout http-request", "timeout client".
10623
10624
Willy Tarreau036fae02008-01-06 13:24:40 +010010625timeout http-request <timeout>
10626 Set the maximum allowed time to wait for a complete HTTP request
10627 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010628 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +010010629 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +010010630 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +010010631 can be in any other unit if the number is suffixed by the unit,
10632 as explained at the top of this document.
10633
10634 In order to offer DoS protection, it may be required to lower the maximum
10635 accepted time to receive a complete HTTP request without affecting the client
10636 timeout. This helps protecting against established connections on which
10637 nothing is sent. The client timeout cannot offer a good protection against
10638 this abuse because it is an inactivity timeout, which means that if the
10639 attacker sends one character every now and then, the timeout will not
10640 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +020010641 types, the request will be aborted if it does not complete in time. When the
10642 timeout expires, an HTTP 408 response is sent to the client to inform it
10643 about the problem, and the connection is closed. The logs will report
10644 termination codes "cR". Some recent browsers are having problems with this
Davor Ocelice9ed2812017-12-25 17:49:28 +010010645 standard, well-documented behavior, so it might be needed to hide the 408
Willy Tarreau0f228a02015-05-01 15:37:53 +020010646 code using "option http-ignore-probes" or "errorfile 408 /dev/null". See
10647 more details in the explanations of the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +010010648
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010649 By default, this timeout only applies to the header part of the request,
10650 and not to any data. As soon as the empty line is received, this timeout is
10651 not used anymore. When combined with "option http-buffer-request", this
10652 timeout also applies to the body of the request..
10653 It is used again on keep-alive connections to wait for a second
Willy Tarreaub16a5742010-01-10 14:46:16 +010010654 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +010010655
10656 Generally it is enough to set it to a few seconds, as most clients send the
10657 full request immediately upon connection. Add 3 or more seconds to cover TCP
Davor Ocelice9ed2812017-12-25 17:49:28 +010010658 retransmits but that's all. Setting it to very low values (e.g. 50 ms) will
Willy Tarreau036fae02008-01-06 13:24:40 +010010659 generally work on local networks as long as there are no packet losses. This
10660 will prevent people from sending bare HTTP requests using telnet.
10661
10662 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +020010663 chunk of the incoming request. It should be set in the frontend to take
10664 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
10665 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +010010666
Willy Tarreau0f228a02015-05-01 15:37:53 +020010667 See also : "errorfile", "http-ignore-probes", "timeout http-keep-alive", and
Baptiste Assmanneccdf432015-10-28 13:49:01 +010010668 "timeout client", "option http-buffer-request".
Willy Tarreau036fae02008-01-06 13:24:40 +010010669
Willy Tarreau844e3c52008-01-11 16:28:18 +010010670
10671timeout queue <timeout>
10672 Set the maximum time to wait in the queue for a connection slot to be free
10673 May be used in sections : defaults | frontend | listen | backend
10674 yes | no | yes | yes
10675 Arguments :
10676 <timeout> is the timeout value specified in milliseconds by default, but
10677 can be in any other unit if the number is suffixed by the unit,
10678 as explained at the top of this document.
10679
10680 When a server's maxconn is reached, connections are left pending in a queue
10681 which may be server-specific or global to the backend. In order not to wait
10682 indefinitely, a timeout is applied to requests pending in the queue. If the
10683 timeout is reached, it is considered that the request will almost never be
10684 served, so it is dropped and a 503 error is returned to the client.
10685
10686 The "timeout queue" statement allows to fix the maximum time for a request to
10687 be left pending in a queue. If unspecified, the same value as the backend's
10688 connection timeout ("timeout connect") is used, for backwards compatibility
10689 with older versions with no "timeout queue" parameter.
10690
10691 See also : "timeout connect", "contimeout".
10692
10693
10694timeout server <timeout>
10695timeout srvtimeout <timeout> (deprecated)
10696 Set the maximum inactivity time on the server side.
10697 May be used in sections : defaults | frontend | listen | backend
10698 yes | no | yes | yes
10699 Arguments :
10700 <timeout> is the timeout value specified in milliseconds by default, but
10701 can be in any other unit if the number is suffixed by the unit,
10702 as explained at the top of this document.
10703
10704 The inactivity timeout applies when the server is expected to acknowledge or
10705 send data. In HTTP mode, this timeout is particularly important to consider
10706 during the first phase of the server's response, when it has to send the
10707 headers, as it directly represents the server's processing time for the
10708 request. To find out what value to put there, it's often good to start with
10709 what would be considered as unacceptable response times, then check the logs
10710 to observe the response time distribution, and adjust the value accordingly.
10711
10712 The value is specified in milliseconds by default, but can be in any other
10713 unit if the number is suffixed by the unit, as specified at the top of this
10714 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
10715 recommended that the client timeout remains equal to the server timeout in
10716 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010717 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +010010718 packet losses by specifying timeouts that are slightly above multiples of 3
Davor Ocelice9ed2812017-12-25 17:49:28 +010010719 seconds (e.g. 4 or 5 seconds minimum). If some long-lived sessions are mixed
10720 with short-lived sessions (e.g. WebSocket and HTTP), it's worth considering
Willy Tarreauce887fd2012-05-12 12:50:00 +020010721 "timeout tunnel", which overrides "timeout client" and "timeout server" for
10722 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010723
10724 This parameter is specific to backends, but can be specified once for all in
10725 "defaults" sections. This is in fact one of the easiest solutions not to
10726 forget about it. An unspecified timeout results in an infinite timeout, which
10727 is not recommended. Such a usage is accepted and works but reports a warning
10728 during startup because it may results in accumulation of expired sessions in
10729 the system if the system's timeouts are not configured either.
10730
10731 This parameter replaces the old, deprecated "srvtimeout". It is recommended
10732 to use it to write new configurations. The form "timeout srvtimeout" is
10733 provided only by backwards compatibility but its use is strongly discouraged.
10734
Willy Tarreauce887fd2012-05-12 12:50:00 +020010735 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +010010736
Willy Tarreau05cdd962014-05-10 14:30:07 +020010737
10738timeout server-fin <timeout>
10739 Set the inactivity timeout on the server side for half-closed connections.
10740 May be used in sections : defaults | frontend | listen | backend
10741 yes | no | yes | yes
10742 Arguments :
10743 <timeout> is the timeout value specified in milliseconds by default, but
10744 can be in any other unit if the number is suffixed by the unit,
10745 as explained at the top of this document.
10746
10747 The inactivity timeout applies when the server is expected to acknowledge or
10748 send data while one direction is already shut down. This timeout is different
10749 from "timeout server" in that it only applies to connections which are closed
10750 in one direction. This is particularly useful to avoid keeping connections in
10751 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
10752 This problem is particularly common long connections such as RDP or WebSocket.
10753 Note that this timeout can override "timeout tunnel" when a connection shuts
10754 down in one direction. This setting was provided for completeness, but in most
10755 situations, it should not be needed.
10756
10757 This parameter is specific to backends, but can be specified once for all in
10758 "defaults" sections. By default it is not set, so half-closed connections
10759 will use the other timeouts (timeout.server or timeout.tunnel).
10760
10761 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
10762
Willy Tarreau844e3c52008-01-11 16:28:18 +010010763
10764timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +010010765 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +010010766 May be used in sections : defaults | frontend | listen | backend
10767 yes | yes | yes | yes
10768 Arguments :
10769 <timeout> is the tarpit duration specified in milliseconds by default, but
10770 can be in any other unit if the number is suffixed by the unit,
10771 as explained at the top of this document.
10772
Jarno Huuskonene5ae7022017-04-03 14:36:21 +030010773 When a connection is tarpitted using "http-request tarpit" or
10774 "reqtarpit", it is maintained open with no activity for a certain
10775 amount of time, then closed. "timeout tarpit" defines how long it will
10776 be maintained open.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010777
10778 The value is specified in milliseconds by default, but can be in any other
10779 unit if the number is suffixed by the unit, as specified at the top of this
10780 document. If unspecified, the same value as the backend's connection timeout
10781 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +010010782 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010783
10784 See also : "timeout connect", "contimeout".
10785
10786
Willy Tarreauce887fd2012-05-12 12:50:00 +020010787timeout tunnel <timeout>
10788 Set the maximum inactivity time on the client and server side for tunnels.
10789 May be used in sections : defaults | frontend | listen | backend
10790 yes | no | yes | yes
10791 Arguments :
10792 <timeout> is the timeout value specified in milliseconds by default, but
10793 can be in any other unit if the number is suffixed by the unit,
10794 as explained at the top of this document.
10795
Jamie Gloudonaaa21002012-08-25 00:18:33 -040010796 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +020010797 between a client and a server, and the connection remains inactive in both
10798 directions. This timeout supersedes both the client and server timeouts once
10799 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
Davor Ocelice9ed2812017-12-25 17:49:28 +010010800 analyzer remains attached to either connection (e.g. tcp content rules are
10801 accepted). In HTTP, this timeout is used when a connection is upgraded (e.g.
Willy Tarreauce887fd2012-05-12 12:50:00 +020010802 when switching to the WebSocket protocol, or forwarding a CONNECT request
10803 to a proxy), or after the first response when no keepalive/close option is
10804 specified.
10805
Willy Tarreau05cdd962014-05-10 14:30:07 +020010806 Since this timeout is usually used in conjunction with long-lived connections,
10807 it usually is a good idea to also set "timeout client-fin" to handle the
10808 situation where a client suddenly disappears from the net and does not
10809 acknowledge a close, or sends a shutdown and does not acknowledge pending
10810 data anymore. This can happen in lossy networks where firewalls are present,
10811 and is detected by the presence of large amounts of sessions in a FIN_WAIT
10812 state.
10813
Willy Tarreauce887fd2012-05-12 12:50:00 +020010814 The value is specified in milliseconds by default, but can be in any other
10815 unit if the number is suffixed by the unit, as specified at the top of this
10816 document. Whatever the expected normal idle time, it is a good practice to
10817 cover at least one or several TCP packet losses by specifying timeouts that
Davor Ocelice9ed2812017-12-25 17:49:28 +010010818 are slightly above multiples of 3 seconds (e.g. 4 or 5 seconds minimum).
Willy Tarreauce887fd2012-05-12 12:50:00 +020010819
10820 This parameter is specific to backends, but can be specified once for all in
10821 "defaults" sections. This is in fact one of the easiest solutions not to
10822 forget about it.
10823
10824 Example :
10825 defaults http
10826 option http-server-close
10827 timeout connect 5s
10828 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +020010829 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +020010830 timeout server 30s
10831 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
10832
Willy Tarreau05cdd962014-05-10 14:30:07 +020010833 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +020010834
10835
Willy Tarreau844e3c52008-01-11 16:28:18 +010010836transparent (deprecated)
10837 Enable client-side transparent proxying
10838 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +010010839 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +010010840 Arguments : none
10841
10842 This keyword was introduced in order to provide layer 7 persistence to layer
10843 3 load balancers. The idea is to use the OS's ability to redirect an incoming
10844 connection for a remote address to a local process (here HAProxy), and let
10845 this process know what address was initially requested. When this option is
10846 used, sessions without cookies will be forwarded to the original destination
10847 IP address of the incoming request (which should match that of another
10848 equipment), while requests with cookies will still be forwarded to the
10849 appropriate server.
10850
10851 The "transparent" keyword is deprecated, use "option transparent" instead.
10852
10853 Note that contrary to a common belief, this option does NOT make HAProxy
10854 present the client's IP to the server when establishing the connection.
10855
Willy Tarreau844e3c52008-01-11 16:28:18 +010010856 See also: "option transparent"
10857
William Lallemanda73203e2012-03-12 12:48:57 +010010858unique-id-format <string>
10859 Generate a unique ID for each request.
10860 May be used in sections : defaults | frontend | listen | backend
10861 yes | yes | yes | no
10862 Arguments :
10863 <string> is a log-format string.
10864
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010865 This keyword creates a ID for each request using the custom log format. A
10866 unique ID is useful to trace a request passing through many components of
10867 a complex infrastructure. The newly created ID may also be logged using the
10868 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +010010869
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010870 The format should be composed from elements that are guaranteed to be
10871 unique when combined together. For instance, if multiple haproxy instances
10872 are involved, it might be important to include the node name. It is often
10873 needed to log the incoming connection's source and destination addresses
10874 and ports. Note that since multiple requests may be performed over the same
10875 connection, including a request counter may help differentiate them.
10876 Similarly, a timestamp may protect against a rollover of the counter.
10877 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +010010878
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010879 It is recommended to use hexadecimal notation for many fields since it
10880 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +010010881
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010882 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010883
Julien Vehentf21be322014-03-07 08:27:34 -050010884 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010885
10886 will generate:
10887
10888 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10889
10890 See also: "unique-id-header"
10891
10892unique-id-header <name>
10893 Add a unique ID header in the HTTP request.
10894 May be used in sections : defaults | frontend | listen | backend
10895 yes | yes | yes | no
10896 Arguments :
10897 <name> is the name of the header.
10898
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010899 Add a unique-id header in the HTTP request sent to the server, using the
10900 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +010010901
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010902 Example:
William Lallemanda73203e2012-03-12 12:48:57 +010010903
Julien Vehentf21be322014-03-07 08:27:34 -050010904 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +010010905 unique-id-header X-Unique-ID
10906
10907 will generate:
10908
10909 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
10910
10911 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +010010912
Willy Tarreauf51658d2014-04-23 01:21:56 +020010913use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010914 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010915 May be used in sections : defaults | frontend | listen | backend
10916 no | yes | yes | no
10917 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010918 <backend> is the name of a valid backend or "listen" section, or a
10919 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010920
Willy Tarreauf51658d2014-04-23 01:21:56 +020010921 <condition> is a condition composed of ACLs, as described in section 7. If
10922 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010923
10924 When doing content-switching, connections arrive on a frontend and are then
10925 dispatched to various backends depending on a number of conditions. The
10926 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010927 "use_backend" keyword. While it is normally used with HTTP processing, it can
Davor Ocelice9ed2812017-12-25 17:49:28 +010010928 also be used in pure TCP, either without content using stateless ACLs (e.g.
Willy Tarreau1d0dfb12009-07-07 15:10:31 +020010929 source address validation) or combined with a "tcp-request" rule to wait for
10930 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +010010931
10932 There may be as many "use_backend" rules as desired. All of these rules are
10933 evaluated in their declaration order, and the first one which matches will
10934 assign the backend.
10935
10936 In the first form, the backend will be used if the condition is met. In the
10937 second form, the backend will be used if the condition is not met. If no
10938 condition is valid, the backend defined with "default_backend" will be used.
10939 If no default backend is defined, either the servers in the same section are
10940 used (in case of a "listen" section) or, in case of a frontend, no server is
10941 used and a 503 service unavailable response is returned.
10942
Willy Tarreau51aecc72009-07-12 09:47:04 +020010943 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010010944 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +020010945 and backend processing will immediately follow, or the backend will wait for
10946 a complete HTTP request to get in. This feature is useful when a frontend
10947 must decode several protocols on a unique port, one of them being HTTP.
10948
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010949 When <backend> is a simple name, it is resolved at configuration time, and an
10950 error is reported if the specified backend does not exist. If <backend> is
10951 a log-format string instead, no check may be done at configuration time, so
10952 the backend name is resolved dynamically at run time. If the resulting
10953 backend name does not correspond to any valid backend, no other rule is
10954 evaluated, and the default_backend directive is applied instead. Note that
10955 when using dynamic backend names, it is highly recommended to use a prefix
10956 that no other backend uses in order to ensure that an unauthorized backend
10957 cannot be forced from the request.
10958
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010959 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +010010960 used to detect the association between frontends and backends to compute the
10961 backend's "fullconn" setting. This cannot be done for dynamic names.
10962
10963 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
10964 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +010010965
Willy Tarreau036fae02008-01-06 13:24:40 +010010966
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010967use-server <server> if <condition>
10968use-server <server> unless <condition>
10969 Only use a specific server if/unless an ACL-based condition is matched.
10970 May be used in sections : defaults | frontend | listen | backend
10971 no | no | yes | yes
10972 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +020010973 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020010974
10975 <condition> is a condition composed of ACLs, as described in section 7.
10976
10977 By default, connections which arrive to a backend are load-balanced across
10978 the available servers according to the configured algorithm, unless a
10979 persistence mechanism such as a cookie is used and found in the request.
10980
10981 Sometimes it is desirable to forward a particular request to a specific
10982 server without having to declare a dedicated backend for this server. This
10983 can be achieved using the "use-server" rules. These rules are evaluated after
10984 the "redirect" rules and before evaluating cookies, and they have precedence
10985 on them. There may be as many "use-server" rules as desired. All of these
10986 rules are evaluated in their declaration order, and the first one which
10987 matches will assign the server.
10988
10989 If a rule designates a server which is down, and "option persist" is not used
10990 and no force-persist rule was validated, it is ignored and evaluation goes on
10991 with the next rules until one matches.
10992
10993 In the first form, the server will be used if the condition is met. In the
10994 second form, the server will be used if the condition is not met. If no
10995 condition is valid, the processing continues and the server will be assigned
10996 according to other persistence mechanisms.
10997
10998 Note that even if a rule is matched, cookie processing is still performed but
10999 does not assign the server. This allows prefixed cookies to have their prefix
11000 stripped.
11001
11002 The "use-server" statement works both in HTTP and TCP mode. This makes it
11003 suitable for use with content-based inspection. For instance, a server could
11004 be selected in a farm according to the TLS SNI field. And if these servers
11005 have their weight set to zero, they will not be used for other traffic.
11006
11007 Example :
11008 # intercept incoming TLS requests based on the SNI field
11009 use-server www if { req_ssl_sni -i www.example.com }
11010 server www 192.168.0.1:443 weight 0
11011 use-server mail if { req_ssl_sni -i mail.example.com }
11012 server mail 192.168.0.1:587 weight 0
11013 use-server imap if { req_ssl_sni -i imap.example.com }
Lukas Tribus98a3e3f2017-03-26 12:55:35 +000011014 server imap 192.168.0.1:993 weight 0
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011015 # all the rest is forwarded to this server
11016 server default 192.168.0.2:443 check
11017
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011018 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011019
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011020
Davor Ocelice9ed2812017-12-25 17:49:28 +0100110215. Bind and server options
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011022--------------------------
11023
11024The "bind", "server" and "default-server" keywords support a number of settings
11025depending on some build options and on the system HAProxy was built on. These
11026settings generally each consist in one word sometimes followed by a value,
11027written on the same line as the "bind" or "server" line. All these options are
11028described in this section.
11029
11030
110315.1. Bind options
11032-----------------
11033
11034The "bind" keyword supports a certain number of settings which are all passed
11035as arguments on the same line. The order in which those arguments appear makes
11036no importance, provided that they appear after the bind address. All of these
11037parameters are optional. Some of them consist in a single words (booleans),
11038while other ones expect a value after them. In this case, the value must be
11039provided immediately after the setting name.
11040
11041The currently supported settings are the following ones.
11042
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011043accept-netscaler-cip <magic number>
11044 Enforces the use of the NetScaler Client IP insertion protocol over any
11045 connection accepted by any of the TCP sockets declared on the same line. The
11046 NetScaler Client IP insertion protocol dictates the layer 3/4 addresses of
11047 the incoming connection to be used everywhere an address is used, with the
11048 only exception of "tcp-request connection" rules which will only see the
11049 real connection address. Logs will reflect the addresses indicated in the
11050 protocol, unless it is violated, in which case the real address will still
11051 be used. This keyword combined with support from external components can be
11052 used as an efficient and reliable alternative to the X-Forwarded-For
Bertrand Jacquin90759682016-06-06 15:35:39 +010011053 mechanism which is not always reliable and not even always usable. See also
11054 "tcp-request connection expect-netscaler-cip" for a finer-grained setting of
11055 which client is allowed to use the protocol.
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010011056
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011057accept-proxy
11058 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +020011059 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
11060 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011061 3/4 addresses of the incoming connection to be used everywhere an address is
11062 used, with the only exception of "tcp-request connection" rules which will
11063 only see the real connection address. Logs will reflect the addresses
11064 indicated in the protocol, unless it is violated, in which case the real
Davor Ocelice9ed2812017-12-25 17:49:28 +010011065 address will still be used. This keyword combined with support from external
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011066 components can be used as an efficient and reliable alternative to the
11067 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +020011068 usable. See also "tcp-request connection expect-proxy" for a finer-grained
11069 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011070
Olivier Houchardc2aae742017-09-22 18:26:28 +020011071allow-0rtt
Bertrand Jacquina25282b2018-08-14 00:56:13 +010011072 Allow receiving early data when using TLSv1.3. This is disabled by default,
Olivier Houchard69752962019-01-08 15:35:32 +010011073 due to security considerations. Because it is vulnerable to replay attacks,
11074 you should only allow if for requests that are safe to replay, ie requests
11075 that are idempotent. You can use the "wait-for-handshake" action for any
11076 request that wouldn't be safe with early data.
Olivier Houchardc2aae742017-09-22 18:26:28 +020011077
Willy Tarreauab861d32013-04-02 02:30:41 +020011078alpn <protocols>
11079 This enables the TLS ALPN extension and advertises the specified protocol
11080 list as supported on top of ALPN. The protocol list consists in a comma-
11081 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
11082 quotes). This requires that the SSL library is build with support for TLS
11083 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
Willy Tarreau95c4e142017-11-26 12:18:55 +010011084 initial NPN extension. ALPN is required to enable HTTP/2 on an HTTP frontend.
11085 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11086 now obsolete NPN extension. At the time of writing this, most browsers still
11087 support both ALPN and NPN for HTTP/2 so a fallback to NPN may still work for
11088 a while. But ALPN must be used whenever possible. If both HTTP/2 and HTTP/1.1
11089 are expected to be supported, both versions can be advertised, in order of
11090 preference, like below :
11091
11092 bind :443 ssl crt pub.pem alpn h2,http/1.1
Willy Tarreauab861d32013-04-02 02:30:41 +020011093
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011094backlog <backlog>
Willy Tarreaue2711c72019-02-27 15:39:41 +010011095 Sets the socket's backlog to this value. If unspecified or 0, the frontend's
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011096 backlog is used instead, which generally defaults to the maxconn value.
11097
Emmanuel Hocdete7f2b732017-01-09 16:15:54 +010011098curves <curves>
11099 This setting is only available when support for OpenSSL was built in. It sets
11100 the string describing the list of elliptic curves algorithms ("curve suite")
11101 that are negotiated during the SSL/TLS handshake with ECDHE. The format of the
11102 string is a colon-delimited list of curve name.
11103 Example: "X25519:P-256" (without quote)
11104 When "curves" is set, "ecdhe" parameter is ignored.
11105
Emeric Brun7fb34422012-09-28 15:26:15 +020011106ecdhe <named curve>
11107 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +010011108 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
11109 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +020011110
Emeric Brunfd33a262012-10-11 16:28:27 +020011111ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011112 This setting is only available when support for OpenSSL was built in. It
11113 designates a PEM file from which to load CA certificates used to verify
11114 client's certificate.
11115
Emeric Brunb6dc9342012-09-28 17:55:37 +020011116ca-ignore-err [all|<errorID>,...]
11117 This setting is only available when support for OpenSSL was built in.
11118 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
11119 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
11120 error is ignored.
11121
Christopher Faulet31af49d2015-06-09 17:29:50 +020011122ca-sign-file <cafile>
11123 This setting is only available when support for OpenSSL was built in. It
11124 designates a PEM file containing both the CA certificate and the CA private
11125 key used to create and sign server's certificates. This is a mandatory
11126 setting when the dynamic generation of certificates is enabled. See
11127 'generate-certificates' for details.
11128
Bertrand Jacquind4d0a232016-11-13 16:37:12 +000011129ca-sign-pass <passphrase>
Christopher Faulet31af49d2015-06-09 17:29:50 +020011130 This setting is only available when support for OpenSSL was built in. It is
11131 the CA private key passphrase. This setting is optional and used only when
11132 the dynamic generation of certificates is enabled. See
11133 'generate-certificates' for details.
11134
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011135ciphers <ciphers>
11136 This setting is only available when support for OpenSSL was built in. It sets
11137 the string describing the list of cipher algorithms ("cipher suite") that are
Bertrand Jacquin8cf7c1e2019-02-03 18:35:25 +000011138 negotiated during the SSL/TLS handshake up to TLSv1.2. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011139 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011140 information and recommendations see e.g.
11141 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11142 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11143 cipher configuration, please check the "ciphersuites" keyword.
11144
11145ciphersuites <ciphersuites>
11146 This setting is only available when support for OpenSSL was built in and
11147 OpenSSL 1.1.1 or later was used to build HAProxy. It sets the string describing
11148 the list of cipher algorithms ("cipher suite") that are negotiated during the
11149 TLSv1.3 handshake. The format of the string is defined in "man 1 ciphers" from
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011150 OpenSSL man pages under the "ciphersuites" section. For cipher configuration
11151 for TLSv1.2 and earlier, please check the "ciphers" keyword.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011152
Emeric Brunfd33a262012-10-11 16:28:27 +020011153crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +020011154 This setting is only available when support for OpenSSL was built in. It
11155 designates a PEM file from which to load certificate revocation list used
11156 to verify client's certificate.
11157
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011158crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +000011159 This setting is only available when support for OpenSSL was built in. It
11160 designates a PEM file containing both the required certificates and any
11161 associated private keys. This file can be built by concatenating multiple
11162 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
11163 requires an intermediate certificate, this can also be concatenated into this
11164 file.
11165
11166 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
11167 are loaded.
11168
11169 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011170 that directory will be loaded in alphabetic order unless their name ends with
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011171 '.issuer', '.ocsp' or '.sctl' (reserved extensions). This directive may be
11172 specified multiple times in order to load certificates from multiple files or
11173 directories. The certificates will be presented to clients who provide a
11174 valid TLS Server Name Indication field matching one of their CN or alt
Davor Ocelice9ed2812017-12-25 17:49:28 +010011175 subjects. Wildcards are supported, where a wildcard character '*' is used
11176 instead of the first hostname component (e.g. *.example.org matches
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011177 www.example.org but not www.sub.example.org).
Alex Davies0fbf0162013-03-02 16:04:50 +000011178
11179 If no SNI is provided by the client or if the SSL library does not support
11180 TLS extensions, or if the client provides an SNI hostname which does not
11181 match any certificate, then the first loaded certificate will be presented.
11182 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +010011183 recommended to load the default one first as a file or to ensure that it will
11184 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +000011185
Emeric Brune032bfa2012-09-28 13:01:45 +020011186 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011187
Davor Ocelice9ed2812017-12-25 17:49:28 +010011188 Some CAs (such as GoDaddy) offer a drop down list of server types that do not
Alex Davies0fbf0162013-03-02 16:04:50 +000011189 include HAProxy when obtaining a certificate. If this happens be sure to
Davor Ocelice9ed2812017-12-25 17:49:28 +010011190 choose a web server that the CA believes requires an intermediate CA (for
11191 GoDaddy, selection Apache Tomcat will get the correct bundle, but many
Alex Davies0fbf0162013-03-02 16:04:50 +000011192 others, e.g. nginx, result in a wrong bundle that will not work for some
11193 clients).
11194
Emeric Brun4147b2e2014-06-16 18:36:30 +020011195 For each PEM file, haproxy checks for the presence of file at the same path
11196 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
11197 Status Request extension (also known as "OCSP stapling") is automatically
11198 enabled. The content of this file is optional. If not empty, it must contain
11199 a valid OCSP Response in DER format. In order to be valid an OCSP Response
11200 must comply with the following rules: it has to indicate a good status,
11201 it has to be a single response for the certificate of the PEM file, and it
11202 has to be valid at the moment of addition. If these rules are not respected
11203 the OCSP Response is ignored and a warning is emitted. In order to identify
11204 which certificate an OCSP Response applies to, the issuer's certificate is
11205 necessary. If the issuer's certificate is not found in the PEM file, it will
11206 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
11207 if it exists otherwise it will fail with an error.
11208
Janusz Dziemidowicz2c701b52015-03-07 23:03:59 +010011209 For each PEM file, haproxy also checks for the presence of file at the same
11210 path suffixed by ".sctl". If such file is found, support for Certificate
11211 Transparency (RFC6962) TLS extension is enabled. The file must contain a
11212 valid Signed Certificate Timestamp List, as described in RFC. File is parsed
11213 to check basic syntax, but no signatures are verified.
11214
yanbzhu6c25e9e2016-01-05 12:52:02 -050011215 There are cases where it is desirable to support multiple key types, e.g. RSA
11216 and ECDSA in the cipher suites offered to the clients. This allows clients
11217 that support EC certificates to be able to use EC ciphers, while
11218 simultaneously supporting older, RSA only clients.
yanbzhud19630c2015-12-14 15:10:25 -050011219
11220 In order to provide this functionality, multiple PEM files, each with a
11221 different key type, are required. To associate these PEM files into a
11222 "cert bundle" that is recognized by haproxy, they must be named in the
11223 following way: All PEM files that are to be bundled must have the same base
11224 name, with a suffix indicating the key type. Currently, three suffixes are
11225 supported: rsa, dsa and ecdsa. For example, if www.example.com has two PEM
11226 files, an RSA file and an ECDSA file, they must be named: "example.pem.rsa"
11227 and "example.pem.ecdsa". The first part of the filename is arbitrary; only the
11228 suffix matters. To load this bundle into haproxy, specify the base name only:
11229
11230 Example : bind :8443 ssl crt example.pem
11231
yanbzhu6c25e9e2016-01-05 12:52:02 -050011232 Note that the suffix is not given to haproxy; this tells haproxy to look for
yanbzhud19630c2015-12-14 15:10:25 -050011233 a cert bundle.
11234
Davor Ocelice9ed2812017-12-25 17:49:28 +010011235 HAProxy will load all PEM files in the bundle at the same time to try to
yanbzhud19630c2015-12-14 15:10:25 -050011236 support multiple key types. PEM files are combined based on Common Name
11237 (CN) and Subject Alternative Name (SAN) to support SNI lookups. This means
11238 that even if you give haproxy a cert bundle, if there are no shared CN/SAN
11239 entries in the certificates in that bundle, haproxy will not be able to
11240 provide multi-cert support.
11241
11242 Assuming bundle in the example above contained the following:
11243
11244 Filename | CN | SAN
11245 -------------------+-----------------+-------------------
11246 example.pem.rsa | www.example.com | rsa.example.com
yanbzhu6c25e9e2016-01-05 12:52:02 -050011247 -------------------+-----------------+-------------------
yanbzhud19630c2015-12-14 15:10:25 -050011248 example.pem.ecdsa | www.example.com | ecdsa.example.com
11249 -------------------+-----------------+-------------------
11250
11251 Users connecting with an SNI of "www.example.com" will be able
11252 to use both RSA and ECDSA cipher suites. Users connecting with an SNI of
11253 "rsa.example.com" will only be able to use RSA cipher suites, and users
11254 connecting with "ecdsa.example.com" will only be able to use ECDSA cipher
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011255 suites. With BoringSSL and Openssl >= 1.1.1 multi-cert is natively supported,
11256 no need to bundle certificates. ECDSA certificate will be preferred if client
11257 support it.
yanbzhud19630c2015-12-14 15:10:25 -050011258
11259 If a directory name is given as the <cert> argument, haproxy will
11260 automatically search and load bundled files in that directory.
11261
11262 OSCP files (.ocsp) and issuer files (.issuer) are supported with multi-cert
11263 bundling. Each certificate can have its own .ocsp and .issuer file. At this
11264 time, sctl is not supported in multi-certificate bundling.
11265
Emeric Brunb6dc9342012-09-28 17:55:37 +020011266crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +000011267 This setting is only available when support for OpenSSL was built in. Sets a
Davor Ocelice9ed2812017-12-25 17:49:28 +010011268 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011269 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +000011270 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +020011271
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011272crt-list <file>
11273 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011274 designates a list of PEM file with an optional ssl configuration and a SNI
11275 filter per certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011276
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011277 <crtfile> [\[<sslbindconf> ...\]] [[!]<snifilter> ...]
11278
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011279 sslbindconf support "npn", "alpn", "verify", "ca-file", "no-ca-names",
11280 crl-file", "ecdhe", "curves", "ciphers" configuration. With BoringSSL
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011281 and Openssl >= 1.1.1 "ssl-min-ver" and "ssl-max-ver" are also supported.
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011282 It override the configuration set in bind line for the certificate.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011283
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +020011284 Wildcards are supported in the SNI filter. Negative filter are also supported,
11285 only useful in combination with a wildcard filter to exclude a particular SNI.
11286 The certificates will be presented to clients who provide a valid TLS Server
11287 Name Indication field matching one of the SNI filters. If no SNI filter is
11288 specified, the CN and alt subjects are used. This directive may be specified
11289 multiple times. See the "crt" option for more information. The default
11290 certificate is still needed to meet OpenSSL expectations. If it is not used,
11291 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +010011292
yanbzhu6c25e9e2016-01-05 12:52:02 -050011293 Multi-cert bundling (see "crt") is supported with crt-list, as long as only
Emmanuel Hocdetd294aea2016-05-13 11:14:06 +020011294 the base name is given in the crt-list. SNI filter will do the same work on
Emmanuel Hocdet84e417d2017-08-16 11:33:17 +020011295 all bundled certificates. With BoringSSL and Openssl >= 1.1.1 multi-cert is
11296 natively supported, avoid multi-cert bundling. RSA and ECDSA certificates can
11297 be declared in a row, and set different ssl and filter parameter.
yanbzhud19630c2015-12-14 15:10:25 -050011298
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011299 crt-list file example:
11300 cert1.pem
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011301 cert2.pem [alpn h2,http/1.1]
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011302 certW.pem *.domain.tld !secure.domain.tld
Emmanuel Hocdet05942112017-02-20 16:11:50 +010011303 certS.pem [curves X25519:P-256 ciphers ECDHE-ECDSA-AES256-GCM-SHA384] secure.domain.tld
Emmanuel Hocdet98263292016-12-29 18:26:15 +010011304
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011305defer-accept
11306 Is an optional keyword which is supported only on certain Linux kernels. It
11307 states that a connection will only be accepted once some data arrive on it,
11308 or at worst after the first retransmit. This should be used only on protocols
Davor Ocelice9ed2812017-12-25 17:49:28 +010011309 for which the client talks first (e.g. HTTP). It can slightly improve
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011310 performance by ensuring that most of the request is already available when
11311 the connection is accepted. On the other hand, it will not be able to detect
11312 connections which don't talk. It is important to note that this option is
11313 broken in all kernels up to 2.6.31, as the connection is never accepted until
11314 the client talks. This can cause issues with front firewalls which would see
11315 an established connection while the proxy will only see it in SYN_RECV. This
11316 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
11317
William Lallemandf6975e92017-05-26 17:42:10 +020011318expose-fd listeners
11319 This option is only usable with the stats socket. It gives your stats socket
11320 the capability to pass listeners FD to another HAProxy process.
William Lallemande202b1e2017-06-01 17:38:56 +020011321 During a reload with the master-worker mode, the process is automatically
11322 reexecuted adding -x and one of the stats socket with this option.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011323 See also "-x" in the management guide.
William Lallemandf6975e92017-05-26 17:42:10 +020011324
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011325force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011326 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011327 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011328 for high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011329 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011330
11331force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011332 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011333 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011334 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011335
11336force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011337 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011338 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011339 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011340
11341force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011342 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011343 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011344 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011345
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011346force-tlsv13
11347 This option enforces use of TLSv1.3 only on SSL connections instantiated from
11348 this listener. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011349 "ssl-default-bind-options". See also "ssl-min-ver" and "ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011350
Christopher Faulet31af49d2015-06-09 17:29:50 +020011351generate-certificates
11352 This setting is only available when support for OpenSSL was built in. It
11353 enables the dynamic SSL certificates generation. A CA certificate and its
11354 private key are necessary (see 'ca-sign-file'). When HAProxy is configured as
11355 a transparent forward proxy, SSL requests generate errors because of a common
11356 name mismatch on the certificate presented to the client. With this option
11357 enabled, HAProxy will try to forge a certificate using the SNI hostname
11358 indicated by the client. This is done only if no certificate matches the SNI
11359 hostname (see 'crt-list'). If an error occurs, the default certificate is
11360 used, else the 'strict-sni' option is set.
11361 It can also be used when HAProxy is configured as a reverse proxy to ease the
11362 deployment of an architecture with many backends.
11363
11364 Creating a SSL certificate is an expensive operation, so a LRU cache is used
11365 to store forged certificates (see 'tune.ssl.ssl-ctx-cache-size'). It
Davor Ocelice9ed2812017-12-25 17:49:28 +010011366 increases the HAProxy's memory footprint to reduce latency when the same
Christopher Faulet31af49d2015-06-09 17:29:50 +020011367 certificate is used many times.
11368
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011369gid <gid>
11370 Sets the group of the UNIX sockets to the designated system gid. It can also
11371 be set by default in the global section's "unix-bind" statement. Note that
11372 some platforms simply ignore this. This setting is equivalent to the "group"
11373 setting except that the group ID is used instead of its name. This setting is
11374 ignored by non UNIX sockets.
11375
11376group <group>
11377 Sets the group of the UNIX sockets to the designated system group. It can
11378 also be set by default in the global section's "unix-bind" statement. Note
11379 that some platforms simply ignore this. This setting is equivalent to the
11380 "gid" setting except that the group name is used instead of its gid. This
11381 setting is ignored by non UNIX sockets.
11382
11383id <id>
11384 Fixes the socket ID. By default, socket IDs are automatically assigned, but
11385 sometimes it is more convenient to fix them to ease monitoring. This value
11386 must be strictly positive and unique within the listener/frontend. This
11387 option can only be used when defining only a single socket.
11388
11389interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +010011390 Restricts the socket to a specific interface. When specified, only packets
11391 received from that particular interface are processed by the socket. This is
11392 currently only supported on Linux. The interface must be a primary system
11393 interface, not an aliased interface. It is also possible to bind multiple
11394 frontends to the same address if they are bound to different interfaces. Note
11395 that binding to a network interface requires root privileges. This parameter
Jérôme Magnin61275192018-02-07 11:39:58 +010011396 is only compatible with TCPv4/TCPv6 sockets. When specified, return traffic
11397 uses the same interface as inbound traffic, and its associated routing table,
11398 even if there are explicit routes through different interfaces configured.
11399 This can prove useful to address asymmetric routing issues when the same
11400 client IP addresses need to be able to reach frontends hosted on different
11401 interfaces.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011402
Willy Tarreauabb175f2012-09-24 12:43:26 +020011403level <level>
11404 This setting is used with the stats sockets only to restrict the nature of
11405 the commands that can be issued on the socket. It is ignored by other
11406 sockets. <level> can be one of :
Davor Ocelice9ed2812017-12-25 17:49:28 +010011407 - "user" is the least privileged level; only non-sensitive stats can be
Willy Tarreauabb175f2012-09-24 12:43:26 +020011408 read, and no change is allowed. It would make sense on systems where it
11409 is not easy to restrict access to the socket.
11410 - "operator" is the default level and fits most common uses. All data can
Davor Ocelice9ed2812017-12-25 17:49:28 +010011411 be read, and only non-sensitive changes are permitted (e.g. clear max
Willy Tarreauabb175f2012-09-24 12:43:26 +020011412 counters).
Davor Ocelice9ed2812017-12-25 17:49:28 +010011413 - "admin" should be used with care, as everything is permitted (e.g. clear
Willy Tarreauabb175f2012-09-24 12:43:26 +020011414 all counters).
11415
Andjelko Iharosc4df59e2017-07-20 11:59:48 +020011416severity-output <format>
11417 This setting is used with the stats sockets only to configure severity
11418 level output prepended to informational feedback messages. Severity
11419 level of messages can range between 0 and 7, conforming to syslog
11420 rfc5424. Valid and successful socket commands requesting data
11421 (i.e. "show map", "get acl foo" etc.) will never have a severity level
11422 prepended. It is ignored by other sockets. <format> can be one of :
11423 - "none" (default) no severity level is prepended to feedback messages.
11424 - "number" severity level is prepended as a number.
11425 - "string" severity level is prepended as a string following the
11426 rfc5424 convention.
11427
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011428maxconn <maxconn>
11429 Limits the sockets to this number of concurrent connections. Extraneous
11430 connections will remain in the system's backlog until a connection is
11431 released. If unspecified, the limit will be the same as the frontend's
11432 maxconn. Note that in case of port ranges or multiple addresses, the same
11433 value will be applied to each socket. This setting enables different
11434 limitations on expensive sockets, for instance SSL entries which may easily
11435 eat all memory.
11436
11437mode <mode>
11438 Sets the octal mode used to define access permissions on the UNIX socket. It
11439 can also be set by default in the global section's "unix-bind" statement.
11440 Note that some platforms simply ignore this. This setting is ignored by non
11441 UNIX sockets.
11442
11443mss <maxseg>
11444 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
11445 connections. This can be used to force a lower MSS for certain specific
11446 ports, for instance for connections passing through a VPN. Note that this
11447 relies on a kernel feature which is theoretically supported under Linux but
11448 was buggy in all versions prior to 2.6.28. It may or may not work on other
11449 operating systems. It may also not change the advertised value but change the
11450 effective size of outgoing segments. The commonly advertised value for TCPv4
11451 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
11452 positive, it will be used as the advertised MSS. If it is negative, it will
11453 indicate by how much to reduce the incoming connection's advertised MSS for
11454 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
11455
11456name <name>
11457 Sets an optional name for these sockets, which will be reported on the stats
11458 page.
11459
Willy Tarreaud72f0f32015-10-13 14:50:22 +020011460namespace <name>
11461 On Linux, it is possible to specify which network namespace a socket will
11462 belong to. This directive makes it possible to explicitly bind a listener to
11463 a namespace different from the default one. Please refer to your operating
11464 system's documentation to find more details about network namespaces.
11465
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011466nice <nice>
11467 Sets the 'niceness' of connections initiated from the socket. Value must be
11468 in the range -1024..1024 inclusive, and defaults to zero. Positive values
11469 means that such connections are more friendly to others and easily offer
11470 their place in the scheduler. On the opposite, negative values mean that
11471 connections want to run with a higher priority than others. The difference
11472 only happens under high loads when the system is close to saturation.
11473 Negative values are appropriate for low-latency or administration services,
11474 and high values are generally recommended for CPU intensive tasks such as SSL
11475 processing or bulk transfers which are less sensible to latency. For example,
11476 it may make sense to use a positive value for an SMTP socket and a negative
11477 one for an RDP socket.
11478
Emmanuel Hocdet174dfe52017-07-28 15:01:05 +020011479no-ca-names
11480 This setting is only available when support for OpenSSL was built in. It
11481 prevents from send CA names in server hello message when ca-file is used.
11482
Emeric Brun9b3009b2012-10-05 11:55:06 +020011483no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011484 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011485 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011486 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011487 be enabled using any configuration option. This option is also available on
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011488 global statement "ssl-default-bind-options". Use "ssl-min-ver" and
11489 "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011490
Emeric Brun90ad8722012-10-02 14:00:59 +020011491no-tls-tickets
11492 This setting is only available when support for OpenSSL was built in. It
11493 disables the stateless session resumption (RFC 5077 TLS Ticket
11494 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011495 session resumption is more expensive in CPU usage. This option is also
11496 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +020011497
Emeric Brun9b3009b2012-10-05 11:55:06 +020011498no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011499 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011500 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011501 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011502 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011503 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11504 and "ssl-max-ver" instead.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011505
Emeric Brun9b3009b2012-10-05 11:55:06 +020011506no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +020011507 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011508 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011509 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011510 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011511 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11512 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011513
Emeric Brun9b3009b2012-10-05 11:55:06 +020011514no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +020011515 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011516 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +020011517 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011518 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011519 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11520 and "ssl-max-ver" instead.
Emeric Brunf5da4932012-09-28 19:42:54 +020011521
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011522no-tlsv13
11523 This setting is only available when support for OpenSSL was built in. It
11524 disables support for TLSv1.3 on any sockets instantiated from the listener
11525 when SSL is supported. Note that SSLv2 is forced disabled in the code and
11526 cannot be enabled using any configuration option. This option is also
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011527 available on global statement "ssl-default-bind-options". Use "ssl-min-ver"
11528 and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011529
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011530npn <protocols>
11531 This enables the NPN TLS extension and advertises the specified protocol list
11532 as supported on top of NPN. The protocol list consists in a comma-delimited
11533 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
11534 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +020011535 enabled (check with haproxy -vv). Note that the NPN extension has been
Willy Tarreau95c4e142017-11-26 12:18:55 +010011536 replaced with the ALPN extension (see the "alpn" keyword), though this one is
11537 only available starting with OpenSSL 1.0.2. If HTTP/2 is desired on an older
11538 version of OpenSSL, NPN might still be used as most clients still support it
11539 at the time of writing this. It is possible to enable both NPN and ALPN
11540 though it probably doesn't make any sense out of testing.
Willy Tarreau6c9a3d52012-10-18 18:57:14 +020011541
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011542prefer-client-ciphers
11543 Use the client's preference when selecting the cipher suite, by default
11544 the server's preference is enforced. This option is also available on
11545 global statement "ssl-default-bind-options".
Lukas Tribus926594f2018-05-18 17:55:57 +020011546 Note that with OpenSSL >= 1.1.1 ChaCha20-Poly1305 is reprioritized anyway
11547 (without setting this option), if a ChaCha20-Poly1305 cipher is at the top of
11548 the client cipher list.
Lukas Tribus53ae85c2017-05-04 15:45:40 +000011549
Christopher Fauletc644fa92017-11-23 22:44:11 +010011550process <process-set>[/<thread-set>]
Willy Tarreaua36b3242019-02-02 13:14:34 +010011551 This restricts the list of processes or threads on which this listener is
Christopher Fauletc644fa92017-11-23 22:44:11 +010011552 allowed to run. It does not enforce any process but eliminates those which do
Davor Ocelice9ed2812017-12-25 17:49:28 +010011553 not match. If the frontend uses a "bind-process" setting, the intersection
Christopher Fauletc644fa92017-11-23 22:44:11 +010011554 between the two is applied. If in the end the listener is not allowed to run
11555 on any remaining process, a warning is emitted, and the listener will either
11556 run on the first process of the listener if a single process was specified,
11557 or on all of its processes if multiple processes were specified. If a thread
Davor Ocelice9ed2812017-12-25 17:49:28 +010011558 set is specified, it limits the threads allowed to process incoming
Willy Tarreaua36b3242019-02-02 13:14:34 +010011559 connections for this listener, for the the process set. If multiple processes
11560 and threads are configured, a warning is emitted, as it either results from a
11561 configuration error or a misunderstanding of these models. For the unlikely
11562 case where several ranges are needed, this directive may be repeated.
11563 <process-set> and <thread-set> must use the format
Christopher Fauletc644fa92017-11-23 22:44:11 +010011564
11565 all | odd | even | number[-[number]]
11566
11567 Ranges can be partially defined. The higher bound can be omitted. In such
11568 case, it is replaced by the corresponding maximum value. The main purpose of
11569 this directive is to be used with the stats sockets and have one different
11570 socket per process. The second purpose is to have multiple bind lines sharing
11571 the same IP:port but not the same process in a listener, so that the system
11572 can distribute the incoming connections into multiple queues and allow a
11573 smoother inter-process load balancing. Currently Linux 3.9 and above is known
11574 for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +020011575
Christopher Fauleta717b992018-04-10 14:43:00 +020011576proto <name>
11577 Forces the multiplexer's protocol to use for the incoming connections. It
11578 must be compatible with the mode of the frontend (TCP or HTTP). It must also
11579 be usable on the frontend side. The list of available protocols is reported
11580 in haproxy -vv.
11581 Idea behind this optipon is to bypass the selection of the best multiplexer's
11582 protocol for all connections instantiated from this listening socket. For
Joseph Herlant71b4b152018-11-13 16:55:16 -080011583 instance, it is possible to force the http/2 on clear TCP by specifying "proto
Christopher Fauleta717b992018-04-10 14:43:00 +020011584 h2" on the bind line.
11585
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011586ssl
11587 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011588 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011589 certificate is necessary (see "crt" above). All contents in the buffers will
11590 appear in clear text, so that ACLs and HTTP processing will only have access
Emmanuel Hocdetbd695fe2017-05-15 15:53:41 +020011591 to deciphered contents. SSLv3 is disabled per default, use "ssl-min-ver SSLv3"
11592 to enable it.
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011593
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011594ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11595 This option enforces use of <version> or lower on SSL connections instantiated
11596 from this listener. This option is also available on global statement
11597 "ssl-default-bind-options". See also "ssl-min-ver".
11598
11599ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
11600 This option enforces use of <version> or upper on SSL connections instantiated
11601 from this listener. This option is also available on global statement
11602 "ssl-default-bind-options". See also "ssl-max-ver".
11603
Emmanuel Hocdet65623372013-01-24 17:17:15 +010011604strict-sni
11605 This setting is only available when support for OpenSSL was built in. The
11606 SSL/TLS negotiation is allow only if the client provided an SNI which match
11607 a certificate. The default certificate is not used.
11608 See the "crt" option for more information.
11609
Willy Tarreau2af207a2015-02-04 00:45:58 +010011610tcp-ut <delay>
Tim Düsterhus4896c442016-11-29 02:15:19 +010011611 Sets the TCP User Timeout for all incoming connections instantiated from this
Willy Tarreau2af207a2015-02-04 00:45:58 +010011612 listening socket. This option is available on Linux since version 2.6.37. It
11613 allows haproxy to configure a timeout for sockets which contain data not
Davor Ocelice9ed2812017-12-25 17:49:28 +010011614 receiving an acknowledgment for the configured delay. This is especially
Willy Tarreau2af207a2015-02-04 00:45:58 +010011615 useful on long-lived connections experiencing long idle periods such as
11616 remote terminals or database connection pools, where the client and server
11617 timeouts must remain high to allow a long period of idle, but where it is
11618 important to detect that the client has disappeared in order to release all
11619 resources associated with its connection (and the server's session). The
11620 argument is a delay expressed in milliseconds by default. This only works
11621 for regular TCP connections, and is ignored for other protocols.
11622
Willy Tarreau1c862c52012-10-05 16:21:00 +020011623tfo
Lukas Tribus0defb902013-02-13 23:35:39 +010011624 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +020011625 enables TCP Fast Open on the listening socket, which means that clients which
11626 support this feature will be able to send a request and receive a response
11627 during the 3-way handshake starting from second connection, thus saving one
11628 round-trip after the first connection. This only makes sense with protocols
11629 that use high connection rates and where each round trip matters. This can
11630 possibly cause issues with many firewalls which do not accept data on SYN
11631 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +020011632 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
11633 need to build HAProxy with USE_TFO=1 if your libc doesn't define
11634 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +020011635
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011636tls-ticket-keys <keyfile>
11637 Sets the TLS ticket keys file to load the keys from. The keys need to be 48
Emeric Brun9e754772019-01-10 17:51:55 +010011638 or 80 bytes long, depending if aes128 or aes256 is used, encoded with base64
11639 with one line per key (ex. openssl rand 80 | openssl base64 -A | xargs echo).
11640 The first key determines the key length used for next keys: you can't mix
11641 aes128 and aes256 keys. Number of keys is specified by the TLS_TICKETS_NO
11642 build option (default 3) and at least as many keys need to be present in
11643 the file. Last TLS_TICKETS_NO keys will be used for decryption and the
11644 penultimate one for encryption. This enables easy key rotation by just
11645 appending new key to the file and reloading the process. Keys must be
11646 periodically rotated (ex. every 12h) or Perfect Forward Secrecy is
11647 compromised. It is also a good idea to keep the keys off any permanent
Nenad Merdanovic188ad3e2015-02-27 19:56:50 +010011648 storage such as hard drives (hint: use tmpfs and don't swap those files).
11649 Lifetime hint can be changed using tune.ssl.timeout.
11650
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011651transparent
11652 Is an optional keyword which is supported only on certain Linux kernels. It
11653 indicates that the addresses will be bound even if they do not belong to the
11654 local machine, and that packets targeting any of these addresses will be
11655 intercepted just as if the addresses were locally configured. This normally
11656 requires that IP forwarding is enabled. Caution! do not use this with the
11657 default address '*', as it would redirect any traffic for the specified port.
11658 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
11659 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
11660 kernel version. Some distribution kernels include backports of the feature,
11661 so check for support with your vendor.
11662
Willy Tarreau77e3af92012-11-24 15:07:23 +010011663v4v6
11664 Is an optional keyword which is supported only on most recent systems
11665 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
11666 and IPv6 when it uses the default address. Doing so is sometimes necessary
11667 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011668 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +010011669
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011670v6only
11671 Is an optional keyword which is supported only on most recent systems
11672 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
11673 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +010011674 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
11675 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +010011676
Willy Tarreaub6205fd2012-09-24 12:27:33 +020011677uid <uid>
11678 Sets the owner of the UNIX sockets to the designated system uid. It can also
11679 be set by default in the global section's "unix-bind" statement. Note that
11680 some platforms simply ignore this. This setting is equivalent to the "user"
11681 setting except that the user numeric ID is used instead of its name. This
11682 setting is ignored by non UNIX sockets.
11683
11684user <user>
11685 Sets the owner of the UNIX sockets to the designated system user. It can also
11686 be set by default in the global section's "unix-bind" statement. Note that
11687 some platforms simply ignore this. This setting is equivalent to the "uid"
11688 setting except that the user name is used instead of its uid. This setting is
11689 ignored by non UNIX sockets.
11690
Emeric Brun1a073b42012-09-28 17:07:34 +020011691verify [none|optional|required]
11692 This setting is only available when support for OpenSSL was built in. If set
11693 to 'none', client certificate is not requested. This is the default. In other
11694 cases, a client certificate is requested. If the client does not provide a
11695 certificate after the request and if 'verify' is set to 'required', then the
11696 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +020011697 certificate provided by the client is always verified using CAs from
11698 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
11699 is aborted, regardless of the 'verify' option, unless the error code exactly
11700 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +020011701
Willy Tarreaub6205fd2012-09-24 12:27:33 +0200117025.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +010011703------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020011704
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011705The "server" and "default-server" keywords support a certain number of settings
11706which are all passed as arguments on the server line. The order in which those
11707arguments appear does not count, and they are all optional. Some of those
11708settings are single words (booleans) while others expect one or several values
11709after them. In this case, the values must immediately follow the setting name.
11710Except default-server, all those settings must be specified after the server's
11711address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +020011712
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011713 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +010011714 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +020011715
Frédéric Lécailled2376272017-03-21 18:52:12 +010011716Note that all these settings are supported both by "server" and "default-server"
11717keywords, except "id" which is only supported by "server".
11718
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011719The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011720
Willy Tarreauceb4ac92012-04-28 00:41:46 +020011721addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011722 Using the "addr" parameter, it becomes possible to use a different IP address
Baptiste Assmann13f83532016-03-06 23:14:36 +010011723 to send health-checks or to probe the agent-check. On some servers, it may be
11724 desirable to dedicate an IP address to specific component able to perform
11725 complex tests which are more suitable to health-checks than the application.
11726 This parameter is ignored if the "check" parameter is not set. See also the
11727 "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +020011728
Simon Hormand60d6912013-11-25 10:46:36 +090011729agent-check
11730 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +010011731 health check. An agent health check is performed by making a TCP connection
Willy Tarreau7a0139e2018-12-16 08:42:56 +010011732 to the port set by the "agent-port" parameter and reading an ASCII string
11733 terminated by the first '\r' or '\n' met. The string is made of a series of
11734 words delimited by spaces, tabs or commas in any order, each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +090011735
Willy Tarreau81f5d942013-12-09 20:51:51 +010011736 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +090011737 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +020011738 weight of a server as configured when haproxy starts. Note that a zero
11739 weight is reported on the stats page as "DRAIN" since it has the same
11740 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +090011741
Davor Ocelice9ed2812017-12-25 17:49:28 +010011742 - The string "maxconn:" followed by an integer (no space between). Values
11743 in this format will set the maxconn of a server. The maximum number of
11744 connections advertised needs to be multiplied by the number of load
11745 balancers and different backends that use this health check to get the
11746 total number of connections the server might receive. Example: maxconn:30
Nenad Merdanovic174dd372016-04-24 23:10:06 +020011747
Willy Tarreau81f5d942013-12-09 20:51:51 +010011748 - The word "ready". This will turn the server's administrative state to the
Davor Ocelice9ed2812017-12-25 17:49:28 +010011749 READY mode, thus canceling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +090011750
Willy Tarreau81f5d942013-12-09 20:51:51 +010011751 - The word "drain". This will turn the server's administrative state to the
11752 DRAIN mode, thus it will not accept any new connections other than those
11753 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +090011754
Willy Tarreau81f5d942013-12-09 20:51:51 +010011755 - The word "maint". This will turn the server's administrative state to the
11756 MAINT mode, thus it will not accept any new connections at all, and health
11757 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +090011758
Willy Tarreau81f5d942013-12-09 20:51:51 +010011759 - The words "down", "failed", or "stopped", optionally followed by a
11760 description string after a sharp ('#'). All of these mark the server's
11761 operating state as DOWN, but since the word itself is reported on the stats
11762 page, the difference allows an administrator to know if the situation was
11763 expected or not : the service may intentionally be stopped, may appear up
Davor Ocelice9ed2812017-12-25 17:49:28 +010011764 but fail some validity tests, or may be seen as down (e.g. missing process,
Willy Tarreau81f5d942013-12-09 20:51:51 +010011765 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +090011766
Willy Tarreau81f5d942013-12-09 20:51:51 +010011767 - The word "up" sets back the server's operating state as UP if health checks
11768 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +090011769
Willy Tarreau81f5d942013-12-09 20:51:51 +010011770 Parameters which are not advertised by the agent are not changed. For
11771 example, an agent might be designed to monitor CPU usage and only report a
11772 relative weight and never interact with the operating status. Similarly, an
11773 agent could be designed as an end-user interface with 3 radio buttons
11774 allowing an administrator to change only the administrative state. However,
11775 it is important to consider that only the agent may revert its own actions,
11776 so if a server is set to DRAIN mode or to DOWN state using the agent, the
11777 agent must implement the other equivalent actions to bring the service into
11778 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +090011779
Simon Horman2f1f9552013-11-25 10:46:37 +090011780 Failure to connect to the agent is not considered an error as connectivity
11781 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +010011782 parameter. Warning though, it is not a good idea to stop an agent after it
11783 reports "down", since only an agent reporting "up" will be able to turn the
11784 server up again. Note that the CLI on the Unix stats socket is also able to
Willy Tarreau989222a2016-01-15 10:26:26 +010011785 force an agent's result in order to work around a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +090011786
Willy Tarreau81f5d942013-12-09 20:51:51 +010011787 Requires the "agent-port" parameter to be set. See also the "agent-inter"
Frédéric Lécailled2376272017-03-21 18:52:12 +010011788 and "no-agent-check" parameters.
Simon Hormand60d6912013-11-25 10:46:36 +090011789
James Brown55f9ff12015-10-21 18:19:05 -070011790agent-send <string>
11791 If this option is specified, haproxy will send the given string (verbatim)
11792 to the agent server upon connection. You could, for example, encode
11793 the backend name into this string, which would enable your agent to send
11794 different responses based on the backend. Make sure to include a '\n' if
11795 you want to terminate your request with a newline.
11796
Simon Hormand60d6912013-11-25 10:46:36 +090011797agent-inter <delay>
11798 The "agent-inter" parameter sets the interval between two agent checks
11799 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
11800
11801 Just as with every other time-based parameter, it may be entered in any
11802 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
11803 parameter also serves as a timeout for agent checks "timeout check" is
11804 not set. In order to reduce "resonance" effects when multiple servers are
11805 hosted on the same hardware, the agent and health checks of all servers
11806 are started with a small time offset between them. It is also possible to
11807 add some random noise in the agent and health checks interval using the
11808 global "spread-checks" keyword. This makes sense for instance when a lot
11809 of backends use the same servers.
11810
11811 See also the "agent-check" and "agent-port" parameters.
11812
Misiek768d8602017-01-09 09:52:43 +010011813agent-addr <addr>
11814 The "agent-addr" parameter sets address for agent check.
11815
11816 You can offload agent-check to another target, so you can make single place
11817 managing status and weights of servers defined in haproxy in case you can't
11818 make self-aware and self-managing services. You can specify both IP or
11819 hostname, it will be resolved.
11820
Simon Hormand60d6912013-11-25 10:46:36 +090011821agent-port <port>
11822 The "agent-port" parameter sets the TCP port used for agent checks.
11823
11824 See also the "agent-check" and "agent-inter" parameters.
11825
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011826allow-0rtt
11827 Allow sending early data to the server when using TLS 1.3.
Olivier Houchard22c9b442019-05-06 19:01:04 +020011828 Note that early data will be sent only if the client used early data, or
11829 if the backend uses "retry-on" with the "0rtt-rejected" keyword.
Olivier Houchard8cb2d2e2019-05-06 18:58:48 +020011830
Olivier Houchardc7566002018-11-20 23:33:50 +010011831alpn <protocols>
11832 This enables the TLS ALPN extension and advertises the specified protocol
11833 list as supported on top of ALPN. The protocol list consists in a comma-
11834 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
11835 quotes). This requires that the SSL library is build with support for TLS
11836 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
11837 initial NPN extension. ALPN is required to connect to HTTP/2 servers.
11838 Versions of OpenSSL prior to 1.0.2 didn't support ALPN and only supposed the
11839 now obsolete NPN extension.
11840 If both HTTP/2 and HTTP/1.1 are expected to be supported, both versions can
11841 be advertised, in order of preference, like below :
11842
11843 server 127.0.0.1:443 ssl crt pub.pem alpn h2,http/1.1
11844
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011845backup
11846 When "backup" is present on a server line, the server is only used in load
11847 balancing when all other non-backup servers are unavailable. Requests coming
11848 with a persistence cookie referencing the server will always be served
11849 though. By default, only the first operational backup server is used, unless
Frédéric Lécailled2376272017-03-21 18:52:12 +010011850 the "allbackups" option is set in the backend. See also the "no-backup" and
11851 "allbackups" options.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011852
Emeric Brunef42d922012-10-11 16:11:36 +020011853ca-file <cafile>
11854 This setting is only available when support for OpenSSL was built in. It
11855 designates a PEM file from which to load CA certificates used to verify
11856 server's certificate.
11857
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011858check
11859 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +010011860 always considered available. If "check" is set, the server is available when
11861 accepting periodic TCP connections, to ensure that it is really able to serve
11862 requests. The default address and port to send the tests to are those of the
11863 server, and the default source is the same as the one defined in the
11864 backend. It is possible to change the address using the "addr" parameter, the
11865 port using the "port" parameter, the source address using the "source"
11866 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +090011867 parameters. The request method is define in the backend using the "httpchk",
11868 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
Frédéric Lécailled2376272017-03-21 18:52:12 +010011869 refer to those options and parameters for more information. See also
11870 "no-check" option.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011871
Willy Tarreau6c16adc2012-10-05 00:04:16 +020011872check-send-proxy
11873 This option forces emission of a PROXY protocol line with outgoing health
11874 checks, regardless of whether the server uses send-proxy or not for the
11875 normal traffic. By default, the PROXY protocol is enabled for health checks
11876 if it is already enabled for normal traffic and if no "port" nor "addr"
11877 directive is present. However, if such a directive is present, the
11878 "check-send-proxy" option needs to be used to force the use of the
11879 protocol. See also the "send-proxy" option for more information.
11880
Olivier Houchard92150142018-12-21 19:47:01 +010011881check-alpn <protocols>
11882 Defines which protocols to advertise with ALPN. The protocol list consists in
11883 a comma-delimited list of protocol names, for instance: "http/1.1,http/1.0"
11884 (without quotes). If it is not set, the server ALPN is used.
11885
Jérôme Magninae9bb762018-12-09 16:08:26 +010011886check-sni <sni>
Olivier Houchard9130a962017-10-17 17:33:43 +020011887 This option allows you to specify the SNI to be used when doing health checks
Jérôme Magninae9bb762018-12-09 16:08:26 +010011888 over SSL. It is only possible to use a string to set <sni>. If you want to
11889 set a SNI for proxied traffic, see "sni".
Olivier Houchard9130a962017-10-17 17:33:43 +020011890
Willy Tarreau763a95b2012-10-04 23:15:39 +020011891check-ssl
11892 This option forces encryption of all health checks over SSL, regardless of
11893 whether the server uses SSL or not for the normal traffic. This is generally
11894 used when an explicit "port" or "addr" directive is specified and SSL health
11895 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011896 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +020011897 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
11898 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
Davor Ocelice9ed2812017-12-25 17:49:28 +010011899 All SSL settings are common to health checks and traffic (e.g. ciphers).
Frédéric Lécailled2376272017-03-21 18:52:12 +010011900 See the "ssl" option for more information and "no-check-ssl" to disable
11901 this option.
Willy Tarreau763a95b2012-10-04 23:15:39 +020011902
Alexander Liu2a54bb72019-05-22 19:44:48 +080011903check-via-socks4
11904 This option enables outgoinng health checks using upstream socks4 proxy. By
11905 default, the health checks won't go through socks tunnel even it was enabled
11906 for normal traffic.
11907
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011908ciphers <ciphers>
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011909 This setting is only available when support for OpenSSL was built in. This
11910 option sets the string describing the list of cipher algorithms that is
11911 negotiated during the SSL/TLS handshake with the server. The format of the
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011912 string is defined in "man 1 ciphers" from OpenSSL man pages. For background
11913 information and recommendations see e.g.
11914 (https://wiki.mozilla.org/Security/Server_Side_TLS) and
11915 (https://mozilla.github.io/server-side-tls/ssl-config-generator/). For TLSv1.3
11916 cipher configuration, please check the "ciphersuites" keyword.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020011917
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011918ciphersuites <ciphersuites>
11919 This setting is only available when support for OpenSSL was built in and
11920 OpenSSL 1.1.1 or later was used to build HAProxy. This option sets the string
11921 describing the list of cipher algorithms that is negotiated during the TLS
11922 1.3 handshake with the server. The format of the string is defined in
Bertrand Jacquin4f03ab02019-02-03 18:48:49 +000011923 "man 1 ciphers" from OpenSSL man pages under the "ciphersuites" section.
11924 For cipher configuration for TLSv1.2 and earlier, please check the "ciphers"
11925 keyword.
Dirkjan Bussink415150f2018-09-14 11:14:21 +020011926
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011927cookie <value>
11928 The "cookie" parameter sets the cookie value assigned to the server to
11929 <value>. This value will be checked in incoming requests, and the first
11930 operational server possessing the same value will be selected. In return, in
11931 cookie insertion or rewrite modes, this value will be assigned to the cookie
11932 sent to the client. There is nothing wrong in having several servers sharing
11933 the same cookie value, and it is in fact somewhat common between normal and
11934 backup servers. See also the "cookie" keyword in backend section.
11935
Emeric Brunef42d922012-10-11 16:11:36 +020011936crl-file <crlfile>
11937 This setting is only available when support for OpenSSL was built in. It
11938 designates a PEM file from which to load certificate revocation list used
11939 to verify server's certificate.
11940
Emeric Bruna7aa3092012-10-26 12:58:00 +020011941crt <cert>
11942 This setting is only available when support for OpenSSL was built in.
11943 It designates a PEM file from which to load both a certificate and the
11944 associated private key. This file can be built by concatenating both PEM
11945 files into one. This certificate will be sent if the server send a client
11946 certificate request.
11947
Willy Tarreau96839092010-03-29 10:02:24 +020011948disabled
11949 The "disabled" keyword starts the server in the "disabled" state. That means
11950 that it is marked down in maintenance mode, and no connection other than the
11951 ones allowed by persist mode will reach it. It is very well suited to setup
11952 new servers, because normal traffic will never reach them, while it is still
11953 possible to test the service by making use of the force-persist mechanism.
Frédéric Lécailled2376272017-03-21 18:52:12 +010011954 See also "enabled" setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011955
Frédéric Lécailled2376272017-03-21 18:52:12 +010011956enabled
11957 This option may be used as 'server' setting to reset any 'disabled'
11958 setting which would have been inherited from 'default-server' directive as
11959 default value.
11960 It may also be used as 'default-server' setting to reset any previous
11961 'default-server' 'disabled' setting.
Willy Tarreau96839092010-03-29 10:02:24 +020011962
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011963error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +010011964 If health observing is enabled, the "error-limit" parameter specifies the
11965 number of consecutive errors that triggers event selected by the "on-error"
11966 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011967
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011968 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010011969
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010011970fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011971 The "fall" parameter states that a server will be considered as dead after
11972 <count> consecutive unsuccessful health checks. This value defaults to 3 if
11973 unspecified. See also the "check", "inter" and "rise" parameters.
11974
Emeric Brun8694b9a2012-10-05 14:39:07 +020011975force-sslv3
11976 This option enforces use of SSLv3 only when SSL is used to communicate with
11977 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011978 high connection rates. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011979 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011980
11981force-tlsv10
11982 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011983 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011984 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011985
11986force-tlsv11
11987 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011988 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011989 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011990
11991force-tlsv12
11992 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +010011993 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011994 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emeric Brun8694b9a2012-10-05 14:39:07 +020011995
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020011996force-tlsv13
11997 This option enforces use of TLSv1.3 only when SSL is used to communicate with
11998 the server. This option is also available on global statement
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020011999 "ssl-default-server-options". See also "ssl-min-ver" and ssl-max-ver".
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012000
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012001id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +020012002 Set a persistent ID for the server. This ID must be positive and unique for
12003 the proxy. An unused ID will automatically be assigned if unset. The first
12004 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012005
Willy Tarreau6a031d12016-11-07 19:42:35 +010012006init-addr {last | libc | none | <ip>},[...]*
12007 Indicate in what order the server's address should be resolved upon startup
12008 if it uses an FQDN. Attempts are made to resolve the address by applying in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012009 turn each of the methods mentioned in the comma-delimited list. The first
Willy Tarreau6a031d12016-11-07 19:42:35 +010012010 method which succeeds is used. If the end of the list is reached without
12011 finding a working method, an error is thrown. Method "last" suggests to pick
12012 the address which appears in the state file (see "server-state-file"). Method
12013 "libc" uses the libc's internal resolver (gethostbyname() or getaddrinfo()
12014 depending on the operating system and build options). Method "none"
12015 specifically indicates that the server should start without any valid IP
12016 address in a down state. It can be useful to ignore some DNS issues upon
12017 startup, waiting for the situation to get fixed later. Finally, an IP address
12018 (IPv4 or IPv6) may be provided. It can be the currently known address of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012019 server (e.g. filled by a configuration generator), or the address of a dummy
Willy Tarreau6a031d12016-11-07 19:42:35 +010012020 server used to catch old sessions and present them with a decent error
12021 message for example. When the "first" load balancing algorithm is used, this
12022 IP address could point to a fake server used to trigger the creation of new
12023 instances on the fly. This option defaults to "last,libc" indicating that the
12024 previous address found in the state file (if any) is used first, otherwise
12025 the libc's resolver is used. This ensures continued compatibility with the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012026 historic behavior.
Willy Tarreau6a031d12016-11-07 19:42:35 +010012027
12028 Example:
12029 defaults
12030 # never fail on address resolution
12031 default-server init-addr last,libc,none
12032
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012033inter <delay>
12034fastinter <delay>
12035downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012036 The "inter" parameter sets the interval between two consecutive health checks
12037 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
12038 It is also possible to use "fastinter" and "downinter" to optimize delays
12039 between checks depending on the server state :
12040
Pieter Baauw44fc9df2015-09-17 21:30:46 +020012041 Server state | Interval used
12042 ----------------------------------------+----------------------------------
12043 UP 100% (non-transitional) | "inter"
12044 ----------------------------------------+----------------------------------
12045 Transitionally UP (going down "fall"), | "fastinter" if set,
12046 Transitionally DOWN (going up "rise"), | "inter" otherwise.
12047 or yet unchecked. |
12048 ----------------------------------------+----------------------------------
12049 DOWN 100% (non-transitional) | "downinter" if set,
12050 | "inter" otherwise.
12051 ----------------------------------------+----------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +010012052
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012053 Just as with every other time-based parameter, they can be entered in any
12054 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
12055 serves as a timeout for health checks sent to servers if "timeout check" is
12056 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +090012057 hosted on the same hardware, the agent and health checks of all servers
12058 are started with a small time offset between them. It is also possible to
12059 add some random noise in the agent and health checks interval using the
12060 global "spread-checks" keyword. This makes sense for instance when a lot
12061 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012062
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012063maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012064 The "maxconn" parameter specifies the maximal number of concurrent
12065 connections that will be sent to this server. If the number of incoming
12066 concurrent requests goes higher than this value, they will be queued, waiting
12067 for a connection to be released. This parameter is very important as it can
12068 save fragile servers from going down under extreme loads. If a "minconn"
12069 parameter is specified, the limit becomes dynamic. The default value is "0"
12070 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
12071 the backend's "fullconn" keyword.
12072
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012073maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012074 The "maxqueue" parameter specifies the maximal number of connections which
12075 will wait in the queue for this server. If this limit is reached, next
12076 requests will be redispatched to other servers instead of indefinitely
12077 waiting to be served. This will break persistence but may allow people to
12078 quickly re-log in when the server they try to connect to is dying. The
12079 default value is "0" which means the queue is unlimited. See also the
12080 "maxconn" and "minconn" parameters.
12081
Willy Tarreau9c538e02019-01-23 10:21:49 +010012082max-reuse <count>
12083 The "max-reuse" argument indicates the HTTP connection processors that they
12084 should not reuse a server connection more than this number of times to send
12085 new requests. Permitted values are -1 (the default), which disables this
12086 limit, or any positive value. Value zero will effectively disable keep-alive.
12087 This is only used to work around certain server bugs which cause them to leak
12088 resources over time. The argument is not necessarily respected by the lower
12089 layers as there might be technical limitations making it impossible to
12090 enforce. At least HTTP/2 connections to servers will respect it.
12091
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012092minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012093 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
12094 limit following the backend's load. The server will always accept at least
12095 <minconn> connections, never more than <maxconn>, and the limit will be on
12096 the ramp between both values when the backend has less than <fullconn>
12097 concurrent connections. This makes it possible to limit the load on the
12098 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012099 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012100 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012101
Willy Tarreaud72f0f32015-10-13 14:50:22 +020012102namespace <name>
12103 On Linux, it is possible to specify which network namespace a socket will
12104 belong to. This directive makes it possible to explicitly bind a server to
12105 a namespace different from the default one. Please refer to your operating
12106 system's documentation to find more details about network namespaces.
12107
Frédéric Lécailled2376272017-03-21 18:52:12 +010012108no-agent-check
12109 This option may be used as "server" setting to reset any "agent-check"
12110 setting which would have been inherited from "default-server" directive as
12111 default value.
12112 It may also be used as "default-server" setting to reset any previous
12113 "default-server" "agent-check" setting.
12114
12115no-backup
12116 This option may be used as "server" setting to reset any "backup"
12117 setting which would have been inherited from "default-server" directive as
12118 default value.
12119 It may also be used as "default-server" setting to reset any previous
12120 "default-server" "backup" setting.
12121
12122no-check
12123 This option may be used as "server" setting to reset any "check"
12124 setting which would have been inherited from "default-server" directive as
12125 default value.
12126 It may also be used as "default-server" setting to reset any previous
12127 "default-server" "check" setting.
12128
12129no-check-ssl
12130 This option may be used as "server" setting to reset any "check-ssl"
12131 setting which would have been inherited from "default-server" directive as
12132 default value.
12133 It may also be used as "default-server" setting to reset any previous
12134 "default-server" "check-ssl" setting.
12135
Frédéric Lécailled2376272017-03-21 18:52:12 +010012136no-send-proxy
12137 This option may be used as "server" setting to reset any "send-proxy"
12138 setting which would have been inherited from "default-server" directive as
12139 default value.
12140 It may also be used as "default-server" setting to reset any previous
12141 "default-server" "send-proxy" setting.
12142
12143no-send-proxy-v2
12144 This option may be used as "server" setting to reset any "send-proxy-v2"
12145 setting which would have been inherited from "default-server" directive as
12146 default value.
12147 It may also be used as "default-server" setting to reset any previous
12148 "default-server" "send-proxy-v2" setting.
12149
12150no-send-proxy-v2-ssl
12151 This option may be used as "server" setting to reset any "send-proxy-v2-ssl"
12152 setting which would have been inherited from "default-server" directive as
12153 default value.
12154 It may also be used as "default-server" setting to reset any previous
12155 "default-server" "send-proxy-v2-ssl" setting.
12156
12157no-send-proxy-v2-ssl-cn
12158 This option may be used as "server" setting to reset any "send-proxy-v2-ssl-cn"
12159 setting which would have been inherited from "default-server" directive as
12160 default value.
12161 It may also be used as "default-server" setting to reset any previous
12162 "default-server" "send-proxy-v2-ssl-cn" setting.
12163
12164no-ssl
12165 This option may be used as "server" setting to reset any "ssl"
12166 setting which would have been inherited from "default-server" directive as
12167 default value.
12168 It may also be used as "default-server" setting to reset any previous
12169 "default-server" "ssl" setting.
12170
Willy Tarreau2a3fb1c2015-02-05 16:47:07 +010012171no-ssl-reuse
12172 This option disables SSL session reuse when SSL is used to communicate with
12173 the server. It will force the server to perform a full handshake for every
12174 new connection. It's probably only useful for benchmarking, troubleshooting,
12175 and for paranoid users.
12176
Emeric Brun9b3009b2012-10-05 11:55:06 +020012177no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012178 This option disables support for SSLv3 when SSL is used to communicate with
12179 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012180 using any configuration option. Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012181
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012182 Supported in default-server: No
12183
Emeric Brunf9c5c472012-10-11 15:28:34 +020012184no-tls-tickets
12185 This setting is only available when support for OpenSSL was built in. It
12186 disables the stateless session resumption (RFC 5077 TLS Ticket
12187 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012188 session resumption is more expensive in CPU usage for servers. This option
12189 is also available on global statement "ssl-default-server-options".
Frédéric Lécailled2376272017-03-21 18:52:12 +010012190 See also "tls-tickets".
Emeric Brunf9c5c472012-10-11 15:28:34 +020012191
Emeric Brun9b3009b2012-10-05 11:55:06 +020012192no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +020012193 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012194 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12195 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012196 often makes sense to disable it when communicating with local servers. This
12197 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012198 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012199
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012200 Supported in default-server: No
12201
Emeric Brun9b3009b2012-10-05 11:55:06 +020012202no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +020012203 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +020012204 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12205 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012206 often makes sense to disable it when communicating with local servers. This
12207 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012208 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012209
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012210 Supported in default-server: No
12211
Emeric Brun9b3009b2012-10-05 11:55:06 +020012212no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +020012213 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012214 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12215 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +010012216 often makes sense to disable it when communicating with local servers. This
12217 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012218 Use "ssl-min-ver" and "ssl-max-ver" instead.
Emmanuel Hocdet42fb9802017-03-30 19:29:39 +020012219
12220 Supported in default-server: No
12221
12222no-tlsv13
12223 This option disables support for TLSv1.3 when SSL is used to communicate with
12224 the server. Note that SSLv2 is disabled in the code and cannot be enabled
12225 using any configuration option. TLSv1 is more expensive than SSLv3 so it
12226 often makes sense to disable it when communicating with local servers. This
12227 option is also available on global statement "ssl-default-server-options".
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012228 Use "ssl-min-ver" and "ssl-max-ver" instead.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012229
Emmanuel Hocdet6cb2d1e2017-03-30 14:43:31 +020012230 Supported in default-server: No
12231
Frédéric Lécailled2376272017-03-21 18:52:12 +010012232no-verifyhost
12233 This option may be used as "server" setting to reset any "verifyhost"
12234 setting which would have been inherited from "default-server" directive as
12235 default value.
12236 It may also be used as "default-server" setting to reset any previous
12237 "default-server" "verifyhost" setting.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012238
Simon Hormanfa461682011-06-25 09:39:49 +090012239non-stick
12240 Never add connections allocated to this sever to a stick-table.
12241 This may be used in conjunction with backup to ensure that
12242 stick-table persistence is disabled for backup servers.
12243
Olivier Houchardc7566002018-11-20 23:33:50 +010012244npn <protocols>
12245 This enables the NPN TLS extension and advertises the specified protocol list
12246 as supported on top of NPN. The protocol list consists in a comma-delimited
12247 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
12248 This requires that the SSL library is build with support for TLS extensions
12249 enabled (check with haproxy -vv). Note that the NPN extension has been
12250 replaced with the ALPN extension (see the "alpn" keyword), though this one is
12251 only available starting with OpenSSL 1.0.2.
12252
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012253observe <mode>
12254 This option enables health adjusting based on observing communication with
12255 the server. By default this functionality is disabled and enabling it also
12256 requires to enable health checks. There are two supported modes: "layer4" and
12257 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
12258 significant. In layer7, which is only allowed for http proxies, responses
12259 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +010012260 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012261
12262 See also the "check", "on-error" and "error-limit".
12263
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012264on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +010012265 Select what should happen when enough consecutive errors are detected.
12266 Currently, four modes are available:
12267 - fastinter: force fastinter
12268 - fail-check: simulate a failed check, also forces fastinter (default)
12269 - sudden-death: simulate a pre-fatal failed health check, one more failed
12270 check will mark a server down, forces fastinter
12271 - mark-down: mark the server immediately down and force fastinter
12272
12273 See also the "check", "observe" and "error-limit".
12274
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012275on-marked-down <action>
12276 Modify what occurs when a server is marked down.
12277 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012278 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
12279 all connections to the server are immediately terminated when the server
12280 goes down. It might be used if the health check detects more complex cases
12281 than a simple connection status, and long timeouts would cause the service
12282 to remain unresponsive for too long a time. For instance, a health check
12283 might detect that a database is stuck and that there's no chance to reuse
12284 existing connections anymore. Connections killed this way are logged with
12285 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +090012286
12287 Actions are disabled by default
12288
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012289on-marked-up <action>
12290 Modify what occurs when a server is marked up.
12291 Currently one action is available:
12292 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
12293 done only if the server is not in backup state and if it is not disabled
12294 (it must have an effective weight > 0). This can be used sometimes to force
12295 an active server to take all the traffic back after recovery when dealing
Davor Ocelice9ed2812017-12-25 17:49:28 +010012296 with long sessions (e.g. LDAP, SQL, ...). Doing this can cause more trouble
12297 than it tries to solve (e.g. incomplete transactions), so use this feature
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070012298 with extreme care. Sessions killed because a server comes up are logged
12299 with an 'U' termination code (for "Up").
12300
12301 Actions are disabled by default
12302
Olivier Houchard006e3102018-12-10 18:30:32 +010012303pool-max-conn <max>
12304 Set the maximum number of idling connections for a server. -1 means unlimited
12305 connections, 0 means no idle connections. The default is -1. When idle
12306 connections are enabled, orphaned idle connections which do not belong to any
12307 client session anymore are moved to a dedicated pool so that they remain
12308 usable by future clients. This only applies to connections that can be shared
12309 according to the same principles as those applying to "http-reuse".
12310
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012311pool-purge-delay <delay>
12312 Sets the delay to start purging idle connections. Each <delay> interval, half
Olivier Houcharda56eebf2019-03-19 16:44:02 +010012313 of the idle connections are closed. 0 means we don't keep any idle connection.
Willy Tarreaufb553652019-06-04 14:06:31 +020012314 The default is 5s.
Olivier Houchardb7b3faa2018-12-14 18:15:36 +010012315
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012316port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012317 Using the "port" parameter, it becomes possible to use a different port to
12318 send health-checks. On some servers, it may be desirable to dedicate a port
12319 to a specific component able to perform complex tests which are more suitable
12320 to health-checks than the application. It is common to run a simple script in
12321 inetd for instance. This parameter is ignored if the "check" parameter is not
12322 set. See also the "addr" parameter.
12323
Christopher Faulet8ed0a3e2018-04-10 14:45:45 +020012324proto <name>
12325
12326 Forces the multiplexer's protocol to use for the outgoing connections to this
12327 server. It must be compatible with the mode of the backend (TCP or HTTP). It
12328 must also be usable on the backend side. The list of available protocols is
12329 reported in haproxy -vv.
12330 Idea behind this optipon is to bypass the selection of the best multiplexer's
12331 protocol for all connections established to this server.
12332
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012333redir <prefix>
12334 The "redir" parameter enables the redirection mode for all GET and HEAD
12335 requests addressing this server. This means that instead of having HAProxy
12336 forward the request to the server, it will send an "HTTP 302" response with
12337 the "Location" header composed of this prefix immediately followed by the
12338 requested URI beginning at the leading '/' of the path component. That means
12339 that no trailing slash should be used after <prefix>. All invalid requests
12340 will be rejected, and all non-GET or HEAD requests will be normally served by
12341 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012342 mangling nor cookie insertion is possible in the response. However, cookies in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012343 requests are still analyzed, making this solution completely usable to direct
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012344 users to a remote location in case of local disaster. Main use consists in
12345 increasing bandwidth for static servers by having the clients directly
12346 connect to them. Note: never use a relative location here, it would cause a
12347 loop between the client and HAProxy!
12348
12349 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
12350
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012351rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012352 The "rise" parameter states that a server will be considered as operational
12353 after <count> consecutive successful health checks. This value defaults to 2
12354 if unspecified. See also the "check", "inter" and "fall" parameters.
12355
Baptiste Assmann8e2d9432018-06-22 15:04:43 +020012356resolve-opts <option>,<option>,...
12357 Comma separated list of options to apply to DNS resolution linked to this
12358 server.
12359
12360 Available options:
12361
12362 * allow-dup-ip
12363 By default, HAProxy prevents IP address duplication in a backend when DNS
12364 resolution at runtime is in operation.
12365 That said, for some cases, it makes sense that two servers (in the same
12366 backend, being resolved by the same FQDN) have the same IP address.
12367 For such case, simply enable this option.
12368 This is the opposite of prevent-dup-ip.
12369
12370 * prevent-dup-ip
12371 Ensure HAProxy's default behavior is enforced on a server: prevent re-using
12372 an IP address already set to a server in the same backend and sharing the
12373 same fqdn.
12374 This is the opposite of allow-dup-ip.
12375
12376 Example:
12377 backend b_myapp
12378 default-server init-addr none resolvers dns
12379 server s1 myapp.example.com:80 check resolve-opts allow-dup-ip
12380 server s2 myapp.example.com:81 check resolve-opts allow-dup-ip
12381
12382 With the option allow-dup-ip set:
12383 * if the nameserver returns a single IP address, then both servers will use
12384 it
12385 * If the nameserver returns 2 IP addresses, then each server will pick up a
12386 different address
12387
12388 Default value: not set
12389
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012390resolve-prefer <family>
12391 When DNS resolution is enabled for a server and multiple IP addresses from
12392 different families are returned, HAProxy will prefer using an IP address
12393 from the family mentioned in the "resolve-prefer" parameter.
12394 Available families: "ipv4" and "ipv6"
12395
Baptiste Assmannc4aabae2015-08-04 22:43:06 +020012396 Default value: ipv6
12397
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012398 Example:
12399
12400 server s1 app1.domain.com:80 resolvers mydns resolve-prefer ipv6
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012401
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012402resolve-net <network>[,<network[,...]]
12403 This options prioritize th choice of an ip address matching a network. This is
12404 useful with clouds to prefer a local ip. In some cases, a cloud high
Tim Düsterhus4896c442016-11-29 02:15:19 +010012405 availability service can be announced with many ip addresses on many
Davor Ocelice9ed2812017-12-25 17:49:28 +010012406 different datacenters. The latency between datacenter is not negligible, so
12407 this patch permits to prefer a local datacenter. If no address matches the
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012408 configured network, another address is selected.
12409
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012410 Example:
12411
12412 server s1 app1.domain.com:80 resolvers mydns resolve-net 10.0.0.0/8
Thierry Fournierac88cfe2016-02-17 22:05:30 +010012413
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012414resolvers <id>
12415 Points to an existing "resolvers" section to resolve current server's
12416 hostname.
12417
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012418 Example:
12419
12420 server s1 app1.domain.com:80 check resolvers mydns
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012421
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012422 See also section 5.3
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012423
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012424send-proxy
12425 The "send-proxy" parameter enforces use of the PROXY protocol over any
12426 connection established to this server. The PROXY protocol informs the other
12427 end about the layer 3/4 addresses of the incoming connection, so that it can
12428 know the client's address or the public address it accessed to, whatever the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010012429 upper layer protocol. For connections accepted by an "accept-proxy" or
12430 "accept-netscaler-cip" listener, the advertised address will be used. Only
12431 TCPv4 and TCPv6 address families are supported. Other families such as
12432 Unix sockets, will report an UNKNOWN family. Servers using this option can
12433 fully be chained to another instance of haproxy listening with an
12434 "accept-proxy" setting. This setting must not be used if the server isn't
12435 aware of the protocol. When health checks are sent to the server, the PROXY
12436 protocol is automatically used when this option is set, unless there is an
12437 explicit "port" or "addr" directive, in which case an explicit
12438 "check-send-proxy" directive would also be needed to use the PROXY protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012439 See also the "no-send-proxy" option of this section and "accept-proxy" and
12440 "accept-netscaler-cip" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +010012441
David Safb76832014-05-08 23:42:08 -040012442send-proxy-v2
12443 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
12444 over any connection established to this server. The PROXY protocol informs
12445 the other end about the layer 3/4 addresses of the incoming connection, so
12446 that it can know the client's address or the public address it accessed to,
Emmanuel Hocdet404d9782017-10-24 10:55:14 +020012447 whatever the upper layer protocol. It also send ALPN information if an alpn
12448 have been negotiated. This setting must not be used if the server isn't aware
12449 of this version of the protocol. See also the "no-send-proxy-v2" option of
12450 this section and send-proxy" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012451
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012452proxy-v2-options <option>[,<option>]*
12453 The "proxy-v2-options" parameter add option to send in PROXY protocol version
12454 2 when "send-proxy-v2" is used. Options available are "ssl" (see also
Emmanuel Hocdetfa8d0f12018-02-01 15:53:52 +010012455 send-proxy-v2-ssl), "cert-cn" (see also "send-proxy-v2-ssl-cn"), "ssl-cipher":
12456 name of the used cipher, "cert-sig": signature algorithm of the used
Emmanuel Hocdet253c3b72018-02-01 18:29:59 +010012457 certificate, "cert-key": key algorithm of the used certificate), "authority":
12458 host name value passed by the client (only sni from a tls connection is
Emmanuel Hocdet4399c752018-02-05 15:26:43 +010012459 supported), "crc32c": checksum of the proxy protocol v2 header.
Emmanuel Hocdetf643b802018-02-01 15:20:32 +010012460
David Safb76832014-05-08 23:42:08 -040012461send-proxy-v2-ssl
12462 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12463 2 over any connection established to this server. The PROXY protocol informs
12464 the other end about the layer 3/4 addresses of the incoming connection, so
12465 that it can know the client's address or the public address it accessed to,
12466 whatever the upper layer protocol. In addition, the SSL information extension
12467 of the PROXY protocol is added to the PROXY protocol header. This setting
12468 must not be used if the server isn't aware of this version of the protocol.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012469 See also the "no-send-proxy-v2-ssl" option of this section and the
12470 "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012471
12472send-proxy-v2-ssl-cn
12473 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
12474 2 over any connection established to this server. The PROXY protocol informs
12475 the other end about the layer 3/4 addresses of the incoming connection, so
12476 that it can know the client's address or the public address it accessed to,
12477 whatever the upper layer protocol. In addition, the SSL information extension
12478 of the PROXY protocol, along along with the Common Name from the subject of
12479 the client certificate (if any), is added to the PROXY protocol header. This
12480 setting must not be used if the server isn't aware of this version of the
Davor Ocelice9ed2812017-12-25 17:49:28 +010012481 protocol. See also the "no-send-proxy-v2-ssl-cn" option of this section and
12482 the "send-proxy-v2" option of the "bind" keyword.
David Safb76832014-05-08 23:42:08 -040012483
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012484slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012485 The "slowstart" parameter for a server accepts a value in milliseconds which
12486 indicates after how long a server which has just come back up will run at
12487 full speed. Just as with every other time-based parameter, it can be entered
12488 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
12489 linearly from 0 to 100% during this time. The limitation applies to two
12490 parameters :
12491
12492 - maxconn: the number of connections accepted by the server will grow from 1
12493 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
12494
12495 - weight: when the backend uses a dynamic weighted algorithm, the weight
12496 grows linearly from 1 to 100%. In this case, the weight is updated at every
12497 health-check. For this reason, it is important that the "inter" parameter
12498 is smaller than the "slowstart", in order to maximize the number of steps.
12499
12500 The slowstart never applies when haproxy starts, otherwise it would cause
12501 trouble to running servers. It only applies when a server has been previously
12502 seen as failed.
12503
Willy Tarreau732eac42015-07-09 11:40:25 +020012504sni <expression>
12505 The "sni" parameter evaluates the sample fetch expression, converts it to a
12506 string and uses the result as the host name sent in the SNI TLS extension to
12507 the server. A typical use case is to send the SNI received from the client in
12508 a bridged HTTPS scenario, using the "ssl_fc_sni" sample fetch for the
Willy Tarreau2ab88672017-07-05 18:23:03 +020012509 expression, though alternatives such as req.hdr(host) can also make sense. If
12510 "verify required" is set (which is the recommended setting), the resulting
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012511 name will also be matched against the server certificate's names. See the
Jérôme Magninb36a6d22018-12-09 16:03:40 +010012512 "verify" directive for more details. If you want to set a SNI for health
12513 checks, see the "check-sni" directive for more details.
Willy Tarreau732eac42015-07-09 11:40:25 +020012514
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012515source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +020012516source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012517source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012518 The "source" parameter sets the source address which will be used when
12519 connecting to the server. It follows the exact same parameters and principle
12520 as the backend "source" keyword, except that it only applies to the server
12521 referencing it. Please consult the "source" keyword for details.
12522
Willy Tarreauc6f4ce82009-06-10 11:09:37 +020012523 Additionally, the "source" statement on a server line allows one to specify a
12524 source port range by indicating the lower and higher bounds delimited by a
12525 dash ('-'). Some operating systems might require a valid IP address when a
12526 source port range is specified. It is permitted to have the same IP/range for
12527 several servers. Doing so makes it possible to bypass the maximum of 64k
12528 total concurrent connections. The limit will then reach 64k connections per
12529 server.
12530
Lukas Tribus7d56c6d2016-09-13 09:51:15 +000012531 Since Linux 4.2/libc 2.23 IP_BIND_ADDRESS_NO_PORT is set for connections
12532 specifying the source address without port(s).
12533
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012534ssl
Willy Tarreau44f65392013-06-25 07:56:20 +020012535 This option enables SSL ciphering on outgoing connections to the server. It
12536 is critical to verify server certificates using "verify" when using SSL to
12537 connect to servers, otherwise the communication is prone to trivial man in
12538 the-middle attacks rendering SSL useless. When this option is used, health
12539 checks are automatically sent in SSL too unless there is a "port" or an
12540 "addr" directive indicating the check should be sent to a different location.
Frédéric Lécailled2376272017-03-21 18:52:12 +010012541 See the "no-ssl" to disable "ssl" option and "check-ssl" option to force
12542 SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +020012543
Emmanuel Hocdete1c722b2017-03-31 15:02:54 +020012544ssl-max-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12545 This option enforces use of <version> or lower when SSL is used to communicate
12546 with the server. This option is also available on global statement
12547 "ssl-default-server-options". See also "ssl-min-ver".
12548
12549ssl-min-ver [ SSLv3 | TLSv1.0 | TLSv1.1 | TLSv1.2 | TLSv1.3 ]
12550 This option enforces use of <version> or upper when SSL is used to communicate
12551 with the server. This option is also available on global statement
12552 "ssl-default-server-options". See also "ssl-max-ver".
12553
Frédéric Lécailled2376272017-03-21 18:52:12 +010012554ssl-reuse
12555 This option may be used as "server" setting to reset any "no-ssl-reuse"
12556 setting which would have been inherited from "default-server" directive as
12557 default value.
12558 It may also be used as "default-server" setting to reset any previous
12559 "default-server" "no-ssl-reuse" setting.
12560
12561stick
12562 This option may be used as "server" setting to reset any "non-stick"
12563 setting which would have been inherited from "default-server" directive as
12564 default value.
12565 It may also be used as "default-server" setting to reset any previous
12566 "default-server" "non-stick" setting.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +020012567
Alexander Liu2a54bb72019-05-22 19:44:48 +080012568socks4 <addr>:<port>
12569 This option enables upstream socks4 tunnel for outgoinng connections to the
12570 server. Using this option won't force the health check to go via socks4 by
12571 default. You will have to use the keyword "check-via-socks4" to enable it.
12572
Willy Tarreau163d4622015-10-13 16:16:41 +020012573tcp-ut <delay>
12574 Sets the TCP User Timeout for all outgoing connections to this server. This
12575 option is available on Linux since version 2.6.37. It allows haproxy to
12576 configure a timeout for sockets which contain data not receiving an
Davor Ocelice9ed2812017-12-25 17:49:28 +010012577 acknowledgment for the configured delay. This is especially useful on
Willy Tarreau163d4622015-10-13 16:16:41 +020012578 long-lived connections experiencing long idle periods such as remote
12579 terminals or database connection pools, where the client and server timeouts
12580 must remain high to allow a long period of idle, but where it is important to
12581 detect that the server has disappeared in order to release all resources
12582 associated with its connection (and the client's session). One typical use
12583 case is also to force dead server connections to die when health checks are
12584 too slow or during a soft reload since health checks are then disabled. The
12585 argument is a delay expressed in milliseconds by default. This only works for
12586 regular TCP connections, and is ignored for other protocols.
12587
Willy Tarreau034c88c2017-01-23 23:36:45 +010012588tfo
12589 This option enables using TCP fast open when connecting to servers, on
12590 systems that support it (currently only the Linux kernel >= 4.11).
12591 See the "tfo" bind option for more information about TCP fast open.
12592 Please note that when using tfo, you should also use the "conn-failure",
12593 "empty-response" and "response-timeout" keywords for "retry-on", or haproxy
12594 won't be able to retry the connection on failure.
12595
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012596track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +020012597 This option enables ability to set the current state of the server by tracking
12598 another one. It is possible to track a server which itself tracks another
12599 server, provided that at the end of the chain, a server has health checks
12600 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012601 used, it has to be enabled on both proxies.
12602
Frédéric Lécailled2376272017-03-21 18:52:12 +010012603tls-tickets
12604 This option may be used as "server" setting to reset any "no-tls-tickets"
12605 setting which would have been inherited from "default-server" directive as
12606 default value.
12607 It may also be used as "default-server" setting to reset any previous
12608 "default-server" "no-tlsv-tickets" setting.
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012609
Emeric Brunef42d922012-10-11 16:11:36 +020012610verify [none|required]
12611 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +010012612 to 'none', server certificate is not verified. In the other case, The
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012613 certificate provided by the server is verified using CAs from 'ca-file' and
12614 optional CRLs from 'crl-file' after having checked that the names provided in
Davor Ocelice9ed2812017-12-25 17:49:28 +010012615 the certificate's subject and subjectAlternateNames attributes match either
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012616 the name passed using the "sni" directive, or if not provided, the static
12617 host name passed using the "verifyhost" directive. When no name is found, the
12618 certificate's names are ignored. For this reason, without SNI it's important
12619 to use "verifyhost". On verification failure the handshake is aborted. It is
12620 critically important to verify server certificates when using SSL to connect
12621 to servers, otherwise the communication is prone to trivial man-in-the-middle
12622 attacks rendering SSL totally useless. Unless "ssl_server_verify" appears in
12623 the global section, "verify" is set to "required" by default.
Emeric Brunef42d922012-10-11 16:11:36 +020012624
Evan Broderbe554312013-06-27 00:05:25 -070012625verifyhost <hostname>
12626 This setting is only available when support for OpenSSL was built in, and
Willy Tarreauad92a9a2017-07-28 11:38:41 +020012627 only takes effect if 'verify required' is also specified. This directive sets
12628 a default static hostname to check the server's certificate against when no
12629 SNI was used to connect to the server. If SNI is not used, this is the only
12630 way to enable hostname verification. This static hostname, when set, will
12631 also be used for health checks (which cannot provide an SNI value). If none
12632 of the hostnames in the certificate match the specified hostname, the
12633 handshake is aborted. The hostnames in the server-provided certificate may
12634 include wildcards. See also "verify", "sni" and "no-verifyhost" options.
Evan Broderbe554312013-06-27 00:05:25 -070012635
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +010012636weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012637 The "weight" parameter is used to adjust the server's weight relative to
12638 other servers. All servers will receive a load proportional to their weight
12639 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +020012640 load. The default weight is 1, and the maximal value is 256. A value of 0
12641 means the server will not participate in load-balancing but will still accept
12642 persistent connections. If this parameter is used to distribute the load
12643 according to server's capacity, it is recommended to start with values which
12644 can both grow and shrink, for instance between 10 and 100 to leave enough
12645 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012646
12647
Cyril Bonté46175dd2015-07-02 22:45:32 +0200126485.3. Server IP address resolution using DNS
12649-------------------------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012650
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012651HAProxy allows using a host name on the server line to retrieve its IP address
12652using name servers. By default, HAProxy resolves the name when parsing the
12653configuration file, at startup and cache the result for the process' life.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012654This is not sufficient in some cases, such as in Amazon where a server's IP
12655can change after a reboot or an ELB Virtual IP can change based on current
12656workload.
12657This chapter describes how HAProxy can be configured to process server's name
12658resolution at run time.
12659Whether run time server name resolution has been enable or not, HAProxy will
12660carry on doing the first resolution when parsing the configuration.
12661
12662
Cyril Bonté46175dd2015-07-02 22:45:32 +0200126635.3.1. Global overview
12664----------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012665
12666As we've seen in introduction, name resolution in HAProxy occurs at two
12667different steps of the process life:
12668
12669 1. when starting up, HAProxy parses the server line definition and matches a
12670 host name. It uses libc functions to get the host name resolved. This
12671 resolution relies on /etc/resolv.conf file.
12672
Christopher Faulet67957bd2017-09-27 11:00:59 +020012673 2. at run time, HAProxy performs periodically name resolutions for servers
12674 requiring DNS resolutions.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012675
12676A few other events can trigger a name resolution at run time:
12677 - when a server's health check ends up in a connection timeout: this may be
12678 because the server has a new IP address. So we need to trigger a name
12679 resolution to know this new IP.
12680
Christopher Faulet67957bd2017-09-27 11:00:59 +020012681When using resolvers, the server name can either be a hostname, or a SRV label.
Davor Ocelice9ed2812017-12-25 17:49:28 +010012682HAProxy considers anything that starts with an underscore as a SRV label. If a
Christopher Faulet67957bd2017-09-27 11:00:59 +020012683SRV label is specified, then the corresponding SRV records will be retrieved
12684from the DNS server, and the provided hostnames will be used. The SRV label
12685will be checked periodically, and if any server are added or removed, haproxy
12686will automatically do the same.
Olivier Houchardecfa18d2017-08-07 17:30:03 +020012687
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012688A few things important to notice:
12689 - all the name servers are queried in the mean time. HAProxy will process the
12690 first valid response.
12691
12692 - a resolution is considered as invalid (NX, timeout, refused), when all the
12693 servers return an error.
12694
12695
Cyril Bonté46175dd2015-07-02 22:45:32 +0200126965.3.2. The resolvers section
12697----------------------------
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012698
12699This section is dedicated to host information related to name resolution in
Christopher Faulet67957bd2017-09-27 11:00:59 +020012700HAProxy. There can be as many as resolvers section as needed. Each section can
12701contain many name servers.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012702
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012703When multiple name servers are configured in a resolvers section, then HAProxy
12704uses the first valid response. In case of invalid responses, only the last one
12705is treated. Purpose is to give the chance to a slow server to deliver a valid
12706answer after a fast faulty or outdated server.
12707
12708When each server returns a different error type, then only the last error is
Christopher Faulet67957bd2017-09-27 11:00:59 +020012709used by HAProxy. The following processing is applied on this error:
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012710
Christopher Faulet67957bd2017-09-27 11:00:59 +020012711 1. HAProxy retries the same DNS query with a new query type. The A queries are
12712 switch to AAAA or the opposite. SRV queries are not concerned here. Timeout
12713 errors are also excluded.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012714
Christopher Faulet67957bd2017-09-27 11:00:59 +020012715 2. When the fallback on the query type was done (or not applicable), HAProxy
12716 retries the original DNS query, with the preferred query type.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012717
Christopher Faulet67957bd2017-09-27 11:00:59 +020012718 3. HAProxy retries previous steps <resolve_retires> times. If no valid
12719 response is received after that, it stops the DNS resolution and reports
12720 the error.
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012721
Christopher Faulet67957bd2017-09-27 11:00:59 +020012722For example, with 2 name servers configured in a resolvers section, the
12723following scenarios are possible:
12724
12725 - First response is valid and is applied directly, second response is
12726 ignored
12727
12728 - First response is invalid and second one is valid, then second response is
12729 applied
12730
12731 - First response is a NX domain and second one a truncated response, then
12732 HAProxy retries the query with a new type
12733
12734 - First response is a NX domain and second one is a timeout, then HAProxy
12735 retries the query with a new type
12736
12737 - Query timed out for both name servers, then HAProxy retries it with the
12738 same query type
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012739
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012740As a DNS server may not answer all the IPs in one DNS request, haproxy keeps
12741a cache of previous answers, an answer will be considered obsolete after
Christopher Faulet67957bd2017-09-27 11:00:59 +020012742<hold obsolete> seconds without the IP returned.
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012743
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012744
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012745resolvers <resolvers id>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012746 Creates a new name server list labeled <resolvers id>
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012747
12748A resolvers section accept the following parameters:
12749
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012750accepted_payload_size <nb>
Davor Ocelice9ed2812017-12-25 17:49:28 +010012751 Defines the maximum payload size accepted by HAProxy and announced to all the
Christopher Faulet67957bd2017-09-27 11:00:59 +020012752 name servers configured in this resolvers section.
Baptiste Assmann2af08fe2017-08-14 00:13:01 +020012753 <nb> is in bytes. If not set, HAProxy announces 512. (minimal value defined
12754 by RFC 6891)
12755
Baptiste Assmann9d8dbbc2017-08-18 23:35:08 +020012756 Note: the maximum allowed value is 8192.
12757
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012758nameserver <id> <ip>:<port>
12759 DNS server description:
12760 <id> : label of the server, should be unique
12761 <ip> : IP address of the server
12762 <port> : port where the DNS service actually runs
12763
Ben Draut44e609b2018-05-29 15:40:08 -060012764parse-resolv-conf
12765 Adds all nameservers found in /etc/resolv.conf to this resolvers nameservers
12766 list. Ordered as if each nameserver in /etc/resolv.conf was individually
12767 placed in the resolvers section in place of this directive.
12768
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012769hold <status> <period>
12770 Defines <period> during which the last name resolution should be kept based
12771 on last resolution <status>
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012772 <status> : last name resolution status. Acceptable values are "nx",
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012773 "other", "refused", "timeout", "valid", "obsolete".
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012774 <period> : interval between two successive name resolution when the last
12775 answer was in <status>. It follows the HAProxy time format.
12776 <period> is in milliseconds by default.
12777
Baptiste Assmann686408b2017-08-18 10:15:42 +020012778 Default value is 10s for "valid", 0s for "obsolete" and 30s for others.
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012779
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012780resolve_retries <nb>
12781 Defines the number <nb> of queries to send to resolve a server name before
12782 giving up.
12783 Default value: 3
12784
Baptiste Assmann62b75b42015-09-09 01:11:36 +020012785 A retry occurs on name server timeout or when the full sequence of DNS query
12786 type failover is over and we need to start up from the default ANY query
12787 type.
12788
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012789timeout <event> <time>
12790 Defines timeouts related to name resolution
12791 <event> : the event on which the <time> timeout period applies to.
12792 events available are:
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012793 - resolve : default time to trigger name resolutions when no
12794 other time applied.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012795 Default value: 1s
12796 - retry : time between two DNS queries, when no valid response
Frédéric Lécaille93d33162019-03-06 09:35:59 +010012797 have been received.
Christopher Faulet67957bd2017-09-27 11:00:59 +020012798 Default value: 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012799 <time> : time related to the event. It follows the HAProxy time format.
12800 <time> is expressed in milliseconds.
12801
Olivier Doucetaa1ea8a2016-08-05 17:15:20 +020012802 Example:
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012803
12804 resolvers mydns
12805 nameserver dns1 10.0.0.1:53
12806 nameserver dns2 10.0.0.2:53
Ben Draut44e609b2018-05-29 15:40:08 -060012807 parse-resolv-conf
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012808 resolve_retries 3
Christopher Faulet67957bd2017-09-27 11:00:59 +020012809 timeout resolve 1s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012810 timeout retry 1s
Baptiste Assmann987e16d2016-11-02 22:23:31 +010012811 hold other 30s
12812 hold refused 30s
12813 hold nx 30s
12814 hold timeout 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012815 hold valid 10s
Olivier Houcharda8c6db82017-07-06 18:46:47 +020012816 hold obsolete 30s
Baptiste Assmann1fa66662015-04-14 00:28:47 +020012817
12818
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200128196. HTTP header manipulation
12820---------------------------
12821
12822In HTTP mode, it is possible to rewrite, add or delete some of the request and
12823response headers based on regular expressions. It is also possible to block a
12824request or a response if a particular header matches a regular expression,
12825which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +010012826against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012827
Willy Tarreau70dffda2014-01-30 03:07:23 +010012828If HAProxy encounters an "Informational Response" (status code 1xx), it is able
12829to process all rsp* rules which can allow, deny, rewrite or delete a header,
12830but it will refuse to add a header to any such messages as this is not
12831HTTP-compliant. The reason for still processing headers in such responses is to
12832stop and/or fix any possible information leak which may happen, for instance
12833because another downstream equipment would unconditionally add a header, or if
12834a server name appears there. When such messages are seen, normal processing
12835still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +020012836
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012837This section covers common usage of the following keywords, described in detail
12838in section 4.2 :
12839
12840 - reqadd <string>
12841 - reqallow <search>
12842 - reqiallow <search>
12843 - reqdel <search>
12844 - reqidel <search>
12845 - reqdeny <search>
12846 - reqideny <search>
12847 - reqpass <search>
12848 - reqipass <search>
12849 - reqrep <search> <replace>
12850 - reqirep <search> <replace>
12851 - reqtarpit <search>
12852 - reqitarpit <search>
12853 - rspadd <string>
12854 - rspdel <search>
12855 - rspidel <search>
12856 - rspdeny <search>
12857 - rspideny <search>
12858 - rsprep <search> <replace>
12859 - rspirep <search> <replace>
12860
12861With all these keywords, the same conventions are used. The <search> parameter
12862is a POSIX extended regular expression (regex) which supports grouping through
12863parenthesis (without the backslash). Spaces and other delimiters must be
12864prefixed with a backslash ('\') to avoid confusion with a field delimiter.
12865Other characters may be prefixed with a backslash to change their meaning :
12866
12867 \t for a tab
12868 \r for a carriage return (CR)
12869 \n for a new line (LF)
12870 \ to mark a space and differentiate it from a delimiter
12871 \# to mark a sharp and differentiate it from a comment
12872 \\ to use a backslash in a regex
12873 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
12874 \xXX to write the ASCII hex code XX as in the C language
12875
12876The <replace> parameter contains the string to be used to replace the largest
12877portion of text matching the regex. It can make use of the special characters
12878above, and can reference a substring which is delimited by parenthesis in the
12879regex, by writing a backslash ('\') immediately followed by one digit from 0 to
128809 indicating the group position (0 designating the entire line). This practice
12881is very common to users of the "sed" program.
12882
12883The <string> parameter represents the string which will systematically be added
12884after the last header line. It can also use special character sequences above.
12885
12886Notes related to these keywords :
12887---------------------------------
12888 - these keywords are not always convenient to allow/deny based on header
12889 contents. It is strongly recommended to use ACLs with the "block" keyword
12890 instead, resulting in far more flexible and manageable rules.
12891
12892 - lines are always considered as a whole. It is not possible to reference
12893 a header name only or a value only. This is important because of the way
12894 headers are written (notably the number of spaces after the colon).
12895
12896 - the first line is always considered as a header, which makes it possible to
12897 rewrite or filter HTTP requests URIs or response codes, but in turn makes
12898 it harder to distinguish between headers and request line. The regex prefix
12899 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
12900 ^[^ \t:]*: matches any header name followed by a colon.
12901
12902 - for performances reasons, the number of characters added to a request or to
12903 a response is limited at build time to values between 1 and 4 kB. This
12904 should normally be far more than enough for most usages. If it is too short
12905 on occasional usages, it is possible to gain some space by removing some
12906 useless headers before adding new ones.
12907
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012908 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012909 without the 'i' letter except that they ignore case when matching patterns.
12910
12911 - when a request passes through a frontend then a backend, all req* rules
12912 from the frontend will be evaluated, then all req* rules from the backend
12913 will be evaluated. The reverse path is applied to responses.
12914
12915 - req* statements are applied after "block" statements, so that "block" is
12916 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +010012917 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012918
12919
Willy Tarreau74ca5042013-06-11 23:12:07 +0200129207. Using ACLs and fetching samples
12921----------------------------------
12922
Davor Ocelice9ed2812017-12-25 17:49:28 +010012923HAProxy is capable of extracting data from request or response streams, from
Willy Tarreau74ca5042013-06-11 23:12:07 +020012924client or server information, from tables, environmental information etc...
12925The action of extracting such data is called fetching a sample. Once retrieved,
12926these samples may be used for various purposes such as a key to a stick-table,
12927but most common usages consist in matching them against predefined constant
12928data called patterns.
12929
12930
129317.1. ACL basics
12932---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012933
12934The use of Access Control Lists (ACL) provides a flexible solution to perform
12935content switching and generally to take decisions based on content extracted
12936from the request, the response or any environmental status. The principle is
12937simple :
12938
Willy Tarreau74ca5042013-06-11 23:12:07 +020012939 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012940 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +020012941 - apply one or multiple pattern matching methods on this sample
12942 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012943
Willy Tarreau74ca5042013-06-11 23:12:07 +020012944The actions generally consist in blocking a request, selecting a backend, or
12945adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012946
12947In order to define a test, the "acl" keyword is used. The syntax is :
12948
Willy Tarreau74ca5042013-06-11 23:12:07 +020012949 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012950
12951This creates a new ACL <aclname> or completes an existing one with new tests.
12952Those tests apply to the portion of request/response specified in <criterion>
12953and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012954an operator which may be specified before the set of values. Optionally some
12955conversion operators may be applied to the sample, and they will be specified
12956as a comma-delimited list of keywords just after the first keyword. The values
12957are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012958
12959ACL names must be formed from upper and lower case letters, digits, '-' (dash),
12960'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
12961which means that "my_acl" and "My_Acl" are two different ACLs.
12962
12963There is no enforced limit to the number of ACLs. The unused ones do not affect
12964performance, they just consume a small amount of memory.
12965
Willy Tarreau74ca5042013-06-11 23:12:07 +020012966The criterion generally is the name of a sample fetch method, or one of its ACL
12967specific declinations. The default test method is implied by the output type of
12968this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012969methods of a same sample fetch method. The sample fetch methods are the only
12970ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012971
12972Sample fetch methods return data which can be of the following types :
12973 - boolean
12974 - integer (signed or unsigned)
12975 - IPv4 or IPv6 address
12976 - string
12977 - data block
12978
Willy Tarreaue6b11e42013-11-26 19:02:32 +010012979Converters transform any of these data into any of these. For example, some
12980converters might convert a string to a lower-case string while other ones
12981would turn a string to an IPv4 address, or apply a netmask to an IP address.
12982The resulting sample is of the type of the last converter applied to the list,
12983which defaults to the type of the sample fetch method.
12984
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020012985Each sample or converter returns data of a specific type, specified with its
12986keyword in this documentation. When an ACL is declared using a standard sample
12987fetch method, certain types automatically involved a default matching method
12988which are summarized in the table below :
12989
12990 +---------------------+-----------------+
12991 | Sample or converter | Default |
12992 | output type | matching method |
12993 +---------------------+-----------------+
12994 | boolean | bool |
12995 +---------------------+-----------------+
12996 | integer | int |
12997 +---------------------+-----------------+
12998 | ip | ip |
12999 +---------------------+-----------------+
13000 | string | str |
13001 +---------------------+-----------------+
13002 | binary | none, use "-m" |
13003 +---------------------+-----------------+
13004
13005Note that in order to match a binary samples, it is mandatory to specify a
13006matching method, see below.
13007
Willy Tarreau74ca5042013-06-11 23:12:07 +020013008The ACL engine can match these types against patterns of the following types :
13009 - boolean
13010 - integer or integer range
13011 - IP address / network
13012 - string (exact, substring, suffix, prefix, subdir, domain)
13013 - regular expression
13014 - hex block
13015
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013016The following ACL flags are currently supported :
13017
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013018 -i : ignore case during matching of all subsequent patterns.
13019 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013020 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013021 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013022 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013023 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +020013024 -- : force end of flags. Useful when a string looks like one of the flags.
13025
Willy Tarreau74ca5042013-06-11 23:12:07 +020013026The "-f" flag is followed by the name of a file from which all lines will be
13027read as individual values. It is even possible to pass multiple "-f" arguments
13028if the patterns are to be loaded from multiple files. Empty lines as well as
13029lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
13030will be stripped. If it is absolutely necessary to insert a valid pattern
13031beginning with a sharp, just prefix it with a space so that it is not taken for
13032a comment. Depending on the data type and match method, haproxy may load the
13033lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
13034exact string matching. In this case, duplicates will automatically be removed.
13035
Thierry FOURNIER9860c412014-01-29 14:23:29 +010013036The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
13037parsed as two column file. The first column contains the patterns used by the
13038ACL, and the second column contain the samples. The sample can be used later by
13039a map. This can be useful in some rare cases where an ACL would just be used to
13040check for the existence of a pattern in a map before a mapping is applied.
13041
Thierry FOURNIER3534d882014-01-20 17:01:44 +010013042The "-u" flag forces the unique id of the ACL. This unique id is used with the
13043socket interface to identify ACL and dynamically change its values. Note that a
13044file is always identified by its name even if an id is set.
13045
Willy Tarreau74ca5042013-06-11 23:12:07 +020013046Also, note that the "-i" flag applies to subsequent entries and not to entries
13047loaded from files preceding it. For instance :
13048
13049 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
13050
13051In this example, each line of "exact-ua.lst" will be exactly matched against
13052the "user-agent" header of the request. Then each line of "generic-ua" will be
13053case-insensitively matched. Then the word "test" will be insensitively matched
13054as well.
13055
13056The "-m" flag is used to select a specific pattern matching method on the input
13057sample. All ACL-specific criteria imply a pattern matching method and generally
13058do not need this flag. However, this flag is useful with generic sample fetch
13059methods to describe how they're going to be matched against the patterns. This
13060is required for sample fetches which return data type for which there is no
Davor Ocelice9ed2812017-12-25 17:49:28 +010013061obvious matching method (e.g. string or binary). When "-m" is specified and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013062followed by a pattern matching method name, this method is used instead of the
13063default one for the criterion. This makes it possible to match contents in ways
13064that were not initially planned, or with sample fetch methods which return a
13065string. The matching method also affects the way the patterns are parsed.
13066
Thierry FOURNIERb7729c92014-02-11 16:24:41 +010013067The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
13068By default, if the parser cannot parse ip address it considers that the parsed
13069string is maybe a domain name and try dns resolution. The flag "-n" disable this
13070resolution. It is useful for detecting malformed ip lists. Note that if the DNS
13071server is not reachable, the haproxy configuration parsing may last many minutes
13072waiting fir the timeout. During this time no error messages are displayed. The
13073flag "-n" disable this behavior. Note also that during the runtime, this
13074function is disabled for the dynamic acl modifications.
13075
Willy Tarreau74ca5042013-06-11 23:12:07 +020013076There are some restrictions however. Not all methods can be used with all
13077sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
13078be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +020013079
13080 - "found" : only check if the requested sample could be found in the stream,
13081 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +020013082 to pass any pattern to avoid confusion. This matching method is
13083 particularly useful to detect presence of certain contents such
13084 as headers, cookies, etc... even if they are empty and without
13085 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013086
13087 - "bool" : check the value as a boolean. It can only be applied to fetches
13088 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013089 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013090
13091 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +020013092 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013093
13094 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +020013095 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013096
Davor Ocelice9ed2812017-12-25 17:49:28 +010013097 - "bin" : match the contents against a hexadecimal string representing a
Willy Tarreau5adeda12013-03-31 22:13:34 +020013098 binary sequence. This may be used with binary or string samples.
13099
13100 - "len" : match the sample's length as an integer. This may be used with
13101 binary or string samples.
13102
Willy Tarreau74ca5042013-06-11 23:12:07 +020013103 - "str" : exact match : match the contents against a string. This may be
13104 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013105
Willy Tarreau74ca5042013-06-11 23:12:07 +020013106 - "sub" : substring match : check that the contents contain at least one of
13107 the provided string patterns. This may be used with binary or
13108 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013109
Willy Tarreau74ca5042013-06-11 23:12:07 +020013110 - "reg" : regex match : match the contents against a list of regular
13111 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013112
Willy Tarreau74ca5042013-06-11 23:12:07 +020013113 - "beg" : prefix match : check that the contents begin like the provided
13114 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013115
Willy Tarreau74ca5042013-06-11 23:12:07 +020013116 - "end" : suffix match : check that the contents end like the provided
13117 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013118
Willy Tarreau74ca5042013-06-11 23:12:07 +020013119 - "dir" : subdir match : check that a slash-delimited portion of the
13120 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013121 This may be used with binary or string samples.
13122
Willy Tarreau74ca5042013-06-11 23:12:07 +020013123 - "dom" : domain match : check that a dot-delimited portion of the contents
13124 exactly match one of the provided string patterns. This may be
13125 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +020013126
13127For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
13128request, it is possible to do :
13129
13130 acl jsess_present cook(JSESSIONID) -m found
13131
13132In order to apply a regular expression on the 500 first bytes of data in the
13133buffer, one would use the following acl :
13134
13135 acl script_tag payload(0,500) -m reg -i <script>
13136
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013137On systems where the regex library is much slower when using "-i", it is
13138possible to convert the sample to lowercase before matching, like this :
13139
13140 acl script_tag payload(0,500),lower -m reg <script>
13141
Willy Tarreau74ca5042013-06-11 23:12:07 +020013142All ACL-specific criteria imply a default matching method. Most often, these
13143criteria are composed by concatenating the name of the original sample fetch
13144method and the matching method. For example, "hdr_beg" applies the "beg" match
13145to samples retrieved using the "hdr" fetch method. Since all ACL-specific
13146criteria rely on a sample fetch method, it is always possible instead to use
13147the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013148
Willy Tarreau74ca5042013-06-11 23:12:07 +020013149If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013150the matching method is simply applied to the underlying sample fetch method.
13151For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013152
Willy Tarreau74ca5042013-06-11 23:12:07 +020013153 acl short_form hdr_beg(host) www.
13154 acl alternate1 hdr_beg(host) -m beg www.
13155 acl alternate2 hdr_dom(host) -m beg www.
13156 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013157
Willy Tarreau2b5285d2010-05-09 23:45:24 +020013158
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013159The table below summarizes the compatibility matrix between sample or converter
13160types and the pattern types to fetch against. It indicates for each compatible
13161combination the name of the matching method to be used, surrounded with angle
13162brackets ">" and "<" when the method is the default one and will work by
13163default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +010013164
Willy Tarreau74ca5042013-06-11 23:12:07 +020013165 +-------------------------------------------------+
13166 | Input sample type |
13167 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013168 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013169 +----------------------+---------+---------+---------+---------+---------+
13170 | none (presence only) | found | found | found | found | found |
13171 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013172 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013173 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013174 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013175 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013176 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013177 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013178 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013179 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +020013180 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013181 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013182 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013183 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013184 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013185 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013186 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013187 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013188 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013189 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013190 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013191 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +010013192 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +020013193 +----------------------+---------+---------+---------+---------+---------+
13194 | hex block | | | | bin | bin |
13195 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +020013196
13197
Willy Tarreau74ca5042013-06-11 23:12:07 +0200131987.1.1. Matching booleans
13199------------------------
13200
13201In order to match a boolean, no value is needed and all values are ignored.
13202Boolean matching is used by default for all fetch methods of type "boolean".
13203When boolean matching is used, the fetched value is returned as-is, which means
13204that a boolean "true" will always match and a boolean "false" will never match.
13205
13206Boolean matching may also be enforced using "-m bool" on fetch methods which
13207return an integer value. Then, integer value 0 is converted to the boolean
13208"false" and all other values are converted to "true".
13209
Willy Tarreau6a06a402007-07-15 20:15:28 +020013210
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132117.1.2. Matching integers
13212------------------------
13213
13214Integer matching applies by default to integer fetch methods. It can also be
13215enforced on boolean fetches using "-m int". In this case, "false" is converted
13216to the integer 0, and "true" is converted to the integer 1.
13217
13218Integer matching also supports integer ranges and operators. Note that integer
13219matching only applies to positive values. A range is a value expressed with a
13220lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013221
13222For instance, "1024:65535" is a valid range to represent a range of
13223unprivileged ports, and "1024:" would also work. "0:1023" is a valid
13224representation of privileged ports, and ":1023" would also work.
13225
Willy Tarreau62644772008-07-16 18:36:06 +020013226As a special case, some ACL functions support decimal numbers which are in fact
13227two integers separated by a dot. This is used with some version checks for
13228instance. All integer properties apply to those decimal numbers, including
13229ranges and operators.
13230
Willy Tarreau6a06a402007-07-15 20:15:28 +020013231For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +010013232operators with ranges does not make much sense and is strongly discouraged.
13233Similarly, it does not make much sense to perform order comparisons with a set
13234of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013235
Willy Tarreau0ba27502007-12-24 16:55:16 +010013236Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013237
13238 eq : true if the tested value equals at least one value
13239 ge : true if the tested value is greater than or equal to at least one value
13240 gt : true if the tested value is greater than at least one value
13241 le : true if the tested value is less than or equal to at least one value
13242 lt : true if the tested value is less than at least one value
13243
Willy Tarreau0ba27502007-12-24 16:55:16 +010013244For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +020013245
13246 acl negative-length hdr_val(content-length) lt 0
13247
Willy Tarreau62644772008-07-16 18:36:06 +020013248This one matches SSL versions between 3.0 and 3.1 (inclusive) :
13249
13250 acl sslv3 req_ssl_ver 3:3.1
13251
Willy Tarreau6a06a402007-07-15 20:15:28 +020013252
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132537.1.3. Matching strings
13254-----------------------
13255
13256String matching applies to string or binary fetch methods, and exists in 6
13257different forms :
13258
13259 - exact match (-m str) : the extracted string must exactly match the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013260 patterns;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013261
13262 - substring match (-m sub) : the patterns are looked up inside the
Davor Ocelice9ed2812017-12-25 17:49:28 +010013263 extracted string, and the ACL matches if any of them is found inside;
Willy Tarreau74ca5042013-06-11 23:12:07 +020013264
13265 - prefix match (-m beg) : the patterns are compared with the beginning of
13266 the extracted string, and the ACL matches if any of them matches.
13267
13268 - suffix match (-m end) : the patterns are compared with the end of the
13269 extracted string, and the ACL matches if any of them matches.
13270
Baptiste Assmann33db6002016-03-06 23:32:10 +010013271 - subdir match (-m dir) : the patterns are looked up inside the extracted
Willy Tarreau74ca5042013-06-11 23:12:07 +020013272 string, delimited with slashes ("/"), and the ACL matches if any of them
13273 matches.
13274
13275 - domain match (-m dom) : the patterns are looked up inside the extracted
13276 string, delimited with dots ("."), and the ACL matches if any of them
13277 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013278
13279String matching applies to verbatim strings as they are passed, with the
13280exception of the backslash ("\") which makes it possible to escape some
13281characters such as the space. If the "-i" flag is passed before the first
13282string, then the matching will be performed ignoring the case. In order
13283to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +010013284before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013285
13286
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132877.1.4. Matching regular expressions (regexes)
13288---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013289
13290Just like with string matching, regex matching applies to verbatim strings as
13291they are passed, with the exception of the backslash ("\") which makes it
13292possible to escape some characters such as the space. If the "-i" flag is
13293passed before the first regex, then the matching will be performed ignoring
13294the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +010013295the "--" flag before the first string. Same principle applies of course to
13296match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +020013297
13298
Willy Tarreau74ca5042013-06-11 23:12:07 +0200132997.1.5. Matching arbitrary data blocks
13300-------------------------------------
13301
13302It is possible to match some extracted samples against a binary block which may
13303not safely be represented as a string. For this, the patterns must be passed as
13304a series of hexadecimal digits in an even number, when the match method is set
13305to binary. Each sequence of two digits will represent a byte. The hexadecimal
13306digits may be used upper or lower case.
13307
13308Example :
13309 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
13310 acl hello payload(0,6) -m bin 48656c6c6f0a
13311
13312
133137.1.6. Matching IPv4 and IPv6 addresses
13314---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020013315
13316IPv4 addresses values can be specified either as plain addresses or with a
13317netmask appended, in which case the IPv4 address matches whenever it is
13318within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010013319host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010013320difficult to read and debug configurations. If hostnames are used, you should
13321at least ensure that they are present in /etc/hosts so that the configuration
13322does not depend on any random DNS match at the moment the configuration is
13323parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020013324
Daniel Schnellereba56342016-04-13 00:26:52 +020013325The dotted IPv4 address notation is supported in both regular as well as the
13326abbreviated form with all-0-octets omitted:
13327
13328 +------------------+------------------+------------------+
13329 | Example 1 | Example 2 | Example 3 |
13330 +------------------+------------------+------------------+
13331 | 192.168.0.1 | 10.0.0.12 | 127.0.0.1 |
13332 | 192.168.1 | 10.12 | 127.1 |
13333 | 192.168.0.1/22 | 10.0.0.12/8 | 127.0.0.1/8 |
13334 | 192.168.1/22 | 10.12/8 | 127.1/8 |
13335 +------------------+------------------+------------------+
13336
13337Notice that this is different from RFC 4632 CIDR address notation in which
13338192.168.42/24 would be equivalent to 192.168.42.0/24.
13339
Willy Tarreauceb4ac92012-04-28 00:41:46 +020013340IPv6 may be entered in their usual form, with or without a netmask appended.
13341Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
13342trouble with randomly resolved IP addresses, host names are never allowed in
13343IPv6 patterns.
13344
13345HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
13346following situations :
13347 - tested address is IPv4, pattern address is IPv4, the match applies
13348 in IPv4 using the supplied mask if any.
13349 - tested address is IPv6, pattern address is IPv6, the match applies
13350 in IPv6 using the supplied mask if any.
13351 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
13352 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
13353 ::IPV4 or ::ffff:IPV4, otherwise it fails.
13354 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
13355 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
13356 applied in IPv6 using the supplied IPv6 mask.
13357
Willy Tarreau74ca5042013-06-11 23:12:07 +020013358
133597.2. Using ACLs to form conditions
13360----------------------------------
13361
13362Some actions are only performed upon a valid condition. A condition is a
13363combination of ACLs with operators. 3 operators are supported :
13364
13365 - AND (implicit)
13366 - OR (explicit with the "or" keyword or the "||" operator)
13367 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020013368
Willy Tarreau74ca5042013-06-11 23:12:07 +020013369A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020013370
Willy Tarreau74ca5042013-06-11 23:12:07 +020013371 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020013372
Willy Tarreau74ca5042013-06-11 23:12:07 +020013373Such conditions are generally used after an "if" or "unless" statement,
13374indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020013375
Willy Tarreau74ca5042013-06-11 23:12:07 +020013376For instance, to block HTTP requests to the "*" URL with methods other than
13377"OPTIONS", as well as POST requests without content-length, and GET or HEAD
13378requests with a content-length greater than 0, and finally every request which
13379is not either GET/HEAD/POST/OPTIONS !
13380
13381 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013382 http-request deny if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
13383 http-request deny if METH_GET HTTP_CONTENT
13384 http-request deny unless METH_GET or METH_POST or METH_OPTIONS
Willy Tarreau74ca5042013-06-11 23:12:07 +020013385
13386To select a different backend for requests to static contents on the "www" site
13387and to every request on the "img", "video", "download" and "ftp" hosts :
13388
13389 acl url_static path_beg /static /images /img /css
13390 acl url_static path_end .gif .png .jpg .css .js
13391 acl host_www hdr_beg(host) -i www
13392 acl host_static hdr_beg(host) -i img. video. download. ftp.
13393
Davor Ocelice9ed2812017-12-25 17:49:28 +010013394 # now use backend "static" for all static-only hosts, and for static URLs
Willy Tarreau74ca5042013-06-11 23:12:07 +020013395 # of host "www". Use backend "www" for the rest.
13396 use_backend static if host_static or host_www url_static
13397 use_backend www if host_www
13398
13399It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
13400expressions that are built on the fly without needing to be declared. They must
13401be enclosed between braces, with a space before and after each brace (because
13402the braces must be seen as independent words). Example :
13403
13404 The following rule :
13405
13406 acl missing_cl hdr_cnt(Content-length) eq 0
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013407 http-request deny if METH_POST missing_cl
Willy Tarreau74ca5042013-06-11 23:12:07 +020013408
13409 Can also be written that way :
13410
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013411 http-request deny if METH_POST { hdr_cnt(Content-length) eq 0 }
Willy Tarreau74ca5042013-06-11 23:12:07 +020013412
13413It is generally not recommended to use this construct because it's a lot easier
13414to leave errors in the configuration when written that way. However, for very
13415simple rules matching only one source IP address for instance, it can make more
13416sense to use them than to declare ACLs with random names. Another example of
13417good use is the following :
13418
13419 With named ACLs :
13420
13421 acl site_dead nbsrv(dynamic) lt 2
13422 acl site_dead nbsrv(static) lt 2
13423 monitor fail if site_dead
13424
13425 With anonymous ACLs :
13426
13427 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
13428
Jarno Huuskonen84c51ec2017-04-03 14:20:34 +030013429See section 4.2 for detailed help on the "http-request deny" and "use_backend"
13430keywords.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013431
13432
134337.3. Fetching samples
13434---------------------
13435
13436Historically, sample fetch methods were only used to retrieve data to match
13437against patterns using ACLs. With the arrival of stick-tables, a new class of
13438sample fetch methods was created, most often sharing the same syntax as their
13439ACL counterpart. These sample fetch methods are also known as "fetches". As
13440of now, ACLs and fetches have converged. All ACL fetch methods have been made
13441available as fetch methods, and ACLs may use any sample fetch method as well.
13442
13443This section details all available sample fetch methods and their output type.
13444Some sample fetch methods have deprecated aliases that are used to maintain
13445compatibility with existing configurations. They are then explicitly marked as
13446deprecated and should not be used in new setups.
13447
13448The ACL derivatives are also indicated when available, with their respective
13449matching methods. These ones all have a well defined default pattern matching
13450method, so it is never necessary (though allowed) to pass the "-m" option to
13451indicate how the sample will be matched using ACLs.
13452
13453As indicated in the sample type versus matching compatibility matrix above,
13454when using a generic sample fetch method in an ACL, the "-m" option is
13455mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
13456the same keyword exists as an ACL keyword and as a standard fetch method, the
13457ACL engine will automatically pick the ACL-only one by default.
13458
13459Some of these keywords support one or multiple mandatory arguments, and one or
13460multiple optional arguments. These arguments are strongly typed and are checked
13461when the configuration is parsed so that there is no risk of running with an
Davor Ocelice9ed2812017-12-25 17:49:28 +010013462incorrect argument (e.g. an unresolved backend name). Fetch function arguments
13463are passed between parenthesis and are delimited by commas. When an argument
Willy Tarreau74ca5042013-06-11 23:12:07 +020013464is optional, it will be indicated below between square brackets ('[ ]'). When
13465all arguments are optional, the parenthesis may be omitted.
13466
13467Thus, the syntax of a standard sample fetch method is one of the following :
13468 - name
13469 - name(arg1)
13470 - name(arg1,arg2)
13471
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013472
134737.3.1. Converters
13474-----------------
13475
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013476Sample fetch methods may be combined with transformations to be applied on top
13477of the fetched sample (also called "converters"). These combinations form what
13478is called "sample expressions" and the result is a "sample". Initially this
13479was only supported by "stick on" and "stick store-request" directives but this
Davor Ocelice9ed2812017-12-25 17:49:28 +010013480has now be extended to all places where samples may be used (ACLs, log-format,
Willy Tarreaue6b11e42013-11-26 19:02:32 +010013481unique-id-format, add-header, ...).
13482
13483These transformations are enumerated as a series of specific keywords after the
13484sample fetch method. These keywords may equally be appended immediately after
13485the fetch keyword's argument, delimited by a comma. These keywords can also
Davor Ocelice9ed2812017-12-25 17:49:28 +010013486support some arguments (e.g. a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010013487
Willy Tarreau97707872015-01-27 15:12:13 +010013488A certain category of converters are bitwise and arithmetic operators which
13489support performing basic operations on integers. Some bitwise operations are
13490supported (and, or, xor, cpl) and some arithmetic operations are supported
13491(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
13492bool) which make it possible to report a match without having to write an ACL.
13493
Willy Tarreau74ca5042013-06-11 23:12:07 +020013494The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010013495
Ben Shillitof25e8e52016-12-02 14:25:37 +00001349651d.single(<prop>[,<prop>*])
13497 Returns values for the properties requested as a string, where values are
13498 separated by the delimiter specified with "51degrees-property-separator".
13499 The device is identified using the User-Agent header passed to the
13500 converter. The function can be passed up to five property names, and if a
13501 property name can't be found, the value "NoData" is returned.
13502
13503 Example :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013504 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request,
13505 # containing values for the three properties requested by using the
Ben Shillitof25e8e52016-12-02 14:25:37 +000013506 # User-Agent passed to the converter.
13507 frontend http-in
13508 bind *:8081
13509 default_backend servers
13510 http-request set-header X-51D-DeviceTypeMobileTablet \
13511 %[req.fhdr(User-Agent),51d.single(DeviceType,IsMobile,IsTablet)]
13512
Willy Tarreau97707872015-01-27 15:12:13 +010013513add(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013514 Adds <value> to the input value of type signed integer, and returns the
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013515 result as a signed integer. <value> can be a numeric value or a variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013516 name. The name of the variable starts with an indication about its scope. The
13517 scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013518 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013519 "sess" : the variable is shared with the whole session
13520 "txn" : the variable is shared with the transaction (request and response)
13521 "req" : the variable is shared only during request processing
13522 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013523 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013524 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013525
Nenad Merdanovicc31499d2019-03-23 11:00:32 +010013526aes_gcm_dec(<bits>,<nonce>,<key>,<aead_tag>)
13527 Decrypts the raw byte input using the AES128-GCM, AES192-GCM or
13528 AES256-GCM algorithm, depending on the <bits> parameter. All other parameters
13529 need to be base64 encoded and the returned result is in raw byte format.
13530 If the <aead_tag> validation fails, the converter doesn't return any data.
13531 The <nonce>, <key> and <aead_tag> can either be strings or variables. This
13532 converter requires at least OpenSSL 1.0.1.
13533
13534 Example:
13535 http-response set-header X-Decrypted-Text %[var(txn.enc),\
13536 aes_gcm_dec(128,txn.nonce,Zm9vb2Zvb29mb29wZm9vbw==,txn.aead_tag)]
13537
Willy Tarreau97707872015-01-27 15:12:13 +010013538and(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013539 Performs a bitwise "AND" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013540 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013541 numeric value or a variable name. The name of the variable starts with an
13542 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013543 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013544 "sess" : the variable is shared with the whole session
13545 "txn" : the variable is shared with the transaction (request and response)
13546 "req" : the variable is shared only during request processing
13547 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013548 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013549 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013550
Holger Just1bfc24b2017-05-06 00:56:53 +020013551b64dec
13552 Converts (decodes) a base64 encoded input string to its binary
13553 representation. It performs the inverse operation of base64().
13554
Emeric Brun53d1a982014-04-30 18:21:37 +020013555base64
13556 Converts a binary input sample to a base64 string. It is used to log or
Davor Ocelice9ed2812017-12-25 17:49:28 +010013557 transfer binary content in a way that can be reliably transferred (e.g.
Emeric Brun53d1a982014-04-30 18:21:37 +020013558 an SSL ID can be copied in a header).
13559
Willy Tarreau97707872015-01-27 15:12:13 +010013560bool
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013561 Returns a boolean TRUE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013562 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013563 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013564 presence of a flag).
13565
Emeric Brun54c4ac82014-11-03 15:32:43 +010013566bytes(<offset>[,<length>])
13567 Extracts some bytes from an input binary sample. The result is a binary
13568 sample starting at an offset (in bytes) of the original sample and
Tim Düsterhus4896c442016-11-29 02:15:19 +010013569 optionally truncated at the given length.
Emeric Brun54c4ac82014-11-03 15:32:43 +010013570
Willy Tarreau280f42b2018-02-19 15:34:12 +010013571concat([<start>],[<var>],[<end>])
13572 Concatenates up to 3 fields after the current sample which is then turned to
13573 a string. The first one, <start>, is a constant string, that will be appended
13574 immediately after the existing sample. It may be omitted if not used. The
13575 second one, <var>, is a variable name. The variable will be looked up, its
13576 contents converted to a string, and it will be appended immediately after the
13577 <first> part. If the variable is not found, nothing is appended. It may be
13578 omitted as well. The third field, <end> is a constant string that will be
13579 appended after the variable. It may also be omitted. Together, these elements
13580 allow to concatenate variables with delimiters to an existing set of
13581 variables. This can be used to build new variables made of a succession of
13582 other variables, such as colon-delimited varlues. Note that due to the config
13583 parser, it is not possible to use a comma nor a closing parenthesis as
13584 delimitors.
13585
13586 Example:
13587 tcp-request session set-var(sess.src) src
13588 tcp-request session set-var(sess.dn) ssl_c_s_dn
13589 tcp-request session set-var(txn.sig) str(),concat(<ip=,sess.ip,>),concat(<dn=,sess.dn,>)
13590 http-request set-header x-hap-sig %[var(txn.sig)]
13591
Willy Tarreau97707872015-01-27 15:12:13 +010013592cpl
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013593 Takes the input value of type signed integer, applies a ones-complement
13594 (flips all bits) and returns the result as an signed integer.
Willy Tarreau97707872015-01-27 15:12:13 +010013595
Willy Tarreau80599772015-01-20 19:35:24 +010013596crc32([<avalanche>])
13597 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
13598 hash function. Optionally, it is possible to apply a full avalanche hash
13599 function to the output if the optional <avalanche> argument equals 1. This
13600 converter uses the same functions as used by the various hash-based load
13601 balancing algorithms, so it will provide exactly the same results. It is
13602 provided for compatibility with other software which want a CRC32 to be
13603 computed on some input keys, so it follows the most common implementation as
13604 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
13605 but may provide a better or at least less predictable distribution. It must
13606 not be used for security purposes as a 32-bit hash is trivial to break. See
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013607 also "djb2", "sdbm", "wt6", "crc32c" and the "hash-type" directive.
13608
13609crc32c([<avalanche>])
13610 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32C
13611 hash function. Optionally, it is possible to apply a full avalanche hash
13612 function to the output if the optional <avalanche> argument equals 1. This
13613 converter uses the same functions as described in RFC4960, Appendix B [8].
13614 It is provided for compatibility with other software which want a CRC32C to be
13615 computed on some input keys. It is slower than the other algorithms and it must
13616 not be used for security purposes as a 32-bit hash is trivial to break. See
13617 also "djb2", "sdbm", "wt6", "crc32" and the "hash-type" directive.
Willy Tarreau80599772015-01-20 19:35:24 +010013618
David Carlier29b3ca32015-09-25 14:09:21 +010013619da-csv-conv(<prop>[,<prop>*])
David Carlier4542b102015-06-01 13:54:29 +020013620 Asks the DeviceAtlas converter to identify the User Agent string passed on
13621 input, and to emit a string made of the concatenation of the properties
13622 enumerated in argument, delimited by the separator defined by the global
13623 keyword "deviceatlas-property-separator", or by default the pipe character
David Carlier840b0242016-03-16 10:09:55 +000013624 ('|'). There's a limit of 12 different properties imposed by the haproxy
David Carlier4542b102015-06-01 13:54:29 +020013625 configuration language.
13626
13627 Example:
13628 frontend www
Cyril Bonté307ee1e2015-09-28 23:16:06 +020013629 bind *:8881
13630 default_backend servers
David Carlier840b0242016-03-16 10:09:55 +000013631 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 +020013632
Thierry FOURNIER9687c772015-05-07 15:46:29 +020013633debug
13634 This converter is used as debug tool. It dumps on screen the content and the
13635 type of the input sample. The sample is returned as is on its output. This
13636 converter only exists when haproxy was built with debugging enabled.
13637
Willy Tarreau97707872015-01-27 15:12:13 +010013638div(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013639 Divides the input value of type signed integer by <value>, and returns the
13640 result as an signed integer. If <value> is null, the largest unsigned
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013641 integer is returned (typically 2^63-1). <value> can be a numeric value or a
Daniel Schneller0b547052016-03-21 20:46:57 +010013642 variable name. The name of the variable starts with an indication about its
13643 scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013644 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013645 "sess" : the variable is shared with the whole session
13646 "txn" : the variable is shared with the transaction (request and response)
13647 "req" : the variable is shared only during request processing
13648 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013649 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013650 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013651
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013652djb2([<avalanche>])
13653 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
13654 hash function. Optionally, it is possible to apply a full avalanche hash
13655 function to the output if the optional <avalanche> argument equals 1. This
13656 converter uses the same functions as used by the various hash-based load
13657 balancing algorithms, so it will provide exactly the same results. It is
13658 mostly intended for debugging, but can be used as a stick-table entry to
13659 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013660 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6", "crc32c",
13661 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013662
Willy Tarreau97707872015-01-27 15:12:13 +010013663even
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013664 Returns a boolean TRUE if the input value of type signed integer is even
Willy Tarreau97707872015-01-27 15:12:13 +010013665 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
13666
Marcin Deranek9631a282018-04-16 14:30:46 +020013667field(<index>,<delimiters>[,<count>])
13668 Extracts the substring at the given index counting from the beginning
13669 (positive index) or from the end (negative index) considering given delimiters
13670 from an input string. Indexes start at 1 or -1 and delimiters are a string
13671 formatted list of chars. Optionally you can specify <count> of fields to
13672 extract (default: 1). Value of 0 indicates extraction of all remaining
13673 fields.
13674
13675 Example :
13676 str(f1_f2_f3__f5),field(5,_) # f5
13677 str(f1_f2_f3__f5),field(2,_,0) # f2_f3__f5
13678 str(f1_f2_f3__f5),field(2,_,2) # f2_f3
13679 str(f1_f2_f3__f5),field(-2,_,3) # f2_f3_
13680 str(f1_f2_f3__f5),field(-3,_,0) # f1_f2_f3
Emeric Brunf399b0d2014-11-03 17:07:03 +010013681
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013682hex
Davor Ocelice9ed2812017-12-25 17:49:28 +010013683 Converts a binary input sample to a hex string containing two hex digits per
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013684 input byte. It is used to log or transfer hex dumps of some binary input data
Davor Ocelice9ed2812017-12-25 17:49:28 +010013685 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 +020013686 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010013687
Dragan Dosen3f957b22017-10-24 09:27:34 +020013688hex2i
13689 Converts a hex string containing two hex digits per input byte to an
13690 integer. If the input value can not be converted, then zero is returned.
13691
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013692http_date([<offset>])
13693 Converts an integer supposed to contain a date since epoch to a string
13694 representing this date in a format suitable for use in HTTP header fields. If
13695 an offset value is specified, then it is a number of seconds that is added to
13696 the date before the conversion is operated. This is particularly useful to
13697 emit Date header fields, Expires values in responses when combined with a
13698 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020013699
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013700in_table(<table>)
13701 Uses the string representation of the input sample to perform a look up in
13702 the specified table. If the key is not found in the table, a boolean false
13703 is returned. Otherwise a boolean true is returned. This can be used to verify
Davor Ocelice9ed2812017-12-25 17:49:28 +010013704 the presence of a certain key in a table tracking some elements (e.g. whether
Willy Tarreaud9f316a2014-07-10 14:03:38 +020013705 or not a source IP address or an Authorization header was already seen).
13706
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013707ipmask(<mask4>, [<mask6>])
13708 Apply a mask to an IP address, and use the result for lookups and storage.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013709 This can be used to make all hosts within a certain mask to share the same
Tim Duesterhus1478aa72018-01-25 16:24:51 +010013710 table entries and as such use the same server. The mask4 can be passed in
13711 dotted form (e.g. 255.255.255.0) or in CIDR form (e.g. 24). The mask6 can
13712 be passed in quadruplet form (e.g. ffff:ffff::) or in CIDR form (e.g. 64).
13713 If no mask6 is given IPv6 addresses will fail to convert for backwards
13714 compatibility reasons.
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013715
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013716json([<input-code>])
Davor Ocelice9ed2812017-12-25 17:49:28 +010013717 Escapes the input string and produces an ASCII output string ready to use as a
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013718 JSON string. The converter tries to decode the input string according to the
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013719 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8p" or
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013720 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
13721 of errors:
13722 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
13723 bytes, ...)
13724 - invalid range (the decoded value is within a UTF-8 prohibited range),
13725 - code overlong (the value is encoded with more bytes than necessary).
13726
13727 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
13728 character is greater than 0xffff because the JSON string escape specification
13729 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
13730 in 4 variants designated by a combination of two suffix letters : "p" for
13731 "permissive" and "s" for "silently ignore". The behaviors of the decoders
13732 are :
Davor Ocelice9ed2812017-12-25 17:49:28 +010013733 - "ascii" : never fails;
13734 - "utf8" : fails on any detected errors;
13735 - "utf8s" : never fails, but removes characters corresponding to errors;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013736 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
Davor Ocelice9ed2812017-12-25 17:49:28 +010013737 error;
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013738 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
13739 characters corresponding to the other errors.
13740
13741 This converter is particularly useful for building properly escaped JSON for
Davor Ocelice9ed2812017-12-25 17:49:28 +010013742 logging to servers which consume JSON-formatted traffic logs.
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013743
13744 Example:
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013745 capture request header Host len 15
Herve COMMOWICK8dfe8632016-08-05 12:01:20 +020013746 capture request header user-agent len 150
13747 log-format '{"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json(utf8s)]"}'
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020013748
13749 Input request from client 127.0.0.1:
13750 GET / HTTP/1.0
13751 User-Agent: Very "Ugly" UA 1/2
13752
13753 Output log:
13754 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
13755
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013756language(<value>[,<default>])
13757 Returns the value with the highest q-factor from a list as extracted from the
13758 "accept-language" header using "req.fhdr". Values with no q-factor have a
13759 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
13760 belong to the list of semi-colon delimited <values> will be considered. The
13761 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
13762 given list and a default value is provided, it is returned. Note that language
13763 names may have a variant after a dash ('-'). If this variant is present in the
13764 list, it will be matched, but if it is not, only the base language is checked.
13765 The match is case-sensitive, and the output string is always one of those
Davor Ocelice9ed2812017-12-25 17:49:28 +010013766 provided in arguments. The ordering of arguments is meaningless, only the
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013767 ordering of the values in the request counts, as the first value among
13768 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020013769
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013770 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013771
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013772 # this configuration switches to the backend matching a
13773 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020013774
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013775 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
13776 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
13777 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
13778 use_backend spanish if es
13779 use_backend french if fr
13780 use_backend english if en
13781 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020013782
Willy Tarreau60a2ee72017-12-15 07:13:48 +010013783length
Etienne Carriereed0d24e2017-12-13 13:41:34 +010013784 Get the length of the string. This can only be placed after a string
13785 sample fetch function or after a transformation keyword returning a string
13786 type. The result is of type integer.
13787
Willy Tarreauffcb2e42014-07-10 16:29:08 +020013788lower
13789 Convert a string sample to lower case. This can only be placed after a string
13790 sample fetch function or after a transformation keyword returning a string
13791 type. The result is of type string.
13792
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013793ltime(<format>[,<offset>])
13794 Converts an integer supposed to contain a date since epoch to a string
13795 representing this date in local time using a format defined by the <format>
13796 string using strftime(3). The purpose is to allow any date format to be used
13797 in logs. An optional <offset> in seconds may be applied to the input date
13798 (positive or negative). See the strftime() man page for the format supported
13799 by your operating system. See also the utime converter.
13800
13801 Example :
13802
13803 # Emit two colons, one with the local time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010013804 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020013805 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
13806
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013807map(<map_file>[,<default_value>])
13808map_<match_type>(<map_file>[,<default_value>])
13809map_<match_type>_<output_type>(<map_file>[,<default_value>])
13810 Search the input value from <map_file> using the <match_type> matching method,
13811 and return the associated value converted to the type <output_type>. If the
13812 input value cannot be found in the <map_file>, the converter returns the
13813 <default_value>. If the <default_value> is not set, the converter fails and
13814 acts as if no input value could be fetched. If the <match_type> is not set, it
13815 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
13816 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
13817 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013818
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013819 It is important to avoid overlapping between the keys : IP addresses and
13820 strings are stored in trees, so the first of the finest match will be used.
13821 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013822
Tim Düsterhus4896c442016-11-29 02:15:19 +010013823 The following array contains the list of all map functions available sorted by
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013824 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013825
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013826 input type | match method | output type str | output type int | output type ip
13827 -----------+--------------+-----------------+-----------------+---------------
13828 str | str | map_str | map_str_int | map_str_ip
13829 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020013830 str | beg | map_beg | map_beg_int | map_end_ip
13831 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013832 str | sub | map_sub | map_sub_int | map_sub_ip
13833 -----------+--------------+-----------------+-----------------+---------------
13834 str | dir | map_dir | map_dir_int | map_dir_ip
13835 -----------+--------------+-----------------+-----------------+---------------
13836 str | dom | map_dom | map_dom_int | map_dom_ip
13837 -----------+--------------+-----------------+-----------------+---------------
13838 str | end | map_end | map_end_int | map_end_ip
13839 -----------+--------------+-----------------+-----------------+---------------
Ruoshan Huang3c5e3742016-12-02 16:25:31 +080013840 str | reg | map_reg | map_reg_int | map_reg_ip
13841 -----------+--------------+-----------------+-----------------+---------------
13842 str | reg | map_regm | map_reg_int | map_reg_ip
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013843 -----------+--------------+-----------------+-----------------+---------------
13844 int | int | map_int | map_int_int | map_int_ip
13845 -----------+--------------+-----------------+-----------------+---------------
13846 ip | ip | map_ip | map_ip_int | map_ip_ip
13847 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013848
Thierry Fournier8feaa662016-02-10 22:55:20 +010013849 The special map called "map_regm" expect matching zone in the regular
13850 expression and modify the output replacing back reference (like "\1") by
13851 the corresponding match text.
13852
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013853 The file contains one key + value per line. Lines which start with '#' are
13854 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
13855 is then the first "word" (series of non-space/tabs characters), and the value
13856 is what follows this series of space/tab till the end of the line excluding
13857 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010013858
Thierry FOURNIER060762e2014-04-23 13:29:15 +020013859 Example :
13860
13861 # this is a comment and is ignored
13862 2.22.246.0/23 United Kingdom \n
13863 <-><-----------><--><------------><---->
13864 | | | | `- trailing spaces ignored
13865 | | | `---------- value
13866 | | `-------------------- middle spaces ignored
13867 | `---------------------------- key
13868 `------------------------------------ leading spaces ignored
13869
Willy Tarreau97707872015-01-27 15:12:13 +010013870mod(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013871 Divides the input value of type signed integer by <value>, and returns the
13872 remainder as an signed integer. If <value> is null, then zero is returned.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013873 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013874 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013875 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013876 "sess" : the variable is shared with the whole session
13877 "txn" : the variable is shared with the transaction (request and response)
13878 "req" : the variable is shared only during request processing
13879 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013880 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013881 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013882
13883mul(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013884 Multiplies the input value of type signed integer by <value>, and returns
Thierry FOURNIER00c005c2015-07-08 01:10:21 +020013885 the product as an signed integer. In case of overflow, the largest possible
13886 value for the sign is returned so that the operation doesn't wrap around.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013887 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010013888 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013889 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013890 "sess" : the variable is shared with the whole session
13891 "txn" : the variable is shared with the transaction (request and response)
13892 "req" : the variable is shared only during request processing
13893 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013894 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013895 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013896
Nenad Merdanovicb7e7c472017-03-12 21:56:55 +010013897nbsrv
13898 Takes an input value of type string, interprets it as a backend name and
13899 returns the number of usable servers in that backend. Can be used in places
13900 where we want to look up a backend from a dynamic name, like a result of a
13901 map lookup.
13902
Willy Tarreau97707872015-01-27 15:12:13 +010013903neg
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013904 Takes the input value of type signed integer, computes the opposite value,
13905 and returns the remainder as an signed integer. 0 is identity. This operator
13906 is provided for reversed subtracts : in order to subtract the input from a
13907 constant, simply perform a "neg,add(value)".
Willy Tarreau97707872015-01-27 15:12:13 +010013908
13909not
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013910 Returns a boolean FALSE if the input value of type signed integer is
Willy Tarreau97707872015-01-27 15:12:13 +010013911 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
Davor Ocelice9ed2812017-12-25 17:49:28 +010013912 used to report true/false for bit testing on input values (e.g. verify the
Willy Tarreau97707872015-01-27 15:12:13 +010013913 absence of a flag).
13914
13915odd
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013916 Returns a boolean TRUE if the input value of type signed integer is odd
Willy Tarreau97707872015-01-27 15:12:13 +010013917 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
13918
13919or(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020013920 Performs a bitwise "OR" between <value> and the input value of type signed
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020013921 integer, and returns the result as an signed integer. <value> can be a
Daniel Schneller0b547052016-03-21 20:46:57 +010013922 numeric value or a variable name. The name of the variable starts with an
13923 indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013924 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010013925 "sess" : the variable is shared with the whole session
13926 "txn" : the variable is shared with the transaction (request and response)
13927 "req" : the variable is shared only during request processing
13928 "res" : the variable is shared only during response processing
Davor Ocelice9ed2812017-12-25 17:49:28 +010013929 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010013930 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010013931
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010013932protobuf(<field_number>,[<field_type>])
13933 This extracts the protocol buffers message field in raw mode of an input binary
13934 sample representation of a protocol buffer message with <field_number> as field
13935 number (dotted notation) if <field_type> is not present, or as an integer sample
13936 if this field is present (see also "ungrpc" below).
13937 The list of the authorized types is the following one: "int32", "int64", "uint32",
13938 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
13939 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
13940 "float" for the wire type 5. Note that "string" is considered as a length-delimited
13941 type, so it does not require any <field_type> argument to be extracted.
13942 More information may be found here about the protocol buffers message field types:
13943 https://developers.google.com/protocol-buffers/docs/encoding
13944
Willy Tarreauc4dc3502015-01-23 20:39:28 +010013945regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010013946 Applies a regex-based substitution to the input string. It does the same
13947 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
13948 default it will replace in the input string the first occurrence of the
13949 largest part matching the regular expression <regex> with the substitution
13950 string <subst>. It is possible to replace all occurrences instead by adding
13951 the flag "g" in the third argument <flags>. It is also possible to make the
13952 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
13953 string, it is made up from the concatenation of all desired flags. Thus if
13954 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
13955 It is important to note that due to the current limitations of the
Baptiste Assmann66025d82016-03-06 23:36:48 +010013956 configuration parser, some characters such as closing parenthesis, closing
13957 square brackets or comma are not possible to use in the arguments. The first
13958 use of this converter is to replace certain characters or sequence of
13959 characters with other ones.
Willy Tarreau7eda8492015-01-20 19:47:06 +010013960
13961 Example :
13962
13963 # de-duplicate "/" in header "x-path".
13964 # input: x-path: /////a///b/c/xzxyz/
13965 # output: x-path: /a/b/c/xzxyz/
13966 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
13967
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013968capture-req(<id>)
13969 Capture the string entry in the request slot <id> and returns the entry as
13970 is. If the slot doesn't exist, the capture fails silently.
13971
13972 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013973 "http-response capture", "capture.req.hdr" and
13974 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013975
13976capture-res(<id>)
13977 Capture the string entry in the response slot <id> and returns the entry as
13978 is. If the slot doesn't exist, the capture fails silently.
13979
13980 See also: "declare capture", "http-request capture",
Baptiste Assmann5ac425c2015-10-21 23:13:46 +020013981 "http-response capture", "capture.req.hdr" and
13982 "capture.res.hdr" (sample fetches).
Thierry FOURNIER35ab2752015-05-28 13:22:03 +020013983
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013984sdbm([<avalanche>])
13985 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
13986 hash function. Optionally, it is possible to apply a full avalanche hash
13987 function to the output if the optional <avalanche> argument equals 1. This
13988 converter uses the same functions as used by the various hash-based load
13989 balancing algorithms, so it will provide exactly the same results. It is
13990 mostly intended for debugging, but can be used as a stick-table entry to
13991 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010013992 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6", "crc32c",
13993 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020013994
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020013995set-var(<var name>)
Davor Ocelice9ed2812017-12-25 17:49:28 +010013996 Sets a variable with the input content and returns the content on the output
13997 as-is. The variable keeps the value and the associated input type. The name of
13998 the variable starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010013999 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014000 "sess" : the variable is shared with the whole session
14001 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014002 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014003 "req" : the variable is shared only during request processing,
14004 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014005 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014006 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014007
Dragan Dosen6e5a9ca2017-10-24 09:18:23 +020014008sha1
14009 Converts a binary input sample to a SHA1 digest. The result is a binary
14010 sample with length of 20 bytes.
14011
Tim Duesterhusca097c12018-04-27 21:18:45 +020014012strcmp(<var>)
14013 Compares the contents of <var> with the input value of type string. Returns
14014 the result as a signed integer compatible with strcmp(3): 0 if both strings
14015 are identical. A value less than 0 if the left string is lexicographically
14016 smaller than the right string or if the left string is shorter. A value greater
14017 than 0 otherwise (right string greater than left string or the right string is
14018 shorter).
14019
14020 Example :
14021
14022 http-request set-var(txn.host) hdr(host)
14023 # Check whether the client is attempting domain fronting.
14024 acl ssl_sni_http_host_match ssl_fc_sni,strcmp(txn.host) eq 0
14025
14026
Willy Tarreau97707872015-01-27 15:12:13 +010014027sub(<value>)
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014028 Subtracts <value> from the input value of type signed integer, and returns
14029 the result as an signed integer. Note: in order to subtract the input from
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014030 a constant, simply perform a "neg,add(value)". <value> can be a numeric value
Daniel Schneller0b547052016-03-21 20:46:57 +010014031 or a variable name. The name of the variable starts with an indication about
14032 its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014033 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014034 "sess" : the variable is shared with the whole session
14035 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014036 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014037 "req" : the variable is shared only during request processing,
14038 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014039 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014040 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014041
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014042table_bytes_in_rate(<table>)
14043 Uses the string representation of the input sample to perform a look up in
14044 the specified table. If the key is not found in the table, integer value zero
14045 is returned. Otherwise the converter returns the average client-to-server
14046 bytes rate associated with the input sample in the designated table, measured
14047 in amount of bytes over the period configured in the table. See also the
14048 sc_bytes_in_rate sample fetch keyword.
14049
14050
14051table_bytes_out_rate(<table>)
14052 Uses the string representation of the input sample to perform a look up in
14053 the specified table. If the key is not found in the table, integer value zero
14054 is returned. Otherwise the converter returns the average server-to-client
14055 bytes rate associated with the input sample in the designated table, measured
14056 in amount of bytes over the period configured in the table. See also the
14057 sc_bytes_out_rate sample fetch keyword.
14058
14059table_conn_cnt(<table>)
14060 Uses the string representation of the input sample to perform a look up in
14061 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014062 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014063 connections associated with the input sample in the designated table. See
14064 also the sc_conn_cnt sample fetch keyword.
14065
14066table_conn_cur(<table>)
14067 Uses the string representation of the input sample to perform a look up in
14068 the specified table. If the key is not found in the table, integer value zero
14069 is returned. Otherwise the converter returns the current amount of concurrent
14070 tracked connections associated with the input sample in the designated table.
14071 See also the sc_conn_cur sample fetch keyword.
14072
14073table_conn_rate(<table>)
14074 Uses the string representation of the input sample to perform a look up in
14075 the specified table. If the key is not found in the table, integer value zero
14076 is returned. Otherwise the converter returns the average incoming connection
14077 rate associated with the input sample in the designated table. See also the
14078 sc_conn_rate sample fetch keyword.
14079
Thierry FOURNIER236657b2015-08-19 08:25:14 +020014080table_gpt0(<table>)
14081 Uses the string representation of the input sample to perform a look up in
14082 the specified table. If the key is not found in the table, boolean value zero
14083 is returned. Otherwise the converter returns the current value of the first
14084 general purpose tag associated with the input sample in the designated table.
14085 See also the sc_get_gpt0 sample fetch keyword.
14086
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014087table_gpc0(<table>)
14088 Uses the string representation of the input sample to perform a look up in
14089 the specified table. If the key is not found in the table, integer value zero
14090 is returned. Otherwise the converter returns the current value of the first
14091 general purpose counter associated with the input sample in the designated
14092 table. See also the sc_get_gpc0 sample fetch keyword.
14093
14094table_gpc0_rate(<table>)
14095 Uses the string representation of the input sample to perform a look up in
14096 the specified table. If the key is not found in the table, integer value zero
14097 is returned. Otherwise the converter returns the frequency which the gpc0
14098 counter was incremented over the configured period in the table, associated
14099 with the input sample in the designated table. See also the sc_get_gpc0_rate
14100 sample fetch keyword.
14101
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014102table_gpc1(<table>)
14103 Uses the string representation of the input sample to perform a look up in
14104 the specified table. If the key is not found in the table, integer value zero
14105 is returned. Otherwise the converter returns the current value of the second
14106 general purpose counter associated with the input sample in the designated
14107 table. See also the sc_get_gpc1 sample fetch keyword.
14108
14109table_gpc1_rate(<table>)
14110 Uses the string representation of the input sample to perform a look up in
14111 the specified table. If the key is not found in the table, integer value zero
14112 is returned. Otherwise the converter returns the frequency which the gpc1
14113 counter was incremented over the configured period in the table, associated
14114 with the input sample in the designated table. See also the sc_get_gpc1_rate
14115 sample fetch keyword.
14116
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014117table_http_err_cnt(<table>)
14118 Uses the string representation of the input sample to perform a look up in
14119 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014120 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014121 errors associated with the input sample in the designated table. See also the
14122 sc_http_err_cnt sample fetch keyword.
14123
14124table_http_err_rate(<table>)
14125 Uses the string representation of the input sample to perform a look up in
14126 the specified table. If the key is not found in the table, integer value zero
14127 is returned. Otherwise the average rate of HTTP errors associated with the
14128 input sample in the designated table, measured in amount of errors over the
14129 period configured in the table. See also the sc_http_err_rate sample fetch
14130 keyword.
14131
14132table_http_req_cnt(<table>)
14133 Uses the string representation of the input sample to perform a look up in
14134 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014135 is returned. Otherwise the converter returns the cumulative number of HTTP
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014136 requests associated with the input sample in the designated table. See also
14137 the sc_http_req_cnt sample fetch keyword.
14138
14139table_http_req_rate(<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
14142 is returned. Otherwise the average rate of HTTP requests associated with the
14143 input sample in the designated table, measured in amount of requests over the
14144 period configured in the table. See also the sc_http_req_rate sample fetch
14145 keyword.
14146
14147table_kbytes_in(<table>)
14148 Uses the string representation of the input sample to perform a look up in
14149 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014150 is returned. Otherwise the converter returns the cumulative number of client-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014151 to-server data associated with the input sample in the designated table,
14152 measured in kilobytes. The test is currently performed on 32-bit integers,
14153 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
14154 keyword.
14155
14156table_kbytes_out(<table>)
14157 Uses the string representation of the input sample to perform a look up in
14158 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014159 is returned. Otherwise the converter returns the cumulative number of server-
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014160 to-client data associated with the input sample in the designated table,
14161 measured in kilobytes. The test is currently performed on 32-bit integers,
14162 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
14163 keyword.
14164
14165table_server_id(<table>)
14166 Uses the string representation of the input sample to perform a look up in
14167 the specified table. If the key is not found in the table, integer value zero
14168 is returned. Otherwise the converter returns the server ID associated with
14169 the input sample in the designated table. A server ID is associated to a
14170 sample by a "stick" rule when a connection to a server succeeds. A server ID
14171 zero means that no server is associated with this key.
14172
14173table_sess_cnt(<table>)
14174 Uses the string representation of the input sample to perform a look up in
14175 the specified table. If the key is not found in the table, integer value zero
Davor Ocelice9ed2812017-12-25 17:49:28 +010014176 is returned. Otherwise the converter returns the cumulative number of incoming
Willy Tarreaud9f316a2014-07-10 14:03:38 +020014177 sessions associated with the input sample in the designated table. Note that
14178 a session here refers to an incoming connection being accepted by the
14179 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
14180 keyword.
14181
14182table_sess_rate(<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 average incoming session
14186 rate associated with the input sample in the designated table. Note that a
14187 session here refers to an incoming connection being accepted by the
14188 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
14189 keyword.
14190
14191table_trackers(<table>)
14192 Uses the string representation of the input sample to perform a look up in
14193 the specified table. If the key is not found in the table, integer value zero
14194 is returned. Otherwise the converter returns the current amount of concurrent
14195 connections tracking the same key as the input sample in the designated
14196 table. It differs from table_conn_cur in that it does not rely on any stored
14197 information but on the table's reference count (the "use" value which is
14198 returned by "show table" on the CLI). This may sometimes be more suited for
14199 layer7 tracking. It can be used to tell a server how many concurrent
14200 connections there are from a given address for example. See also the
14201 sc_trackers sample fetch keyword.
14202
Willy Tarreauffcb2e42014-07-10 16:29:08 +020014203upper
14204 Convert a string sample to upper case. This can only be placed after a string
14205 sample fetch function or after a transformation keyword returning a string
14206 type. The result is of type string.
14207
Thierry FOURNIER82ff3c92015-05-07 15:46:20 +020014208url_dec
14209 Takes an url-encoded string provided as input and returns the decoded
14210 version as output. The input and the output are of type string.
14211
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014212ungrpc(<field_number>,[<field_type>])
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014213 This extracts the protocol buffers message field in raw mode of an input binary
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014214 sample representation of a gRPC message with <field_number> as field number
14215 (dotted notation) if <field_type> is not present, or as an integer sample if this
14216 field is present.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014217 The list of the authorized types is the following one: "int32", "int64", "uint32",
14218 "uint64", "sint32", "sint64", "bool", "enum" for the "varint" wire type 0
14219 "fixed64", "sfixed64", "double" for the 64bit wire type 1, "fixed32", "sfixed32",
14220 "float" for the wire type 5. Note that "string" is considered as a length-delimited
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014221 type, so it does not require any <field_type> argument to be extracted.
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014222 More information may be found here about the protocol buffers message field types:
14223 https://developers.google.com/protocol-buffers/docs/encoding
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014224
14225 Example:
14226 // with such a protocol buffer .proto file content adapted from
14227 // https://github.com/grpc/grpc/blob/master/examples/protos/route_guide.proto
14228
14229 message Point {
14230 int32 latitude = 1;
14231 int32 longitude = 2;
14232 }
14233
14234 message PPoint {
14235 Point point = 59;
14236 }
14237
14238 message Rectangle {
14239 // One corner of the rectangle.
14240 PPoint lo = 48;
14241 // The other corner of the rectangle.
14242 PPoint hi = 49;
14243 }
14244
14245 let's say a body request is made of a "Rectangle" object value (two PPoint
14246 protocol buffers messages), the four protocol buffers fields could be
14247 extracted with these "ungrpc" directives:
14248
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014249 req.body,ungrpc(48.59.1,int32) # "latitude" of "lo" first PPoint
14250 req.body,ungrpc(48.59.2,int32) # "longitude" of "lo" first PPoint
14251 req.body,ungrpc(49.59.1,int32) # "latidude" of "hi" second PPoint
14252 req.body,ungrpc(49.59.2,int32) # "longitude" of "hi" second PPoint
14253
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014254 We could also extract the intermediary 48.59 field as a binary sample as follows:
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014255
Frédéric Lécaille93d33162019-03-06 09:35:59 +010014256 req.body,ungrpc(48.59)
Frédéric Lécaille756d97f2019-03-04 19:03:48 +010014257
Frédéric Lécaillebfe61382019-03-06 14:34:36 +010014258 As a gRPC message is alway made of a gRPC header followed by protocol buffers
14259 messages, in the previous example the "latitude" of "lo" first PPoint
14260 could be extracted with these equivalent directives:
14261
14262 req.body,ungrpc(48.59),protobuf(1,int32)
14263 req.body,ungrpc(48),protobuf(59.1,int32)
14264 req.body,ungrpc(48),protobuf(59),protobuf(1,int32)
14265
14266 Note that the first convert must be "ungrpc", the remaining ones must be
14267 "protobuf" and only the last one may have or not a second argument to
14268 interpret the previous binary sample.
14269
Frédéric Lécaille50290fb2019-02-27 14:34:51 +010014270
Christopher Faulet85d79c92016-11-09 16:54:56 +010014271unset-var(<var name>)
14272 Unsets a variable if the input content is defined. The name of the variable
14273 starts with an indication about its scope. The scopes allowed are:
14274 "proc" : the variable is shared with the whole process
14275 "sess" : the variable is shared with the whole session
14276 "txn" : the variable is shared with the transaction (request and
14277 response),
14278 "req" : the variable is shared only during request processing,
14279 "res" : the variable is shared only during response processing.
14280 This prefix is followed by a name. The separator is a '.'. The name may only
14281 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
14282
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014283utime(<format>[,<offset>])
14284 Converts an integer supposed to contain a date since epoch to a string
14285 representing this date in UTC time using a format defined by the <format>
14286 string using strftime(3). The purpose is to allow any date format to be used
14287 in logs. An optional <offset> in seconds may be applied to the input date
14288 (positive or negative). See the strftime() man page for the format supported
14289 by your operating system. See also the ltime converter.
14290
14291 Example :
14292
14293 # Emit two colons, one with the UTC time and another with ip:port
Davor Ocelice9ed2812017-12-25 17:49:28 +010014294 # e.g. 20140710162350 127.0.0.1:57325
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020014295 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
14296
Marcin Deranek9631a282018-04-16 14:30:46 +020014297word(<index>,<delimiters>[,<count>])
14298 Extracts the nth word counting from the beginning (positive index) or from
14299 the end (negative index) considering given delimiters from an input string.
14300 Indexes start at 1 or -1 and delimiters are a string formatted list of chars.
14301 Optionally you can specify <count> of words to extract (default: 1).
14302 Value of 0 indicates extraction of all remaining words.
14303
14304 Example :
14305 str(f1_f2_f3__f5),word(4,_) # f5
14306 str(f1_f2_f3__f5),word(2,_,0) # f2_f3__f5
14307 str(f1_f2_f3__f5),word(3,_,2) # f3__f5
14308 str(f1_f2_f3__f5),word(-2,_,3) # f1_f2_f3
14309 str(f1_f2_f3__f5),word(-3,_,0) # f1_f2
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010014310
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014311wt6([<avalanche>])
14312 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
14313 hash function. Optionally, it is possible to apply a full avalanche hash
14314 function to the output if the optional <avalanche> argument equals 1. This
14315 converter uses the same functions as used by the various hash-based load
14316 balancing algorithms, so it will provide exactly the same results. It is
14317 mostly intended for debugging, but can be used as a stick-table entry to
14318 collect rough statistics. It must not be used for security purposes as a
Emmanuel Hocdet50791a72018-03-21 11:19:01 +010014319 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", "crc32c",
14320 and the "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020014321
Willy Tarreau97707872015-01-27 15:12:13 +010014322xor(<value>)
14323 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014324 of type signed integer, and returns the result as an signed integer.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014325 <value> can be a numeric value or a variable name. The name of the variable
Daniel Schneller0b547052016-03-21 20:46:57 +010014326 starts with an indication about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014327 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014328 "sess" : the variable is shared with the whole session
14329 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014330 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014331 "req" : the variable is shared only during request processing,
14332 "res" : the variable is shared only during response processing.
Thierry FOURNIER5d86fae2015-07-07 21:10:16 +020014333 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014334 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Willy Tarreau97707872015-01-27 15:12:13 +010014335
Thierry FOURNIER01e09742016-12-26 11:46:11 +010014336xxh32([<seed>])
14337 Hashes a binary input sample into an unsigned 32-bit quantity using the 32-bit
14338 variant of the XXHash hash function. This hash supports a seed which defaults
14339 to zero but a different value maybe passed as the <seed> argument. This hash
14340 is known to be very good and very fast so it can be used to hash URLs and/or
14341 URL parameters for use as stick-table keys to collect statistics with a low
14342 collision rate, though care must be taken as the algorithm is not considered
14343 as cryptographically secure.
14344
14345xxh64([<seed>])
14346 Hashes a binary input sample into a signed 64-bit quantity using the 64-bit
14347 variant of the XXHash hash function. This hash supports a seed which defaults
14348 to zero but a different value maybe passed as the <seed> argument. This hash
14349 is known to be very good and very fast so it can be used to hash URLs and/or
14350 URL parameters for use as stick-table keys to collect statistics with a low
14351 collision rate, though care must be taken as the algorithm is not considered
14352 as cryptographically secure.
14353
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010014354
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200143557.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020014356--------------------------------------------
14357
14358A first set of sample fetch methods applies to internal information which does
14359not even relate to any client information. These ones are sometimes used with
14360"monitor-fail" directives to report an internal status to external watchers.
14361The sample fetch methods described in this section are usable anywhere.
14362
14363always_false : boolean
14364 Always returns the boolean "false" value. It may be used with ACLs as a
14365 temporary replacement for another one when adjusting configurations.
14366
14367always_true : boolean
14368 Always returns the boolean "true" value. It may be used with ACLs as a
14369 temporary replacement for another one when adjusting configurations.
14370
14371avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014372 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020014373 divided by the number of active servers. The current backend is used if no
14374 backend is specified. This is very similar to "queue" except that the size of
14375 the farm is considered, in order to give a more accurate measurement of the
14376 time it may take for a new connection to be processed. The main usage is with
14377 ACL to return a sorry page to new users when it becomes certain they will get
14378 a degraded service, or to pass to the backend servers in a header so that
14379 they decide to work in degraded mode or to disable some functions to speed up
14380 the processing a bit. Note that in the event there would not be any active
14381 server anymore, twice the number of queued connections would be considered as
14382 the measured value. This is a fair estimate, as we expect one server to get
14383 back soon anyway, but we still prefer to send new traffic to another backend
14384 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
14385 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010014386
Willy Tarreau74ca5042013-06-11 23:12:07 +020014387be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020014388 Applies to the number of currently established connections on the backend,
14389 possibly including the connection being evaluated. If no backend name is
14390 specified, the current one is used. But it is also possible to check another
14391 backend. It can be used to use a specific farm when the nominal one is full.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014392 See also the "fe_conn", "queue", "be_conn_free", and "be_sess_rate" criteria.
14393
14394be_conn_free([<backend>]) : integer
14395 Returns an integer value corresponding to the number of available connections
14396 across available servers in the backend. Queue slots are not included. Backup
14397 servers are also not included, unless all other servers are down. If no
14398 backend name is specified, the current one is used. But it is also possible
14399 to check another backend. It can be used to use a specific farm when the
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014400 nominal one is full. See also the "be_conn", "connslots", and "srv_conn_free"
14401 criteria.
Patrick Hemmer4cdf3ab2018-06-14 17:10:27 -040014402
14403 OTHER CAVEATS AND NOTES: if any of the server maxconn, or maxqueue is 0
14404 (meaning unlimited), then this fetch clearly does not make sense, in which
14405 case the value returned will be -1.
Willy Tarreau6a06a402007-07-15 20:15:28 +020014406
Willy Tarreau74ca5042013-06-11 23:12:07 +020014407be_sess_rate([<backend>]) : integer
14408 Returns an integer value corresponding to the sessions creation rate on the
14409 backend, in number of new sessions per second. This is used with ACLs to
14410 switch to an alternate backend when an expensive or fragile one reaches too
Davor Ocelice9ed2812017-12-25 17:49:28 +010014411 high a session rate, or to limit abuse of service (e.g. prevent sucking of an
Willy Tarreau74ca5042013-06-11 23:12:07 +020014412 online dictionary). It can also be useful to add this element to logs using a
14413 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014414
14415 Example :
14416 # Redirect to an error page if the dictionary is requested too often
14417 backend dynamic
14418 mode http
14419 acl being_scanned be_sess_rate gt 100
14420 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010014421
Davor Ocelice9ed2812017-12-25 17:49:28 +010014422bin(<hex>) : bin
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014423 Returns a binary chain. The input is the hexadecimal representation
14424 of the string.
14425
14426bool(<bool>) : bool
14427 Returns a boolean value. <bool> can be 'true', 'false', '1' or '0'.
14428 'false' and '0' are the same. 'true' and '1' are the same.
14429
Willy Tarreau74ca5042013-06-11 23:12:07 +020014430connslots([<backend>]) : integer
14431 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014432 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020014433 connections on all servers and the maximum queue size. This is probably only
14434 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050014435
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014436 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020014437 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014438 usage; see "use_backend" keyword) can be redirected to a different backend.
14439
Willy Tarreau55165fe2009-05-10 12:02:55 +020014440 'connslots' = number of available server connection slots, + number of
14441 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014442
Willy Tarreaua36af912009-10-10 12:02:45 +020014443 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020014444 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020014445 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020014446 you want to be able to differentiate between different backends, and their
Davor Ocelice9ed2812017-12-25 17:49:28 +010014447 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020014448 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020014449 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014450
Willy Tarreau55165fe2009-05-10 12:02:55 +020014451 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
14452 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020014453 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020014454 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080014455
Willy Tarreau70fe9442018-11-22 16:07:39 +010014456cpu_calls : integer
14457 Returns the number of calls to the task processing the stream or current
14458 request since it was allocated. This number is reset for each new request on
14459 the same connections in case of HTTP keep-alive. This value should usually be
14460 low and stable (around 2 calls for a typically simple request) but may become
14461 high if some processing (compression, caching or analysis) is performed. This
14462 is purely for performance monitoring purposes.
14463
14464cpu_ns_avg : integer
14465 Returns the average number of nanoseconds spent in each call to the task
14466 processing the stream or current request. This number is reset for each new
14467 request on the same connections in case of HTTP keep-alive. This value
14468 indicates the overall cost of processing the request or the connection for
14469 each call. There is no good nor bad value but the time spent in a call
14470 automatically causes latency for other processing (see lat_ns_avg below),
14471 and may affect other connection's apparent response time. Certain operations
14472 like compression, complex regex matching or heavy Lua operations may directly
14473 affect this value, and having it in the logs will make it easier to spot the
14474 faulty processing that needs to be fixed to recover decent performance.
14475 Note: this value is exactly cpu_ns_tot divided by cpu_calls.
14476
14477cpu_ns_tot : integer
14478 Returns the total number of nanoseconds spent in each call to the task
14479 processing the stream or current request. This number is reset for each new
14480 request on the same connections in case of HTTP keep-alive. This value
14481 indicates the overall cost of processing the request or the connection for
14482 each call. There is no good nor bad value but the time spent in a call
14483 automatically causes latency for other processing (see lat_ns_avg below),
14484 induces CPU costs on the machine, and may affect other connection's apparent
14485 response time. Certain operations like compression, complex regex matching or
14486 heavy Lua operations may directly affect this value, and having it in the
14487 logs will make it easier to spot the faulty processing that needs to be fixed
14488 to recover decent performance. The value may be artificially high due to a
14489 high cpu_calls count, for example when processing many HTTP chunks, and for
14490 this reason it is often preferred to log cpu_ns_avg instead.
14491
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014492date([<offset>]) : integer
14493 Returns the current date as the epoch (number of seconds since 01/01/1970).
14494 If an offset value is specified, then it is a number of seconds that is added
14495 to the current date before returning the value. This is particularly useful
14496 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020014497 It is useful combined with the http_date converter.
14498
14499 Example :
14500
14501 # set an expires header to now+1 hour in every response
14502 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020014503
Etienne Carrierea792a0a2018-01-17 13:43:24 +010014504date_us : integer
14505 Return the microseconds part of the date (the "second" part is returned by
14506 date sample). This sample is coherent with the date sample as it is comes
14507 from the same timeval structure.
14508
Willy Tarreaud716f9b2017-10-13 11:03:15 +020014509distcc_body(<token>[,<occ>]) : binary
14510 Parses a distcc message and returns the body associated to occurrence #<occ>
14511 of the token <token>. Occurrences start at 1, and when unspecified, any may
14512 match though in practice only the first one is checked for now. This can be
14513 used to extract file names or arguments in files built using distcc through
14514 haproxy. Please refer to distcc's protocol documentation for the complete
14515 list of supported tokens.
14516
14517distcc_param(<token>[,<occ>]) : integer
14518 Parses a distcc message and returns the parameter associated to occurrence
14519 #<occ> of the token <token>. Occurrences start at 1, and when unspecified,
14520 any may match though in practice only the first one is checked for now. This
14521 can be used to extract certain information such as the protocol version, the
14522 file size or the argument in files built using distcc through haproxy.
14523 Another use case consists in waiting for the start of the preprocessed file
14524 contents before connecting to the server to avoid keeping idle connections.
14525 Please refer to distcc's protocol documentation for the complete list of
14526 supported tokens.
14527
14528 Example :
14529 # wait up to 20s for the pre-processed file to be uploaded
14530 tcp-request inspect-delay 20s
14531 tcp-request content accept if { distcc_param(DOTI) -m found }
14532 # send large files to the big farm
14533 use_backend big_farm if { distcc_param(DOTI) gt 1000000 }
14534
Willy Tarreau595ec542013-06-12 21:34:28 +020014535env(<name>) : string
14536 Returns a string containing the value of environment variable <name>. As a
14537 reminder, environment variables are per-process and are sampled when the
14538 process starts. This can be useful to pass some information to a next hop
14539 server, or with ACLs to take specific action when the process is started a
14540 certain way.
14541
14542 Examples :
14543 # Pass the Via header to next hop with the local hostname in it
14544 http-request add-header Via 1.1\ %[env(HOSTNAME)]
14545
14546 # reject cookie-less requests when the STOP environment variable is set
14547 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
14548
Willy Tarreau74ca5042013-06-11 23:12:07 +020014549fe_conn([<frontend>]) : integer
14550 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010014551 possibly including the connection being evaluated. If no frontend name is
14552 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014553 frontend. It can be used to return a sorry page before hard-blocking, or to
14554 use a specific backend to drain new requests when the farm is considered
Davor Ocelice9ed2812017-12-25 17:49:28 +010014555 full. This is mostly used with ACLs but can also be used to pass some
Willy Tarreau74ca5042013-06-11 23:12:07 +020014556 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
14557 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020014558
Nenad Merdanovicad9a7e92016-10-03 04:57:37 +020014559fe_req_rate([<frontend>]) : integer
14560 Returns an integer value corresponding to the number of HTTP requests per
14561 second sent to a frontend. This number can differ from "fe_sess_rate" in
14562 situations where client-side keep-alive is enabled.
14563
Willy Tarreau74ca5042013-06-11 23:12:07 +020014564fe_sess_rate([<frontend>]) : integer
14565 Returns an integer value corresponding to the sessions creation rate on the
14566 frontend, in number of new sessions per second. This is used with ACLs to
14567 limit the incoming session rate to an acceptable range in order to prevent
14568 abuse of service at the earliest moment, for example when combined with other
14569 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
14570 down below the limit. It can also be useful to add this element to logs using
14571 a log-format directive. See also the "rate-limit sessions" directive for use
14572 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014573
14574 Example :
14575 # This frontend limits incoming mails to 10/s with a max of 100
14576 # concurrent connections. We accept any connection below 10/s, and
14577 # force excess clients to wait for 100 ms. Since clients are limited to
14578 # 100 max, there cannot be more than 10 incoming mails per second.
14579 frontend mail
14580 bind :25
14581 mode tcp
14582 maxconn 100
14583 acl too_fast fe_sess_rate ge 10
14584 tcp-request inspect-delay 100ms
14585 tcp-request content accept if ! too_fast
14586 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010014587
Nenad Merdanovic807a6e72017-03-12 22:00:00 +010014588hostname : string
14589 Returns the system hostname.
14590
Thierry FOURNIER07ee64e2015-07-06 23:43:03 +020014591int(<integer>) : signed integer
14592 Returns a signed integer.
14593
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014594ipv4(<ipv4>) : ipv4
14595 Returns an ipv4.
14596
14597ipv6(<ipv6>) : ipv6
14598 Returns an ipv6.
14599
Willy Tarreau70fe9442018-11-22 16:07:39 +010014600lat_ns_avg : integer
14601 Returns the average number of nanoseconds spent between the moment the task
14602 handling the stream is woken up and the moment it is effectively called. This
14603 number is reset for each new request on the same connections in case of HTTP
14604 keep-alive. This value indicates the overall latency inflicted to the current
14605 request by all other requests being processed in parallel, and is a direct
14606 indicator of perceived performance due to noisy neighbours. In order to keep
14607 the value low, it is possible to reduce the scheduler's run queue depth using
14608 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14609 once using "tune.maxpollevents", to decrease the stream's nice value using
14610 the "nice" option on the "bind" lines or in the frontend, or to look for
14611 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14612 whose processing needs to be adjusted or fixed. Compression of large buffers
14613 could be a culprit, like heavy regex or long lists of regex.
14614 Note: this value is exactly lat_ns_tot divided by cpu_calls.
14615
14616lat_ns_tot : integer
14617 Returns the total number of nanoseconds spent between the moment the task
14618 handling the stream is woken up and the moment it is effectively called. This
14619 number is reset for each new request on the same connections in case of HTTP
14620 keep-alive. This value indicates the overall latency inflicted to the current
14621 request by all other requests being processed in parallel, and is a direct
14622 indicator of perceived performance due to noisy neighbours. In order to keep
14623 the value low, it is possible to reduce the scheduler's run queue depth using
14624 "tune.runqueue-depth", to reduce the number of concurrent events processed at
14625 once using "tune.maxpollevents", to decrease the stream's nice value using
14626 the "nice" option on the "bind" lines or in the frontend, or to look for
14627 other heavy requests in logs (those exhibiting large values of "cpu_ns_avg"),
14628 whose processing needs to be adjusted or fixed. Compression of large buffers
14629 could be a culprit, like heavy regex or long lists of regex. Note: while it
14630 may intuitively seem that the total latency adds to a transfer time, it is
14631 almost never true because while a task waits for the CPU, network buffers
14632 continue to fill up and the next call will process more at once. The value
14633 may be artificially high due to a high cpu_calls count, for example when
14634 processing many HTTP chunks, and for this reason it is often preferred to log
14635 lat_ns_avg instead, which is a more relevant performance indicator.
14636
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014637meth(<method>) : method
14638 Returns a method.
14639
Willy Tarreau0f30d262014-11-24 16:02:05 +010014640nbproc : integer
14641 Returns an integer value corresponding to the number of processes that were
14642 started (it equals the global "nbproc" setting). This is useful for logging
14643 and debugging purposes.
14644
Willy Tarreau74ca5042013-06-11 23:12:07 +020014645nbsrv([<backend>]) : integer
14646 Returns an integer value corresponding to the number of usable servers of
14647 either the current backend or the named backend. This is mostly used with
14648 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010014649 switch to an alternate backend when the number of servers is too low to
14650 to handle some load. It is useful to report a failure when combined with
14651 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010014652
Patrick Hemmerfabb24f2018-08-13 14:07:57 -040014653prio_class : integer
14654 Returns the priority class of the current session for http mode or connection
14655 for tcp mode. The value will be that set by the last call to "http-request
14656 set-priority-class" or "tcp-request content set-priority-class".
14657
14658prio_offset : integer
14659 Returns the priority offset of the current session for http mode or
14660 connection for tcp mode. The value will be that set by the last call to
14661 "http-request set-priority-offset" or "tcp-request content
14662 set-priority-offset".
14663
Willy Tarreau0f30d262014-11-24 16:02:05 +010014664proc : integer
14665 Returns an integer value corresponding to the position of the process calling
14666 the function, between 1 and global.nbproc. This is useful for logging and
14667 debugging purposes.
14668
Willy Tarreau74ca5042013-06-11 23:12:07 +020014669queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010014670 Returns the total number of queued connections of the designated backend,
14671 including all the connections in server queues. If no backend name is
14672 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020014673 one. This is useful with ACLs or to pass statistics to backend servers. This
14674 can be used to take actions when queuing goes above a known level, generally
14675 indicating a surge of traffic or a massive slowdown on the servers. One
14676 possible action could be to reject new users but still accept old ones. See
14677 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
14678
Willy Tarreau84310e22014-02-14 11:59:04 +010014679rand([<range>]) : integer
14680 Returns a random integer value within a range of <range> possible values,
14681 starting at zero. If the range is not specified, it defaults to 2^32, which
14682 gives numbers between 0 and 4294967295. It can be useful to pass some values
14683 needed to take some routing decisions for example, or just for debugging
14684 purposes. This random must not be used for security purposes.
14685
Willy Tarreau74ca5042013-06-11 23:12:07 +020014686srv_conn([<backend>/]<server>) : integer
14687 Returns an integer value corresponding to the number of currently established
14688 connections on the designated server, possibly including the connection being
14689 evaluated. If <backend> is omitted, then the server is looked up in the
14690 current backend. It can be used to use a specific farm when one server is
14691 full, or to inform the server about our view of the number of active
Patrick Hemmer155e93e2018-06-14 18:01:35 -040014692 connections with it. See also the "fe_conn", "be_conn", "queue", and
14693 "srv_conn_free" fetch methods.
14694
14695srv_conn_free([<backend>/]<server>) : integer
14696 Returns an integer value corresponding to the number of available connections
14697 on the designated server, possibly including the connection being evaluated.
14698 The value does not include queue slots. If <backend> is omitted, then the
14699 server is looked up in the current backend. It can be used to use a specific
14700 farm when one server is full, or to inform the server about our view of the
14701 number of active connections with it. See also the "be_conn_free" and
14702 "srv_conn" fetch methods.
14703
14704 OTHER CAVEATS AND NOTES: If the server maxconn is 0, then this fetch clearly
14705 does not make sense, in which case the value returned will be -1.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014706
14707srv_is_up([<backend>/]<server>) : boolean
14708 Returns true when the designated server is UP, and false when it is either
14709 DOWN or in maintenance mode. If <backend> is omitted, then the server is
14710 looked up in the current backend. It is mainly used to take action based on
Davor Ocelice9ed2812017-12-25 17:49:28 +010014711 an external status reported via a health check (e.g. a geographical site's
Willy Tarreau74ca5042013-06-11 23:12:07 +020014712 availability). Another possible use which is more of a hack consists in
14713 using dummy servers as boolean variables that can be enabled or disabled from
14714 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
14715
Willy Tarreauff2b7af2017-10-13 11:46:26 +020014716srv_queue([<backend>/]<server>) : integer
14717 Returns an integer value corresponding to the number of connections currently
14718 pending in the designated server's queue. If <backend> is omitted, then the
14719 server is looked up in the current backend. It can sometimes be used together
14720 with the "use-server" directive to force to use a known faster server when it
14721 is not much loaded. See also the "srv_conn", "avg_queue" and "queue" sample
14722 fetch methods.
14723
Willy Tarreau74ca5042013-06-11 23:12:07 +020014724srv_sess_rate([<backend>/]<server>) : integer
14725 Returns an integer corresponding to the sessions creation rate on the
14726 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030014727 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020014728 used with ACLs but can make sense with logs too. This is used to switch to an
14729 alternate backend when an expensive or fragile one reaches too high a session
Davor Ocelice9ed2812017-12-25 17:49:28 +010014730 rate, or to limit abuse of service (e.g. prevent latent requests from
Willy Tarreau74ca5042013-06-11 23:12:07 +020014731 overloading servers).
14732
14733 Example :
14734 # Redirect to a separate back
14735 acl srv1_full srv_sess_rate(be1/srv1) gt 50
14736 acl srv2_full srv_sess_rate(be1/srv2) gt 50
14737 use_backend be2 if srv1_full or srv2_full
14738
Willy Tarreau0f30d262014-11-24 16:02:05 +010014739stopping : boolean
14740 Returns TRUE if the process calling the function is currently stopping. This
14741 can be useful for logging, or for relaxing certain checks or helping close
14742 certain connections upon graceful shutdown.
14743
Thierry FOURNIERcc103292015-06-06 19:30:17 +020014744str(<string>) : string
14745 Returns a string.
14746
Willy Tarreau74ca5042013-06-11 23:12:07 +020014747table_avl([<table>]) : integer
14748 Returns the total number of available entries in the current proxy's
14749 stick-table or in the designated stick-table. See also table_cnt.
14750
14751table_cnt([<table>]) : integer
14752 Returns the total number of entries currently in use in the current proxy's
14753 stick-table or in the designated stick-table. See also src_conn_cnt and
14754 table_avl for other entry counting methods.
14755
Christopher Faulet34adb2a2017-11-21 21:45:38 +010014756thread : integer
14757 Returns an integer value corresponding to the position of the thread calling
14758 the function, between 0 and (global.nbthread-1). This is useful for logging
14759 and debugging purposes.
14760
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014761var(<var-name>) : undefined
14762 Returns a variable with the stored type. If the variable is not set, the
Daniel Schneller0b547052016-03-21 20:46:57 +010014763 sample fetch fails. The name of the variable starts with an indication
14764 about its scope. The scopes allowed are:
Christopher Fauletff2613e2016-11-09 11:36:17 +010014765 "proc" : the variable is shared with the whole process
Daniel Schneller0b547052016-03-21 20:46:57 +010014766 "sess" : the variable is shared with the whole session
14767 "txn" : the variable is shared with the transaction (request and
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014768 response),
Daniel Schneller0b547052016-03-21 20:46:57 +010014769 "req" : the variable is shared only during request processing,
14770 "res" : the variable is shared only during response processing.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014771 This prefix is followed by a name. The separator is a '.'. The name may only
Christopher Fauletb71557a2016-10-31 10:49:03 +010014772 contain characters 'a-z', 'A-Z', '0-9', '.' and '_'.
Thierry FOURNIER4834bc72015-06-06 19:29:07 +020014773
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200147747.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020014775----------------------------------
14776
14777The layer 4 usually describes just the transport layer which in haproxy is
14778closest to the connection, where no content is yet made available. The fetch
14779methods described here are usable as low as the "tcp-request connection" rule
14780sets unless they require some future information. Those generally include
14781TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014782the incoming connection. For retrieving a value from a sticky counters, the
14783counter number can be explicitly set as 0, 1, or 2 using the pre-defined
Moemen MHEDHBI9cf46342018-09-25 17:50:53 +020014784"sc0_", "sc1_", or "sc2_" prefix. These three pre-defined prefixes can only be
14785used if MAX_SESS_STKCTR value does not exceed 3, otherwise the counter number
14786can be specified as the first integer argument when using the "sc_" prefix.
14787Starting from "sc_0" to "sc_N" where N is (MAX_SESS_STKCTR-1). An optional
14788table may be specified with the "sc*" form, in which case the currently
14789tracked key will be looked up into this alternate table instead of the table
14790currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014791
Jérôme Magnin35e53a62019-01-16 14:38:37 +010014792bc_http_major : integer
Jérôme Magnin86577422018-12-07 09:03:11 +010014793 Returns the backend connection's HTTP major version encoding, which may be 1
14794 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14795 encoding and not the version present in the request header.
14796
Willy Tarreau74ca5042013-06-11 23:12:07 +020014797be_id : integer
14798 Returns an integer containing the current backend's id. It can be used in
14799 frontends with responses to check which backend processed the request.
14800
Marcin Deranekd2471c22016-12-12 14:08:05 +010014801be_name : string
14802 Returns a string containing the current backend's name. It can be used in
14803 frontends with responses to check which backend processed the request.
14804
Willy Tarreau74ca5042013-06-11 23:12:07 +020014805dst : ip
14806 This is the destination IPv4 address of the connection on the client side,
14807 which is the address the client connected to. It can be useful when running
14808 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
14809 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
Willy Tarreau64ded3d2019-01-23 10:02:15 +010014810 RFC 4291. When the incoming connection passed through address translation or
14811 redirection involving connection tracking, the original destination address
14812 before the redirection will be reported. On Linux systems, the source and
14813 destination may seldom appear reversed if the nf_conntrack_tcp_loose sysctl
14814 is set, because a late response may reopen a timed out connection and switch
14815 what is believed to be the source and the destination.
Willy Tarreau74ca5042013-06-11 23:12:07 +020014816
14817dst_conn : integer
14818 Returns an integer value corresponding to the number of currently established
14819 connections on the same socket including the one being evaluated. It is
14820 normally used with ACLs but can as well be used to pass the information to
14821 servers in an HTTP header or in logs. It can be used to either return a sorry
14822 page before hard-blocking, or to use a specific backend to drain new requests
14823 when the socket is considered saturated. This offers the ability to assign
14824 different limits to different listening ports or addresses. See also the
14825 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010014826
Willy Tarreau16e01562016-08-09 16:46:18 +020014827dst_is_local : boolean
14828 Returns true if the destination address of the incoming connection is local
14829 to the system, or false if the address doesn't exist on the system, meaning
14830 that it was intercepted in transparent mode. It can be useful to apply
14831 certain rules by default to forwarded traffic and other rules to the traffic
Davor Ocelice9ed2812017-12-25 17:49:28 +010014832 targeting the real address of the machine. For example the stats page could
Willy Tarreau16e01562016-08-09 16:46:18 +020014833 be delivered only on this address, or SSH access could be locally redirected.
14834 Please note that the check involves a few system calls, so it's better to do
14835 it only once per connection.
14836
Willy Tarreau74ca5042013-06-11 23:12:07 +020014837dst_port : integer
14838 Returns an integer value corresponding to the destination TCP port of the
14839 connection on the client side, which is the port the client connected to.
14840 This might be used when running in transparent mode, when assigning dynamic
14841 ports to some clients for a whole application session, to stick all users to
14842 a same server, or to pass the destination port information to a server using
14843 an HTTP header.
14844
Willy Tarreau60ca10a2017-08-18 15:26:54 +020014845fc_http_major : integer
14846 Reports the front connection's HTTP major version encoding, which may be 1
14847 for HTTP/0.9 to HTTP/1.1 or 2 for HTTP/2. Note, this is based on the on-wire
14848 encoding and not on the version present in the request header.
14849
Emeric Brun4f603012017-01-05 15:11:44 +010014850fc_rcvd_proxy : boolean
14851 Returns true if the client initiated the connection with a PROXY protocol
14852 header.
14853
Thierry Fournier / OZON.IO6310bef2016-07-24 20:16:50 +020014854fc_rtt(<unit>) : integer
14855 Returns the Round Trip Time (RTT) measured by the kernel for the client
14856 connection. <unit> is facultative, by default the unit is milliseconds. <unit>
14857 can be set to "ms" for milliseconds or "us" for microseconds. If the server
14858 connection is not established, if the connection is not TCP or if the
14859 operating system does not support TCP_INFO, for example Linux kernels before
14860 2.4, the sample fetch fails.
14861
14862fc_rttvar(<unit>) : integer
14863 Returns the Round Trip Time (RTT) variance measured by the kernel for the
14864 client connection. <unit> is facultative, by default the unit is milliseconds.
14865 <unit> can be set to "ms" for milliseconds or "us" for microseconds. If the
14866 server connection is not established, if the connection is not TCP or if the
14867 operating system does not support TCP_INFO, for example Linux kernels before
14868 2.4, the sample fetch fails.
14869
Joe Williams30fcd392016-08-10 07:06:44 -070014870fc_unacked(<unit>) : integer
14871 Returns the unacked counter measured by the kernel for the client connection.
14872 If the server connection is not established, if the connection is not TCP or
14873 if the operating system does not support TCP_INFO, for example Linux kernels
14874 before 2.4, the sample fetch fails.
14875
14876fc_sacked(<unit>) : integer
14877 Returns the sacked counter measured by the kernel for the client connection.
14878 If the server connection is not established, if the connection is not TCP or
14879 if the operating system does not support TCP_INFO, for example Linux kernels
14880 before 2.4, the sample fetch fails.
14881
14882fc_retrans(<unit>) : integer
14883 Returns the retransmits counter measured by the kernel for the client
14884 connection. If the server connection is not established, if the connection is
14885 not TCP or if the operating system does not support TCP_INFO, for example
14886 Linux kernels before 2.4, the sample fetch fails.
14887
14888fc_fackets(<unit>) : integer
14889 Returns the fack counter measured by the kernel for the client
14890 connection. If the server connection is not established, if the connection is
14891 not TCP or if the operating system does not support TCP_INFO, for example
14892 Linux kernels before 2.4, the sample fetch fails.
14893
14894fc_lost(<unit>) : integer
14895 Returns the lost counter measured by the kernel for the client
14896 connection. If the server connection is not established, if the connection is
14897 not TCP or if the operating system does not support TCP_INFO, for example
14898 Linux kernels before 2.4, the sample fetch fails.
14899
14900fc_reordering(<unit>) : integer
14901 Returns the reordering counter measured by the kernel for the client
14902 connection. If the server connection is not established, if the connection is
14903 not TCP or if the operating system does not support TCP_INFO, for example
14904 Linux kernels before 2.4, the sample fetch fails.
14905
Marcin Deranek9a66dfb2018-04-13 14:37:50 +020014906fe_defbe : string
14907 Returns a string containing the frontend's default backend name. It can be
14908 used in frontends to check which backend will handle requests by default.
14909
Willy Tarreau74ca5042013-06-11 23:12:07 +020014910fe_id : integer
14911 Returns an integer containing the current frontend's id. It can be used in
Marcin Deranek6e413ed2016-12-13 12:40:01 +010014912 backends to check from which frontend it was called, or to stick all users
Willy Tarreau74ca5042013-06-11 23:12:07 +020014913 coming via a same frontend to the same server.
14914
Marcin Deranekd2471c22016-12-12 14:08:05 +010014915fe_name : string
14916 Returns a string containing the current frontend's name. It can be used in
14917 backends to check from which frontend it was called, or to stick all users
14918 coming via a same frontend to the same server.
14919
Cyril Bonté62ba8702014-04-22 23:52:25 +020014920sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014921sc0_bytes_in_rate([<table>]) : integer
14922sc1_bytes_in_rate([<table>]) : integer
14923sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014924 Returns the average client-to-server bytes rate from the currently tracked
14925 counters, measured in amount of bytes over the period configured in the
14926 table. See also src_bytes_in_rate.
14927
Cyril Bonté62ba8702014-04-22 23:52:25 +020014928sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014929sc0_bytes_out_rate([<table>]) : integer
14930sc1_bytes_out_rate([<table>]) : integer
14931sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014932 Returns the average server-to-client bytes rate from the currently tracked
14933 counters, measured in amount of bytes over the period configured in the
14934 table. See also src_bytes_out_rate.
14935
Cyril Bonté62ba8702014-04-22 23:52:25 +020014936sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014937sc0_clr_gpc0([<table>]) : integer
14938sc1_clr_gpc0([<table>]) : integer
14939sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020014940 Clears the first General Purpose Counter associated to the currently tracked
14941 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010014942 stored value is zero, so first invocation will always return zero. This is
14943 typically used as a second ACL in an expression in order to mark a connection
14944 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020014945
Jarno Huuskonen676f6222017-03-30 09:19:45 +030014946 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020014947 # block if 5 consecutive requests continue to come faster than 10 sess
14948 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020014949 acl abuse sc0_http_req_rate gt 10
14950 acl kill sc0_inc_gpc0 gt 5
14951 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020014952 tcp-request connection accept if !abuse save
14953 tcp-request connection reject if abuse kill
14954
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014955sc_clr_gpc1(<ctr>[,<table>]) : integer
14956sc0_clr_gpc1([<table>]) : integer
14957sc1_clr_gpc1([<table>]) : integer
14958sc2_clr_gpc1([<table>]) : integer
14959 Clears the second General Purpose Counter associated to the currently tracked
14960 counters, and returns its previous value. Before the first invocation, the
14961 stored value is zero, so first invocation will always return zero. This is
14962 typically used as a second ACL in an expression in order to mark a connection
14963 when a first ACL was verified.
14964
Cyril Bonté62ba8702014-04-22 23:52:25 +020014965sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014966sc0_conn_cnt([<table>]) : integer
14967sc1_conn_cnt([<table>]) : integer
14968sc2_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010014969 Returns the cumulative number of incoming connections from currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020014970 counters. See also src_conn_cnt.
14971
Cyril Bonté62ba8702014-04-22 23:52:25 +020014972sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014973sc0_conn_cur([<table>]) : integer
14974sc1_conn_cur([<table>]) : integer
14975sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014976 Returns the current amount of concurrent connections tracking the same
14977 tracked counters. This number is automatically incremented when tracking
14978 begins and decremented when tracking stops. See also src_conn_cur.
14979
Cyril Bonté62ba8702014-04-22 23:52:25 +020014980sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014981sc0_conn_rate([<table>]) : integer
14982sc1_conn_rate([<table>]) : integer
14983sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014984 Returns the average connection rate from the currently tracked counters,
14985 measured in amount of connections over the period configured in the table.
14986 See also src_conn_rate.
14987
Cyril Bonté62ba8702014-04-22 23:52:25 +020014988sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020014989sc0_get_gpc0([<table>]) : integer
14990sc1_get_gpc0([<table>]) : integer
14991sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020014992 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020014993 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020014994
Frédéric Lécaille6778b272018-01-29 15:22:53 +010014995sc_get_gpc1(<ctr>[,<table>]) : integer
14996sc0_get_gpc1([<table>]) : integer
14997sc1_get_gpc1([<table>]) : integer
14998sc2_get_gpc1([<table>]) : integer
14999 Returns the value of the second General Purpose Counter associated to the
15000 currently tracked counters. See also src_get_gpc1 and sc/sc0/sc1/sc2_inc_gpc1.
15001
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015002sc_get_gpt0(<ctr>[,<table>]) : integer
15003sc0_get_gpt0([<table>]) : integer
15004sc1_get_gpt0([<table>]) : integer
15005sc2_get_gpt0([<table>]) : integer
15006 Returns the value of the first General Purpose Tag associated to the
15007 currently tracked counters. See also src_get_gpt0.
15008
Cyril Bonté62ba8702014-04-22 23:52:25 +020015009sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015010sc0_gpc0_rate([<table>]) : integer
15011sc1_gpc0_rate([<table>]) : integer
15012sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015013 Returns the average increment rate of the first General Purpose Counter
15014 associated to the currently tracked counters. It reports the frequency
15015 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015016 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15017 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15018 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020015019
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015020sc_gpc1_rate(<ctr>[,<table>]) : integer
15021sc0_gpc1_rate([<table>]) : integer
15022sc1_gpc1_rate([<table>]) : integer
15023sc2_gpc1_rate([<table>]) : integer
15024 Returns the average increment rate of the second General Purpose Counter
15025 associated to the currently tracked counters. It reports the frequency
15026 which the gpc1 counter was incremented over the configured period. See also
15027 src_gpcA_rate, sc/sc0/sc1/sc2_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15028 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15029 be returned, as "gpc1" only holds the event count.
15030
Cyril Bonté62ba8702014-04-22 23:52:25 +020015031sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015032sc0_http_err_cnt([<table>]) : integer
15033sc1_http_err_cnt([<table>]) : integer
15034sc2_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015035 Returns the cumulative number of HTTP errors from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015036 counters. This includes the both request errors and 4xx error responses.
15037 See also src_http_err_cnt.
15038
Cyril Bonté62ba8702014-04-22 23:52:25 +020015039sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015040sc0_http_err_rate([<table>]) : integer
15041sc1_http_err_rate([<table>]) : integer
15042sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015043 Returns the average rate of HTTP errors from the currently tracked counters,
15044 measured in amount of errors over the period configured in the table. This
15045 includes the both request errors and 4xx error responses. See also
15046 src_http_err_rate.
15047
Cyril Bonté62ba8702014-04-22 23:52:25 +020015048sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015049sc0_http_req_cnt([<table>]) : integer
15050sc1_http_req_cnt([<table>]) : integer
15051sc2_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015052 Returns the cumulative number of HTTP requests from the currently tracked
Willy Tarreaue9656522010-08-17 15:40:09 +020015053 counters. This includes every started request, valid or not. See also
15054 src_http_req_cnt.
15055
Cyril Bonté62ba8702014-04-22 23:52:25 +020015056sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015057sc0_http_req_rate([<table>]) : integer
15058sc1_http_req_rate([<table>]) : integer
15059sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015060 Returns the average rate of HTTP requests from the currently tracked
15061 counters, measured in amount of requests over the period configured in
15062 the table. This includes every started request, valid or not. See also
15063 src_http_req_rate.
15064
Cyril Bonté62ba8702014-04-22 23:52:25 +020015065sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015066sc0_inc_gpc0([<table>]) : integer
15067sc1_inc_gpc0([<table>]) : integer
15068sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015069 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010015070 tracked counters, and returns its new value. Before the first invocation,
15071 the stored value is zero, so first invocation will increase it to 1 and will
15072 return 1. This is typically used as a second ACL in an expression in order
15073 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020015074
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015075 Example:
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015076 acl abuse sc0_http_req_rate gt 10
15077 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015078 tcp-request connection reject if abuse kill
15079
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015080sc_inc_gpc1(<ctr>[,<table>]) : integer
15081sc0_inc_gpc1([<table>]) : integer
15082sc1_inc_gpc1([<table>]) : integer
15083sc2_inc_gpc1([<table>]) : integer
15084 Increments the second General Purpose Counter associated to the currently
15085 tracked counters, and returns its new value. Before the first invocation,
15086 the stored value is zero, so first invocation will increase it to 1 and will
15087 return 1. This is typically used as a second ACL in an expression in order
15088 to mark a connection when a first ACL was verified.
15089
Cyril Bonté62ba8702014-04-22 23:52:25 +020015090sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015091sc0_kbytes_in([<table>]) : integer
15092sc1_kbytes_in([<table>]) : integer
15093sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015094 Returns the total amount of client-to-server data from the currently tracked
15095 counters, measured in kilobytes. The test is currently performed on 32-bit
15096 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020015097
Cyril Bonté62ba8702014-04-22 23:52:25 +020015098sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015099sc0_kbytes_out([<table>]) : integer
15100sc1_kbytes_out([<table>]) : integer
15101sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015102 Returns the total amount of server-to-client data from the currently tracked
15103 counters, measured in kilobytes. The test is currently performed on 32-bit
15104 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020015105
Cyril Bonté62ba8702014-04-22 23:52:25 +020015106sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015107sc0_sess_cnt([<table>]) : integer
15108sc1_sess_cnt([<table>]) : integer
15109sc2_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015110 Returns the cumulative number of incoming connections that were transformed
Willy Tarreaue9656522010-08-17 15:40:09 +020015111 into sessions, which means that they were accepted by a "tcp-request
15112 connection" rule, from the currently tracked counters. A backend may count
15113 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015114 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020015115 with the client. See also src_sess_cnt.
15116
Cyril Bonté62ba8702014-04-22 23:52:25 +020015117sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015118sc0_sess_rate([<table>]) : integer
15119sc1_sess_rate([<table>]) : integer
15120sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020015121 Returns the average session rate from the currently tracked counters,
15122 measured in amount of sessions over the period configured in the table. A
15123 session is a connection that got past the early "tcp-request connection"
15124 rules. A backend may count more sessions than connections because each
15125 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040015126 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020015127
Cyril Bonté62ba8702014-04-22 23:52:25 +020015128sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020015129sc0_tracked([<table>]) : boolean
15130sc1_tracked([<table>]) : boolean
15131sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020015132 Returns true if the designated session counter is currently being tracked by
15133 the current session. This can be useful when deciding whether or not we want
15134 to set some values in a header passed to the server.
15135
Cyril Bonté62ba8702014-04-22 23:52:25 +020015136sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020015137sc0_trackers([<table>]) : integer
15138sc1_trackers([<table>]) : integer
15139sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010015140 Returns the current amount of concurrent connections tracking the same
15141 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015142 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010015143 that it does not rely on any stored information but on the table's reference
15144 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015145 may sometimes be more suited for layer7 tracking. It can be used to tell a
15146 server how many concurrent connections there are from a given address for
15147 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010015148
Willy Tarreau74ca5042013-06-11 23:12:07 +020015149so_id : integer
15150 Returns an integer containing the current listening socket's id. It is useful
15151 in frontends involving many "bind" lines, or to stick all users coming via a
15152 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015153
Willy Tarreau74ca5042013-06-11 23:12:07 +020015154src : ip
Davor Ocelice9ed2812017-12-25 17:49:28 +010015155 This is the source IPv4 address of the client of the session. It is of type
Willy Tarreau74ca5042013-06-11 23:12:07 +020015156 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
15157 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
15158 TCP-level source address which is used, and not the address of a client
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010015159 behind a proxy. However if the "accept-proxy" or "accept-netscaler-cip" bind
15160 directive is used, it can be the address of a client behind another
15161 PROXY-protocol compatible component for all rule sets except
Willy Tarreau64ded3d2019-01-23 10:02:15 +010015162 "tcp-request connection" which sees the real address. When the incoming
15163 connection passed through address translation or redirection involving
15164 connection tracking, the original destination address before the redirection
15165 will be reported. On Linux systems, the source and destination may seldom
15166 appear reversed if the nf_conntrack_tcp_loose sysctl is set, because a late
15167 response may reopen a timed out connection and switch what is believed to be
15168 the source and the destination.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015169
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010015170 Example:
15171 # add an HTTP header in requests with the originating address' country
15172 http-request set-header X-Country %[src,map_ip(geoip.lst)]
15173
Willy Tarreau74ca5042013-06-11 23:12:07 +020015174src_bytes_in_rate([<table>]) : integer
15175 Returns the average bytes rate from the incoming connection's source address
15176 in the current proxy's stick-table or in the designated stick-table, measured
15177 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015178 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015179
Willy Tarreau74ca5042013-06-11 23:12:07 +020015180src_bytes_out_rate([<table>]) : integer
15181 Returns the average bytes rate to the incoming connection's source address in
15182 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015183 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015184 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015185
Willy Tarreau74ca5042013-06-11 23:12:07 +020015186src_clr_gpc0([<table>]) : integer
15187 Clears the first General Purpose Counter associated to the incoming
15188 connection's source address in the current proxy's stick-table or in the
15189 designated stick-table, and returns its previous value. If the address is not
15190 found, an entry is created and 0 is returned. This is typically used as a
15191 second ACL in an expression in order to mark a connection when a first ACL
15192 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020015193
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015194 Example:
Willy Tarreauf73cd112011-08-13 01:45:16 +020015195 # block if 5 consecutive requests continue to come faster than 10 sess
15196 # per second, and reset the counter as soon as the traffic slows down.
15197 acl abuse src_http_req_rate gt 10
15198 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010015199 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020015200 tcp-request connection accept if !abuse save
15201 tcp-request connection reject if abuse kill
15202
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015203src_clr_gpc1([<table>]) : integer
15204 Clears the second General Purpose Counter associated to the incoming
15205 connection's source address in the current proxy's stick-table or in the
15206 designated stick-table, and returns its previous value. If the address is not
15207 found, an entry is created and 0 is returned. This is typically used as a
15208 second ACL in an expression in order to mark a connection when a first ACL
15209 was verified.
15210
Willy Tarreau74ca5042013-06-11 23:12:07 +020015211src_conn_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015212 Returns the cumulative number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020015213 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015214 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015215 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015216
Willy Tarreau74ca5042013-06-11 23:12:07 +020015217src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015218 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015219 current incoming connection's source address in the current proxy's
15220 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015221 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015222
Willy Tarreau74ca5042013-06-11 23:12:07 +020015223src_conn_rate([<table>]) : integer
15224 Returns the average connection rate from the incoming connection's source
15225 address in the current proxy's stick-table or in the designated stick-table,
15226 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015227 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015228
Willy Tarreau74ca5042013-06-11 23:12:07 +020015229src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020015230 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015231 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020015232 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015233 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015234
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015235src_get_gpc1([<table>]) : integer
15236 Returns the value of the second General Purpose Counter associated to the
15237 incoming connection's source address in the current proxy's stick-table or in
15238 the designated stick-table. If the address is not found, zero is returned.
15239 See also sc/sc0/sc1/sc2_get_gpc1 and src_inc_gpc1.
15240
Thierry FOURNIER236657b2015-08-19 08:25:14 +020015241src_get_gpt0([<table>]) : integer
15242 Returns the value of the first General Purpose Tag associated to the
15243 incoming connection's source address in the current proxy's stick-table or in
15244 the designated stick-table. If the address is not found, zero is returned.
15245 See also sc/sc0/sc1/sc2_get_gpt0.
15246
Willy Tarreau74ca5042013-06-11 23:12:07 +020015247src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015248 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020015249 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015250 stick-table or in the designated stick-table. It reports the frequency
15251 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015252 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
15253 that the "gpc0_rate" counter must be stored in the stick-table for a value to
15254 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020015255
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015256src_gpc1_rate([<table>]) : integer
15257 Returns the average increment rate of the second General Purpose Counter
15258 associated to the incoming connection's source address in the current proxy's
15259 stick-table or in the designated stick-table. It reports the frequency
15260 which the gpc1 counter was incremented over the configured period. See also
15261 sc/sc0/sc1/sc2_gpc1_rate, src_get_gpc1, and sc/sc0/sc1/sc2_inc_gpc1. Note
15262 that the "gpc1_rate" counter must be stored in the stick-table for a value to
15263 be returned, as "gpc1" only holds the event count.
15264
Willy Tarreau74ca5042013-06-11 23:12:07 +020015265src_http_err_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015266 Returns the cumulative number of HTTP errors from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015267 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020015268 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015269 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020015270 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015271
Willy Tarreau74ca5042013-06-11 23:12:07 +020015272src_http_err_rate([<table>]) : integer
15273 Returns the average rate of HTTP errors from the incoming connection's source
15274 address in the current proxy's stick-table or in the designated stick-table,
15275 measured in amount of errors over the period configured in the table. This
15276 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015277 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015278
Willy Tarreau74ca5042013-06-11 23:12:07 +020015279src_http_req_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015280 Returns the cumulative number of HTTP requests from the incoming connection's
Willy Tarreau74ca5042013-06-11 23:12:07 +020015281 source address in the current proxy's stick-table or in the designated stick-
15282 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015283 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015284
Willy Tarreau74ca5042013-06-11 23:12:07 +020015285src_http_req_rate([<table>]) : integer
15286 Returns the average rate of HTTP requests from the incoming connection's
15287 source address in the current proxy's stick-table or in the designated stick-
15288 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020015289 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015290 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015291
Willy Tarreau74ca5042013-06-11 23:12:07 +020015292src_inc_gpc0([<table>]) : integer
15293 Increments the first General Purpose Counter associated to the incoming
15294 connection's source address in the current proxy's stick-table or in the
15295 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015296 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015297 This is typically used as a second ACL in an expression in order to mark a
15298 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020015299
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015300 Example:
Willy Tarreauc9705a12010-07-27 20:05:50 +020015301 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010015302 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020015303 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020015304
Frédéric Lécaille6778b272018-01-29 15:22:53 +010015305src_inc_gpc1([<table>]) : integer
15306 Increments the second General Purpose Counter associated to the incoming
15307 connection's source address in the current proxy's stick-table or in the
15308 designated stick-table, and returns its new value. If the address is not
15309 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc1.
15310 This is typically used as a second ACL in an expression in order to mark a
15311 connection when a first ACL was verified.
15312
Willy Tarreau16e01562016-08-09 16:46:18 +020015313src_is_local : boolean
15314 Returns true if the source address of the incoming connection is local to the
15315 system, or false if the address doesn't exist on the system, meaning that it
15316 comes from a remote machine. Note that UNIX addresses are considered local.
15317 It can be useful to apply certain access restrictions based on where the
Davor Ocelice9ed2812017-12-25 17:49:28 +010015318 client comes from (e.g. require auth or https for remote machines). Please
Willy Tarreau16e01562016-08-09 16:46:18 +020015319 note that the check involves a few system calls, so it's better to do it only
15320 once per connection.
15321
Willy Tarreau74ca5042013-06-11 23:12:07 +020015322src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015323 Returns the total amount of data received from the incoming connection's
15324 source address in the current proxy's stick-table or in the designated
15325 stick-table, measured in kilobytes. If the address is not found, zero is
15326 returned. The test is currently performed on 32-bit integers, which limits
15327 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015328
Willy Tarreau74ca5042013-06-11 23:12:07 +020015329src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020015330 Returns the total amount of data sent to the incoming connection's source
15331 address in the current proxy's stick-table or in the designated stick-table,
15332 measured in kilobytes. If the address is not found, zero is returned. The
15333 test is currently performed on 32-bit integers, which limits values to 4
15334 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015335
Willy Tarreau74ca5042013-06-11 23:12:07 +020015336src_port : integer
15337 Returns an integer value corresponding to the TCP source port of the
15338 connection on the client side, which is the port the client connected from.
15339 Usage of this function is very limited as modern protocols do not care much
15340 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010015341
Willy Tarreau74ca5042013-06-11 23:12:07 +020015342src_sess_cnt([<table>]) : integer
Davor Ocelice9ed2812017-12-25 17:49:28 +010015343 Returns the cumulative number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020015344 connection's source IPv4 address in the current proxy's stick-table or in the
15345 designated stick-table, that were transformed into sessions, which means that
15346 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015347 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015348
Willy Tarreau74ca5042013-06-11 23:12:07 +020015349src_sess_rate([<table>]) : integer
15350 Returns the average session rate from the incoming connection's source
15351 address in the current proxy's stick-table or in the designated stick-table,
15352 measured in amount of sessions over the period configured in the table. A
15353 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020015354 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020015355
Willy Tarreau74ca5042013-06-11 23:12:07 +020015356src_updt_conn_cnt([<table>]) : integer
15357 Creates or updates the entry associated to the incoming connection's source
15358 address in the current proxy's stick-table or in the designated stick-table.
15359 This table must be configured to store the "conn_cnt" data type, otherwise
15360 the match will be ignored. The current count is incremented by one, and the
15361 expiration timer refreshed. The updated count is returned, so this match
15362 can't return zero. This was used to reject service abusers based on their
15363 source address. Note: it is recommended to use the more complete "track-sc*"
15364 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015365
15366 Example :
15367 # This frontend limits incoming SSH connections to 3 per 10 second for
15368 # each source address, and rejects excess connections until a 10 second
15369 # silence is observed. At most 20 addresses are tracked.
15370 listen ssh
15371 bind :22
15372 mode tcp
15373 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020015374 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020015375 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020015376 server local 127.0.0.1:22
15377
Willy Tarreau74ca5042013-06-11 23:12:07 +020015378srv_id : integer
15379 Returns an integer containing the server's id when processing the response.
15380 While it's almost only used with ACLs, it may be used for logging or
15381 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020015382
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200153837.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020015384----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020015385
Willy Tarreau74ca5042013-06-11 23:12:07 +020015386The layer 5 usually describes just the session layer which in haproxy is
15387closest to the session once all the connection handshakes are finished, but
15388when no content is yet made available. The fetch methods described here are
15389usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015390future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020015391
Ben Shillitof25e8e52016-12-02 14:25:37 +00001539251d.all(<prop>[,<prop>*]) : string
15393 Returns values for the properties requested as a string, where values are
15394 separated by the delimiter specified with "51degrees-property-separator".
15395 The device is identified using all the important HTTP headers from the
15396 request. The function can be passed up to five property names, and if a
15397 property name can't be found, the value "NoData" is returned.
15398
15399 Example :
15400 # Here the header "X-51D-DeviceTypeMobileTablet" is added to the request
15401 # containing the three properties requested using all relevant headers from
15402 # the request.
15403 frontend http-in
15404 bind *:8081
15405 default_backend servers
15406 http-request set-header X-51D-DeviceTypeMobileTablet \
15407 %[51d.all(DeviceType,IsMobile,IsTablet)]
15408
Emeric Brun645ae792014-04-30 14:21:06 +020015409ssl_bc : boolean
15410 Returns true when the back connection was made via an SSL/TLS transport
15411 layer and is locally deciphered. This means the outgoing connection was made
15412 other a server with the "ssl" option.
15413
15414ssl_bc_alg_keysize : integer
15415 Returns the symmetric cipher key size supported in bits when the outgoing
15416 connection was made over an SSL/TLS transport layer.
15417
Olivier Houchard6b77f492018-11-22 18:18:29 +010015418ssl_bc_alpn : string
15419 This extracts the Application Layer Protocol Negotiation field from an
15420 outgoing connection made via a TLS transport layer.
Michael Prokop4438c602019-05-24 10:25:45 +020015421 The result is a string containing the protocol name negotiated with the
Olivier Houchard6b77f492018-11-22 18:18:29 +010015422 server. The SSL library must have been built with support for TLS
15423 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15424 not advertised unless the "alpn" keyword on the "server" line specifies a
15425 protocol list. Also, nothing forces the server to pick a protocol from this
15426 list, any other one may be requested. The TLS ALPN extension is meant to
15427 replace the TLS NPN extension. See also "ssl_bc_npn".
15428
Emeric Brun645ae792014-04-30 14:21:06 +020015429ssl_bc_cipher : string
15430 Returns the name of the used cipher when the outgoing connection was made
15431 over an SSL/TLS transport layer.
15432
Emeric Brun74f7ffa2018-02-19 16:14:12 +010015433ssl_bc_is_resumed : boolean
15434 Returns true when the back connection was made over an SSL/TLS transport
15435 layer and the newly created SSL session was resumed using a cached
15436 session or a TLS ticket.
15437
Olivier Houchard6b77f492018-11-22 18:18:29 +010015438ssl_bc_npn : string
15439 This extracts the Next Protocol Negotiation field from an outgoing connection
15440 made via a TLS transport layer. The result is a string containing the
Michael Prokop4438c602019-05-24 10:25:45 +020015441 protocol name negotiated with the server . The SSL library must have been
Olivier Houchard6b77f492018-11-22 18:18:29 +010015442 built with support for TLS extensions enabled (check haproxy -vv). Note that
15443 the TLS NPN extension is not advertised unless the "npn" keyword on the
15444 "server" line specifies a protocol list. Also, nothing forces the server to
15445 pick a protocol from this list, any other one may be used. Please note that
15446 the TLS NPN extension was replaced with ALPN.
15447
Emeric Brun645ae792014-04-30 14:21:06 +020015448ssl_bc_protocol : string
15449 Returns the name of the used protocol when the outgoing connection was made
15450 over an SSL/TLS transport layer.
15451
Emeric Brunb73a9b02014-04-30 18:49:19 +020015452ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020015453 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015454 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15455 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020015456
15457ssl_bc_session_id : binary
15458 Returns the SSL ID of the back connection when the outgoing connection was
15459 made over an SSL/TLS transport layer. It is useful to log if we want to know
15460 if session was reused or not.
15461
Patrick Hemmere0275472018-04-28 19:15:51 -040015462ssl_bc_session_key : binary
15463 Returns the SSL session master key of the back connection when the outgoing
15464 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15465 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15466 BoringSSL.
15467
Emeric Brun645ae792014-04-30 14:21:06 +020015468ssl_bc_use_keysize : integer
15469 Returns the symmetric cipher key size used in bits when the outgoing
15470 connection was made over an SSL/TLS transport layer.
15471
Willy Tarreau74ca5042013-06-11 23:12:07 +020015472ssl_c_ca_err : integer
15473 When the incoming connection was made over an SSL/TLS transport layer,
15474 returns the ID of the first error detected during verification of the client
15475 certificate at depth > 0, or 0 if no error was encountered during this
15476 verification process. Please refer to your SSL library's documentation to
15477 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020015478
Willy Tarreau74ca5042013-06-11 23:12:07 +020015479ssl_c_ca_err_depth : integer
15480 When the incoming connection was made over an SSL/TLS transport layer,
15481 returns the depth in the CA chain of the first error detected during the
15482 verification of the client certificate. If no error is encountered, 0 is
15483 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010015484
Emeric Brun43e79582014-10-29 19:03:26 +010015485ssl_c_der : binary
15486 Returns the DER formatted certificate presented by the client when the
15487 incoming connection was made over an SSL/TLS transport layer. When used for
15488 an ACL, the value(s) to match against can be passed in hexadecimal form.
15489
Willy Tarreau74ca5042013-06-11 23:12:07 +020015490ssl_c_err : integer
15491 When the incoming connection was made over an SSL/TLS transport layer,
15492 returns the ID of the first error detected during verification at depth 0, or
15493 0 if no error was encountered during this verification process. Please refer
15494 to your SSL library's documentation to find the exhaustive list of error
15495 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020015496
Willy Tarreau74ca5042013-06-11 23:12:07 +020015497ssl_c_i_dn([<entry>[,<occ>]]) : string
15498 When the incoming connection was made over an SSL/TLS transport layer,
15499 returns the full distinguished name of the issuer of the certificate
15500 presented by the client when no <entry> is specified, or the value of the
15501 first given entry found from the beginning of the DN. If a positive/negative
15502 occurrence number is specified as the optional second argument, it returns
15503 the value of the nth given entry value from the beginning/end of the DN.
15504 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
15505 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020015506
Willy Tarreau74ca5042013-06-11 23:12:07 +020015507ssl_c_key_alg : string
15508 Returns the name of the algorithm used to generate the key of the certificate
15509 presented by the client when the incoming connection was made over an SSL/TLS
15510 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020015511
Willy Tarreau74ca5042013-06-11 23:12:07 +020015512ssl_c_notafter : string
15513 Returns the end date presented by the client as a formatted string
15514 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15515 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020015516
Willy Tarreau74ca5042013-06-11 23:12:07 +020015517ssl_c_notbefore : string
15518 Returns the start date presented by the client as a formatted string
15519 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15520 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015521
Willy Tarreau74ca5042013-06-11 23:12:07 +020015522ssl_c_s_dn([<entry>[,<occ>]]) : string
15523 When the incoming connection was made over an SSL/TLS transport layer,
15524 returns the full distinguished name of the subject of the certificate
15525 presented by the client when no <entry> is specified, or the value of the
15526 first given entry found from the beginning of the DN. If a positive/negative
15527 occurrence number is specified as the optional second argument, it returns
15528 the value of the nth given entry value from the beginning/end of the DN.
15529 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
15530 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010015531
Willy Tarreau74ca5042013-06-11 23:12:07 +020015532ssl_c_serial : binary
15533 Returns the serial of the certificate presented by the client when the
15534 incoming connection was made over an SSL/TLS transport layer. When used for
15535 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015536
Willy Tarreau74ca5042013-06-11 23:12:07 +020015537ssl_c_sha1 : binary
15538 Returns the SHA-1 fingerprint of the certificate presented by the client when
15539 the incoming connection was made over an SSL/TLS transport layer. This can be
15540 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015541 Note that the output is binary, so if you want to pass that signature to the
15542 server, you need to encode it in hex or base64, such as in the example below:
15543
Jarno Huuskonen676f6222017-03-30 09:19:45 +030015544 Example:
Willy Tarreau2d0caa32014-07-02 19:01:22 +020015545 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020015546
Willy Tarreau74ca5042013-06-11 23:12:07 +020015547ssl_c_sig_alg : string
15548 Returns the name of the algorithm used to sign the certificate presented by
15549 the client when the incoming connection was made over an SSL/TLS transport
15550 layer.
Emeric Brun87855892012-10-17 17:39:35 +020015551
Willy Tarreau74ca5042013-06-11 23:12:07 +020015552ssl_c_used : boolean
15553 Returns true if current SSL session uses a client certificate even if current
15554 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020015555
Willy Tarreau74ca5042013-06-11 23:12:07 +020015556ssl_c_verify : integer
15557 Returns the verify result error ID when the incoming connection was made over
15558 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
15559 refer to your SSL library's documentation for an exhaustive list of error
15560 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020015561
Willy Tarreau74ca5042013-06-11 23:12:07 +020015562ssl_c_version : integer
15563 Returns the version of the certificate presented by the client when the
15564 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020015565
Emeric Brun43e79582014-10-29 19:03:26 +010015566ssl_f_der : binary
15567 Returns the DER formatted certificate presented by the frontend when the
15568 incoming connection was made over an SSL/TLS transport layer. When used for
15569 an ACL, the value(s) to match against can be passed in hexadecimal form.
15570
Willy Tarreau74ca5042013-06-11 23:12:07 +020015571ssl_f_i_dn([<entry>[,<occ>]]) : string
15572 When the incoming connection was made over an SSL/TLS transport layer,
15573 returns the full distinguished name of the issuer of the certificate
15574 presented by the frontend when no <entry> is specified, or the value of the
15575 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020015576 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020015577 the value of the nth given entry value from the beginning/end of the DN.
15578 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
15579 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020015580
Willy Tarreau74ca5042013-06-11 23:12:07 +020015581ssl_f_key_alg : string
15582 Returns the name of the algorithm used to generate the key of the certificate
15583 presented by the frontend when the incoming connection was made over an
15584 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015585
Willy Tarreau74ca5042013-06-11 23:12:07 +020015586ssl_f_notafter : string
15587 Returns the end date presented by the frontend as a formatted string
15588 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15589 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015590
Willy Tarreau74ca5042013-06-11 23:12:07 +020015591ssl_f_notbefore : string
15592 Returns the start date presented by the frontend as a formatted string
15593 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
15594 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020015595
Willy Tarreau74ca5042013-06-11 23:12:07 +020015596ssl_f_s_dn([<entry>[,<occ>]]) : string
15597 When the incoming connection was made over an SSL/TLS transport layer,
15598 returns the full distinguished name of the subject of the certificate
15599 presented by the frontend when no <entry> is specified, or the value of the
15600 first given entry found from the beginning of the DN. If a positive/negative
15601 occurrence number is specified as the optional second argument, it returns
15602 the value of the nth given entry value from the beginning/end of the DN.
15603 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
15604 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020015605
Willy Tarreau74ca5042013-06-11 23:12:07 +020015606ssl_f_serial : binary
15607 Returns the serial of the certificate presented by the frontend when the
15608 incoming connection was made over an SSL/TLS transport layer. When used for
15609 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020015610
Emeric Brun55f4fa82014-04-30 17:11:25 +020015611ssl_f_sha1 : binary
15612 Returns the SHA-1 fingerprint of the certificate presented by the frontend
15613 when the incoming connection was made over an SSL/TLS transport layer. This
15614 can be used to know which certificate was chosen using SNI.
15615
Willy Tarreau74ca5042013-06-11 23:12:07 +020015616ssl_f_sig_alg : string
15617 Returns the name of the algorithm used to sign the certificate presented by
15618 the frontend when the incoming connection was made over an SSL/TLS transport
15619 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020015620
Willy Tarreau74ca5042013-06-11 23:12:07 +020015621ssl_f_version : integer
15622 Returns the version of the certificate presented by the frontend when the
15623 incoming connection was made over an SSL/TLS transport layer.
15624
15625ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015626 Returns true when the front connection was made via an SSL/TLS transport
15627 layer and is locally deciphered. This means it has matched a socket declared
15628 with a "bind" line having the "ssl" option.
15629
Willy Tarreau74ca5042013-06-11 23:12:07 +020015630 Example :
15631 # This passes "X-Proto: https" to servers when client connects over SSL
15632 listen http-https
15633 bind :80
15634 bind :443 ssl crt /etc/haproxy.pem
15635 http-request add-header X-Proto https if { ssl_fc }
15636
15637ssl_fc_alg_keysize : integer
15638 Returns the symmetric cipher key size supported in bits when the incoming
15639 connection was made over an SSL/TLS transport layer.
15640
15641ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015642 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020015643 incoming connection made via a TLS transport layer and locally deciphered by
15644 haproxy. The result is a string containing the protocol name advertised by
15645 the client. The SSL library must have been built with support for TLS
15646 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
15647 not advertised unless the "alpn" keyword on the "bind" line specifies a
15648 protocol list. Also, nothing forces the client to pick a protocol from this
15649 list, any other one may be requested. The TLS ALPN extension is meant to
15650 replace the TLS NPN extension. See also "ssl_fc_npn".
15651
Willy Tarreau74ca5042013-06-11 23:12:07 +020015652ssl_fc_cipher : string
15653 Returns the name of the used cipher when the incoming connection was made
15654 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020015655
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015656ssl_fc_cipherlist_bin : binary
15657 Returns the binary form of the client hello cipher list. The maximum returned
15658 value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015659 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015660
15661ssl_fc_cipherlist_hex : string
15662 Returns the binary form of the client hello cipher list encoded as
15663 hexadecimal. The maximum returned value length is according with the value of
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015664 "tune.ssl.capture-cipherlist-size".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015665
15666ssl_fc_cipherlist_str : string
15667 Returns the decoded text form of the client hello cipher list. The maximum
15668 number of ciphers returned is according with the value of
15669 "tune.ssl.capture-cipherlist-size". Note that this sample-fetch is only
Davor Ocelice9ed2812017-12-25 17:49:28 +010015670 available with OpenSSL >= 1.0.2. If the function is not enabled, this
Emmanuel Hocdetddcde192017-09-01 17:32:08 +020015671 sample-fetch returns the hash like "ssl_fc_cipherlist_xxh".
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015672
15673ssl_fc_cipherlist_xxh : integer
15674 Returns a xxh64 of the cipher list. This hash can be return only is the value
15675 "tune.ssl.capture-cipherlist-size" is set greater than 0, however the hash
Emmanuel Hocdetaaee7502017-03-07 18:34:58 +010015676 take in account all the data of the cipher list.
Thierry FOURNIER5bf77322017-02-25 12:45:22 +010015677
Willy Tarreau74ca5042013-06-11 23:12:07 +020015678ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020015679 Returns true if a client certificate is present in an incoming connection over
15680 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010015681 Note: on SSL session resumption with Session ID or TLS ticket, client
15682 certificate is not present in the current connection but may be retrieved
15683 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
15684 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020015685
Olivier Houchardccaa7de2017-10-02 11:51:03 +020015686ssl_fc_has_early : boolean
15687 Returns true if early data were sent, and the handshake didn't happen yet. As
15688 it has security implications, it is useful to be able to refuse those, or
15689 wait until the handshake happened.
15690
Willy Tarreau74ca5042013-06-11 23:12:07 +020015691ssl_fc_has_sni : boolean
15692 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020015693 in an incoming connection was made over an SSL/TLS transport layer. Returns
15694 true when the incoming connection presents a TLS SNI field. This requires
15695 that the SSL library is build with support for TLS extensions enabled (check
15696 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020015697
Nenad Merdanovic1516fe32016-05-17 03:31:21 +020015698ssl_fc_is_resumed : boolean
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015699 Returns true if the SSL/TLS session has been resumed through the use of
Jérôme Magnin4a326cb2018-01-15 14:01:17 +010015700 SSL session cache or TLS tickets on an incoming connection over an SSL/TLS
15701 transport layer.
Nenad Merdanovic26ea8222015-05-18 02:28:57 +020015702
Willy Tarreau74ca5042013-06-11 23:12:07 +020015703ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030015704 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020015705 made via a TLS transport layer and locally deciphered by haproxy. The result
15706 is a string containing the protocol name advertised by the client. The SSL
15707 library must have been built with support for TLS extensions enabled (check
15708 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
15709 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
15710 forces the client to pick a protocol from this list, any other one may be
15711 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020015712
Willy Tarreau74ca5042013-06-11 23:12:07 +020015713ssl_fc_protocol : string
15714 Returns the name of the used protocol when the incoming connection was made
15715 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020015716
Emeric Brunb73a9b02014-04-30 18:49:19 +020015717ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040015718 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020015719 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
15720 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040015721
Willy Tarreau74ca5042013-06-11 23:12:07 +020015722ssl_fc_session_id : binary
15723 Returns the SSL ID of the front connection when the incoming connection was
15724 made over an SSL/TLS transport layer. It is useful to stick a given client to
15725 a server. It is important to note that some browsers refresh their session ID
15726 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020015727
Patrick Hemmere0275472018-04-28 19:15:51 -040015728ssl_fc_session_key : binary
15729 Returns the SSL session master key of the front connection when the incoming
15730 connection was made over an SSL/TLS transport layer. It is useful to decrypt
15731 traffic sent using ephemeral ciphers. This requires OpenSSL >= 1.1.0, or
15732 BoringSSL.
15733
15734
Willy Tarreau74ca5042013-06-11 23:12:07 +020015735ssl_fc_sni : string
15736 This extracts the Server Name Indication TLS extension (SNI) field from an
15737 incoming connection made via an SSL/TLS transport layer and locally
15738 deciphered by haproxy. The result (when present) typically is a string
15739 matching the HTTPS host name (253 chars or less). The SSL library must have
15740 been built with support for TLS extensions enabled (check haproxy -vv).
15741
15742 This fetch is different from "req_ssl_sni" above in that it applies to the
15743 connection being deciphered by haproxy and not to SSL contents being blindly
15744 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020015745 requires that the SSL library is build with support for TLS extensions
15746 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020015747
Willy Tarreau74ca5042013-06-11 23:12:07 +020015748 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020015749 ssl_fc_sni_end : suffix match
15750 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020015751
Willy Tarreau74ca5042013-06-11 23:12:07 +020015752ssl_fc_use_keysize : integer
15753 Returns the symmetric cipher key size used in bits when the incoming
15754 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015755
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015756
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200157577.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020015758------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020015759
Willy Tarreau74ca5042013-06-11 23:12:07 +020015760Fetching samples from buffer contents is a bit different from the previous
15761sample fetches above because the sampled data are ephemeral. These data can
15762only be used when they're available and will be lost when they're forwarded.
15763For this reason, samples fetched from buffer contents during a request cannot
15764be used in a response for example. Even while the data are being fetched, they
15765can change. Sometimes it is necessary to set some delays or combine multiple
15766sample fetch methods to ensure that the expected data are complete and usable,
15767for example through TCP request content inspection. Please see the "tcp-request
15768content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020015769
Willy Tarreau74ca5042013-06-11 23:12:07 +020015770payload(<offset>,<length>) : binary (deprecated)
Davor Ocelice9ed2812017-12-25 17:49:28 +010015771 This is an alias for "req.payload" when used in the context of a request (e.g.
Willy Tarreau74ca5042013-06-11 23:12:07 +020015772 "stick on", "stick match"), and for "res.payload" when used in the context of
15773 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015774
Willy Tarreau74ca5042013-06-11 23:12:07 +020015775payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
15776 This is an alias for "req.payload_lv" when used in the context of a request
Davor Ocelice9ed2812017-12-25 17:49:28 +010015777 (e.g. "stick on", "stick match"), and for "res.payload_lv" when used in the
Willy Tarreau74ca5042013-06-11 23:12:07 +020015778 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010015779
Thierry FOURNIERd7d88812017-04-19 15:15:14 +020015780req.hdrs : string
15781 Returns the current request headers as string including the last empty line
15782 separating headers from the request body. The last empty line can be used to
15783 detect a truncated header block. This sample fetch is useful for some SPOE
15784 headers analyzers and for advanced logging.
15785
Thierry FOURNIER5617dce2017-04-09 05:38:19 +020015786req.hdrs_bin : binary
15787 Returns the current request headers contained in preparsed binary form. This
15788 is useful for offloading some processing with SPOE. Each string is described
15789 by a length followed by the number of bytes indicated in the length. The
15790 length is represented using the variable integer encoding detailed in the
15791 SPOE documentation. The end of the list is marked by a couple of empty header
15792 names and values (length of 0 for both).
15793
15794 *(<str:header-name><str:header-value>)<empty string><empty string>
15795
15796 int: refer to the SPOE documentation for the encoding
15797 str: <int:length><bytes>
15798
Willy Tarreau74ca5042013-06-11 23:12:07 +020015799req.len : integer
15800req_len : integer (deprecated)
15801 Returns an integer value corresponding to the number of bytes present in the
15802 request buffer. This is mostly used in ACL. It is important to understand
15803 that this test does not return false as long as the buffer is changing. This
15804 means that a check with equality to zero will almost always immediately match
15805 at the beginning of the session, while a test for more data will wait for
15806 that data to come in and return false only when haproxy is certain that no
15807 more data will come in. This test was designed to be used with TCP request
15808 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015809
Willy Tarreau74ca5042013-06-11 23:12:07 +020015810req.payload(<offset>,<length>) : binary
15811 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015812 in the request buffer. As a special case, if the <length> argument is zero,
15813 the the whole buffer from <offset> to the end is extracted. This can be used
15814 with ACLs in order to check for the presence of some content in a buffer at
15815 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015816
Willy Tarreau74ca5042013-06-11 23:12:07 +020015817 ACL alternatives :
15818 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015819
Willy Tarreau74ca5042013-06-11 23:12:07 +020015820req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
15821 This extracts a binary block whose size is specified at <offset1> for <length>
15822 bytes, and which starts at <offset2> if specified or just after the length in
15823 the request buffer. The <offset2> parameter also supports relative offsets if
15824 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015825
Willy Tarreau74ca5042013-06-11 23:12:07 +020015826 ACL alternatives :
15827 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015828
Willy Tarreau74ca5042013-06-11 23:12:07 +020015829 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015830
Willy Tarreau74ca5042013-06-11 23:12:07 +020015831req.proto_http : boolean
15832req_proto_http : boolean (deprecated)
15833 Returns true when data in the request buffer look like HTTP and correctly
15834 parses as such. It is the same parser as the common HTTP request parser which
15835 is used so there should be no surprises. The test does not match until the
15836 request is complete, failed or timed out. This test may be used to report the
15837 protocol in TCP logs, but the biggest use is to block TCP request analysis
15838 until a complete HTTP request is present in the buffer, for example to track
15839 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015840
Willy Tarreau74ca5042013-06-11 23:12:07 +020015841 Example:
15842 # track request counts per "base" (concatenation of Host+URL)
15843 tcp-request inspect-delay 10s
15844 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020015845 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020015846
Willy Tarreau74ca5042013-06-11 23:12:07 +020015847req.rdp_cookie([<name>]) : string
15848rdp_cookie([<name>]) : string (deprecated)
15849 When the request buffer looks like the RDP protocol, extracts the RDP cookie
15850 <name>, or any cookie if unspecified. The parser only checks for the first
15851 cookie, as illustrated in the RDP protocol specification. The cookie name is
15852 case insensitive. Generally the "MSTS" cookie name will be used, as it can
15853 contain the user name of the client connecting to the server if properly
15854 configured on the client. The "MSTSHASH" cookie is often used as well for
15855 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015856
Willy Tarreau74ca5042013-06-11 23:12:07 +020015857 This differs from "balance rdp-cookie" in that any balancing algorithm may be
15858 used and thus the distribution of clients to backend servers is not linked to
15859 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
15860 such as "balance roundrobin" or "balance leastconn" will lead to a more even
15861 distribution of clients to backend servers than the hash used by "balance
15862 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015863
Willy Tarreau74ca5042013-06-11 23:12:07 +020015864 ACL derivatives :
15865 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015866
Willy Tarreau74ca5042013-06-11 23:12:07 +020015867 Example :
15868 listen tse-farm
15869 bind 0.0.0.0:3389
15870 # wait up to 5s for an RDP cookie in the request
15871 tcp-request inspect-delay 5s
15872 tcp-request content accept if RDP_COOKIE
15873 # apply RDP cookie persistence
15874 persist rdp-cookie
15875 # Persist based on the mstshash cookie
15876 # This is only useful makes sense if
15877 # balance rdp-cookie is not used
15878 stick-table type string size 204800
15879 stick on req.rdp_cookie(mstshash)
15880 server srv1 1.1.1.1:3389
15881 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015882
Willy Tarreau74ca5042013-06-11 23:12:07 +020015883 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
15884 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015885
Willy Tarreau74ca5042013-06-11 23:12:07 +020015886req.rdp_cookie_cnt([name]) : integer
15887rdp_cookie_cnt([name]) : integer (deprecated)
15888 Tries to parse the request buffer as RDP protocol, then returns an integer
15889 corresponding to the number of RDP cookies found. If an optional cookie name
15890 is passed, only cookies matching this name are considered. This is mostly
15891 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015892
Willy Tarreau74ca5042013-06-11 23:12:07 +020015893 ACL derivatives :
15894 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015895
Alex Zorin4afdd132018-12-30 13:56:28 +110015896req.ssl_alpn : string
15897 Returns a string containing the values of the Application-Layer Protocol
15898 Negotiation (ALPN) TLS extension (RFC7301), sent by the client within the SSL
15899 ClientHello message. Note that this only applies to raw contents found in the
15900 request buffer and not to the contents deciphered via an SSL data layer, so
15901 this will not work with "bind" lines having the "ssl" option. This is useful
15902 in ACL to make a routing decision based upon the ALPN preferences of a TLS
Jarno Huuskonene504f812019-01-03 07:56:49 +020015903 client, like in the example below. See also "ssl_fc_alpn".
Alex Zorin4afdd132018-12-30 13:56:28 +110015904
15905 Examples :
15906 # Wait for a client hello for at most 5 seconds
15907 tcp-request inspect-delay 5s
15908 tcp-request content accept if { req_ssl_hello_type 1 }
Jarno Huuskonene504f812019-01-03 07:56:49 +020015909 use_backend bk_acme if { req.ssl_alpn acme-tls/1 }
Alex Zorin4afdd132018-12-30 13:56:28 +110015910 default_backend bk_default
15911
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020015912req.ssl_ec_ext : boolean
15913 Returns a boolean identifying if client sent the Supported Elliptic Curves
15914 Extension as defined in RFC4492, section 5.1. within the SSL ClientHello
Cyril Bonté307ee1e2015-09-28 23:16:06 +020015915 message. This can be used to present ECC compatible clients with EC
15916 certificate and to use RSA for all others, on the same IP address. Note that
15917 this only applies to raw contents found in the request buffer and not to
15918 contents deciphered via an SSL data layer, so this will not work with "bind"
15919 lines having the "ssl" option.
Nenad Merdanovic5fc7d7e2015-07-07 22:00:17 +020015920
Willy Tarreau74ca5042013-06-11 23:12:07 +020015921req.ssl_hello_type : integer
15922req_ssl_hello_type : integer (deprecated)
15923 Returns an integer value containing the type of the SSL hello message found
15924 in the request buffer if the buffer contains data that parse as a complete
15925 SSL (v3 or superior) client hello message. Note that this only applies to raw
15926 contents found in the request buffer and not to contents deciphered via an
15927 SSL data layer, so this will not work with "bind" lines having the "ssl"
15928 option. This is mostly used in ACL to detect presence of an SSL hello message
15929 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015930
Willy Tarreau74ca5042013-06-11 23:12:07 +020015931req.ssl_sni : string
15932req_ssl_sni : string (deprecated)
15933 Returns a string containing the value of the Server Name TLS extension sent
15934 by a client in a TLS stream passing through the request buffer if the buffer
15935 contains data that parse as a complete SSL (v3 or superior) client hello
15936 message. Note that this only applies to raw contents found in the request
15937 buffer and not to contents deciphered via an SSL data layer, so this will not
15938 work with "bind" lines having the "ssl" option. SNI normally contains the
15939 name of the host the client tries to connect to (for recent browsers). SNI is
15940 useful for allowing or denying access to certain hosts when SSL/TLS is used
15941 by the client. This test was designed to be used with TCP request content
15942 inspection. If content switching is needed, it is recommended to first wait
15943 for a complete client hello (type 1), like in the example below. See also
15944 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015945
Willy Tarreau74ca5042013-06-11 23:12:07 +020015946 ACL derivatives :
15947 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015948
Willy Tarreau74ca5042013-06-11 23:12:07 +020015949 Examples :
15950 # Wait for a client hello for at most 5 seconds
15951 tcp-request inspect-delay 5s
15952 tcp-request content accept if { req_ssl_hello_type 1 }
15953 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
15954 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020015955
Pradeep Jindalbb2acf52015-09-29 10:12:57 +053015956req.ssl_st_ext : integer
15957 Returns 0 if the client didn't send a SessionTicket TLS Extension (RFC5077)
15958 Returns 1 if the client sent SessionTicket TLS Extension
15959 Returns 2 if the client also sent non-zero length TLS SessionTicket
15960 Note that this only applies to raw contents found in the request buffer and
15961 not to contents deciphered via an SSL data layer, so this will not work with
15962 "bind" lines having the "ssl" option. This can for example be used to detect
15963 whether the client sent a SessionTicket or not and stick it accordingly, if
15964 no SessionTicket then stick on SessionID or don't stick as there's no server
15965 side state is there when SessionTickets are in use.
15966
Willy Tarreau74ca5042013-06-11 23:12:07 +020015967req.ssl_ver : integer
15968req_ssl_ver : integer (deprecated)
15969 Returns an integer value containing the version of the SSL/TLS protocol of a
15970 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
15971 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
15972 composed of the major version multiplied by 65536, added to the minor
15973 version. Note that this only applies to raw contents found in the request
15974 buffer and not to contents deciphered via an SSL data layer, so this will not
15975 work with "bind" lines having the "ssl" option. The ACL version of the test
Davor Ocelice9ed2812017-12-25 17:49:28 +010015976 matches against a decimal notation in the form MAJOR.MINOR (e.g. 3.1). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020015977 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015978
Willy Tarreau74ca5042013-06-11 23:12:07 +020015979 ACL derivatives :
15980 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010015981
Willy Tarreau47e8eba2013-09-11 23:28:46 +020015982res.len : integer
15983 Returns an integer value corresponding to the number of bytes present in the
15984 response buffer. This is mostly used in ACL. It is important to understand
15985 that this test does not return false as long as the buffer is changing. This
15986 means that a check with equality to zero will almost always immediately match
15987 at the beginning of the session, while a test for more data will wait for
15988 that data to come in and return false only when haproxy is certain that no
15989 more data will come in. This test was designed to be used with TCP response
15990 content inspection.
15991
Willy Tarreau74ca5042013-06-11 23:12:07 +020015992res.payload(<offset>,<length>) : binary
15993 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020015994 in the response buffer. As a special case, if the <length> argument is zero,
15995 the the whole buffer from <offset> to the end is extracted. This can be used
15996 with ACLs in order to check for the presence of some content in a buffer at
15997 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010015998
Willy Tarreau74ca5042013-06-11 23:12:07 +020015999res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
16000 This extracts a binary block whose size is specified at <offset1> for <length>
16001 bytes, and which starts at <offset2> if specified or just after the length in
16002 the response buffer. The <offset2> parameter also supports relative offsets
16003 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016004
Willy Tarreau74ca5042013-06-11 23:12:07 +020016005 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016006
Willy Tarreau971f7b62015-09-29 14:06:59 +020016007res.ssl_hello_type : integer
16008rep_ssl_hello_type : integer (deprecated)
16009 Returns an integer value containing the type of the SSL hello message found
16010 in the response buffer if the buffer contains data that parses as a complete
16011 SSL (v3 or superior) hello message. Note that this only applies to raw
16012 contents found in the response buffer and not to contents deciphered via an
16013 SSL data layer, so this will not work with "server" lines having the "ssl"
16014 option. This is mostly used in ACL to detect presence of an SSL hello message
16015 that is supposed to contain an SSL session ID usable for stickiness.
16016
Willy Tarreau74ca5042013-06-11 23:12:07 +020016017wait_end : boolean
16018 This fetch either returns true when the inspection period is over, or does
16019 not fetch. It is only used in ACLs, in conjunction with content analysis to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016020 avoid returning a wrong verdict early. It may also be used to delay some
Willy Tarreau74ca5042013-06-11 23:12:07 +020016021 actions, such as a delayed reject for some special addresses. Since it either
16022 stops the rules evaluation or immediately returns true, it is recommended to
Davor Ocelice9ed2812017-12-25 17:49:28 +010016023 use this acl as the last one in a rule. Please note that the default ACL
Willy Tarreau74ca5042013-06-11 23:12:07 +020016024 "WAIT_END" is always usable without prior declaration. This test was designed
16025 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016026
Willy Tarreau74ca5042013-06-11 23:12:07 +020016027 Examples :
16028 # delay every incoming request by 2 seconds
16029 tcp-request inspect-delay 2s
16030 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010016031
Willy Tarreau74ca5042013-06-11 23:12:07 +020016032 # don't immediately tell bad guys they are rejected
16033 tcp-request inspect-delay 10s
16034 acl goodguys src 10.0.0.0/24
16035 acl badguys src 10.0.1.0/24
16036 tcp-request content accept if goodguys
16037 tcp-request content reject if badguys WAIT_END
16038 tcp-request content reject
16039
16040
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200160417.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020016042--------------------------------------
16043
16044It is possible to fetch samples from HTTP contents, requests and responses.
16045This application layer is also called layer 7. It is only possible to fetch the
16046data in this section when a full HTTP request or response has been parsed from
16047its respective request or response buffer. This is always the case with all
16048HTTP specific rules and for sections running with "mode http". When using TCP
16049content inspection, it may be necessary to support an inspection delay in order
16050to let the request or response come in first. These fetches may require a bit
16051more CPU resources than the layer 4 ones, but not much since the request and
16052response are indexed.
16053
16054base : string
16055 This returns the concatenation of the first Host header and the path part of
16056 the request, which starts at the first slash and ends before the question
16057 mark. It can be useful in virtual hosted environments to detect URL abuses as
16058 well as to improve shared caches efficiency. Using this with a limited size
16059 stick table also allows one to collect statistics about most commonly
16060 requested objects by host/path. With ACLs it can allow simple content
16061 switching rules involving the host and the path at the same time, such as
16062 "www.example.com/favicon.ico". See also "path" and "uri".
16063
16064 ACL derivatives :
16065 base : exact string match
16066 base_beg : prefix match
16067 base_dir : subdir match
16068 base_dom : domain match
16069 base_end : suffix match
16070 base_len : length match
16071 base_reg : regex match
16072 base_sub : substring match
16073
16074base32 : integer
16075 This returns a 32-bit hash of the value returned by the "base" fetch method
16076 above. This is useful to track per-URL activity on high traffic sites without
16077 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020016078 memory. The output type is an unsigned integer. The hash function used is
16079 SDBM with full avalanche on the output. Technically, base32 is exactly equal
16080 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020016081
16082base32+src : binary
16083 This returns the concatenation of the base32 fetch above and the src fetch
16084 below. The resulting type is of type binary, with a size of 8 or 20 bytes
16085 depending on the source address family. This can be used to track per-IP,
16086 per-URL counters.
16087
William Lallemand65ad6e12014-01-31 15:08:02 +010016088capture.req.hdr(<idx>) : string
16089 This extracts the content of the header captured by the "capture request
16090 header", idx is the position of the capture keyword in the configuration.
16091 The first entry is an index of 0. See also: "capture request header".
16092
16093capture.req.method : string
16094 This extracts the METHOD of an HTTP request. It can be used in both request
16095 and response. Unlike "method", it can be used in both request and response
16096 because it's allocated.
16097
16098capture.req.uri : string
16099 This extracts the request's URI, which starts at the first slash and ends
16100 before the first space in the request (without the host part). Unlike "path"
16101 and "url", it can be used in both request and response because it's
16102 allocated.
16103
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016104capture.req.ver : string
16105 This extracts the request's HTTP version and returns either "HTTP/1.0" or
16106 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
16107 logs because it relies on a persistent flag.
16108
William Lallemand65ad6e12014-01-31 15:08:02 +010016109capture.res.hdr(<idx>) : string
16110 This extracts the content of the header captured by the "capture response
16111 header", idx is the position of the capture keyword in the configuration.
16112 The first entry is an index of 0.
16113 See also: "capture response header"
16114
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020016115capture.res.ver : string
16116 This extracts the response's HTTP version and returns either "HTTP/1.0" or
16117 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
16118 persistent flag.
16119
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016120req.body : binary
16121 This returns the HTTP request's available body as a block of data. It
16122 requires that the request body has been buffered made available using
16123 "option http-buffer-request". In case of chunked-encoded body, currently only
16124 the first chunk is analyzed.
16125
Thierry FOURNIER9826c772015-05-20 15:50:54 +020016126req.body_param([<name>) : string
16127 This fetch assumes that the body of the POST request is url-encoded. The user
16128 can check if the "content-type" contains the value
16129 "application/x-www-form-urlencoded". This extracts the first occurrence of the
16130 parameter <name> in the body, which ends before '&'. The parameter name is
16131 case-sensitive. If no name is given, any parameter will match, and the first
16132 one will be returned. The result is a string corresponding to the value of the
16133 parameter <name> as presented in the request body (no URL decoding is
16134 performed). Note that the ACL version of this fetch iterates over multiple
16135 parameters and will iteratively report all parameters values if no name is
16136 given.
16137
Willy Tarreaua5910cc2015-05-02 00:46:08 +020016138req.body_len : integer
16139 This returns the length of the HTTP request's available body in bytes. It may
16140 be lower than the advertised length if the body is larger than the buffer. It
16141 requires that the request body has been buffered made available using
16142 "option http-buffer-request".
16143
16144req.body_size : integer
16145 This returns the advertised length of the HTTP request's body in bytes. It
16146 will represent the advertised Content-Length header, or the size of the first
16147 chunk in case of chunked encoding. In order to parse the chunks, it requires
16148 that the request body has been buffered made available using
16149 "option http-buffer-request".
16150
Willy Tarreau74ca5042013-06-11 23:12:07 +020016151req.cook([<name>]) : string
16152cook([<name>]) : string (deprecated)
16153 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16154 header line from the request, and returns its value as string. If no name is
16155 specified, the first cookie value is returned. When used with ACLs, all
16156 matching cookies are evaluated. Spaces around the name and the value are
16157 ignored as requested by the Cookie header specification (RFC6265). The cookie
16158 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
16159 well return an empty value if it is present. Use the "found" match to detect
16160 presence. Use the res.cook() variant for response cookies sent by the server.
16161
16162 ACL derivatives :
16163 cook([<name>]) : exact string match
16164 cook_beg([<name>]) : prefix match
16165 cook_dir([<name>]) : subdir match
16166 cook_dom([<name>]) : domain match
16167 cook_end([<name>]) : suffix match
16168 cook_len([<name>]) : length match
16169 cook_reg([<name>]) : regex match
16170 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010016171
Willy Tarreau74ca5042013-06-11 23:12:07 +020016172req.cook_cnt([<name>]) : integer
16173cook_cnt([<name>]) : integer (deprecated)
16174 Returns an integer value representing the number of occurrences of the cookie
16175 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016176
Willy Tarreau74ca5042013-06-11 23:12:07 +020016177req.cook_val([<name>]) : integer
16178cook_val([<name>]) : integer (deprecated)
16179 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16180 header line from the request, and converts its value to an integer which is
16181 returned. If no name is specified, the first cookie value is returned. When
16182 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020016183
Willy Tarreau74ca5042013-06-11 23:12:07 +020016184cookie([<name>]) : string (deprecated)
16185 This extracts the last occurrence of the cookie name <name> on a "Cookie"
16186 header line from the request, or a "Set-Cookie" header from the response, and
16187 returns its value as a string. A typical use is to get multiple clients
16188 sharing a same profile use the same server. This can be similar to what
Willy Tarreau294d0f02015-08-10 19:40:12 +020016189 "appsession" did with the "request-learn" statement, but with support for
Willy Tarreau74ca5042013-06-11 23:12:07 +020016190 multi-peer synchronization and state keeping across restarts. If no name is
16191 specified, the first cookie value is returned. This fetch should not be used
16192 anymore and should be replaced by req.cook() or res.cook() instead as it
16193 ambiguously uses the direction based on the context where it is used.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016194
Willy Tarreau74ca5042013-06-11 23:12:07 +020016195hdr([<name>[,<occ>]]) : string
16196 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
16197 used on responses. Please refer to these respective fetches for more details.
16198 In case of doubt about the fetch direction, please use the explicit ones.
16199 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016200 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016201
Willy Tarreau74ca5042013-06-11 23:12:07 +020016202req.fhdr(<name>[,<occ>]) : string
16203 This extracts the last occurrence of header <name> in an HTTP request. When
16204 used from an ACL, all occurrences are iterated over until a match is found.
16205 Optionally, a specific occurrence might be specified as a position number.
16206 Positive values indicate a position from the first occurrence, with 1 being
16207 the first one. Negative values indicate positions relative to the last one,
16208 with -1 being the last one. It differs from req.hdr() in that any commas
16209 present in the value are returned and are not used as delimiters. This is
16210 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016211
Willy Tarreau74ca5042013-06-11 23:12:07 +020016212req.fhdr_cnt([<name>]) : integer
16213 Returns an integer value representing the number of occurrences of request
16214 header field name <name>, or the total number of header fields if <name> is
16215 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
16216 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016217
Willy Tarreau74ca5042013-06-11 23:12:07 +020016218req.hdr([<name>[,<occ>]]) : string
16219 This extracts the last occurrence of header <name> in an HTTP request. When
16220 used from an ACL, all occurrences are iterated over until a match is found.
16221 Optionally, a specific occurrence might be specified as a position number.
16222 Positive values indicate a position from the first occurrence, with 1 being
16223 the first one. Negative values indicate positions relative to the last one,
16224 with -1 being the last one. A typical use is with the X-Forwarded-For header
16225 once converted to IP, associated with an IP stick-table. The function
16226 considers any comma as a delimiter for distinct values. If full-line headers
Lukas Tribus23953682017-04-28 13:24:30 +000016227 are desired instead, use req.fhdr(). Please carefully check RFC7231 to know
Willy Tarreau74ca5042013-06-11 23:12:07 +020016228 how certain headers are supposed to be parsed. Also, some of them are case
Davor Ocelice9ed2812017-12-25 17:49:28 +010016229 insensitive (e.g. Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010016230
Willy Tarreau74ca5042013-06-11 23:12:07 +020016231 ACL derivatives :
16232 hdr([<name>[,<occ>]]) : exact string match
16233 hdr_beg([<name>[,<occ>]]) : prefix match
16234 hdr_dir([<name>[,<occ>]]) : subdir match
16235 hdr_dom([<name>[,<occ>]]) : domain match
16236 hdr_end([<name>[,<occ>]]) : suffix match
16237 hdr_len([<name>[,<occ>]]) : length match
16238 hdr_reg([<name>[,<occ>]]) : regex match
16239 hdr_sub([<name>[,<occ>]]) : substring match
16240
16241req.hdr_cnt([<name>]) : integer
16242hdr_cnt([<header>]) : integer (deprecated)
16243 Returns an integer value representing the number of occurrences of request
16244 header field name <name>, or the total number of header field values if
16245 <name> is not specified. It is important to remember that one header line may
16246 count as several headers if it has several values. The function considers any
16247 comma as a delimiter for distinct values. If full-line headers are desired
16248 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
16249 detect presence, absence or abuse of a specific header, as well as to block
16250 request smuggling attacks by rejecting requests which contain more than one
16251 of certain headers. See "req.hdr" for more information on header matching.
16252
16253req.hdr_ip([<name>[,<occ>]]) : ip
16254hdr_ip([<name>[,<occ>]]) : ip (deprecated)
16255 This extracts the last occurrence of header <name> in an HTTP request,
16256 converts it to an IPv4 or IPv6 address and returns this address. When used
16257 with ACLs, all occurrences are checked, and if <name> is omitted, every value
16258 of every header is checked. Optionally, a specific occurrence might be
16259 specified as a position number. Positive values indicate a position from the
Davor Ocelice9ed2812017-12-25 17:49:28 +010016260 first occurrence, with 1 being the first one. Negative values indicate
Willy Tarreau74ca5042013-06-11 23:12:07 +020016261 positions relative to the last one, with -1 being the last one. A typical use
16262 is with the X-Forwarded-For and X-Client-IP headers.
16263
16264req.hdr_val([<name>[,<occ>]]) : integer
16265hdr_val([<name>[,<occ>]]) : integer (deprecated)
16266 This extracts the last occurrence of header <name> in an HTTP request, and
16267 converts it to an integer value. When used with ACLs, all occurrences are
16268 checked, and if <name> is omitted, every value of every header is checked.
16269 Optionally, a specific occurrence might be specified as a position number.
16270 Positive values indicate a position from the first occurrence, with 1 being
16271 the first one. Negative values indicate positions relative to the last one,
16272 with -1 being the last one. A typical use is with the X-Forwarded-For header.
16273
Frédéric Lécailleec891192019-02-26 15:02:35 +010016274
16275
Willy Tarreau74ca5042013-06-11 23:12:07 +020016276http_auth(<userlist>) : boolean
16277 Returns a boolean indicating whether the authentication data received from
16278 the client match a username & password stored in the specified userlist. This
16279 fetch function is not really useful outside of ACLs. Currently only http
16280 basic auth is supported.
16281
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016282http_auth_group(<userlist>) : string
16283 Returns a string corresponding to the user name found in the authentication
16284 data received from the client if both the user name and password are valid
16285 according to the specified userlist. The main purpose is to use it in ACLs
16286 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016287 This fetch function is not really useful outside of ACLs. Currently only http
16288 basic auth is supported.
16289
16290 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010016291 http_auth_group(<userlist>) : group ...
16292 Returns true when the user extracted from the request and whose password is
16293 valid according to the specified userlist belongs to at least one of the
16294 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020016295
16296http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020016297 Returns true when the request being processed is the first one of the
16298 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020016299 from some requests when a request is not the first one, or to help grouping
16300 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020016301
Willy Tarreau74ca5042013-06-11 23:12:07 +020016302method : integer + string
16303 Returns an integer value corresponding to the method in the HTTP request. For
16304 example, "GET" equals 1 (check sources to establish the matching). Value 9
16305 means "other method" and may be converted to a string extracted from the
16306 stream. This should not be used directly as a sample, this is only meant to
16307 be used from ACLs, which transparently convert methods from patterns to these
16308 integer + string values. Some predefined ACL already check for most common
16309 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016310
Willy Tarreau74ca5042013-06-11 23:12:07 +020016311 ACL derivatives :
16312 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016313
Willy Tarreau74ca5042013-06-11 23:12:07 +020016314 Example :
16315 # only accept GET and HEAD requests
16316 acl valid_method method GET HEAD
16317 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020016318
Willy Tarreau74ca5042013-06-11 23:12:07 +020016319path : string
16320 This extracts the request's URL path, which starts at the first slash and
16321 ends before the question mark (without the host part). A typical use is with
16322 prefetch-capable caches, and with portals which need to aggregate multiple
16323 information from databases and keep them in caches. Note that with outgoing
16324 caches, it would be wiser to use "url" instead. With ACLs, it's typically
Davor Ocelice9ed2812017-12-25 17:49:28 +010016325 used to match exact file names (e.g. "/login.php"), or directory parts using
Willy Tarreau74ca5042013-06-11 23:12:07 +020016326 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016327
Willy Tarreau74ca5042013-06-11 23:12:07 +020016328 ACL derivatives :
16329 path : exact string match
16330 path_beg : prefix match
16331 path_dir : subdir match
16332 path_dom : domain match
16333 path_end : suffix match
16334 path_len : length match
16335 path_reg : regex match
16336 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020016337
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016338query : string
16339 This extracts the request's query string, which starts after the first
16340 question mark. If no question mark is present, this fetch returns nothing. If
16341 a question mark is present but nothing follows, it returns an empty string.
16342 This means it's possible to easily know whether a query string is present
Tim Düsterhus4896c442016-11-29 02:15:19 +010016343 using the "found" matching method. This fetch is the complement of "path"
Willy Tarreau49ad95c2015-01-19 15:06:26 +010016344 which stops before the question mark.
16345
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016346req.hdr_names([<delim>]) : string
16347 This builds a string made from the concatenation of all header names as they
16348 appear in the request when the rule is evaluated. The default delimiter is
16349 the comma (',') but it may be overridden as an optional argument <delim>. In
16350 this case, only the first character of <delim> is considered.
16351
Willy Tarreau74ca5042013-06-11 23:12:07 +020016352req.ver : string
16353req_ver : string (deprecated)
16354 Returns the version string from the HTTP request, for example "1.1". This can
16355 be useful for logs, but is mostly there for ACL. Some predefined ACL already
16356 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016357
Willy Tarreau74ca5042013-06-11 23:12:07 +020016358 ACL derivatives :
16359 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020016360
Willy Tarreau74ca5042013-06-11 23:12:07 +020016361res.comp : boolean
16362 Returns the boolean "true" value if the response has been compressed by
16363 HAProxy, otherwise returns boolean "false". This may be used to add
16364 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016365
Willy Tarreau74ca5042013-06-11 23:12:07 +020016366res.comp_algo : string
16367 Returns a string containing the name of the algorithm used if the response
16368 was compressed by HAProxy, for example : "deflate". This may be used to add
16369 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016370
Willy Tarreau74ca5042013-06-11 23:12:07 +020016371res.cook([<name>]) : string
16372scook([<name>]) : string (deprecated)
16373 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16374 header line from the response, and returns its value as string. If no name is
16375 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016376
Willy Tarreau74ca5042013-06-11 23:12:07 +020016377 ACL derivatives :
16378 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020016379
Willy Tarreau74ca5042013-06-11 23:12:07 +020016380res.cook_cnt([<name>]) : integer
16381scook_cnt([<name>]) : integer (deprecated)
16382 Returns an integer value representing the number of occurrences of the cookie
16383 <name> in the response, or all cookies if <name> is not specified. This is
16384 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016385
Willy Tarreau74ca5042013-06-11 23:12:07 +020016386res.cook_val([<name>]) : integer
16387scook_val([<name>]) : integer (deprecated)
16388 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16389 header line from the response, and converts its value to an integer which is
16390 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010016391
Willy Tarreau74ca5042013-06-11 23:12:07 +020016392res.fhdr([<name>[,<occ>]]) : string
16393 This extracts the last occurrence of header <name> in an HTTP response, or of
16394 the last header if no <name> is specified. Optionally, a specific occurrence
16395 might be specified as a position number. Positive values indicate a position
16396 from the first occurrence, with 1 being the first one. Negative values
16397 indicate positions relative to the last one, with -1 being the last one. It
16398 differs from res.hdr() in that any commas present in the value are returned
16399 and are not used as delimiters. If this is not desired, the res.hdr() fetch
16400 should be used instead. This is sometimes useful with headers such as Date or
16401 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016402
Willy Tarreau74ca5042013-06-11 23:12:07 +020016403res.fhdr_cnt([<name>]) : integer
16404 Returns an integer value representing the number of occurrences of response
16405 header field name <name>, or the total number of header fields if <name> is
16406 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
16407 the number of full line headers and does not stop on commas. If this is not
16408 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016409
Willy Tarreau74ca5042013-06-11 23:12:07 +020016410res.hdr([<name>[,<occ>]]) : string
16411shdr([<name>[,<occ>]]) : string (deprecated)
16412 This extracts the last occurrence of header <name> in an HTTP response, or of
16413 the last header if no <name> is specified. Optionally, a specific occurrence
16414 might be specified as a position number. Positive values indicate a position
16415 from the first occurrence, with 1 being the first one. Negative values
16416 indicate positions relative to the last one, with -1 being the last one. This
16417 can be useful to learn some data into a stick-table. The function considers
16418 any comma as a delimiter for distinct values. If this is not desired, the
16419 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016420
Willy Tarreau74ca5042013-06-11 23:12:07 +020016421 ACL derivatives :
16422 shdr([<name>[,<occ>]]) : exact string match
16423 shdr_beg([<name>[,<occ>]]) : prefix match
16424 shdr_dir([<name>[,<occ>]]) : subdir match
16425 shdr_dom([<name>[,<occ>]]) : domain match
16426 shdr_end([<name>[,<occ>]]) : suffix match
16427 shdr_len([<name>[,<occ>]]) : length match
16428 shdr_reg([<name>[,<occ>]]) : regex match
16429 shdr_sub([<name>[,<occ>]]) : substring match
16430
16431res.hdr_cnt([<name>]) : integer
16432shdr_cnt([<name>]) : integer (deprecated)
16433 Returns an integer value representing the number of occurrences of response
16434 header field name <name>, or the total number of header fields if <name> is
16435 not specified. The function considers any comma as a delimiter for distinct
16436 values. If this is not desired, the res.fhdr_cnt() fetch should be used
16437 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016438
Willy Tarreau74ca5042013-06-11 23:12:07 +020016439res.hdr_ip([<name>[,<occ>]]) : ip
16440shdr_ip([<name>[,<occ>]]) : ip (deprecated)
16441 This extracts the last occurrence of header <name> in an HTTP response,
16442 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
16443 specific occurrence might be specified as a position number. Positive values
16444 indicate a position from the first occurrence, with 1 being the first one.
16445 Negative values indicate positions relative to the last one, with -1 being
16446 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020016447
Willy Tarreaueb27ec72015-02-20 13:55:29 +010016448res.hdr_names([<delim>]) : string
16449 This builds a string made from the concatenation of all header names as they
16450 appear in the response when the rule is evaluated. The default delimiter is
16451 the comma (',') but it may be overridden as an optional argument <delim>. In
16452 this case, only the first character of <delim> is considered.
16453
Willy Tarreau74ca5042013-06-11 23:12:07 +020016454res.hdr_val([<name>[,<occ>]]) : integer
16455shdr_val([<name>[,<occ>]]) : integer (deprecated)
16456 This extracts the last occurrence of header <name> in an HTTP response, and
16457 converts it to an integer value. Optionally, a specific occurrence might be
16458 specified as a position number. Positive values indicate a position from the
16459 first occurrence, with 1 being the first one. Negative values indicate
16460 positions relative to the last one, with -1 being the last one. This can be
16461 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016462
Willy Tarreau74ca5042013-06-11 23:12:07 +020016463res.ver : string
16464resp_ver : string (deprecated)
16465 Returns the version string from the HTTP response, for example "1.1". This
16466 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020016467
Willy Tarreau74ca5042013-06-11 23:12:07 +020016468 ACL derivatives :
16469 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010016470
Willy Tarreau74ca5042013-06-11 23:12:07 +020016471set-cookie([<name>]) : string (deprecated)
16472 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
16473 header line from the response and uses the corresponding value to match. This
Willy Tarreau294d0f02015-08-10 19:40:12 +020016474 can be comparable to what "appsession" did with default options, but with
Willy Tarreau74ca5042013-06-11 23:12:07 +020016475 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016476
Willy Tarreau74ca5042013-06-11 23:12:07 +020016477 This fetch function is deprecated and has been superseded by the "res.cook"
16478 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010016479
Willy Tarreau74ca5042013-06-11 23:12:07 +020016480status : integer
16481 Returns an integer containing the HTTP status code in the HTTP response, for
16482 example, 302. It is mostly used within ACLs and integer ranges, for example,
16483 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016484
Thierry Fournier0e00dca2016-04-07 15:47:40 +020016485unique-id : string
16486 Returns the unique-id attached to the request. The directive
16487 "unique-id-format" must be set. If it is not set, the unique-id sample fetch
16488 fails. Note that the unique-id is usually used with HTTP requests, however this
16489 sample fetch can be used with other protocols. Obviously, if it is used with
16490 other protocols than HTTP, the unique-id-format directive must not contain
16491 HTTP parts. See: unique-id-format and unique-id-header
16492
Willy Tarreau74ca5042013-06-11 23:12:07 +020016493url : string
16494 This extracts the request's URL as presented in the request. A typical use is
16495 with prefetch-capable caches, and with portals which need to aggregate
16496 multiple information from databases and keep them in caches. With ACLs, using
16497 "path" is preferred over using "url", because clients may send a full URL as
16498 is normally done with proxies. The only real use is to match "*" which does
16499 not match in "path", and for which there is already a predefined ACL. See
16500 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016501
Willy Tarreau74ca5042013-06-11 23:12:07 +020016502 ACL derivatives :
16503 url : exact string match
16504 url_beg : prefix match
16505 url_dir : subdir match
16506 url_dom : domain match
16507 url_end : suffix match
16508 url_len : length match
16509 url_reg : regex match
16510 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016511
Willy Tarreau74ca5042013-06-11 23:12:07 +020016512url_ip : ip
16513 This extracts the IP address from the request's URL when the host part is
16514 presented as an IP address. Its use is very limited. For instance, a
16515 monitoring system might use this field as an alternative for the source IP in
16516 order to test what path a given source address would follow, or to force an
16517 entry in a table for a given source address. With ACLs it can be used to
16518 restrict access to certain systems through a proxy, for example when combined
16519 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016520
Willy Tarreau74ca5042013-06-11 23:12:07 +020016521url_port : integer
16522 This extracts the port part from the request's URL. Note that if the port is
16523 not specified in the request, port 80 is assumed. With ACLs it can be used to
16524 restrict access to certain systems through a proxy, for example when combined
16525 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016526
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016527urlp([<name>[,<delim>]]) : string
16528url_param([<name>[,<delim>]]) : string
Willy Tarreau74ca5042013-06-11 23:12:07 +020016529 This extracts the first occurrence of the parameter <name> in the query
16530 string, which begins after either '?' or <delim>, and which ends before '&',
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016531 ';' or <delim>. The parameter name is case-sensitive. If no name is given,
16532 any parameter will match, and the first one will be returned. The result is
16533 a string corresponding to the value of the parameter <name> as presented in
16534 the request (no URL decoding is performed). This can be used for session
Willy Tarreau74ca5042013-06-11 23:12:07 +020016535 stickiness based on a client ID, to extract an application cookie passed as a
16536 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
Willy Tarreau1ede1da2015-05-07 16:06:18 +020016537 this fetch iterates over multiple parameters and will iteratively report all
16538 parameters values if no name is given
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016539
Willy Tarreau74ca5042013-06-11 23:12:07 +020016540 ACL derivatives :
16541 urlp(<name>[,<delim>]) : exact string match
16542 urlp_beg(<name>[,<delim>]) : prefix match
16543 urlp_dir(<name>[,<delim>]) : subdir match
16544 urlp_dom(<name>[,<delim>]) : domain match
16545 urlp_end(<name>[,<delim>]) : suffix match
16546 urlp_len(<name>[,<delim>]) : length match
16547 urlp_reg(<name>[,<delim>]) : regex match
16548 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016549
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016550
Willy Tarreau74ca5042013-06-11 23:12:07 +020016551 Example :
16552 # match http://example.com/foo?PHPSESSIONID=some_id
16553 stick on urlp(PHPSESSIONID)
16554 # match http://example.com/foo;JSESSIONID=some_id
16555 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020016556
Jarno Huuskonen676f6222017-03-30 09:19:45 +030016557urlp_val([<name>[,<delim>]]) : integer
Willy Tarreau74ca5042013-06-11 23:12:07 +020016558 See "urlp" above. This one extracts the URL parameter <name> in the request
16559 and converts it to an integer value. This can be used for session stickiness
16560 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020016561
Dragan Dosen0070cd52016-06-16 12:19:49 +020016562url32 : integer
16563 This returns a 32-bit hash of the value obtained by concatenating the first
16564 Host header and the whole URL including parameters (not only the path part of
16565 the request, as in the "base32" fetch above). This is useful to track per-URL
16566 activity. A shorter hash is stored, saving a lot of memory. The output type
16567 is an unsigned integer.
16568
16569url32+src : binary
16570 This returns the concatenation of the "url32" fetch and the "src" fetch. The
16571 resulting type is of type binary, with a size of 8 or 20 bytes depending on
16572 the source address family. This can be used to track per-IP, per-URL counters.
16573
Willy Tarreau198a7442008-01-17 12:05:32 +010016574
Willy Tarreau74ca5042013-06-11 23:12:07 +0200165757.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016576---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016577
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016578Some predefined ACLs are hard-coded so that they do not have to be declared in
16579every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020016580order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010016581
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016582ACL name Equivalent to Usage
16583---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016584FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020016585HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016586HTTP_1.0 req_ver 1.0 match HTTP version 1.0
16587HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010016588HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
16589HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
16590HTTP_URL_SLASH url_beg / match URL beginning with "/"
16591HTTP_URL_STAR url * match URL equal to "*"
16592LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016593METH_CONNECT method CONNECT match HTTP CONNECT method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016594METH_DELETE method DELETE match HTTP DELETE method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016595METH_GET method GET HEAD match HTTP GET or HEAD method
16596METH_HEAD method HEAD match HTTP HEAD method
16597METH_OPTIONS method OPTIONS match HTTP OPTIONS method
16598METH_POST method POST match HTTP POST method
Daniel Schneller9ff96c72016-04-11 17:45:29 +020016599METH_PUT method PUT match HTTP PUT method
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016600METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020016601RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016602REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010016603TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016604WAIT_END wait_end wait for end of content analysis
16605---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010016606
Willy Tarreaub937b7e2010-01-12 15:27:54 +010016607
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166088. Logging
16609----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010016610
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016611One of HAProxy's strong points certainly lies is its precise logs. It probably
16612provides the finest level of information available for such a product, which is
16613very important for troubleshooting complex environments. Standard information
16614provided in logs include client ports, TCP/HTTP state timers, precise session
16615state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010016616to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016617headers.
16618
16619In order to improve administrators reactivity, it offers a great transparency
16620about encountered problems, both internal and external, and it is possible to
16621send logs to different sources at the same time with different level filters :
16622
16623 - global process-level logs (system errors, start/stop, etc..)
16624 - per-instance system and internal errors (lack of resource, bugs, ...)
16625 - per-instance external troubles (servers up/down, max connections)
16626 - per-instance activity (client connections), either at the establishment or
16627 at the termination.
Davor Ocelice9ed2812017-12-25 17:49:28 +010016628 - per-request control of log-level, e.g.
Jim Freeman9e8714b2015-05-26 09:16:34 -060016629 http-request set-log-level silent if sensitive_request
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016630
16631The ability to distribute different levels of logs to different log servers
16632allow several production teams to interact and to fix their problems as soon
16633as possible. For example, the system team might monitor system-wide errors,
16634while the application team might be monitoring the up/down for their servers in
16635real time, and the security team might analyze the activity logs with one hour
16636delay.
16637
16638
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166398.1. Log levels
16640---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016641
Simon Hormandf791f52011-05-29 15:01:10 +090016642TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016643source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090016644HTTP request, HTTP return code, number of bytes transmitted, conditions
16645in which the session ended, and even exchanged cookies values. For example
16646track a particular user's problems. All messages may be sent to up to two
16647syslog servers. Check the "log" keyword in section 4.2 for more information
16648about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016649
16650
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166518.2. Log formats
16652----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016653
William Lallemand48940402012-01-30 16:47:22 +010016654HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090016655and will be detailed in the following sections. A few of them may vary
16656slightly with the configuration, due to indicators specific to certain
16657options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016658
16659 - the default format, which is very basic and very rarely used. It only
16660 provides very basic information about the incoming connection at the moment
16661 it is accepted : source IP:port, destination IP:port, and frontend-name.
16662 This mode will eventually disappear so it will not be described to great
16663 extents.
16664
16665 - the TCP format, which is more advanced. This format is enabled when "option
16666 tcplog" is set on the frontend. HAProxy will then usually wait for the
16667 connection to terminate before logging. This format provides much richer
16668 information, such as timers, connection counts, queue size, etc... This
16669 format is recommended for pure TCP proxies.
16670
16671 - the HTTP format, which is the most advanced for HTTP proxying. This format
16672 is enabled when "option httplog" is set on the frontend. It provides the
16673 same information as the TCP format with some HTTP-specific fields such as
16674 the request, the status code, and captures of headers and cookies. This
16675 format is recommended for HTTP proxies.
16676
Emeric Brun3a058f32009-06-30 18:26:00 +020016677 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
16678 fields arranged in the same order as the CLF format. In this mode, all
16679 timers, captures, flags, etc... appear one per field after the end of the
16680 common fields, in the same order they appear in the standard HTTP format.
16681
William Lallemand48940402012-01-30 16:47:22 +010016682 - the custom log format, allows you to make your own log line.
16683
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016684Next sections will go deeper into details for each of these formats. Format
16685specification will be performed on a "field" basis. Unless stated otherwise, a
16686field is a portion of text delimited by any number of spaces. Since syslog
16687servers are susceptible of inserting fields at the beginning of a line, it is
16688always assumed that the first field is the one containing the process name and
16689identifier.
16690
16691Note : Since log lines may be quite long, the log examples in sections below
16692 might be broken into multiple lines. The example log lines will be
16693 prefixed with 3 closing angle brackets ('>>>') and each time a log is
16694 broken into multiple lines, each non-final line will end with a
16695 backslash ('\') and the next line will start indented by two characters.
16696
16697
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200166988.2.1. Default log format
16699-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016700
16701This format is used when no specific option is set. The log is emitted as soon
16702as the connection is accepted. One should note that this currently is the only
16703format which logs the request's destination IP and ports.
16704
16705 Example :
16706 listen www
16707 mode http
16708 log global
16709 server srv1 127.0.0.1:8000
16710
16711 >>> Feb 6 12:12:09 localhost \
16712 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
16713 (www/HTTP)
16714
16715 Field Format Extract from the example above
16716 1 process_name '[' pid ']:' haproxy[14385]:
16717 2 'Connect from' Connect from
16718 3 source_ip ':' source_port 10.0.1.2:33312
16719 4 'to' to
16720 5 destination_ip ':' destination_port 10.0.3.31:8012
16721 6 '(' frontend_name '/' mode ')' (www/HTTP)
16722
16723Detailed fields description :
16724 - "source_ip" is the IP address of the client which initiated the connection.
16725 - "source_port" is the TCP port of the client which initiated the connection.
16726 - "destination_ip" is the IP address the client connected to.
16727 - "destination_port" is the TCP port the client connected to.
16728 - "frontend_name" is the name of the frontend (or listener) which received
16729 and processed the connection.
16730 - "mode is the mode the frontend is operating (TCP or HTTP).
16731
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016732In case of a UNIX socket, the source and destination addresses are marked as
16733"unix:" and the ports reflect the internal ID of the socket which accepted the
16734connection (the same ID as reported in the stats).
16735
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016736It is advised not to use this deprecated format for newer installations as it
16737will eventually disappear.
16738
16739
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200167408.2.2. TCP log format
16741---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016742
16743The TCP format is used when "option tcplog" is specified in the frontend, and
16744is the recommended format for pure TCP proxies. It provides a lot of precious
16745information for troubleshooting. Since this format includes timers and byte
16746counts, the log is normally emitted at the end of the session. It can be
16747emitted earlier if "option logasap" is specified, which makes sense in most
16748environments with long sessions such as remote terminals. Sessions which match
16749the "monitor" rules are never logged. It is also possible not to emit logs for
16750sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016751specifying "option dontlognull" in the frontend. Successful connections will
16752not be logged if "option dontlog-normal" is specified in the frontend. A few
16753fields may slightly vary depending on some configuration options, those are
16754marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016755
16756 Example :
16757 frontend fnt
16758 mode tcp
16759 option tcplog
16760 log global
16761 default_backend bck
16762
16763 backend bck
16764 server srv1 127.0.0.1:8000
16765
16766 >>> Feb 6 12:12:56 localhost \
16767 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
16768 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
16769
16770 Field Format Extract from the example above
16771 1 process_name '[' pid ']:' haproxy[14387]:
16772 2 client_ip ':' client_port 10.0.1.2:33313
16773 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
16774 4 frontend_name fnt
16775 5 backend_name '/' server_name bck/srv1
16776 6 Tw '/' Tc '/' Tt* 0/0/5007
16777 7 bytes_read* 212
16778 8 termination_state --
16779 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
16780 10 srv_queue '/' backend_queue 0/0
16781
16782Detailed fields description :
16783 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016784 connection to haproxy. If the connection was accepted on a UNIX socket
16785 instead, the IP address would be replaced with the word "unix". Note that
16786 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016787 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016788 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016789 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016790
16791 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016792 If the connection was accepted on a UNIX socket instead, the port would be
16793 replaced with the ID of the accepting socket, which is also reported in the
16794 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016795
16796 - "accept_date" is the exact date when the connection was received by haproxy
16797 (which might be very slightly different from the date observed on the
16798 network if there was some queuing in the system's backlog). This is usually
Willy Tarreau590a0512018-09-05 11:56:48 +020016799 the same date which may appear in any upstream firewall's log. When used in
16800 HTTP mode, the accept_date field will be reset to the first moment the
16801 connection is ready to receive a new request (end of previous response for
16802 HTTP/1, immediately after previous request for HTTP/2).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016803
16804 - "frontend_name" is the name of the frontend (or listener) which received
16805 and processed the connection.
16806
16807 - "backend_name" is the name of the backend (or listener) which was selected
16808 to manage the connection to the server. This will be the same as the
16809 frontend if no switching rule has been applied, which is common for TCP
16810 applications.
16811
16812 - "server_name" is the name of the last server to which the connection was
16813 sent, which might differ from the first one if there were connection errors
16814 and a redispatch occurred. Note that this server belongs to the backend
16815 which processed the request. If the connection was aborted before reaching
16816 a server, "<NOSRV>" is indicated instead of a server name.
16817
16818 - "Tw" is the total time in milliseconds spent waiting in the various queues.
16819 It can be "-1" if the connection was aborted before reaching the queue.
16820 See "Timers" below for more details.
16821
16822 - "Tc" is the total time in milliseconds spent waiting for the connection to
16823 establish to the final server, including retries. It can be "-1" if the
16824 connection was aborted before a connection could be established. See
16825 "Timers" below for more details.
16826
16827 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030016828 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016829 "option logasap" was specified, then the time counting stops at the moment
16830 the log is emitted. In this case, a '+' sign is prepended before the value,
16831 indicating that the final one will be larger. See "Timers" below for more
16832 details.
16833
16834 - "bytes_read" is the total number of bytes transmitted from the server to
16835 the client when the log is emitted. If "option logasap" is specified, the
16836 this value will be prefixed with a '+' sign indicating that the final one
16837 may be larger. Please note that this value is a 64-bit counter, so log
16838 analysis tools must be able to handle it without overflowing.
16839
16840 - "termination_state" is the condition the session was in when the session
16841 ended. This indicates the session state, which side caused the end of
16842 session to happen, and for what reason (timeout, error, ...). The normal
16843 flags should be "--", indicating the session was closed by either end with
16844 no data remaining in buffers. See below "Session state at disconnection"
16845 for more details.
16846
16847 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040016848 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016849 limits have been reached. For instance, if actconn is close to 512 when
16850 multiple connection errors occur, chances are high that the system limits
16851 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020016852 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016853
16854 - "feconn" is the total number of concurrent connections on the frontend when
16855 the session was logged. It is useful to estimate the amount of resource
16856 required to sustain high loads, and to detect when the frontend's "maxconn"
16857 has been reached. Most often when this value increases by huge jumps, it is
16858 because there is congestion on the backend servers, but sometimes it can be
16859 caused by a denial of service attack.
16860
16861 - "beconn" is the total number of concurrent connections handled by the
16862 backend when the session was logged. It includes the total number of
16863 concurrent connections active on servers as well as the number of
16864 connections pending in queues. It is useful to estimate the amount of
16865 additional servers needed to support high loads for a given application.
16866 Most often when this value increases by huge jumps, it is because there is
16867 congestion on the backend servers, but sometimes it can be caused by a
16868 denial of service attack.
16869
16870 - "srv_conn" is the total number of concurrent connections still active on
16871 the server when the session was logged. It can never exceed the server's
16872 configured "maxconn" parameter. If this value is very often close or equal
16873 to the server's "maxconn", it means that traffic regulation is involved a
16874 lot, meaning that either the server's maxconn value is too low, or that
16875 there aren't enough servers to process the load with an optimal response
16876 time. When only one of the server's "srv_conn" is high, it usually means
16877 that this server has some trouble causing the connections to take longer to
16878 be processed than on other servers.
16879
16880 - "retries" is the number of connection retries experienced by this session
16881 when trying to connect to the server. It must normally be zero, unless a
16882 server is being stopped at the same moment the connection was attempted.
16883 Frequent retries generally indicate either a network problem between
16884 haproxy and the server, or a misconfigured system backlog on the server
16885 preventing new connections from being queued. This field may optionally be
16886 prefixed with a '+' sign, indicating that the session has experienced a
16887 redispatch after the maximal retry count has been reached on the initial
16888 server. In this case, the server name appearing in the log is the one the
16889 connection was redispatched to, and not the first one, though both may
16890 sometimes be the same in case of hashing for instance. So as a general rule
16891 of thumb, when a '+' is present in front of the retry count, this count
16892 should not be attributed to the logged server.
16893
16894 - "srv_queue" is the total number of requests which were processed before
16895 this one in the server queue. It is zero when the request has not gone
16896 through the server queue. It makes it possible to estimate the approximate
16897 server's response time by dividing the time spent in queue by the number of
16898 requests in the queue. It is worth noting that if a session experiences a
16899 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010016900 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016901 backend queue unless a redispatch occurs.
16902
16903 - "backend_queue" is the total number of requests which were processed before
16904 this one in the backend's global queue. It is zero when the request has not
16905 gone through the global queue. It makes it possible to estimate the average
16906 queue length, which easily translates into a number of missing servers when
16907 divided by a server's "maxconn" parameter. It is worth noting that if a
16908 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010016909 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016910 through both the server queue and the backend queue unless a redispatch
16911 occurs.
16912
16913
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200169148.2.3. HTTP log format
16915----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016916
16917The HTTP format is the most complete and the best suited for HTTP proxies. It
16918is enabled by when "option httplog" is specified in the frontend. It provides
16919the same level of information as the TCP format with additional features which
16920are specific to the HTTP protocol. Just like the TCP format, the log is usually
16921emitted at the end of the session, unless "option logasap" is specified, which
16922generally only makes sense for download sites. A session which matches the
16923"monitor" rules will never logged. It is also possible not to log sessions for
16924which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020016925frontend. Successful connections will not be logged if "option dontlog-normal"
16926is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016927
16928Most fields are shared with the TCP log, some being different. A few fields may
16929slightly vary depending on some configuration options. Those ones are marked
16930with a star ('*') after the field name below.
16931
16932 Example :
16933 frontend http-in
16934 mode http
16935 option httplog
16936 log global
16937 default_backend bck
16938
16939 backend static
16940 server srv1 127.0.0.1:8000
16941
16942 >>> Feb 6 12:14:14 localhost \
16943 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
16944 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 +010016945 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016946
16947 Field Format Extract from the example above
16948 1 process_name '[' pid ']:' haproxy[14389]:
16949 2 client_ip ':' client_port 10.0.1.2:33317
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016950 3 '[' request_date ']' [06/Feb/2009:12:14:14.655]
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016951 4 frontend_name http-in
16952 5 backend_name '/' server_name static/srv1
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016953 6 TR '/' Tw '/' Tc '/' Tr '/' Ta* 10/0/30/69/109
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016954 7 status_code 200
16955 8 bytes_read* 2750
16956 9 captured_request_cookie -
16957 10 captured_response_cookie -
16958 11 termination_state ----
16959 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
16960 13 srv_queue '/' backend_queue 0/0
16961 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
16962 15 '{' captured_response_headers* '}' {}
16963 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010016964
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016965Detailed fields description :
16966 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016967 connection to haproxy. If the connection was accepted on a UNIX socket
16968 instead, the IP address would be replaced with the word "unix". Note that
16969 when the connection is accepted on a socket configured with "accept-proxy"
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016970 and the PROXY protocol is correctly used, or with a "accept-netscaler-cip"
Davor Ocelice9ed2812017-12-25 17:49:28 +010016971 and the NetScaler Client IP insertion protocol is correctly used, then the
Bertrand Jacquin93b227d2016-06-04 15:11:10 +010016972 logs will reflect the forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016973
16974 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010016975 If the connection was accepted on a UNIX socket instead, the port would be
16976 replaced with the ID of the accepting socket, which is also reported in the
16977 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016978
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016979 - "request_date" is the exact date when the first byte of the HTTP request
16980 was received by haproxy (log field %tr).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010016981
16982 - "frontend_name" is the name of the frontend (or listener) which received
16983 and processed the connection.
16984
16985 - "backend_name" is the name of the backend (or listener) which was selected
16986 to manage the connection to the server. This will be the same as the
16987 frontend if no switching rule has been applied.
16988
16989 - "server_name" is the name of the last server to which the connection was
16990 sent, which might differ from the first one if there were connection errors
16991 and a redispatch occurred. Note that this server belongs to the backend
16992 which processed the request. If the request was aborted before reaching a
16993 server, "<NOSRV>" is indicated instead of a server name. If the request was
16994 intercepted by the stats subsystem, "<STATS>" is indicated instead.
16995
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020016996 - "TR" is the total time in milliseconds spent waiting for a full HTTP
16997 request from the client (not counting body) after the first byte was
16998 received. It can be "-1" if the connection was aborted before a complete
16999 request could be received or the a bad request was received. It should
17000 always be very small because a request generally fits in one single packet.
17001 Large times here generally indicate network issues between the client and
Willy Tarreau590a0512018-09-05 11:56:48 +020017002 haproxy or requests being typed by hand. See section 8.4 "Timing Events"
17003 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017004
17005 - "Tw" is the total time in milliseconds spent waiting in the various queues.
17006 It can be "-1" if the connection was aborted before reaching the queue.
Willy Tarreau590a0512018-09-05 11:56:48 +020017007 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017008
17009 - "Tc" is the total time in milliseconds spent waiting for the connection to
17010 establish to the final server, including retries. It can be "-1" if the
Willy Tarreau590a0512018-09-05 11:56:48 +020017011 request was aborted before a connection could be established. See section
17012 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017013
17014 - "Tr" is the total time in milliseconds spent waiting for the server to send
17015 a full HTTP response, not counting data. It can be "-1" if the request was
17016 aborted before a complete response could be received. It generally matches
17017 the server's processing time for the request, though it may be altered by
17018 the amount of data sent by the client to the server. Large times here on
Willy Tarreau590a0512018-09-05 11:56:48 +020017019 "GET" requests generally indicate an overloaded server. See section 8.4
17020 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017021
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017022 - "Ta" is the time the request remained active in haproxy, which is the total
17023 time in milliseconds elapsed between the first byte of the request was
17024 received and the last byte of response was sent. It covers all possible
17025 processing except the handshake (see Th) and idle time (see Ti). There is
17026 one exception, if "option logasap" was specified, then the time counting
17027 stops at the moment the log is emitted. In this case, a '+' sign is
17028 prepended before the value, indicating that the final one will be larger.
Willy Tarreau590a0512018-09-05 11:56:48 +020017029 See section 8.4 "Timing Events" for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017030
17031 - "status_code" is the HTTP status code returned to the client. This status
17032 is generally set by the server, but it might also be set by haproxy when
17033 the server cannot be reached or when its response is blocked by haproxy.
17034
17035 - "bytes_read" is the total number of bytes transmitted to the client when
17036 the log is emitted. This does include HTTP headers. If "option logasap" is
17037 specified, the this value will be prefixed with a '+' sign indicating that
17038 the final one may be larger. Please note that this value is a 64-bit
17039 counter, so log analysis tools must be able to handle it without
17040 overflowing.
17041
17042 - "captured_request_cookie" is an optional "name=value" entry indicating that
17043 the client had this cookie in the request. The cookie name and its maximum
17044 length are defined by the "capture cookie" statement in the frontend
17045 configuration. The field is a single dash ('-') when the option is not
17046 set. Only one cookie may be captured, it is generally used to track session
17047 ID exchanges between a client and a server to detect session crossing
17048 between clients due to application bugs. For more details, please consult
17049 the section "Capturing HTTP headers and cookies" below.
17050
17051 - "captured_response_cookie" is an optional "name=value" entry indicating
17052 that the server has returned a cookie with its response. The cookie name
17053 and its maximum length are defined by the "capture cookie" statement in the
17054 frontend configuration. The field is a single dash ('-') when the option is
17055 not set. Only one cookie may be captured, it is generally used to track
17056 session ID exchanges between a client and a server to detect session
17057 crossing between clients due to application bugs. For more details, please
17058 consult the section "Capturing HTTP headers and cookies" below.
17059
17060 - "termination_state" is the condition the session was in when the session
17061 ended. This indicates the session state, which side caused the end of
17062 session to happen, for what reason (timeout, error, ...), just like in TCP
17063 logs, and information about persistence operations on cookies in the last
17064 two characters. The normal flags should begin with "--", indicating the
17065 session was closed by either end with no data remaining in buffers. See
17066 below "Session state at disconnection" for more details.
17067
17068 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017069 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017070 limits have been reached. For instance, if actconn is close to 512 or 1024
17071 when multiple connection errors occur, chances are high that the system
17072 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017073 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017074 system.
17075
17076 - "feconn" is the total number of concurrent connections on the frontend when
17077 the session was logged. It is useful to estimate the amount of resource
17078 required to sustain high loads, and to detect when the frontend's "maxconn"
17079 has been reached. Most often when this value increases by huge jumps, it is
17080 because there is congestion on the backend servers, but sometimes it can be
17081 caused by a denial of service attack.
17082
17083 - "beconn" is the total number of concurrent connections handled by the
17084 backend when the session was logged. It includes the total number of
17085 concurrent connections active on servers as well as the number of
17086 connections pending in queues. It is useful to estimate the amount of
17087 additional servers needed to support high loads for a given application.
17088 Most often when this value increases by huge jumps, it is because there is
17089 congestion on the backend servers, but sometimes it can be caused by a
17090 denial of service attack.
17091
17092 - "srv_conn" is the total number of concurrent connections still active on
17093 the server when the session was logged. It can never exceed the server's
17094 configured "maxconn" parameter. If this value is very often close or equal
17095 to the server's "maxconn", it means that traffic regulation is involved a
17096 lot, meaning that either the server's maxconn value is too low, or that
17097 there aren't enough servers to process the load with an optimal response
17098 time. When only one of the server's "srv_conn" is high, it usually means
17099 that this server has some trouble causing the requests to take longer to be
17100 processed than on other servers.
17101
17102 - "retries" is the number of connection retries experienced by this session
17103 when trying to connect to the server. It must normally be zero, unless a
17104 server is being stopped at the same moment the connection was attempted.
17105 Frequent retries generally indicate either a network problem between
17106 haproxy and the server, or a misconfigured system backlog on the server
17107 preventing new connections from being queued. This field may optionally be
17108 prefixed with a '+' sign, indicating that the session has experienced a
17109 redispatch after the maximal retry count has been reached on the initial
17110 server. In this case, the server name appearing in the log is the one the
17111 connection was redispatched to, and not the first one, though both may
17112 sometimes be the same in case of hashing for instance. So as a general rule
17113 of thumb, when a '+' is present in front of the retry count, this count
17114 should not be attributed to the logged server.
17115
17116 - "srv_queue" is the total number of requests which were processed before
17117 this one in the server queue. It is zero when the request has not gone
17118 through the server queue. It makes it possible to estimate the approximate
17119 server's response time by dividing the time spent in queue by the number of
17120 requests in the queue. It is worth noting that if a session experiences a
17121 redispatch and passes through two server queues, their positions will be
Davor Ocelice9ed2812017-12-25 17:49:28 +010017122 cumulative. A request should not pass through both the server queue and the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017123 backend queue unless a redispatch occurs.
17124
17125 - "backend_queue" is the total number of requests which were processed before
17126 this one in the backend's global queue. It is zero when the request has not
17127 gone through the global queue. It makes it possible to estimate the average
17128 queue length, which easily translates into a number of missing servers when
17129 divided by a server's "maxconn" parameter. It is worth noting that if a
17130 session experiences a redispatch, it may pass twice in the backend's queue,
Davor Ocelice9ed2812017-12-25 17:49:28 +010017131 and then both positions will be cumulative. A request should not pass
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017132 through both the server queue and the backend queue unless a redispatch
17133 occurs.
17134
17135 - "captured_request_headers" is a list of headers captured in the request due
17136 to the presence of the "capture request header" statement in the frontend.
17137 Multiple headers can be captured, they will be delimited by a vertical bar
17138 ('|'). When no capture is enabled, the braces do not appear, causing a
17139 shift of remaining fields. It is important to note that this field may
17140 contain spaces, and that using it requires a smarter log parser than when
17141 it's not used. Please consult the section "Capturing HTTP headers and
17142 cookies" below for more details.
17143
17144 - "captured_response_headers" is a list of headers captured in the response
17145 due to the presence of the "capture response header" statement in the
17146 frontend. Multiple headers can be captured, they will be delimited by a
17147 vertical bar ('|'). When no capture is enabled, the braces do not appear,
17148 causing a shift of remaining fields. It is important to note that this
17149 field may contain spaces, and that using it requires a smarter log parser
17150 than when it's not used. Please consult the section "Capturing HTTP headers
17151 and cookies" below for more details.
17152
17153 - "http_request" is the complete HTTP request line, including the method,
17154 request and HTTP version string. Non-printable characters are encoded (see
17155 below the section "Non-printable characters"). This is always the last
17156 field, and it is always delimited by quotes and is the only one which can
17157 contain quotes. If new fields are added to the log format, they will be
17158 added before this field. This field might be truncated if the request is
17159 huge and does not fit in the standard syslog buffer (1024 characters). This
17160 is the reason why this field must always remain the last one.
17161
17162
Cyril Bontédc4d9032012-04-08 21:57:39 +0200171638.2.4. Custom log format
17164------------------------
William Lallemand48940402012-01-30 16:47:22 +010017165
Willy Tarreau2beef582012-12-20 17:22:52 +010017166The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010017167mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010017168
Davor Ocelice9ed2812017-12-25 17:49:28 +010017169HAProxy understands some log format variables. % precedes log format variables.
William Lallemand48940402012-01-30 16:47:22 +010017170Variables can take arguments using braces ('{}'), and multiple arguments are
17171separated by commas within the braces. Flags may be added or removed by
17172prefixing them with a '+' or '-' sign.
17173
17174Special variable "%o" may be used to propagate its flags to all other
17175variables on the same format string. This is particularly handy with quoted
Dragan Dosen835b9212016-02-12 13:23:03 +010017176("Q") and escaped ("E") string formats.
William Lallemand48940402012-01-30 16:47:22 +010017177
Willy Tarreauc8368452012-12-21 00:09:23 +010017178If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020017179as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010017180less common information such as the client's SSL certificate's DN, or to log
17181the key that would be used to store an entry into a stick table.
17182
William Lallemand48940402012-01-30 16:47:22 +010017183Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017184In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010017185in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010017186
Dragan Dosen835b9212016-02-12 13:23:03 +010017187Note: when using the RFC5424 syslog message format, the characters '"',
17188'\' and ']' inside PARAM-VALUE should be escaped with '\' as prefix (see
17189https://tools.ietf.org/html/rfc5424#section-6.3.3 for more details). In
17190such cases, the use of the flag "E" should be considered.
17191
William Lallemand48940402012-01-30 16:47:22 +010017192Flags are :
17193 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017194 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
Dragan Dosen835b9212016-02-12 13:23:03 +010017195 * E: escape characters '"', '\' and ']' in a string with '\' as prefix
17196 (intended purpose is for the RFC5424 structured-data log formats)
William Lallemand48940402012-01-30 16:47:22 +010017197
17198 Example:
17199
17200 log-format %T\ %t\ Some\ Text
17201 log-format %{+Q}o\ %t\ %s\ %{-Q}r
17202
Dragan Dosen835b9212016-02-12 13:23:03 +010017203 log-format-sd %{+Q,+E}o\ [exampleSDID@1234\ header=%[capture.req.hdr(0)]]
17204
William Lallemand48940402012-01-30 16:47:22 +010017205At the moment, the default HTTP format is defined this way :
17206
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017207 log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC \
17208 %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"
William Lallemand48940402012-01-30 16:47:22 +010017209
William Lallemandbddd4fd2012-02-27 11:23:10 +010017210the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010017211
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017212 log-format "%{+Q}o %{-Q}ci - - [%trg] %r %ST %B \"\" \"\" %cp \
17213 %ms %ft %b %s %TR %Tw %Tc %Tr %Ta %tsc %ac %fc \
17214 %bc %sc %rc %sq %bq %CC %CS %hrl %hsl"
William Lallemand48940402012-01-30 16:47:22 +010017215
William Lallemandbddd4fd2012-02-27 11:23:10 +010017216and the default TCP format is defined this way :
17217
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017218 log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts \
17219 %ac/%fc/%bc/%sc/%rc %sq/%bq"
William Lallemandbddd4fd2012-02-27 11:23:10 +010017220
William Lallemand48940402012-01-30 16:47:22 +010017221Please refer to the table below for currently defined variables :
17222
William Lallemandbddd4fd2012-02-27 11:23:10 +010017223 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017224 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017225 +---+------+-----------------------------------------------+-------------+
17226 | | %o | special variable, apply flags on all next var | |
17227 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010017228 | | %B | bytes_read (from server to client) | numeric |
17229 | H | %CC | captured_request_cookie | string |
17230 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020017231 | | %H | hostname | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017232 | H | %HM | HTTP method (ex: POST) | string |
17233 | H | %HP | HTTP request URI without query string (path) | string |
Andrew Hayworthe63ac872015-07-31 16:14:16 +000017234 | H | %HQ | HTTP request URI query string (ex: ?bar=baz) | string |
Andrew Hayworth0ebc55f2015-04-27 21:37:03 +000017235 | H | %HU | HTTP request URI (ex: /foo?bar=baz) | string |
17236 | H | %HV | HTTP version (ex: HTTP/1.0) | string |
William Lallemanda73203e2012-03-12 12:48:57 +010017237 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020017238 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017239 | | %T | gmt_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017240 | | %Ta | Active time of the request (from TR to end) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017241 | | %Tc | Tc | numeric |
Willy Tarreau27b639d2016-05-17 17:55:27 +020017242 | | %Td | Td = Tt - (Tq + Tw + Tc + Tr) | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080017243 | | %Tl | local_date_time | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017244 | | %Th | connection handshake time (SSL, PROXY proto) | numeric |
17245 | H | %Ti | idle time before the HTTP request | numeric |
17246 | H | %Tq | Th + Ti + TR | numeric |
17247 | H | %TR | time to receive the full request from 1st byte| numeric |
17248 | H | %Tr | Tr (response time) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017249 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017250 | | %Tt | Tt | numeric |
17251 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017252 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017253 | | %ac | actconn | numeric |
17254 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017255 | | %bc | beconn (backend concurrent connections) | numeric |
17256 | | %bi | backend_source_ip (connecting address) | IP |
17257 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017258 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010017259 | | %ci | client_ip (accepted address) | IP |
17260 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017261 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017262 | | %fc | feconn (frontend concurrent connections) | numeric |
17263 | | %fi | frontend_ip (accepting address) | IP |
17264 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020017265 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020017266 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020017267 | | %hr | captured_request_headers default style | string |
17268 | | %hrl | captured_request_headers CLF style | string list |
17269 | | %hs | captured_response_headers default style | string |
17270 | | %hsl | captured_response_headers CLF style | string list |
Willy Tarreau812c88e2015-08-09 10:56:35 +020017271 | | %ms | accept date milliseconds (left-padded with 0) | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020017272 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017273 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017274 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010017275 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017276 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017277 | | %sc | srv_conn (server concurrent connections) | numeric |
17278 | | %si | server_IP (target address) | IP |
17279 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017280 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017281 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
17282 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010017283 | | %t | date_time (with millisecond resolution) | date |
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017284 | H | %tr | date_time of HTTP request | date |
17285 | H | %trg | gmt_date_time of start of HTTP request | date |
Jens Bissinger15c64ff2018-08-23 14:11:27 +020017286 | H | %trl | local_date_time of start of HTTP request | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017287 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017288 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010017289 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010017290
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020017291 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010017292
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017293
172948.2.5. Error log format
17295-----------------------
17296
17297When an incoming connection fails due to an SSL handshake or an invalid PROXY
17298protocol header, haproxy will log the event using a shorter, fixed line format.
17299By default, logs are emitted at the LOG_INFO level, unless the option
17300"log-separate-errors" is set in the backend, in which case the LOG_ERR level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017301will be used. Connections on which no data are exchanged (e.g. probes) are not
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010017302logged if the "dontlognull" option is set.
17303
17304The format looks like this :
17305
17306 >>> Dec 3 18:27:14 localhost \
17307 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
17308 Connection error during SSL handshake
17309
17310 Field Format Extract from the example above
17311 1 process_name '[' pid ']:' haproxy[6103]:
17312 2 client_ip ':' client_port 127.0.0.1:56059
17313 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
17314 4 frontend_name "/" bind_name ":" frt/f1:
17315 5 message Connection error during SSL handshake
17316
17317These fields just provide minimal information to help debugging connection
17318failures.
17319
17320
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173218.3. Advanced logging options
17322-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017323
17324Some advanced logging options are often looked for but are not easy to find out
17325just by looking at the various options. Here is an entry point for the few
17326options which can enable better logging. Please refer to the keywords reference
17327for more information about their usage.
17328
17329
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173308.3.1. Disabling logging of external tests
17331------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017332
17333It is quite common to have some monitoring tools perform health checks on
17334haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
17335commercial load-balancer, and sometimes it will simply be a more complete
17336monitoring system such as Nagios. When the tests are very frequent, users often
17337ask how to disable logging for those checks. There are three possibilities :
17338
17339 - if connections come from everywhere and are just TCP probes, it is often
17340 desired to simply disable logging of connections without data exchange, by
17341 setting "option dontlognull" in the frontend. It also disables logging of
17342 port scans, which may or may not be desired.
17343
17344 - if the connection come from a known source network, use "monitor-net" to
17345 declare this network as monitoring only. Any host in this network will then
17346 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017347 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017348 such as other load-balancers.
17349
17350 - if the tests are performed on a known URI, use "monitor-uri" to declare
17351 this URI as dedicated to monitoring. Any host sending this request will
17352 only get the result of a health-check, and the request will not be logged.
17353
17354
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173558.3.2. Logging before waiting for the session to terminate
17356----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017357
17358The problem with logging at end of connection is that you have no clue about
17359what is happening during very long sessions, such as remote terminal sessions
17360or large file downloads. This problem can be worked around by specifying
Davor Ocelice9ed2812017-12-25 17:49:28 +010017361"option logasap" in the frontend. HAProxy will then log as soon as possible,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017362just before data transfer begins. This means that in case of TCP, it will still
17363log the connection status to the server, and in case of HTTP, it will log just
17364after processing the server headers. In this case, the number of bytes reported
17365is the number of header bytes sent to the client. In order to avoid confusion
17366with normal logs, the total time field and the number of bytes are prefixed
17367with a '+' sign which means that real numbers are certainly larger.
17368
17369
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173708.3.3. Raising log level upon errors
17371------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017372
17373Sometimes it is more convenient to separate normal traffic from errors logs,
17374for instance in order to ease error monitoring from log files. When the option
17375"log-separate-errors" is used, connections which experience errors, timeouts,
17376retries, redispatches or HTTP status codes 5xx will see their syslog level
17377raised from "info" to "err". This will help a syslog daemon store the log in
17378a separate file. It is very important to keep the errors in the normal traffic
17379file too, so that log ordering is not altered. You should also be careful if
17380you already have configured your syslog daemon to store all logs higher than
17381"notice" in an "admin" file, because the "err" level is higher than "notice".
17382
17383
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173848.3.4. Disabling logging of successful connections
17385--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020017386
17387Although this may sound strange at first, some large sites have to deal with
17388multiple thousands of logs per second and are experiencing difficulties keeping
17389them intact for a long time or detecting errors within them. If the option
17390"dontlog-normal" is set on the frontend, all normal connections will not be
17391logged. In this regard, a normal connection is defined as one without any
17392error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
17393and a response with a status 5xx is not considered normal and will be logged
17394too. Of course, doing is is really discouraged as it will remove most of the
17395useful information from the logs. Do this only if you have no other
17396alternative.
17397
17398
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200173998.4. Timing events
17400------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017401
17402Timers provide a great help in troubleshooting network problems. All values are
17403reported in milliseconds (ms). These timers should be used in conjunction with
17404the session termination flags. In TCP mode with "option tcplog" set on the
17405frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017406mode, 5 control points are reported under the form "TR/Tw/Tc/Tr/Ta". In
17407addition, three other measures are provided, "Th", "Ti", and "Tq".
17408
Guillaume de Lafondf27cddc2016-12-23 17:32:43 +010017409Timings events in HTTP mode:
17410
17411 first request 2nd request
17412 |<-------------------------------->|<-------------- ...
17413 t tr t tr ...
17414 ---|----|----|----|----|----|----|----|----|--
17415 : Th Ti TR Tw Tc Tr Td : Ti ...
17416 :<---- Tq ---->: :
17417 :<-------------- Tt -------------->:
17418 :<--------- Ta --------->:
17419
17420Timings events in TCP mode:
17421
17422 TCP session
17423 |<----------------->|
17424 t t
17425 ---|----|----|----|----|---
17426 | Th Tw Tc Td |
17427 |<------ Tt ------->|
17428
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017429 - Th: total time to accept tcp connection and execute handshakes for low level
Davor Ocelice9ed2812017-12-25 17:49:28 +010017430 protocols. Currently, these protocols are proxy-protocol and SSL. This may
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017431 only happen once during the whole connection's lifetime. A large time here
17432 may indicate that the client only pre-established the connection without
17433 speaking, that it is experiencing network issues preventing it from
Davor Ocelice9ed2812017-12-25 17:49:28 +010017434 completing a handshake in a reasonable time (e.g. MTU issues), or that an
Willy Tarreau590a0512018-09-05 11:56:48 +020017435 SSL handshake was very expensive to compute. Please note that this time is
17436 reported only before the first request, so it is safe to average it over
17437 all request to calculate the amortized value. The second and subsequent
17438 request will always report zero here.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017439
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017440 - Ti: is the idle time before the HTTP request (HTTP mode only). This timer
17441 counts between the end of the handshakes and the first byte of the HTTP
17442 request. When dealing with a second request in keep-alive mode, it starts
Willy Tarreau590a0512018-09-05 11:56:48 +020017443 to count after the end of the transmission the previous response. When a
17444 multiplexed protocol such as HTTP/2 is used, it starts to count immediately
17445 after the previous request. Some browsers pre-establish connections to a
17446 server in order to reduce the latency of a future request, and keep them
17447 pending until they need it. This delay will be reported as the idle time. A
17448 value of -1 indicates that nothing was received on the connection.
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017449
17450 - TR: total time to get the client request (HTTP mode only). It's the time
17451 elapsed between the first bytes received and the moment the proxy received
17452 the empty line marking the end of the HTTP headers. The value "-1"
17453 indicates that the end of headers has never been seen. This happens when
17454 the client closes prematurely or times out. This time is usually very short
17455 since most requests fit in a single packet. A large time may indicate a
17456 request typed by hand during a test.
17457
17458 - Tq: total time to get the client request from the accept date or since the
17459 emission of the last byte of the previous response (HTTP mode only). It's
Davor Ocelice9ed2812017-12-25 17:49:28 +010017460 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 +020017461 returns -1 as well. This timer used to be very useful before the arrival of
17462 HTTP keep-alive and browsers' pre-connect feature. It's recommended to drop
17463 it in favor of TR nowadays, as the idle time adds a lot of noise to the
17464 reports.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017465
17466 - Tw: total time spent in the queues waiting for a connection slot. It
17467 accounts for backend queue as well as the server queues, and depends on the
17468 queue size, and the time needed for the server to complete previous
17469 requests. The value "-1" means that the request was killed before reaching
17470 the queue, which is generally what happens with invalid or denied requests.
17471
17472 - Tc: total time to establish the TCP connection to the server. It's the time
17473 elapsed between the moment the proxy sent the connection request, and the
17474 moment it was acknowledged by the server, or between the TCP SYN packet and
17475 the matching SYN/ACK packet in return. The value "-1" means that the
17476 connection never established.
17477
17478 - Tr: server response time (HTTP mode only). It's the time elapsed between
17479 the moment the TCP connection was established to the server and the moment
17480 the server sent its complete response headers. It purely shows its request
17481 processing time, without the network overhead due to the data transmission.
17482 It is worth noting that when the client has data to send to the server, for
17483 instance during a POST request, the time already runs, and this can distort
17484 apparent response time. For this reason, it's generally wise not to trust
17485 too much this field for POST requests initiated from clients behind an
17486 untrusted network. A value of "-1" here means that the last the response
17487 header (empty line) was never seen, most likely because the server timeout
17488 stroke before the server managed to process the request.
17489
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017490 - Ta: total active time for the HTTP request, between the moment the proxy
17491 received the first byte of the request header and the emission of the last
17492 byte of the response body. The exception is when the "logasap" option is
17493 specified. In this case, it only equals (TR+Tw+Tc+Tr), and is prefixed with
17494 a '+' sign. From this field, we can deduce "Td", the data transmission time,
17495 by subtracting other timers when valid :
17496
17497 Td = Ta - (TR + Tw + Tc + Tr)
17498
17499 Timers with "-1" values have to be excluded from this equation. Note that
17500 "Ta" can never be negative.
17501
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017502 - Tt: total session duration time, between the moment the proxy accepted it
17503 and the moment both ends were closed. The exception is when the "logasap"
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017504 option is specified. In this case, it only equals (Th+Ti+TR+Tw+Tc+Tr), and
17505 is prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017506 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017507
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017508 Td = Tt - (Th + Ti + TR + Tw + Tc + Tr)
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017509
17510 Timers with "-1" values have to be excluded from this equation. In TCP
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017511 mode, "Ti", "Tq" and "Tr" have to be excluded too. Note that "Tt" can never
17512 be negative and that for HTTP, Tt is simply equal to (Th+Ti+Ta).
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017513
17514These timers provide precious indications on trouble causes. Since the TCP
17515protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
17516that timers close to multiples of 3s are nearly always related to lost packets
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017517due to network problems (wires, negotiation, congestion). Moreover, if "Ta" or
17518"Tt" is close to a timeout value specified in the configuration, it often means
17519that a session has been aborted on timeout.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017520
17521Most common cases :
17522
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017523 - If "Th" or "Ti" are close to 3000, a packet has probably been lost between
17524 the client and the proxy. This is very rare on local networks but might
17525 happen when clients are on far remote networks and send large requests. It
17526 may happen that values larger than usual appear here without any network
17527 cause. Sometimes, during an attack or just after a resource starvation has
17528 ended, haproxy may accept thousands of connections in a few milliseconds.
17529 The time spent accepting these connections will inevitably slightly delay
17530 processing of other connections, and it can happen that request times in the
17531 order of a few tens of milliseconds are measured after a few thousands of
17532 new connections have been accepted at once. Using one of the keep-alive
17533 modes may display larger idle times since "Ti" measures the time spent
Patrick Mezard105faca2010-06-12 17:02:46 +020017534 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017535
17536 - If "Tc" is close to 3000, a packet has probably been lost between the
17537 server and the proxy during the server connection phase. This value should
17538 always be very low, such as 1 ms on local networks and less than a few tens
17539 of ms on remote networks.
17540
Willy Tarreau55165fe2009-05-10 12:02:55 +020017541 - If "Tr" is nearly always lower than 3000 except some rare values which seem
17542 to be the average majored by 3000, there are probably some packets lost
17543 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017544
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017545 - If "Ta" is large even for small byte counts, it generally is because
17546 neither the client nor the server decides to close the connection while
17547 haproxy is running in tunnel mode and both have agreed on a keep-alive
17548 connection mode. In order to solve this issue, it will be needed to specify
17549 one of the HTTP options to manipulate keep-alive or close options on either
17550 the frontend or the backend. Having the smallest possible 'Ta' or 'Tt' is
17551 important when connection regulation is used with the "maxconn" option on
17552 the servers, since no new connection will be sent to the server until
17553 another one is released.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017554
17555Other noticeable HTTP log cases ('xx' means any value to be ignored) :
17556
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017557 TR/Tw/Tc/Tr/+Ta The "option logasap" is present on the frontend and the log
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017558 was emitted before the data phase. All the timers are valid
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017559 except "Ta" which is shorter than reality.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017560
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017561 -1/xx/xx/xx/Ta The client was not able to send a complete request in time
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017562 or it aborted too early. Check the session termination flags
17563 then "timeout http-request" and "timeout client" settings.
17564
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017565 TR/-1/xx/xx/Ta It was not possible to process the request, maybe because
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017566 servers were out of order, because the request was invalid
17567 or forbidden by ACL rules. Check the session termination
17568 flags.
17569
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017570 TR/Tw/-1/xx/Ta The connection could not establish on the server. Either it
17571 actively refused it or it timed out after Ta-(TR+Tw) ms.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017572 Check the session termination flags, then check the
17573 "timeout connect" setting. Note that the tarpit action might
17574 return similar-looking patterns, with "Tw" equal to the time
17575 the client connection was maintained open.
17576
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017577 TR/Tw/Tc/-1/Ta The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017578 a complete response in time, or it closed its connection
Thierry FOURNIER / OZON.IO4cac3592016-07-28 17:19:45 +020017579 unexpectedly after Ta-(TR+Tw+Tc) ms. Check the session
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017580 termination flags, then check the "timeout server" setting.
17581
17582
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200175838.5. Session state at disconnection
17584-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017585
17586TCP and HTTP logs provide a session termination indicator in the
17587"termination_state" field, just before the number of active connections. It is
175882-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
17589each of which has a special meaning :
17590
17591 - On the first character, a code reporting the first event which caused the
17592 session to terminate :
17593
17594 C : the TCP session was unexpectedly aborted by the client.
17595
17596 S : the TCP session was unexpectedly aborted by the server, or the
17597 server explicitly refused it.
17598
17599 P : the session was prematurely aborted by the proxy, because of a
17600 connection limit enforcement, because a DENY filter was matched,
17601 because of a security check which detected and blocked a dangerous
17602 error in server response which might have caused information leak
Davor Ocelice9ed2812017-12-25 17:49:28 +010017603 (e.g. cacheable cookie).
Willy Tarreau570f2212013-06-10 16:42:09 +020017604
17605 L : the session was locally processed by haproxy and was not passed to
17606 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017607
17608 R : a resource on the proxy has been exhausted (memory, sockets, source
17609 ports, ...). Usually, this appears during the connection phase, and
17610 system logs should contain a copy of the precise error. If this
17611 happens, it must be considered as a very serious anomaly which
17612 should be fixed as soon as possible by any means.
17613
17614 I : an internal error was identified by the proxy during a self-check.
17615 This should NEVER happen, and you are encouraged to report any log
17616 containing this, because this would almost certainly be a bug. It
17617 would be wise to preventively restart the process after such an
17618 event too, in case it would be caused by memory corruption.
17619
Simon Horman752dc4a2011-06-21 14:34:59 +090017620 D : the session was killed by haproxy because the server was detected
17621 as down and was configured to kill all connections when going down.
17622
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070017623 U : the session was killed by haproxy on this backup server because an
17624 active server was detected as up and was configured to kill all
17625 backup connections when going up.
17626
Willy Tarreaua2a64e92011-09-07 23:01:56 +020017627 K : the session was actively killed by an admin operating on haproxy.
17628
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017629 c : the client-side timeout expired while waiting for the client to
17630 send or receive data.
17631
17632 s : the server-side timeout expired while waiting for the server to
17633 send or receive data.
17634
17635 - : normal session completion, both the client and the server closed
17636 with nothing left in the buffers.
17637
17638 - on the second character, the TCP or HTTP session state when it was closed :
17639
Willy Tarreauf7b30a92010-12-06 22:59:17 +010017640 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017641 (HTTP mode only). Nothing was sent to any server.
17642
17643 Q : the proxy was waiting in the QUEUE for a connection slot. This can
17644 only happen when servers have a 'maxconn' parameter set. It can
17645 also happen in the global queue after a redispatch consecutive to
17646 a failed attempt to connect to a dying server. If no redispatch is
17647 reported, then no connection attempt was made to any server.
17648
17649 C : the proxy was waiting for the CONNECTION to establish on the
17650 server. The server might at most have noticed a connection attempt.
17651
17652 H : the proxy was waiting for complete, valid response HEADERS from the
17653 server (HTTP only).
17654
17655 D : the session was in the DATA phase.
17656
17657 L : the proxy was still transmitting LAST data to the client while the
17658 server had already finished. This one is very rare as it can only
17659 happen when the client dies while receiving the last packets.
17660
17661 T : the request was tarpitted. It has been held open with the client
17662 during the whole "timeout tarpit" duration or until the client
17663 closed, both of which will be reported in the "Tw" timer.
17664
17665 - : normal session completion after end of data transfer.
17666
17667 - the third character tells whether the persistence cookie was provided by
17668 the client (only in HTTP mode) :
17669
17670 N : the client provided NO cookie. This is usually the case for new
17671 visitors, so counting the number of occurrences of this flag in the
17672 logs generally indicate a valid trend for the site frequentation.
17673
17674 I : the client provided an INVALID cookie matching no known server.
17675 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020017676 cookies between HTTP/HTTPS sites, persistence conditionally
17677 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017678
17679 D : the client provided a cookie designating a server which was DOWN,
17680 so either "option persist" was used and the client was sent to
17681 this server, or it was not set and the client was redispatched to
17682 another server.
17683
Willy Tarreau996a92c2010-10-13 19:30:47 +020017684 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017685 server.
17686
Willy Tarreau996a92c2010-10-13 19:30:47 +020017687 E : the client provided a valid cookie, but with a last date which was
17688 older than what is allowed by the "maxidle" cookie parameter, so
17689 the cookie is consider EXPIRED and is ignored. The request will be
17690 redispatched just as if there was no cookie.
17691
17692 O : the client provided a valid cookie, but with a first date which was
17693 older than what is allowed by the "maxlife" cookie parameter, so
17694 the cookie is consider too OLD and is ignored. The request will be
17695 redispatched just as if there was no cookie.
17696
Willy Tarreauc89ccb62012-04-05 21:18:22 +020017697 U : a cookie was present but was not used to select the server because
17698 some other server selection mechanism was used instead (typically a
17699 "use-server" rule).
17700
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017701 - : does not apply (no cookie set in configuration).
17702
17703 - the last character reports what operations were performed on the persistence
17704 cookie returned by the server (only in HTTP mode) :
17705
17706 N : NO cookie was provided by the server, and none was inserted either.
17707
17708 I : no cookie was provided by the server, and the proxy INSERTED one.
17709 Note that in "cookie insert" mode, if the server provides a cookie,
17710 it will still be overwritten and reported as "I" here.
17711
Willy Tarreau996a92c2010-10-13 19:30:47 +020017712 U : the proxy UPDATED the last date in the cookie that was presented by
17713 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030017714 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020017715 date indicated in the cookie. If any other change happens, such as
17716 a redispatch, then the cookie will be marked as inserted instead.
17717
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017718 P : a cookie was PROVIDED by the server and transmitted as-is.
17719
17720 R : the cookie provided by the server was REWRITTEN by the proxy, which
17721 happens in "cookie rewrite" or "cookie prefix" modes.
17722
17723 D : the cookie provided by the server was DELETED by the proxy.
17724
17725 - : does not apply (no cookie set in configuration).
17726
Willy Tarreau996a92c2010-10-13 19:30:47 +020017727The combination of the two first flags gives a lot of information about what
17728was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017729helpful to detect server saturation, network troubles, local system resource
17730starvation, attacks, etc...
17731
17732The most common termination flags combinations are indicated below. They are
17733alphabetically sorted, with the lowercase set just after the upper case for
17734easier finding and understanding.
17735
17736 Flags Reason
17737
17738 -- Normal termination.
17739
17740 CC The client aborted before the connection could be established to the
17741 server. This can happen when haproxy tries to connect to a recently
17742 dead (or unchecked) server, and the client aborts while haproxy is
17743 waiting for the server to respond or for "timeout connect" to expire.
17744
17745 CD The client unexpectedly aborted during data transfer. This can be
17746 caused by a browser crash, by an intermediate equipment between the
17747 client and haproxy which decided to actively break the connection,
17748 by network routing issues between the client and haproxy, or by a
17749 keep-alive session between the server and the client terminated first
17750 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010017751
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017752 cD The client did not send nor acknowledge any data for as long as the
17753 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020017754 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017755
17756 CH The client aborted while waiting for the server to start responding.
17757 It might be the server taking too long to respond or the client
17758 clicking the 'Stop' button too fast.
17759
17760 cH The "timeout client" stroke while waiting for client data during a
17761 POST request. This is sometimes caused by too large TCP MSS values
17762 for PPPoE networks which cannot transport full-sized packets. It can
17763 also happen when client timeout is smaller than server timeout and
17764 the server takes too long to respond.
17765
17766 CQ The client aborted while its session was queued, waiting for a server
17767 with enough empty slots to accept it. It might be that either all the
17768 servers were saturated or that the assigned server was taking too
17769 long a time to respond.
17770
17771 CR The client aborted before sending a full HTTP request. Most likely
17772 the request was typed by hand using a telnet client, and aborted
17773 too early. The HTTP status code is likely a 400 here. Sometimes this
17774 might also be caused by an IDS killing the connection between haproxy
Willy Tarreau0f228a02015-05-01 15:37:53 +020017775 and the client. "option http-ignore-probes" can be used to ignore
17776 connections without any data transfer.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017777
17778 cR The "timeout http-request" stroke before the client sent a full HTTP
17779 request. This is sometimes caused by too large TCP MSS values on the
17780 client side for PPPoE networks which cannot transport full-sized
17781 packets, or by clients sending requests by hand and not typing fast
17782 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020017783 request. The HTTP status code is likely a 408 here. Note: recently,
Willy Tarreau0f228a02015-05-01 15:37:53 +020017784 some browsers started to implement a "pre-connect" feature consisting
17785 in speculatively connecting to some recently visited web sites just
17786 in case the user would like to visit them. This results in many
17787 connections being established to web sites, which end up in 408
17788 Request Timeout if the timeout strikes first, or 400 Bad Request when
17789 the browser decides to close them first. These ones pollute the log
17790 and feed the error counters. Some versions of some browsers have even
17791 been reported to display the error code. It is possible to work
Davor Ocelice9ed2812017-12-25 17:49:28 +010017792 around the undesirable effects of this behavior by adding "option
Willy Tarreau0f228a02015-05-01 15:37:53 +020017793 http-ignore-probes" in the frontend, resulting in connections with
17794 zero data transfer to be totally ignored. This will definitely hide
17795 the errors of people experiencing connectivity issues though.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017796
17797 CT The client aborted while its session was tarpitted. It is important to
17798 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020017799 wrong tarpit rules have been written. If a lot of them happen, it
17800 might make sense to lower the "timeout tarpit" value to something
17801 closer to the average reported "Tw" timer, in order not to consume
17802 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017803
Willy Tarreau570f2212013-06-10 16:42:09 +020017804 LR The request was intercepted and locally handled by haproxy. Generally
17805 it means that this was a redirect or a stats request.
17806
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017807 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017808 the TCP connection (the proxy received a TCP RST or an ICMP message
17809 in return). Under some circumstances, it can also be the network
Davor Ocelice9ed2812017-12-25 17:49:28 +010017810 stack telling the proxy that the server is unreachable (e.g. no route,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017811 or no ARP response on local network). When this happens in HTTP mode,
17812 the status code is likely a 502 or 503 here.
17813
17814 sC The "timeout connect" stroke before a connection to the server could
17815 complete. When this happens in HTTP mode, the status code is likely a
17816 503 or 504 here.
17817
17818 SD The connection to the server died with an error during the data
17819 transfer. This usually means that haproxy has received an RST from
17820 the server or an ICMP message from an intermediate equipment while
17821 exchanging data with the server. This can be caused by a server crash
17822 or by a network issue on an intermediate equipment.
17823
17824 sD The server did not send nor acknowledge any data for as long as the
17825 "timeout server" setting during the data phase. This is often caused
Davor Ocelice9ed2812017-12-25 17:49:28 +010017826 by too short timeouts on L4 equipment before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017827 load-balancers, ...), as well as keep-alive sessions maintained
17828 between the client and the server expiring first on haproxy.
17829
17830 SH The server aborted before sending its full HTTP response headers, or
17831 it crashed while processing the request. Since a server aborting at
17832 this moment is very rare, it would be wise to inspect its logs to
17833 control whether it crashed and why. The logged request may indicate a
17834 small set of faulty requests, demonstrating bugs in the application.
17835 Sometimes this might also be caused by an IDS killing the connection
17836 between haproxy and the server.
17837
17838 sH The "timeout server" stroke before the server could return its
17839 response headers. This is the most common anomaly, indicating too
17840 long transactions, probably caused by server or database saturation.
17841 The immediate workaround consists in increasing the "timeout server"
17842 setting, but it is important to keep in mind that the user experience
17843 will suffer from these long response times. The only long term
17844 solution is to fix the application.
17845
17846 sQ The session spent too much time in queue and has been expired. See
17847 the "timeout queue" and "timeout connect" settings to find out how to
17848 fix this if it happens too often. If it often happens massively in
17849 short periods, it may indicate general problems on the affected
17850 servers due to I/O or database congestion, or saturation caused by
17851 external attacks.
17852
17853 PC The proxy refused to establish a connection to the server because the
17854 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020017855 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017856 so that it does not happen anymore. This status is very rare and
17857 might happen when the global "ulimit-n" parameter is forced by hand.
17858
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017859 PD The proxy blocked an incorrectly formatted chunked encoded message in
17860 a request or a response, after the server has emitted its headers. In
17861 most cases, this will indicate an invalid message from the server to
Davor Ocelice9ed2812017-12-25 17:49:28 +010017862 the client. HAProxy supports chunk sizes of up to 2GB - 1 (2147483647
Willy Tarreauf3a3e132013-08-31 08:16:26 +020017863 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017864
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017865 PH The proxy blocked the server's response, because it was invalid,
17866 incomplete, dangerous (cache control), or matched a security filter.
17867 In any case, an HTTP 502 error is sent to the client. One possible
17868 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010017869 containing unauthorized characters. It is also possible but quite
17870 rare, that the proxy blocked a chunked-encoding request from the
17871 client due to an invalid syntax, before the server responded. In this
17872 case, an HTTP 400 error is sent to the client and reported in the
17873 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017874
17875 PR The proxy blocked the client's HTTP request, either because of an
17876 invalid HTTP syntax, in which case it returned an HTTP 400 error to
17877 the client, or because a deny filter matched, in which case it
17878 returned an HTTP 403 error.
17879
17880 PT The proxy blocked the client's request and has tarpitted its
17881 connection before returning it a 500 server error. Nothing was sent
17882 to the server. The connection was maintained open for as long as
17883 reported by the "Tw" timer field.
17884
17885 RC A local resource has been exhausted (memory, sockets, source ports)
17886 preventing the connection to the server from establishing. The error
17887 logs will tell precisely what was missing. This is very rare and can
17888 only be solved by proper system tuning.
17889
Willy Tarreau996a92c2010-10-13 19:30:47 +020017890The combination of the two last flags gives a lot of information about how
17891persistence was handled by the client, the server and by haproxy. This is very
17892important to troubleshoot disconnections, when users complain they have to
17893re-authenticate. The commonly encountered flags are :
17894
17895 -- Persistence cookie is not enabled.
17896
17897 NN No cookie was provided by the client, none was inserted in the
17898 response. For instance, this can be in insert mode with "postonly"
17899 set on a GET request.
17900
17901 II A cookie designating an invalid server was provided by the client,
17902 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040017903 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020017904 value can be presented by a client when no other server knows it.
17905
17906 NI No cookie was provided by the client, one was inserted in the
17907 response. This typically happens for first requests from every user
17908 in "insert" mode, which makes it an easy way to count real users.
17909
17910 VN A cookie was provided by the client, none was inserted in the
17911 response. This happens for most responses for which the client has
17912 already got a cookie.
17913
17914 VU A cookie was provided by the client, with a last visit date which is
17915 not completely up-to-date, so an updated cookie was provided in
17916 response. This can also happen if there was no date at all, or if
17917 there was a date but the "maxidle" parameter was not set, so that the
17918 cookie can be switched to unlimited time.
17919
17920 EI A cookie was provided by the client, with a last visit date which is
17921 too old for the "maxidle" parameter, so the cookie was ignored and a
17922 new cookie was inserted in the response.
17923
17924 OI A cookie was provided by the client, with a first visit date which is
17925 too old for the "maxlife" parameter, so the cookie was ignored and a
17926 new cookie was inserted in the response.
17927
17928 DI The server designated by the cookie was down, a new server was
17929 selected and a new cookie was emitted in the response.
17930
17931 VI The server designated by the cookie was not marked dead but could not
17932 be reached. A redispatch happened and selected another one, which was
17933 then advertised in the response.
17934
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017935
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179368.6. Non-printable characters
17937-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017938
17939In order not to cause trouble to log analysis tools or terminals during log
17940consulting, non-printable characters are not sent as-is into log files, but are
17941converted to the two-digits hexadecimal representation of their ASCII code,
17942prefixed by the character '#'. The only characters that can be logged without
17943being escaped are comprised between 32 and 126 (inclusive). Obviously, the
17944escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
17945is the same for the character '"' which becomes "#22", as well as '{', '|' and
17946'}' when logging headers.
17947
17948Note that the space character (' ') is not encoded in headers, which can cause
17949issues for tools relying on space count to locate fields. A typical header
17950containing spaces is "User-Agent".
17951
17952Last, it has been observed that some syslog daemons such as syslog-ng escape
17953the quote ('"') with a backslash ('\'). The reverse operation can safely be
17954performed since no quote may appear anywhere else in the logs.
17955
17956
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179578.7. Capturing HTTP cookies
17958---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017959
17960Cookie capture simplifies the tracking a complete user session. This can be
17961achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017962section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017963cookie will simultaneously be checked in the request ("Cookie:" header) and in
17964the response ("Set-Cookie:" header). The respective values will be reported in
17965the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017966locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017967not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
17968user switches to a new session for example, because the server will reassign it
17969a new cookie. It is also possible to detect if a server unexpectedly sets a
17970wrong cookie to a client, leading to session crossing.
17971
17972 Examples :
17973 # capture the first cookie whose name starts with "ASPSESSION"
17974 capture cookie ASPSESSION len 32
17975
17976 # capture the first cookie whose name is exactly "vgnvisitor"
17977 capture cookie vgnvisitor= len 32
17978
17979
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200179808.8. Capturing HTTP headers
17981---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017982
17983Header captures are useful to track unique request identifiers set by an upper
17984proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
17985the response, one can search for information about the response length, how the
17986server asked the cache to behave, or an object location during a redirection.
17987
17988Header captures are performed using the "capture request header" and "capture
17989response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020017990section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017991
17992It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010017993time. Non-existent headers are logged as empty strings, and if one header
17994appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010017995are grouped within braces '{' and '}' in the same order as they were declared,
17996and delimited with a vertical bar '|' without any space. Response headers
17997follow the same representation, but are displayed after a space following the
17998request headers block. These blocks are displayed just before the HTTP request
17999in the logs.
18000
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020018001As a special case, it is possible to specify an HTTP header capture in a TCP
18002frontend. The purpose is to enable logging of headers which will be parsed in
18003an HTTP backend if the request is then switched to this HTTP backend.
18004
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018005 Example :
18006 # This instance chains to the outgoing proxy
18007 listen proxy-out
18008 mode http
18009 option httplog
18010 option logasap
18011 log global
18012 server cache1 192.168.1.1:3128
18013
18014 # log the name of the virtual server
18015 capture request header Host len 20
18016
18017 # log the amount of data uploaded during a POST
18018 capture request header Content-Length len 10
18019
18020 # log the beginning of the referrer
18021 capture request header Referer len 20
18022
18023 # server name (useful for outgoing proxies only)
18024 capture response header Server len 20
18025
18026 # logging the content-length is useful with "option logasap"
18027 capture response header Content-Length len 10
18028
Davor Ocelice9ed2812017-12-25 17:49:28 +010018029 # log the expected cache behavior on the response
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018030 capture response header Cache-Control len 8
18031
18032 # the Via header will report the next proxy's name
18033 capture response header Via len 20
18034
18035 # log the URL location during a redirection
18036 capture response header Location len 20
18037
18038 >>> Aug 9 20:26:09 localhost \
18039 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
18040 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
18041 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
18042 "GET http://fr.adserver.yahoo.com/"
18043
18044 >>> Aug 9 20:30:46 localhost \
18045 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
18046 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
18047 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018048 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018049
18050 >>> Aug 9 20:30:46 localhost \
18051 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
18052 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
18053 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
18054 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018055 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018056
18057
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180588.9. Examples of logs
18059---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018060
18061These are real-world examples of logs accompanied with an explanation. Some of
18062them have been made up by hand. The syslog part has been removed for better
18063reading. Their sole purpose is to explain how to decipher them.
18064
18065 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
18066 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
18067 "HEAD / HTTP/1.0"
18068
18069 => long request (6.5s) entered by hand through 'telnet'. The server replied
18070 in 147 ms, and the session ended normally ('----')
18071
18072 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
18073 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
18074 0/9 "HEAD / HTTP/1.0"
18075
18076 => Idem, but the request was queued in the global queue behind 9 other
18077 requests, and waited there for 1230 ms.
18078
18079 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
18080 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
18081 "GET /image.iso HTTP/1.0"
18082
18083 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010018084 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018085 14 ms, 243 bytes of headers were sent to the client, and total time from
18086 accept to first data byte is 30 ms.
18087
18088 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
18089 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
18090 "GET /cgi-bin/bug.cgi? HTTP/1.0"
18091
18092 => the proxy blocked a server response either because of an "rspdeny" or
18093 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020018094 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018095 risked being cached. In this case, the response is replaced with a "502
18096 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
18097 to return the 502 and not the server.
18098
18099 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018100 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 +010018101
18102 => the client never completed its request and aborted itself ("C---") after
18103 8.5s, while the proxy was waiting for the request headers ("-R--").
18104 Nothing was sent to any server.
18105
18106 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
18107 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
18108
18109 => The client never completed its request, which was aborted by the
18110 time-out ("c---") after 50s, while the proxy was waiting for the request
Davor Ocelice9ed2812017-12-25 17:49:28 +010018111 headers ("-R--"). Nothing was sent to any server, but the proxy could
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018112 send a 408 return code to the client.
18113
18114 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
18115 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
18116
18117 => This log was produced with "option tcplog". The client timed out after
18118 5 seconds ("c----").
18119
18120 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
18121 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010018122 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018123
18124 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020018125 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010018126 (config says 'retries 3'), and no redispatch (otherwise we would have
18127 seen "/+3"). Status code 503 was returned to the client. There were 115
18128 connections on this server, 202 connections on this proxy, and 205 on
18129 the global process. It is possible that the server refused the
18130 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010018131
Willy Tarreau52b2d222011-09-07 23:48:48 +020018132
Christopher Fauletc3fe5332016-04-07 15:30:10 +0200181339. Supported filters
18134--------------------
18135
18136Here are listed officially supported filters with the list of parameters they
18137accept. Depending on compile options, some of these filters might be
18138unavailable. The list of available filters is reported in haproxy -vv.
18139
18140See also : "filter"
18141
181429.1. Trace
18143----------
18144
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018145filter trace [name <name>] [random-parsing] [random-forwarding] [hexdump]
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018146
18147 Arguments:
18148 <name> is an arbitrary name that will be reported in
18149 messages. If no name is provided, "TRACE" is used.
18150
18151 <random-parsing> enables the random parsing of data exchanged between
18152 the client and the server. By default, this filter
18153 parses all available data. With this parameter, it
18154 only parses a random amount of the available data.
18155
Davor Ocelice9ed2812017-12-25 17:49:28 +010018156 <random-forwarding> enables the random forwarding of parsed data. By
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018157 default, this filter forwards all previously parsed
18158 data. With this parameter, it only forwards a random
18159 amount of the parsed data.
18160
Davor Ocelice9ed2812017-12-25 17:49:28 +010018161 <hexdump> dumps all forwarded data to the server and the client.
Christopher Faulet31bfe1f2016-12-09 17:42:38 +010018162
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018163This filter can be used as a base to develop new filters. It defines all
18164callbacks and print a message on the standard error stream (stderr) with useful
18165information for all of them. It may be useful to debug the activity of other
18166filters or, quite simply, HAProxy's activity.
18167
18168Using <random-parsing> and/or <random-forwarding> parameters is a good way to
18169tests the behavior of a filter that parses data exchanged between a client and
18170a server by adding some latencies in the processing.
18171
18172
181739.2. HTTP compression
18174---------------------
18175
18176filter compression
18177
18178The HTTP compression has been moved in a filter in HAProxy 1.7. "compression"
18179keyword must still be used to enable and configure the HTTP compression. And
Christopher Faulet27d93c32018-12-15 22:32:02 +010018180when no other filter is used, it is enough. When used with the cache enabled,
18181it is also enough. In this case, the compression is always done after the
18182response is stored in the cache. But it is mandatory to explicitly use a filter
18183line to enable the HTTP compression when at least one filter other than the
18184cache is used for the same listener/frontend/backend. This is important to know
18185the filters evaluation order.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018186
Christopher Faulet27d93c32018-12-15 22:32:02 +010018187See also : "compression" and section 9.4 about the cache filter.
Christopher Fauletc3fe5332016-04-07 15:30:10 +020018188
18189
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +0200181909.3. Stream Processing Offload Engine (SPOE)
18191--------------------------------------------
18192
18193filter spoe [engine <name>] config <file>
18194
18195 Arguments :
18196
18197 <name> is the engine name that will be used to find the right scope in
18198 the configuration file. If not provided, all the file will be
18199 parsed.
18200
18201 <file> is the path of the engine configuration file. This file can
18202 contain configuration of several engines. In this case, each
18203 part must be placed in its own scope.
18204
18205The Stream Processing Offload Engine (SPOE) is a filter communicating with
18206external components. It allows the offload of some specifics processing on the
Davor Ocelice9ed2812017-12-25 17:49:28 +010018207streams in tiered applications. These external components and information
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018208exchanged with them are configured in dedicated files, for the main part. It
18209also requires dedicated backends, defined in HAProxy configuration.
18210
18211SPOE communicates with external components using an in-house binary protocol,
18212the Stream Processing Offload Protocol (SPOP).
18213
Tim Düsterhus4896c442016-11-29 02:15:19 +010018214For all information about the SPOE configuration and the SPOP specification, see
Christopher Fauletf7e4e7e2016-10-27 22:29:49 +020018215"doc/SPOE.txt".
18216
18217Important note:
18218 The SPOE filter is highly experimental for now and was not heavily
18219 tested. It is really not production ready. So use it carefully.
18220
Christopher Faulet99a17a22018-12-11 09:18:27 +0100182219.4. Cache
18222----------
18223
18224filter cache <name>
18225
18226 Arguments :
18227
18228 <name> is name of the cache section this filter will use.
18229
18230The cache uses a filter to store cacheable responses. The HTTP rules
18231"cache-store" and "cache-use" must be used to define how and when to use a
18232cache. By default the correpsonding filter is implicitly defined. And when no
Christopher Faulet27d93c32018-12-15 22:32:02 +010018233other filters than cache or compression are used, it is enough. In such case,
18234the compression filter is always evaluated after the cache filter. But it is
18235mandatory to explicitly use a filter line to use a cache when at least one
18236filter other than the compression is used for the same
18237listener/frontend/backend. This is important to know the filters evaluation
18238order.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018239
Christopher Faulet27d93c32018-12-15 22:32:02 +010018240See also : section 9.2 about the compression filter and section 10 about cache.
Christopher Faulet99a17a22018-12-11 09:18:27 +010018241
William Lallemand86d0df02017-11-24 21:36:45 +01001824210. Cache
18243---------
18244
18245HAProxy provides a cache, which was designed to perform cache on small objects
18246(favicon, css...). This is a minimalist low-maintenance cache which runs in
18247RAM.
18248
18249The cache is based on a memory which is shared between processes and threads,
Cyril Bonté7b888f12017-11-26 22:24:31 +010018250this memory is split in blocks of 1k.
William Lallemand86d0df02017-11-24 21:36:45 +010018251
18252If an object is not used anymore, it can be deleted to store a new object
18253independently of its expiration date. The oldest objects are deleted first
18254when we try to allocate a new one.
18255
Cyril Bonté7b888f12017-11-26 22:24:31 +010018256The cache uses a hash of the host header and the URI as the key.
William Lallemand86d0df02017-11-24 21:36:45 +010018257
18258It's possible to view the status of a cache using the Unix socket command
18259"show cache" consult section 9.3 "Unix Socket commands" of Management Guide
18260for more details.
18261
18262When an object is delivered from the cache, the server name in the log is
18263replaced by "<CACHE>".
18264
Cyril Bonté7b888f12017-11-26 22:24:31 +01001826510.1. Limitation
18266----------------
William Lallemand86d0df02017-11-24 21:36:45 +010018267
18268The cache won't store and won't deliver objects in these cases:
18269
18270- If the response is not a 200
18271- If the response contains a Vary header
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018272- If the Content-Length + the headers size is greater than "max-object-size"
William Lallemand86d0df02017-11-24 21:36:45 +010018273- If the response is not cacheable
18274
18275- If the request is not a GET
18276- If the HTTP version of the request is smaller than 1.1
William Lallemand8a16fe02018-05-22 11:04:33 +020018277- If the request contains an Authorization header
William Lallemand86d0df02017-11-24 21:36:45 +010018278
Christopher Faulet99a17a22018-12-11 09:18:27 +010018279Caution!: For HAProxy version prior to 1.9, due to the limitation of the
18280filters, it is not recommended to use the cache with other filters. Using them
18281can cause undefined behavior if they modify the response (compression for
18282example). For HAProxy 1.9 and greater, it is safe, for HTX proxies only (see
18283"option http-use-htx" for details).
William Lallemand86d0df02017-11-24 21:36:45 +010018284
Cyril Bonté7b888f12017-11-26 22:24:31 +01001828510.2. Setup
18286-----------
William Lallemand86d0df02017-11-24 21:36:45 +010018287
18288To setup a cache, you must define a cache section and use it in a proxy with
18289the corresponding http-request and response actions.
18290
Cyril Bonté7b888f12017-11-26 22:24:31 +01001829110.2.1. Cache section
18292---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018293
18294cache <name>
18295 Declare a cache section, allocate a shared cache memory named <name>, the
18296 size of cache is mandatory.
18297
18298total-max-size <megabytes>
Davor Ocelice9ed2812017-12-25 17:49:28 +010018299 Define the size in RAM of the cache in megabytes. This size is split in
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018300 blocks of 1kB which are used by the cache entries. Its maximum value is 4095.
William Lallemand86d0df02017-11-24 21:36:45 +010018301
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018302max-object-size <bytes>
Frédéric Lécaillee3c83d82018-10-25 10:46:40 +020018303 Define the maximum size of the objects to be cached. Must not be greater than
18304 an half of "total-max-size". If not set, it equals to a 256th of the cache size.
18305 All objects with sizes larger than "max-object-size" will not be cached.
Frédéric Lécaille5f8bea62018-10-23 10:09:19 +020018306
William Lallemand86d0df02017-11-24 21:36:45 +010018307max-age <seconds>
18308 Define the maximum expiration duration. The expiration is set has the lowest
18309 value between the s-maxage or max-age (in this order) directive in the
18310 Cache-Control response header and this value. The default value is 60
18311 seconds, which means that you can't cache an object more than 60 seconds by
18312 default.
18313
Cyril Bonté7b888f12017-11-26 22:24:31 +01001831410.2.2. Proxy section
18315---------------------
William Lallemand86d0df02017-11-24 21:36:45 +010018316
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018317http-request cache-use <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018318 Try to deliver a cached object from the cache <name>. This directive is also
18319 mandatory to store the cache as it calculates the cache hash. If you want to
18320 use a condition for both storage and delivering that's a good idea to put it
18321 after this one.
18322
Jarno Huuskonen251a6b72019-01-04 14:05:02 +020018323http-response cache-store <name> [ { if | unless } <condition> ]
William Lallemand86d0df02017-11-24 21:36:45 +010018324 Store an http-response within the cache. The storage of the response headers
18325 is done at this step, which means you can use others http-response actions
18326 to modify headers before or after the storage of the response. This action
18327 is responsible for the setup of the cache storage filter.
18328
18329
18330Example:
18331
18332 backend bck1
18333 mode http
18334
18335 http-request cache-use foobar
18336 http-response cache-store foobar
18337 server srv1 127.0.0.1:80
18338
18339 cache foobar
18340 total-max-size 4
18341 max-age 240
18342
Willy Tarreau0ba27502007-12-24 16:55:16 +010018343/*
18344 * Local variables:
18345 * fill-column: 79
18346 * End:
18347 */