blob: c8295905360a3acdcb5777f8b1e6a2b8dc72966a [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 Tarreau15480d72014-06-19 21:10:58 +02005 version 1.6
Willy Tarreau6a06a402007-07-15 20:15:28 +02006 willy tarreau
Willy Tarreau9229f122014-06-19 21:01:06 +02007 2014/06/19
Willy Tarreau6a06a402007-07-15 20:15:28 +02008
9
10This document covers the configuration language as implemented in the version
11specified above. It does not provide any hint, example or advice. For such
Willy Tarreau0ba27502007-12-24 16:55:16 +010012documentation, please refer to the Reference Manual or the Architecture Manual.
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013The summary below is meant to help you search sections by name and navigate
14through 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
22 sometimes useful to prefix all output lines (logs, console outs) with 3
23 closing angle brackets ('>>>') in order to help get the difference between
24 inputs and outputs when it can become ambiguous. If you add sections,
25 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
341.2.1. The Request line
351.2.2. The request headers
361.3. HTTP response
371.3.1. The Response line
381.3.2. The response headers
39
402. Configuring HAProxy
412.1. Configuration file format
422.2. Time format
Patrick Mezard35da19c2010-06-12 17:02:47 +0200432.3. Examples
Willy Tarreauc57f0e22009-05-10 13:12:33 +020044
453. Global parameters
463.1. Process management and security
473.2. Performance tuning
483.3. Debugging
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +0100493.4. Userlists
Cyril Bontédc4d9032012-04-08 21:57:39 +0200503.5. Peers
Willy Tarreauc57f0e22009-05-10 13:12:33 +020051
524. Proxies
534.1. Proxy keywords matrix
544.2. Alphabetically sorted keywords reference
55
Willy Tarreau086fbf52012-09-24 20:34:51 +0200565. Bind and Server options
575.1. Bind options
585.2. Server and default-server options
Willy Tarreauc57f0e22009-05-10 13:12:33 +020059
606. HTTP header manipulation
61
Willy Tarreau74ca5042013-06-11 23:12:07 +0200627. Using ACLs and fetching samples
637.1. ACL basics
647.1.1. Matching booleans
657.1.2. Matching integers
667.1.3. Matching strings
677.1.4. Matching regular expressions (regexes)
687.1.5. Matching arbitrary data blocks
697.1.6. Matching IPv4 and IPv6 addresses
707.2. Using ACLs to form conditions
717.3. Fetching samples
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200727.3.1. Converters
737.3.2. Fetching samples from internal states
747.3.3. Fetching samples at Layer 4
757.3.4. Fetching samples at Layer 5
767.3.5. Fetching samples from buffer contents (Layer 6)
777.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +0200787.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020079
808. Logging
818.1. Log levels
828.2. Log formats
838.2.1. Default log format
848.2.2. TCP log format
858.2.3. HTTP log format
William Lallemand48940402012-01-30 16:47:22 +0100868.2.4. Custom log format
Willy Tarreau5f51e1a2012-12-03 18:40:10 +0100878.2.5. Error log format
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200888.3. Advanced logging options
898.3.1. Disabling logging of external tests
908.3.2. Logging before waiting for the session to terminate
918.3.3. Raising log level upon errors
928.3.4. Disabling logging of successful connections
938.4. Timing events
948.5. Session state at disconnection
958.6. Non-printable characters
968.7. Capturing HTTP cookies
978.8. Capturing HTTP headers
988.9. Examples of logs
99
1009. Statistics and monitoring
1019.1. CSV format
1029.2. Unix Socket commands
103
104
1051. Quick reminder about HTTP
106----------------------------
107
108When haproxy is running in HTTP mode, both the request and the response are
109fully analyzed and indexed, thus it becomes possible to build matching criteria
110on almost anything found in the contents.
111
112However, it is important to understand how HTTP requests and responses are
113formed, and how HAProxy decomposes them. It will then become easier to write
114correct rules and to debug existing configurations.
115
116
1171.1. The HTTP transaction model
118-------------------------------
119
120The HTTP protocol is transaction-driven. This means that each request will lead
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100121to one and only one response. Traditionally, a TCP connection is established
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200122from the client to the server, a request is sent by the client on the
123connection, the server responds and the connection is closed. A new request
124will involve a new connection :
125
126 [CON1] [REQ1] ... [RESP1] [CLO1] [CON2] [REQ2] ... [RESP2] [CLO2] ...
127
128In this mode, called the "HTTP close" mode, there are as many connection
129establishments as there are HTTP transactions. Since the connection is closed
130by the server after the response, the client does not need to know the content
131length.
132
133Due to the transactional nature of the protocol, it was possible to improve it
134to avoid closing a connection between two subsequent transactions. In this mode
135however, it is mandatory that the server indicates the content length for each
136response so that the client does not wait indefinitely. For this, a special
137header is used: "Content-length". This mode is called the "keep-alive" mode :
138
139 [CON] [REQ1] ... [RESP1] [REQ2] ... [RESP2] [CLO] ...
140
141Its advantages are a reduced latency between transactions, and less processing
142power required on the server side. It is generally better than the close mode,
143but not always because the clients often limit their concurrent connections to
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200144a smaller value.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200145
146A last improvement in the communications is the pipelining mode. It still uses
147keep-alive, but the client does not wait for the first response to send the
148second request. This is useful for fetching large number of images composing a
149page :
150
151 [CON] [REQ1] [REQ2] ... [RESP1] [RESP2] [CLO] ...
152
153This can obviously have a tremendous benefit on performance because the network
154latency is eliminated between subsequent requests. Many HTTP agents do not
155correctly support pipelining since there is no way to associate a response with
156the corresponding request in HTTP. For this reason, it is mandatory for the
Cyril Bonté78caf842010-03-10 22:41:43 +0100157server to reply in the exact same order as the requests were received.
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200158
Willy Tarreau70dffda2014-01-30 03:07:23 +0100159By default HAProxy operates in keep-alive mode with regards to persistent
160connections: for each connection it processes each request and response, and
161leaves the connection idle on both sides between the end of a response and the
162start of a new request.
Patrick Mezard9ec2ec42010-06-12 17:02:45 +0200163
Willy Tarreau70dffda2014-01-30 03:07:23 +0100164HAProxy supports 5 connection modes :
165 - keep alive : all requests and responses are processed (default)
166 - tunnel : only the first request and response are processed,
167 everything else is forwarded with no analysis.
168 - passive close : tunnel with "Connection: close" added in both directions.
169 - server close : the server-facing connection is closed after the response.
170 - forced close : the connection is actively closed after end of response.
171
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200172
1731.2. HTTP request
174-----------------
175
176First, let's consider this HTTP request :
177
178 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100179 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200180 1 GET /serv/login.php?lang=en&profile=2 HTTP/1.1
181 2 Host: www.mydomain.com
182 3 User-agent: my small browser
183 4 Accept: image/jpeg, image/gif
184 5 Accept: image/png
185
186
1871.2.1. The Request line
188-----------------------
189
190Line 1 is the "request line". It is always composed of 3 fields :
191
192 - a METHOD : GET
193 - a URI : /serv/login.php?lang=en&profile=2
194 - a version tag : HTTP/1.1
195
196All of them are delimited by what the standard calls LWS (linear white spaces),
197which are commonly spaces, but can also be tabs or line feeds/carriage returns
198followed by spaces/tabs. The method itself cannot contain any colon (':') and
199is limited to alphabetic letters. All those various combinations make it
200desirable that HAProxy performs the splitting itself rather than leaving it to
201the user to write a complex or inaccurate regular expression.
202
203The URI itself can have several forms :
204
205 - A "relative URI" :
206
207 /serv/login.php?lang=en&profile=2
208
209 It is a complete URL without the host part. This is generally what is
210 received by servers, reverse proxies and transparent proxies.
211
212 - An "absolute URI", also called a "URL" :
213
214 http://192.168.0.12:8080/serv/login.php?lang=en&profile=2
215
216 It is composed of a "scheme" (the protocol name followed by '://'), a host
217 name or address, optionally a colon (':') followed by a port number, then
218 a relative URI beginning at the first slash ('/') after the address part.
219 This is generally what proxies receive, but a server supporting HTTP/1.1
220 must accept this form too.
221
222 - a star ('*') : this form is only accepted in association with the OPTIONS
223 method and is not relayable. It is used to inquiry a next hop's
224 capabilities.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100225
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200226 - an address:port combination : 192.168.0.12:80
227 This is used with the CONNECT method, which is used to establish TCP
228 tunnels through HTTP proxies, generally for HTTPS, but sometimes for
229 other protocols too.
230
231In a relative URI, two sub-parts are identified. The part before the question
232mark is called the "path". It is typically the relative path to static objects
233on the server. The part after the question mark is called the "query string".
234It is mostly used with GET requests sent to dynamic scripts and is very
235specific to the language, framework or application in use.
236
237
2381.2.2. The request headers
239--------------------------
240
241The headers start at the second line. They are composed of a name at the
242beginning of the line, immediately followed by a colon (':'). Traditionally,
243an LWS is added after the colon but that's not required. Then come the values.
244Multiple identical headers may be folded into one single line, delimiting the
245values with commas, provided that their order is respected. This is commonly
246encountered in the "Cookie:" field. A header may span over multiple lines if
247the subsequent lines begin with an LWS. In the example in 1.2, lines 4 and 5
248define a total of 3 values for the "Accept:" header.
249
250Contrary to a common mis-conception, header names are not case-sensitive, and
251their values are not either if they refer to other header names (such as the
252"Connection:" header).
253
254The end of the headers is indicated by the first empty line. People often say
255that it's a double line feed, which is not exact, even if a double line feed
256is one valid form of empty line.
257
258Fortunately, HAProxy takes care of all these complex combinations when indexing
259headers, checking values and counting them, so there is no reason to worry
260about the way they could be written, but it is important not to accuse an
261application of being buggy if it does unusual, valid things.
262
263Important note:
264 As suggested by RFC2616, HAProxy normalizes headers by replacing line breaks
265 in the middle of headers by LWS in order to join multi-line headers. This
266 is necessary for proper analysis and helps less capable HTTP parsers to work
267 correctly and not to be fooled by such complex constructs.
268
269
2701.3. HTTP response
271------------------
272
273An HTTP response looks very much like an HTTP request. Both are called HTTP
274messages. Let's consider this HTTP response :
275
276 Line Contents
Willy Tarreaud72758d2010-01-12 10:42:19 +0100277 number
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200278 1 HTTP/1.1 200 OK
279 2 Content-length: 350
280 3 Content-Type: text/html
281
Willy Tarreau816b9792009-09-15 21:25:21 +0200282As a special case, HTTP supports so called "Informational responses" as status
283codes 1xx. These messages are special in that they don't convey any part of the
284response, they're just used as sort of a signaling message to ask a client to
Willy Tarreau5843d1a2010-02-01 15:13:32 +0100285continue to post its request for instance. In the case of a status 100 response
286the requested information will be carried by the next non-100 response message
287following the informational one. This implies that multiple responses may be
288sent to a single request, and that this only works when keep-alive is enabled
289(1xx messages are HTTP/1.1 only). HAProxy handles these messages and is able to
290correctly forward and skip them, and only process the next non-100 response. As
291such, these messages are neither logged nor transformed, unless explicitly
292state otherwise. Status 101 messages indicate that the protocol is changing
293over the same connection and that haproxy must switch to tunnel mode, just as
294if a CONNECT had occurred. Then the Upgrade header would contain additional
295information about the type of protocol the connection is switching to.
Willy Tarreau816b9792009-09-15 21:25:21 +0200296
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200297
2981.3.1. The Response line
299------------------------
300
301Line 1 is the "response line". It is always composed of 3 fields :
302
303 - a version tag : HTTP/1.1
304 - a status code : 200
305 - a reason : OK
306
307The status code is always 3-digit. The first digit indicates a general status :
Willy Tarreau816b9792009-09-15 21:25:21 +0200308 - 1xx = informational message to be skipped (eg: 100, 101)
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200309 - 2xx = OK, content is following (eg: 200, 206)
310 - 3xx = OK, no content following (eg: 302, 304)
311 - 4xx = error caused by the client (eg: 401, 403, 404)
312 - 5xx = error caused by the server (eg: 500, 502, 503)
313
314Please refer to RFC2616 for the detailed meaning of all such codes. The
Willy Tarreaud72758d2010-01-12 10:42:19 +0100315"reason" field is just a hint, but is not parsed by clients. Anything can be
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200316found there, but it's a common practice to respect the well-established
317messages. It can be composed of one or multiple words, such as "OK", "Found",
318or "Authentication Required".
319
320Haproxy may emit the following status codes by itself :
321
322 Code When / reason
323 200 access to stats page, and when replying to monitoring requests
324 301 when performing a redirection, depending on the configured code
325 302 when performing a redirection, depending on the configured code
326 303 when performing a redirection, depending on the configured code
Willy Tarreaub67fdc42013-03-29 19:28:11 +0100327 307 when performing a redirection, depending on the configured code
328 308 when performing a redirection, depending on the configured code
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200329 400 for an invalid or too large request
330 401 when an authentication is required to perform the action (when
331 accessing the stats page)
332 403 when a request is forbidden by a "block" ACL or "reqdeny" filter
333 408 when the request timeout strikes before the request is complete
334 500 when haproxy encounters an unrecoverable internal error, such as a
335 memory allocation failure, which should never happen
336 502 when the server returns an empty, invalid or incomplete response, or
337 when an "rspdeny" filter blocks the response.
338 503 when no server was available to handle the request, or in response to
339 monitoring requests which match the "monitor fail" condition
340 504 when the response timeout strikes before the server responds
341
342The error 4xx and 5xx codes above may be customized (see "errorloc" in section
3434.2).
344
345
3461.3.2. The response headers
347---------------------------
348
349Response headers work exactly like request headers, and as such, HAProxy uses
350the same parsing function for both. Please refer to paragraph 1.2.2 for more
351details.
352
353
3542. Configuring HAProxy
355----------------------
356
3572.1. Configuration file format
358------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +0200359
360HAProxy's configuration process involves 3 major sources of parameters :
361
362 - the arguments from the command-line, which always take precedence
363 - the "global" section, which sets process-wide parameters
364 - the proxies sections which can take form of "defaults", "listen",
365 "frontend" and "backend".
366
Willy Tarreau0ba27502007-12-24 16:55:16 +0100367The configuration file syntax consists in lines beginning with a keyword
368referenced in this manual, optionally followed by one or several parameters
369delimited by spaces. If spaces have to be entered in strings, then they must be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100370preceded by a backslash ('\') to be escaped. Backslashes also have to be
Willy Tarreau0ba27502007-12-24 16:55:16 +0100371escaped by doubling them.
372
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200373
3742.2. Time format
375----------------
376
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100377Some parameters involve values representing time, such as timeouts. These
Willy Tarreau0ba27502007-12-24 16:55:16 +0100378values are generally expressed in milliseconds (unless explicitly stated
379otherwise) but may be expressed in any other unit by suffixing the unit to the
380numeric value. It is important to consider this because it will not be repeated
381for every keyword. Supported units are :
382
383 - us : microseconds. 1 microsecond = 1/1000000 second
384 - ms : milliseconds. 1 millisecond = 1/1000 second. This is the default.
385 - s : seconds. 1s = 1000ms
386 - m : minutes. 1m = 60s = 60000ms
387 - h : hours. 1h = 60m = 3600s = 3600000ms
388 - d : days. 1d = 24h = 1440m = 86400s = 86400000ms
389
390
Patrick Mezard35da19c2010-06-12 17:02:47 +02003912.3. Examples
392-------------
393
394 # Simple configuration for an HTTP proxy listening on port 80 on all
395 # interfaces and forwarding requests to a single backend "servers" with a
396 # single server "server1" listening on 127.0.0.1:8000
397 global
398 daemon
399 maxconn 256
400
401 defaults
402 mode http
403 timeout connect 5000ms
404 timeout client 50000ms
405 timeout server 50000ms
406
407 frontend http-in
408 bind *:80
409 default_backend servers
410
411 backend servers
412 server server1 127.0.0.1:8000 maxconn 32
413
414
415 # The same configuration defined with a single listen block. Shorter but
416 # less expressive, especially in HTTP mode.
417 global
418 daemon
419 maxconn 256
420
421 defaults
422 mode http
423 timeout connect 5000ms
424 timeout client 50000ms
425 timeout server 50000ms
426
427 listen http-in
428 bind *:80
429 server server1 127.0.0.1:8000 maxconn 32
430
431
432Assuming haproxy is in $PATH, test these configurations in a shell with:
433
Willy Tarreauccb289d2010-12-11 20:19:38 +0100434 $ sudo haproxy -f configuration.conf -c
Patrick Mezard35da19c2010-06-12 17:02:47 +0200435
436
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004373. Global parameters
Willy Tarreau6a06a402007-07-15 20:15:28 +0200438--------------------
439
440Parameters in the "global" section are process-wide and often OS-specific. They
441are generally set once for all and do not need being changed once correct. Some
442of them have command-line equivalents.
443
444The following keywords are supported in the "global" section :
445
446 * Process management and security
Emeric Brunc8e8d122012-10-02 18:42:10 +0200447 - ca-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200448 - chroot
Emeric Brunc8e8d122012-10-02 18:42:10 +0200449 - crt-base
Willy Tarreau6a06a402007-07-15 20:15:28 +0200450 - daemon
Simon Horman98637e52014-06-20 12:30:16 +0900451 - external-check
Willy Tarreau6a06a402007-07-15 20:15:28 +0200452 - gid
453 - group
454 - log
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100455 - log-send-hostname
Willy Tarreau6a06a402007-07-15 20:15:28 +0200456 - nbproc
457 - pidfile
458 - uid
459 - ulimit-n
460 - user
Willy Tarreaufbee7132007-10-18 13:53:22 +0200461 - stats
Emeric Brun850efd52014-01-29 12:24:34 +0100462 - ssl-server-verify
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200463 - node
464 - description
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100465 - unix-bind
Willy Tarreaud72758d2010-01-12 10:42:19 +0100466
Willy Tarreau6a06a402007-07-15 20:15:28 +0200467 * Performance tuning
Willy Tarreau1746eec2014-04-25 10:46:47 +0200468 - max-spread-checks
Willy Tarreau6a06a402007-07-15 20:15:28 +0200469 - maxconn
Willy Tarreau81c25d02011-09-07 15:17:21 +0200470 - maxconnrate
William Lallemandd85f9172012-11-09 17:05:39 +0100471 - maxcomprate
William Lallemand072a2bf2012-11-20 17:01:01 +0100472 - maxcompcpuusage
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100473 - maxpipes
Willy Tarreau93e7c002013-10-07 18:51:07 +0200474 - maxsessrate
Willy Tarreau403edff2012-09-06 11:58:37 +0200475 - maxsslconn
Willy Tarreaue43d5322013-10-07 20:01:52 +0200476 - maxsslrate
Willy Tarreau6a06a402007-07-15 20:15:28 +0200477 - noepoll
478 - nokqueue
479 - nopoll
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100480 - nosplice
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300481 - nogetaddrinfo
Willy Tarreaufe255b72007-10-14 23:09:26 +0200482 - spread-checks
Willy Tarreau27a674e2009-08-17 07:23:33 +0200483 - tune.bufsize
Willy Tarreau43961d52010-10-04 20:39:20 +0200484 - tune.chksize
William Lallemandf3747832012-11-09 12:33:10 +0100485 - tune.comp.maxlevel
Willy Tarreau193b8c62012-11-22 00:17:38 +0100486 - tune.http.cookielen
Willy Tarreauac1932d2011-10-24 19:14:41 +0200487 - tune.http.maxhdr
Willy Tarreau7e312732014-02-12 16:35:14 +0100488 - tune.idletimer
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100489 - tune.maxaccept
490 - tune.maxpollevents
Willy Tarreau27a674e2009-08-17 07:23:33 +0200491 - tune.maxrewrite
Willy Tarreaubd9a0a72011-10-23 21:14:29 +0200492 - tune.pipesize
Willy Tarreaue803de22010-01-21 17:43:04 +0100493 - tune.rcvbuf.client
494 - tune.rcvbuf.server
495 - tune.sndbuf.client
496 - tune.sndbuf.server
Willy Tarreau6ec58db2012-11-16 16:32:15 +0100497 - tune.ssl.cachesize
Willy Tarreaubfd59462013-02-21 07:46:09 +0100498 - tune.ssl.lifetime
Emeric Brun8dc60392014-05-09 13:52:00 +0200499 - tune.ssl.force-private-cache
Willy Tarreaubfd59462013-02-21 07:46:09 +0100500 - tune.ssl.maxrecord
Remi Gacognef46cd6e2014-06-12 14:58:40 +0200501 - tune.ssl.default-dh-param
William Lallemanda509e4c2012-11-07 16:54:34 +0100502 - tune.zlib.memlevel
503 - tune.zlib.windowsize
Willy Tarreaud72758d2010-01-12 10:42:19 +0100504
Willy Tarreau6a06a402007-07-15 20:15:28 +0200505 * Debugging
506 - debug
507 - quiet
Willy Tarreau6a06a402007-07-15 20:15:28 +0200508
509
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005103.1. Process management and security
Willy Tarreau6a06a402007-07-15 20:15:28 +0200511------------------------------------
512
Emeric Brunc8e8d122012-10-02 18:42:10 +0200513ca-base <dir>
514 Assigns a default directory to fetch SSL CA certificates and CRLs from when a
Emeric Brunfd33a262012-10-11 16:28:27 +0200515 relative path is used with "ca-file" or "crl-file" directives. Absolute
516 locations specified in "ca-file" and "crl-file" prevail and ignore "ca-base".
Emeric Brunc8e8d122012-10-02 18:42:10 +0200517
Willy Tarreau6a06a402007-07-15 20:15:28 +0200518chroot <jail dir>
519 Changes current directory to <jail dir> and performs a chroot() there before
520 dropping privileges. This increases the security level in case an unknown
521 vulnerability would be exploited, since it would make it very hard for the
522 attacker to exploit the system. This only works when the process is started
523 with superuser privileges. It is important to ensure that <jail_dir> is both
524 empty and unwritable to anyone.
Willy Tarreaud72758d2010-01-12 10:42:19 +0100525
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100526cpu-map <"all"|"odd"|"even"|process_num> <cpu-set>...
527 On Linux 2.6 and above, it is possible to bind a process to a specific CPU
528 set. This means that the process will never run on other CPUs. The "cpu-map"
529 directive specifies CPU sets for process sets. The first argument is the
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100530 process number to bind. This process must have a number between 1 and 32 or
531 64, depending on the machine's word size, and any process IDs above nbproc
532 are ignored. It is possible to specify all processes at once using "all",
533 only odd numbers using "odd" or even numbers using "even", just like with the
534 "bind-process" directive. The second and forthcoming arguments are CPU sets.
535 Each CPU set is either a unique number between 0 and 31 or 63 or a range with
536 two such numbers delimited by a dash ('-'). Multiple CPU numbers or ranges
537 may be specified, and the processes will be allowed to bind to all of them.
538 Obviously, multiple "cpu-map" directives may be specified. Each "cpu-map"
539 directive will replace the previous ones when they overlap.
Willy Tarreaufc6c0322012-11-16 16:12:27 +0100540
Emeric Brunc8e8d122012-10-02 18:42:10 +0200541crt-base <dir>
542 Assigns a default directory to fetch SSL certificates from when a relative
543 path is used with "crtfile" directives. Absolute locations specified after
544 "crtfile" prevail and ignore "crt-base".
545
Willy Tarreau6a06a402007-07-15 20:15:28 +0200546daemon
547 Makes the process fork into background. This is the recommended mode of
548 operation. It is equivalent to the command line "-D" argument. It can be
549 disabled by the command line "-db" argument.
550
Simon Horman98637e52014-06-20 12:30:16 +0900551external-check
552 Allows the use of an external agent to perform health checks.
553 This is disabled by default as a security precaution.
554 See "option external-check".
555
Willy Tarreau6a06a402007-07-15 20:15:28 +0200556gid <number>
557 Changes the process' group ID to <number>. It is recommended that the group
558 ID is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
559 be started with a user belonging to this group, or with superuser privileges.
Michael Schererab012dd2013-01-12 18:35:19 +0100560 Note that if haproxy is started from a user having supplementary groups, it
561 will only be able to drop these groups if started with superuser privileges.
Willy Tarreau6a06a402007-07-15 20:15:28 +0200562 See also "group" and "uid".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100563
Willy Tarreau6a06a402007-07-15 20:15:28 +0200564group <group name>
565 Similar to "gid" but uses the GID of group name <group name> from /etc/group.
566 See also "gid" and "user".
Willy Tarreaud72758d2010-01-12 10:42:19 +0100567
Willy Tarreau18324f52014-06-27 18:10:07 +0200568log <address> [len <length>] <facility> [max level [min level]]
Willy Tarreau6a06a402007-07-15 20:15:28 +0200569 Adds a global syslog server. Up to two global servers can be defined. They
570 will receive logs for startups and exits, as well as all logs from proxies
Robert Tsai81ae1952007-12-05 10:47:29 +0100571 configured with "log global".
572
573 <address> can be one of:
574
Willy Tarreau2769aa02007-12-27 18:26:09 +0100575 - An IPv4 address optionally followed by a colon and a UDP port. If
Robert Tsai81ae1952007-12-05 10:47:29 +0100576 no port is specified, 514 is used by default (the standard syslog
577 port).
578
David du Colombier24bb5f52011-03-17 10:40:23 +0100579 - An IPv6 address followed by a colon and optionally a UDP port. If
580 no port is specified, 514 is used by default (the standard syslog
581 port).
582
Robert Tsai81ae1952007-12-05 10:47:29 +0100583 - A filesystem path to a UNIX domain socket, keeping in mind
584 considerations for chroot (be sure the path is accessible inside
585 the chroot) and uid/gid (be sure the path is appropriately
586 writeable).
587
Willy Tarreaudad36a32013-03-11 01:20:04 +0100588 Any part of the address string may reference any number of environment
589 variables by preceding their name with a dollar sign ('$') and
590 optionally enclosing them with braces ('{}'), similarly to what is done
591 in Bourne shell.
592
Willy Tarreau18324f52014-06-27 18:10:07 +0200593 <length> is an optional maximum line length. Log lines larger than this value
594 will be truncated before being sent. The reason is that syslog
595 servers act differently on log line length. All servers support the
596 default value of 1024, but some servers simply drop larger lines
597 while others do log them. If a server supports long lines, it may
598 make sense to set this value here in order to avoid truncating long
599 lines. Similarly, if a server drops long lines, it is preferable to
600 truncate them before sending them. Accepted values are 80 to 65535
601 inclusive. The default value of 1024 is generally fine for all
602 standard usages. Some specific cases of long captures or
603 JSON-formated logs may require larger values.
604
Robert Tsai81ae1952007-12-05 10:47:29 +0100605 <facility> must be one of the 24 standard syslog facilities :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200606
607 kern user mail daemon auth syslog lpr news
608 uucp cron auth2 ftp ntp audit alert cron2
609 local0 local1 local2 local3 local4 local5 local6 local7
610
611 An optional level can be specified to filter outgoing messages. By default,
Willy Tarreauf7edefa2009-05-10 17:20:05 +0200612 all messages are sent. If a maximum level is specified, only messages with a
613 severity at least as important as this level will be sent. An optional minimum
614 level can be specified. If it is set, logs emitted with a more severe level
615 than this one will be capped to this level. This is used to avoid sending
616 "emerg" messages on all terminals on some default syslog configurations.
617 Eight levels are known :
Willy Tarreau6a06a402007-07-15 20:15:28 +0200618
Cyril Bontédc4d9032012-04-08 21:57:39 +0200619 emerg alert crit err warning notice info debug
Willy Tarreau6a06a402007-07-15 20:15:28 +0200620
Joe Williamsdf5b38f2010-12-29 17:05:48 +0100621log-send-hostname [<string>]
622 Sets the hostname field in the syslog header. If optional "string" parameter
623 is set the header is set to the string contents, otherwise uses the hostname
624 of the system. Generally used if one is not relaying logs through an
625 intermediate syslog server or for simply customizing the hostname printed in
626 the logs.
627
Kevinm48936af2010-12-22 16:08:21 +0000628log-tag <string>
629 Sets the tag field in the syslog header to this string. It defaults to the
630 program name as launched from the command line, which usually is "haproxy".
631 Sometimes it can be useful to differentiate between multiple processes
Willy Tarreau094af4e2015-01-07 15:03:42 +0100632 running on the same host. See also the per-proxy "log-tag" directive.
Kevinm48936af2010-12-22 16:08:21 +0000633
Willy Tarreau6a06a402007-07-15 20:15:28 +0200634nbproc <number>
635 Creates <number> processes when going daemon. This requires the "daemon"
636 mode. By default, only one process is created, which is the recommended mode
637 of operation. For systems limited to small sets of file descriptors per
638 process, it may be needed to fork multiple daemons. USING MULTIPLE PROCESSES
639 IS HARDER TO DEBUG AND IS REALLY DISCOURAGED. See also "daemon".
640
641pidfile <pidfile>
642 Writes pids of all daemons into file <pidfile>. This option is equivalent to
643 the "-p" command line argument. The file must be accessible to the user
644 starting the process. See also "daemon".
645
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100646stats bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau35b7b162012-10-22 23:17:18 +0200647 Limits the stats socket to a certain set of processes numbers. By default the
648 stats socket is bound to all processes, causing a warning to be emitted when
649 nbproc is greater than 1 because there is no way to select the target process
650 when connecting. However, by using this setting, it becomes possible to pin
651 the stats socket to a specific set of processes, typically the first one. The
652 warning will automatically be disabled when this setting is used, whatever
Willy Tarreaua9db57e2013-01-18 11:29:29 +0100653 the number of processes used. The maximum process ID depends on the machine's
Willy Tarreauae302532014-05-07 19:22:24 +0200654 word size (32 or 64). A better option consists in using the "process" setting
655 of the "stats socket" line to force the process on each line.
Willy Tarreau35b7b162012-10-22 23:17:18 +0200656
Willy Tarreau610f04b2014-02-13 11:36:41 +0100657ssl-default-bind-ciphers <ciphers>
658 This setting is only available when support for OpenSSL was built in. It sets
659 the default string describing the list of cipher algorithms ("cipher suite")
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300660 that are negotiated during the SSL/TLS handshake for all "bind" lines which
Willy Tarreau610f04b2014-02-13 11:36:41 +0100661 do not explicitly define theirs. The format of the string is defined in
662 "man 1 ciphers" from OpenSSL man pages, and can be for instance a string such
663 as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes). Please check the
664 "bind" keyword for more information.
665
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100666ssl-default-bind-options [<option>]...
667 This setting is only available when support for OpenSSL was built in. It sets
668 default ssl-options to force on all "bind" lines. Please check the "bind"
669 keyword to see available options.
670
671 Example:
672 global
673 ssl-default-bind-options no-sslv3 no-tls-tickets
674
Willy Tarreau610f04b2014-02-13 11:36:41 +0100675ssl-default-server-ciphers <ciphers>
676 This setting is only available when support for OpenSSL was built in. It
677 sets the default string describing the list of cipher algorithms that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300678 negotiated during the SSL/TLS handshake with the server, for all "server"
Willy Tarreau610f04b2014-02-13 11:36:41 +0100679 lines which do not explicitly define theirs. The format of the string is
680 defined in "man 1 ciphers". Please check the "server" keyword for more
681 information.
682
Emeric Brun2c86cbf2014-10-30 15:56:50 +0100683ssl-default-server-options [<option>]...
684 This setting is only available when support for OpenSSL was built in. It sets
685 default ssl-options to force on all "server" lines. Please check the "server"
686 keyword to see available options.
687
Emeric Brun850efd52014-01-29 12:24:34 +0100688ssl-server-verify [none|required]
689 The default behavior for SSL verify on servers side. If specified to 'none',
690 servers certificates are not verified. The default is 'required' except if
691 forced using cmdline option '-dV'.
692
Willy Tarreauabb175f2012-09-24 12:43:26 +0200693stats socket [<address:port>|<path>] [param*]
694 Binds a UNIX socket to <path> or a TCPv4/v6 address to <address:port>.
695 Connections to this socket will return various statistics outputs and even
696 allow some commands to be issued to change some runtime settings. Please
697 consult section 9.2 "Unix Socket commands" for more details.
Willy Tarreau6162db22009-10-10 17:13:00 +0200698
Willy Tarreauabb175f2012-09-24 12:43:26 +0200699 All parameters supported by "bind" lines are supported, for instance to
700 restrict access to some users or their access rights. Please consult
701 section 5.1 for more information.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200702
703stats timeout <timeout, in milliseconds>
704 The default timeout on the stats socket is set to 10 seconds. It is possible
705 to change this value with "stats timeout". The value must be passed in
Willy Tarreaubefdff12007-12-02 22:27:38 +0100706 milliseconds, or be suffixed by a time unit among { us, ms, s, m, h, d }.
Willy Tarreaufbee7132007-10-18 13:53:22 +0200707
708stats maxconn <connections>
709 By default, the stats socket is limited to 10 concurrent connections. It is
710 possible to change this value with "stats maxconn".
711
Willy Tarreau6a06a402007-07-15 20:15:28 +0200712uid <number>
713 Changes the process' user ID to <number>. It is recommended that the user ID
714 is dedicated to HAProxy or to a small set of similar daemons. HAProxy must
715 be started with superuser privileges in order to be able to switch to another
716 one. See also "gid" and "user".
717
718ulimit-n <number>
719 Sets the maximum number of per-process file-descriptors to <number>. By
720 default, it is automatically computed, so it is recommended not to use this
721 option.
722
Willy Tarreauceb24bc2010-11-09 12:46:41 +0100723unix-bind [ prefix <prefix> ] [ mode <mode> ] [ user <user> ] [ uid <uid> ]
724 [ group <group> ] [ gid <gid> ]
725
726 Fixes common settings to UNIX listening sockets declared in "bind" statements.
727 This is mainly used to simplify declaration of those UNIX sockets and reduce
728 the risk of errors, since those settings are most commonly required but are
729 also process-specific. The <prefix> setting can be used to force all socket
730 path to be relative to that directory. This might be needed to access another
731 component's chroot. Note that those paths are resolved before haproxy chroots
732 itself, so they are absolute. The <mode>, <user>, <uid>, <group> and <gid>
733 all have the same meaning as their homonyms used by the "bind" statement. If
734 both are specified, the "bind" statement has priority, meaning that the
735 "unix-bind" settings may be seen as process-wide default settings.
736
Willy Tarreau6a06a402007-07-15 20:15:28 +0200737user <user name>
738 Similar to "uid" but uses the UID of user name <user name> from /etc/passwd.
739 See also "uid" and "group".
740
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +0200741node <name>
742 Only letters, digits, hyphen and underscore are allowed, like in DNS names.
743
744 This statement is useful in HA configurations where two or more processes or
745 servers share the same IP address. By setting a different node-name on all
746 nodes, it becomes easy to immediately spot what server is handling the
747 traffic.
748
749description <text>
750 Add a text that describes the instance.
751
752 Please note that it is required to escape certain characters (# for example)
753 and this text is inserted into a html page so you should avoid using
754 "<" and ">" characters.
755
Willy Tarreau6a06a402007-07-15 20:15:28 +0200756
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007573.2. Performance tuning
Willy Tarreau6a06a402007-07-15 20:15:28 +0200758-----------------------
759
Willy Tarreau1746eec2014-04-25 10:46:47 +0200760max-spread-checks <delay in milliseconds>
761 By default, haproxy tries to spread the start of health checks across the
762 smallest health check interval of all the servers in a farm. The principle is
763 to avoid hammering services running on the same server. But when using large
764 check intervals (10 seconds or more), the last servers in the farm take some
765 time before starting to be tested, which can be a problem. This parameter is
766 used to enforce an upper bound on delay between the first and the last check,
767 even if the servers' check intervals are larger. When servers run with
768 shorter intervals, their intervals will be respected though.
769
Willy Tarreau6a06a402007-07-15 20:15:28 +0200770maxconn <number>
771 Sets the maximum per-process number of concurrent connections to <number>. It
772 is equivalent to the command-line argument "-n". Proxies will stop accepting
773 connections when this limit is reached. The "ulimit-n" parameter is
Willy Tarreau8274e102014-06-19 15:31:25 +0200774 automatically adjusted according to this value. See also "ulimit-n". Note:
775 the "select" poller cannot reliably use more than 1024 file descriptors on
776 some platforms. If your platform only supports select and reports "select
777 FAILED" on startup, you need to reduce maxconn until it works (slightly
Willy Tarreaud0256482015-01-15 21:45:22 +0100778 below 500 in general). If this value is not set, it will default to the value
779 set in DEFAULT_MAXCONN at build time (reported in haproxy -vv) if no memory
780 limit is enforced, or will be computed based on the memory limit, the buffer
781 size, memory allocated to compression, SSL cache size, and use or not of SSL
782 and the associated maxsslconn (which can also be automatic).
Willy Tarreau6a06a402007-07-15 20:15:28 +0200783
Willy Tarreau81c25d02011-09-07 15:17:21 +0200784maxconnrate <number>
785 Sets the maximum per-process number of connections per second to <number>.
786 Proxies will stop accepting connections when this limit is reached. It can be
787 used to limit the global capacity regardless of each frontend capacity. It is
788 important to note that this can only be used as a service protection measure,
789 as there will not necessarily be a fair share between frontends when the
790 limit is reached, so it's a good idea to also limit each frontend to some
791 value close to its expected share. Also, lowering tune.maxaccept can improve
792 fairness.
793
William Lallemandd85f9172012-11-09 17:05:39 +0100794maxcomprate <number>
795 Sets the maximum per-process input compression rate to <number> kilobytes
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300796 per second. For each session, if the maximum is reached, the compression
William Lallemandd85f9172012-11-09 17:05:39 +0100797 level will be decreased during the session. If the maximum is reached at the
798 beginning of a session, the session will not compress at all. If the maximum
799 is not reached, the compression level will be increased up to
800 tune.comp.maxlevel. A value of zero means there is no limit, this is the
801 default value.
802
William Lallemand072a2bf2012-11-20 17:01:01 +0100803maxcompcpuusage <number>
804 Sets the maximum CPU usage HAProxy can reach before stopping the compression
805 for new requests or decreasing the compression level of current requests.
806 It works like 'maxcomprate' but measures CPU usage instead of incoming data
807 bandwidth. The value is expressed in percent of the CPU used by haproxy. In
808 case of multiple processes (nbproc > 1), each process manages its individual
809 usage. A value of 100 disable the limit. The default value is 100. Setting
810 a lower value will prevent the compression work from slowing the whole
811 process down and from introducing high latencies.
812
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100813maxpipes <number>
814 Sets the maximum per-process number of pipes to <number>. Currently, pipes
815 are only used by kernel-based tcp splicing. Since a pipe contains two file
816 descriptors, the "ulimit-n" value will be increased accordingly. The default
817 value is maxconn/4, which seems to be more than enough for most heavy usages.
818 The splice code dynamically allocates and releases pipes, and can fall back
819 to standard copy, so setting this value too low may only impact performance.
820
Willy Tarreau93e7c002013-10-07 18:51:07 +0200821maxsessrate <number>
822 Sets the maximum per-process number of sessions per second to <number>.
823 Proxies will stop accepting connections when this limit is reached. It can be
824 used to limit the global capacity regardless of each frontend capacity. It is
825 important to note that this can only be used as a service protection measure,
826 as there will not necessarily be a fair share between frontends when the
827 limit is reached, so it's a good idea to also limit each frontend to some
828 value close to its expected share. Also, lowering tune.maxaccept can improve
829 fairness.
830
Willy Tarreau403edff2012-09-06 11:58:37 +0200831maxsslconn <number>
832 Sets the maximum per-process number of concurrent SSL connections to
833 <number>. By default there is no SSL-specific limit, which means that the
834 global maxconn setting will apply to all connections. Setting this limit
835 avoids having openssl use too much memory and crash when malloc returns NULL
836 (since it unfortunately does not reliably check for such conditions). Note
837 that the limit applies both to incoming and outgoing connections, so one
838 connection which is deciphered then ciphered accounts for 2 SSL connections.
Willy Tarreaud0256482015-01-15 21:45:22 +0100839 If this value is not set, but a memory limit is enforced, this value will be
840 automatically computed based on the memory limit, maxconn, the buffer size,
841 memory allocated to compression, SSL cache size, and use of SSL in either
842 frontends, backends or both. If neither maxconn nor maxsslconn are specified
843 when there is a memory limit, haproxy will automatically adjust these values
844 so that 100% of the connections can be made over SSL with no risk, and will
845 consider the sides where it is enabled (frontend, backend, both).
Willy Tarreau403edff2012-09-06 11:58:37 +0200846
Willy Tarreaue43d5322013-10-07 20:01:52 +0200847maxsslrate <number>
848 Sets the maximum per-process number of SSL sessions per second to <number>.
849 SSL listeners will stop accepting connections when this limit is reached. It
850 can be used to limit the global SSL CPU usage regardless of each frontend
851 capacity. It is important to note that this can only be used as a service
852 protection measure, as there will not necessarily be a fair share between
853 frontends when the limit is reached, so it's a good idea to also limit each
854 frontend to some value close to its expected share. It is also important to
855 note that the sessions are accounted before they enter the SSL stack and not
856 after, which also protects the stack against bad handshakes. Also, lowering
857 tune.maxaccept can improve fairness.
858
William Lallemand9d5f5482012-11-07 16:12:57 +0100859maxzlibmem <number>
860 Sets the maximum amount of RAM in megabytes per process usable by the zlib.
861 When the maximum amount is reached, future sessions will not compress as long
862 as RAM is unavailable. When sets to 0, there is no limit.
William Lallemande3a7d992012-11-20 11:25:20 +0100863 The default value is 0. The value is available in bytes on the UNIX socket
864 with "show info" on the line "MaxZlibMemUsage", the memory used by zlib is
865 "ZlibMemUsage" in bytes.
866
Willy Tarreau6a06a402007-07-15 20:15:28 +0200867noepoll
868 Disables the use of the "epoll" event polling system on Linux. It is
869 equivalent to the command-line argument "-de". The next polling system
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100870 used will generally be "poll". See also "nopoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200871
872nokqueue
873 Disables the use of the "kqueue" event polling system on BSD. It is
874 equivalent to the command-line argument "-dk". The next polling system
875 used will generally be "poll". See also "nopoll".
876
877nopoll
878 Disables the use of the "poll" event polling system. It is equivalent to the
879 command-line argument "-dp". The next polling system used will be "select".
Willy Tarreau0ba27502007-12-24 16:55:16 +0100880 It should never be needed to disable "poll" since it's available on all
Willy Tarreaue9f49e72012-11-11 17:42:00 +0100881 platforms supported by HAProxy. See also "nokqueue" and "noepoll".
Willy Tarreau6a06a402007-07-15 20:15:28 +0200882
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100883nosplice
884 Disables the use of kernel tcp splicing between sockets on Linux. It is
885 equivalent to the command line argument "-dS". Data will then be copied
886 using conventional and more portable recv/send calls. Kernel tcp splicing is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +0100887 limited to some very recent instances of kernel 2.6. Most versions between
Willy Tarreauff4f82d2009-02-06 11:28:13 +0100888 2.6.25 and 2.6.28 are buggy and will forward corrupted data, so they must not
889 be used. This option makes it easier to globally disable kernel splicing in
890 case of doubt. See also "option splice-auto", "option splice-request" and
891 "option splice-response".
892
Jarno Huuskonen0e82b922014-04-12 18:22:19 +0300893nogetaddrinfo
894 Disables the use of getaddrinfo(3) for name resolving. It is equivalent to
895 the command line argument "-dG". Deprecated gethostbyname(3) will be used.
896
Willy Tarreaufe255b72007-10-14 23:09:26 +0200897spread-checks <0..50, in percent>
Simon Hormand60d6912013-11-25 10:46:36 +0900898 Sometimes it is desirable to avoid sending agent and health checks to
899 servers at exact intervals, for instance when many logical servers are
900 located on the same physical server. With the help of this parameter, it
901 becomes possible to add some randomness in the check interval between 0
902 and +/- 50%. A value between 2 and 5 seems to show good results. The
903 default value remains at 0.
Willy Tarreaufe255b72007-10-14 23:09:26 +0200904
Willy Tarreau33cb0652014-12-23 22:52:37 +0100905tune.buffers.limit <number>
906 Sets a hard limit on the number of buffers which may be allocated per process.
907 The default value is zero which means unlimited. The minimum non-zero value
908 will always be greater than "tune.buffers.reserve" and should ideally always
909 be about twice as large. Forcing this value can be particularly useful to
910 limit the amount of memory a process may take, while retaining a sane
911 behaviour. When this limit is reached, sessions which need a buffer wait for
912 another one to be released by another session. Since buffers are dynamically
913 allocated and released, the waiting time is very short and not perceptible
914 provided that limits remain reasonable. In fact sometimes reducing the limit
915 may even increase performance by increasing the CPU cache's efficiency. Tests
916 have shown good results on average HTTP traffic with a limit to 1/10 of the
917 expected global maxconn setting, which also significantly reduces memory
918 usage. The memory savings come from the fact that a number of connections
919 will not allocate 2*tune.bufsize. It is best not to touch this value unless
920 advised to do so by an haproxy core developer.
921
Willy Tarreau1058ae72014-12-23 22:40:40 +0100922tune.buffers.reserve <number>
923 Sets the number of buffers which are pre-allocated and reserved for use only
924 during memory shortage conditions resulting in failed memory allocations. The
925 minimum value is 2 and is also the default. There is no reason a user would
926 want to change this value, it's mostly aimed at haproxy core developers.
927
Willy Tarreau27a674e2009-08-17 07:23:33 +0200928tune.bufsize <number>
929 Sets the buffer size to this size (in bytes). Lower values allow more
930 sessions to coexist in the same amount of RAM, and higher values allow some
931 applications with very large cookies to work. The default value is 16384 and
932 can be changed at build time. It is strongly recommended not to change this
933 from the default value, as very low values will break some services such as
934 statistics, and values larger than default size will increase memory usage,
935 possibly causing the system to run out of memory. At least the global maxconn
936 parameter should be decreased by the same factor as this one is increased.
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +0400937 If HTTP request is larger than (tune.bufsize - tune.maxrewrite), haproxy will
938 return HTTP 400 (Bad Request) error. Similarly if an HTTP response is larger
939 than this size, haproxy will return HTTP 502 (Bad Gateway).
Willy Tarreau27a674e2009-08-17 07:23:33 +0200940
Willy Tarreau43961d52010-10-04 20:39:20 +0200941tune.chksize <number>
942 Sets the check buffer size to this size (in bytes). Higher values may help
943 find string or regex patterns in very large pages, though doing so may imply
944 more memory and CPU usage. The default value is 16384 and can be changed at
945 build time. It is not recommended to change this value, but to use better
946 checks whenever possible.
947
William Lallemandf3747832012-11-09 12:33:10 +0100948tune.comp.maxlevel <number>
949 Sets the maximum compression level. The compression level affects CPU
950 usage during compression. This value affects CPU usage during compression.
951 Each session using compression initializes the compression algorithm with
952 this value. The default value is 1.
953
Willy Tarreau193b8c62012-11-22 00:17:38 +0100954tune.http.cookielen <number>
955 Sets the maximum length of captured cookies. This is the maximum value that
956 the "capture cookie xxx len yyy" will be allowed to take, and any upper value
957 will automatically be truncated to this one. It is important not to set too
958 high a value because all cookie captures still allocate this size whatever
959 their configured value (they share a same pool). This value is per request
960 per response, so the memory allocated is twice this value per connection.
961 When not specified, the limit is set to 63 characters. It is recommended not
962 to change this value.
963
Willy Tarreauac1932d2011-10-24 19:14:41 +0200964tune.http.maxhdr <number>
965 Sets the maximum number of headers in a request. When a request comes with a
966 number of headers greater than this value (including the first line), it is
967 rejected with a "400 Bad Request" status code. Similarly, too large responses
968 are blocked with "502 Bad Gateway". The default value is 101, which is enough
969 for all usages, considering that the widely deployed Apache server uses the
970 same limit. It can be useful to push this limit further to temporarily allow
971 a buggy application to work by the time it gets fixed. Keep in mind that each
972 new header consumes 32bits of memory for each session, so don't push this
973 limit too high.
974
Willy Tarreau7e312732014-02-12 16:35:14 +0100975tune.idletimer <timeout>
976 Sets the duration after which haproxy will consider that an empty buffer is
977 probably associated with an idle stream. This is used to optimally adjust
978 some packet sizes while forwarding large and small data alternatively. The
979 decision to use splice() or to send large buffers in SSL is modulated by this
980 parameter. The value is in milliseconds between 0 and 65535. A value of zero
981 means that haproxy will not try to detect idle streams. The default is 1000,
982 which seems to correctly detect end user pauses (eg: read a page before
983 clicking). There should be not reason for changing this value. Please check
984 tune.ssl.maxrecord below.
985
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100986tune.maxaccept <number>
Willy Tarreau16a21472012-11-19 12:39:59 +0100987 Sets the maximum number of consecutive connections a process may accept in a
988 row before switching to other work. In single process mode, higher numbers
989 give better performance at high connection rates. However in multi-process
990 modes, keeping a bit of fairness between processes generally is better to
991 increase performance. This value applies individually to each listener, so
992 that the number of processes a listener is bound to is taken into account.
993 This value defaults to 64. In multi-process mode, it is divided by twice
994 the number of processes the listener is bound to. Setting this value to -1
995 completely disables the limitation. It should normally not be needed to tweak
996 this value.
Willy Tarreaua0250ba2008-01-06 11:22:57 +0100997
998tune.maxpollevents <number>
999 Sets the maximum amount of events that can be processed at once in a call to
1000 the polling system. The default value is adapted to the operating system. It
1001 has been noticed that reducing it below 200 tends to slightly decrease
1002 latency at the expense of network bandwidth, and increasing it above 200
1003 tends to trade latency for slightly increased bandwidth.
1004
Willy Tarreau27a674e2009-08-17 07:23:33 +02001005tune.maxrewrite <number>
1006 Sets the reserved buffer space to this size in bytes. The reserved space is
1007 used for header rewriting or appending. The first reads on sockets will never
1008 fill more than bufsize-maxrewrite. Historically it has defaulted to half of
1009 bufsize, though that does not make much sense since there are rarely large
1010 numbers of headers to add. Setting it too high prevents processing of large
1011 requests or responses. Setting it too low prevents addition of new headers
1012 to already large requests or to POST requests. It is generally wise to set it
1013 to about 1024. It is automatically readjusted to half of bufsize if it is
1014 larger than that. This means you don't have to worry about it when changing
1015 bufsize.
1016
Willy Tarreaubd9a0a72011-10-23 21:14:29 +02001017tune.pipesize <number>
1018 Sets the kernel pipe buffer size to this size (in bytes). By default, pipes
1019 are the default size for the system. But sometimes when using TCP splicing,
1020 it can improve performance to increase pipe sizes, especially if it is
1021 suspected that pipes are not filled and that many calls to splice() are
1022 performed. This has an impact on the kernel's memory footprint, so this must
1023 not be changed if impacts are not understood.
1024
Willy Tarreaue803de22010-01-21 17:43:04 +01001025tune.rcvbuf.client <number>
1026tune.rcvbuf.server <number>
1027 Forces the kernel socket receive buffer size on the client or the server side
1028 to the specified value in bytes. This value applies to all TCP/HTTP frontends
1029 and backends. It should normally never be set, and the default size (0) lets
1030 the kernel autotune this value depending on the amount of available memory.
1031 However it can sometimes help to set it to very low values (eg: 4096) in
1032 order to save kernel memory by preventing it from buffering too large amounts
1033 of received data. Lower values will significantly increase CPU usage though.
1034
1035tune.sndbuf.client <number>
1036tune.sndbuf.server <number>
1037 Forces the kernel socket send buffer size on the client or the server side to
1038 the specified value in bytes. This value applies to all TCP/HTTP frontends
1039 and backends. It should normally never be set, and the default size (0) lets
1040 the kernel autotune this value depending on the amount of available memory.
1041 However it can sometimes help to set it to very low values (eg: 4096) in
1042 order to save kernel memory by preventing it from buffering too large amounts
1043 of received data. Lower values will significantly increase CPU usage though.
1044 Another use case is to prevent write timeouts with extremely slow clients due
1045 to the kernel waiting for a large part of the buffer to be read before
1046 notifying haproxy again.
1047
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001048tune.ssl.cachesize <number>
Emeric Brunaf9619d2012-11-28 18:47:52 +01001049 Sets the size of the global SSL session cache, in a number of blocks. A block
1050 is large enough to contain an encoded session without peer certificate.
1051 An encoded session with peer certificate is stored in multiple blocks
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001052 depending on the size of the peer certificate. A block uses approximately
Emeric Brunaf9619d2012-11-28 18:47:52 +01001053 200 bytes of memory. The default value may be forced at build time, otherwise
1054 defaults to 20000. When the cache is full, the most idle entries are purged
1055 and reassigned. Higher values reduce the occurrence of such a purge, hence
1056 the number of CPU-intensive SSL handshakes by ensuring that all users keep
1057 their session as long as possible. All entries are pre-allocated upon startup
Emeric Brun22890a12012-12-28 14:41:32 +01001058 and are shared between all processes if "nbproc" is greater than 1. Setting
1059 this value to 0 disables the SSL session cache.
Willy Tarreau6ec58db2012-11-16 16:32:15 +01001060
Emeric Brun8dc60392014-05-09 13:52:00 +02001061tune.ssl.force-private-cache
1062 This boolean disables SSL session cache sharing between all processes. It
1063 should normally not be used since it will force many renegotiations due to
1064 clients hitting a random process. But it may be required on some operating
1065 systems where none of the SSL cache synchronization method may be used. In
1066 this case, adding a first layer of hash-based load balancing before the SSL
1067 layer might limit the impact of the lack of session sharing.
1068
Emeric Brun4f65bff2012-11-16 15:11:00 +01001069tune.ssl.lifetime <timeout>
1070 Sets how long a cached SSL session may remain valid. This time is expressed
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001071 in seconds and defaults to 300 (5 min). It is important to understand that it
Emeric Brun4f65bff2012-11-16 15:11:00 +01001072 does not guarantee that sessions will last that long, because if the cache is
1073 full, the longest idle sessions will be purged despite their configured
1074 lifetime. The real usefulness of this setting is to prevent sessions from
1075 being used for too long.
1076
Willy Tarreaubfd59462013-02-21 07:46:09 +01001077tune.ssl.maxrecord <number>
1078 Sets the maximum amount of bytes passed to SSL_write() at a time. Default
1079 value 0 means there is no limit. Over SSL/TLS, the client can decipher the
1080 data only once it has received a full record. With large records, it means
1081 that clients might have to download up to 16kB of data before starting to
1082 process them. Limiting the value can improve page load times on browsers
1083 located over high latency or low bandwidth networks. It is suggested to find
1084 optimal values which fit into 1 or 2 TCP segments (generally 1448 bytes over
1085 Ethernet with TCP timestamps enabled, or 1460 when timestamps are disabled),
1086 keeping in mind that SSL/TLS add some overhead. Typical values of 1419 and
1087 2859 gave good results during tests. Use "strace -e trace=write" to find the
Willy Tarreau7e312732014-02-12 16:35:14 +01001088 best value. Haproxy will automatically switch to this setting after an idle
1089 stream has been detected (see tune.idletimer above).
Willy Tarreaubfd59462013-02-21 07:46:09 +01001090
Remi Gacognef46cd6e2014-06-12 14:58:40 +02001091tune.ssl.default-dh-param <number>
1092 Sets the maximum size of the Diffie-Hellman parameters used for generating
1093 the ephemeral/temporary Diffie-Hellman key in case of DHE key exchange. The
1094 final size will try to match the size of the server's RSA (or DSA) key (e.g,
1095 a 2048 bits temporary DH key for a 2048 bits RSA key), but will not exceed
1096 this maximum value. Default value if 1024. Only 1024 or higher values are
1097 allowed. Higher values will increase the CPU load, and values greater than
1098 1024 bits are not supported by Java 7 and earlier clients. This value is not
1099 used if static Diffie-Hellman parameters are supplied via the certificate file.
1100
William Lallemanda509e4c2012-11-07 16:54:34 +01001101tune.zlib.memlevel <number>
1102 Sets the memLevel parameter in zlib initialization for each session. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001103 defines how much memory should be allocated for the internal compression
William Lallemanda509e4c2012-11-07 16:54:34 +01001104 state. A value of 1 uses minimum memory but is slow and reduces compression
1105 ratio, a value of 9 uses maximum memory for optimal speed. Can be a value
1106 between 1 and 9. The default value is 8.
1107
1108tune.zlib.windowsize <number>
1109 Sets the window size (the size of the history buffer) as a parameter of the
1110 zlib initialization for each session. Larger values of this parameter result
1111 in better compression at the expense of memory usage. Can be a value between
1112 8 and 15. The default value is 15.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001113
Willy Tarreauc57f0e22009-05-10 13:12:33 +020011143.3. Debugging
1115--------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02001116
1117debug
1118 Enables debug mode which dumps to stdout all exchanges, and disables forking
1119 into background. It is the equivalent of the command-line argument "-d". It
1120 should never be used in a production configuration since it may prevent full
1121 system startup.
1122
1123quiet
1124 Do not display any message during startup. It is equivalent to the command-
1125 line argument "-q".
1126
Emeric Brunf099e792010-09-27 12:05:28 +02001127
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010011283.4. Userlists
1129--------------
1130It is possible to control access to frontend/backend/listen sections or to
1131http stats by allowing only authenticated and authorized users. To do this,
1132it is required to create at least one userlist and to define users.
1133
1134userlist <listname>
Cyril Bonté78caf842010-03-10 22:41:43 +01001135 Creates new userlist with name <listname>. Many independent userlists can be
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001136 used to store authentication & authorization data for independent customers.
1137
1138group <groupname> [users <user>,<user>,(...)]
Cyril Bonté78caf842010-03-10 22:41:43 +01001139 Adds group <groupname> to the current userlist. It is also possible to
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001140 attach users to this group by using a comma separated list of names
1141 proceeded by "users" keyword.
1142
Cyril Bontéf0c60612010-02-06 14:44:47 +01001143user <username> [password|insecure-password <password>]
1144 [groups <group>,<group>,(...)]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001145 Adds user <username> to the current userlist. Both secure (encrypted) and
1146 insecure (unencrypted) passwords can be used. Encrypted passwords are
Cyril Bonté78caf842010-03-10 22:41:43 +01001147 evaluated using the crypt(3) function so depending of the system's
1148 capabilities, different algorithms are supported. For example modern Glibc
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001149 based Linux system supports MD5, SHA-256, SHA-512 and of course classic,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001150 DES-based method of encrypting passwords.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001151
1152
1153 Example:
Cyril Bontéf0c60612010-02-06 14:44:47 +01001154 userlist L1
1155 group G1 users tiger,scott
1156 group G2 users xdb,scott
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001157
Cyril Bontéf0c60612010-02-06 14:44:47 +01001158 user tiger password $6$k6y3o.eP$JlKBx9za9667qe4(...)xHSwRv6J.C0/D7cV91
1159 user scott insecure-password elgato
1160 user xdb insecure-password hello
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001161
Cyril Bontéf0c60612010-02-06 14:44:47 +01001162 userlist L2
1163 group G1
1164 group G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001165
Cyril Bontéf0c60612010-02-06 14:44:47 +01001166 user tiger password $6$k6y3o.eP$JlKBx(...)xHSwRv6J.C0/D7cV91 groups G1
1167 user scott insecure-password elgato groups G1,G2
1168 user xdb insecure-password hello groups G2
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01001169
1170 Please note that both lists are functionally identical.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001171
Emeric Brunf099e792010-09-27 12:05:28 +02001172
11733.5. Peers
Cyril Bontédc4d9032012-04-08 21:57:39 +02001174----------
Emeric Brunf099e792010-09-27 12:05:28 +02001175It is possible to synchronize server entries in stick tables between several
1176haproxy instances over TCP connections in a multi-master fashion. Each instance
1177pushes its local updates and insertions to remote peers. Server IDs are used to
1178identify servers remotely, so it is important that configurations look similar
1179or at least that the same IDs are forced on each server on all participants.
1180Interrupted exchanges are automatically detected and recovered from the last
1181known point. In addition, during a soft restart, the old process connects to
1182the new one using such a TCP connection to push all its entries before the new
1183process tries to connect to other peers. That ensures very fast replication
1184during a reload, it typically takes a fraction of a second even for large
1185tables.
1186
1187peers <peersect>
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001188 Creates a new peer list with name <peersect>. It is an independent section,
Emeric Brunf099e792010-09-27 12:05:28 +02001189 which is referenced by one or more stick-tables.
1190
1191peer <peername> <ip>:<port>
1192 Defines a peer inside a peers section.
1193 If <peername> is set to the local peer name (by default hostname, or forced
1194 using "-L" command line option), haproxy will listen for incoming remote peer
1195 connection on <ip>:<port>. Otherwise, <ip>:<port> defines where to connect to
1196 to join the remote peer, and <peername> is used at the protocol level to
1197 identify and validate the remote peer on the server side.
1198
1199 During a soft restart, local peer <ip>:<port> is used by the old instance to
1200 connect the new one and initiate a complete replication (teaching process).
1201
1202 It is strongly recommended to have the exact same peers declaration on all
1203 peers and to only rely on the "-L" command line argument to change the local
1204 peer name. This makes it easier to maintain coherent configuration files
1205 across all peers.
1206
Willy Tarreaudad36a32013-03-11 01:20:04 +01001207 Any part of the address string may reference any number of environment
1208 variables by preceding their name with a dollar sign ('$') and optionally
1209 enclosing them with braces ('{}'), similarly to what is done in Bourne shell.
1210
Cyril Bontédc4d9032012-04-08 21:57:39 +02001211 Example:
Emeric Brunf099e792010-09-27 12:05:28 +02001212 peers mypeers
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001213 peer haproxy1 192.168.0.1:1024
1214 peer haproxy2 192.168.0.2:1024
1215 peer haproxy3 10.2.0.1:1024
Emeric Brunf099e792010-09-27 12:05:28 +02001216
1217 backend mybackend
1218 mode tcp
1219 balance roundrobin
1220 stick-table type ip size 20k peers mypeers
1221 stick on src
1222
Willy Tarreauf7b30a92010-12-06 22:59:17 +01001223 server srv1 192.168.0.30:80
1224 server srv2 192.168.0.31:80
Emeric Brunf099e792010-09-27 12:05:28 +02001225
1226
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012274. Proxies
Willy Tarreau6a06a402007-07-15 20:15:28 +02001228----------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001229
Willy Tarreau6a06a402007-07-15 20:15:28 +02001230Proxy configuration can be located in a set of sections :
1231 - defaults <name>
1232 - frontend <name>
1233 - backend <name>
1234 - listen <name>
1235
1236A "defaults" section sets default parameters for all other sections following
1237its declaration. Those default parameters are reset by the next "defaults"
1238section. See below for the list of parameters which can be set in a "defaults"
Willy Tarreau0ba27502007-12-24 16:55:16 +01001239section. The name is optional but its use is encouraged for better readability.
Willy Tarreau6a06a402007-07-15 20:15:28 +02001240
1241A "frontend" section describes a set of listening sockets accepting client
1242connections.
1243
1244A "backend" section describes a set of servers to which the proxy will connect
1245to forward incoming connections.
1246
1247A "listen" section defines a complete proxy with its frontend and backend
1248parts combined in one section. It is generally useful for TCP-only traffic.
1249
Willy Tarreau0ba27502007-12-24 16:55:16 +01001250All proxy names must be formed from upper and lower case letters, digits,
1251'-' (dash), '_' (underscore) , '.' (dot) and ':' (colon). ACL names are
1252case-sensitive, which means that "www" and "WWW" are two different proxies.
1253
1254Historically, all proxy names could overlap, it just caused troubles in the
1255logs. Since the introduction of content switching, it is mandatory that two
1256proxies with overlapping capabilities (frontend/backend) have different names.
1257However, it is still permitted that a frontend and a backend share the same
1258name, as this configuration seems to be commonly encountered.
1259
1260Right now, two major proxy modes are supported : "tcp", also known as layer 4,
1261and "http", also known as layer 7. In layer 4 mode, HAProxy simply forwards
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001262bidirectional traffic between two sides. In layer 7 mode, HAProxy analyzes the
Willy Tarreau0ba27502007-12-24 16:55:16 +01001263protocol, and can interact with it by allowing, blocking, switching, adding,
1264modifying, or removing arbitrary contents in requests or responses, based on
1265arbitrary criteria.
1266
Willy Tarreau70dffda2014-01-30 03:07:23 +01001267In HTTP mode, the processing applied to requests and responses flowing over
1268a connection depends in the combination of the frontend's HTTP options and
1269the backend's. HAProxy supports 5 connection modes :
1270
1271 - KAL : keep alive ("option http-keep-alive") which is the default mode : all
1272 requests and responses are processed, and connections remain open but idle
1273 between responses and new requests.
1274
1275 - TUN: tunnel ("option http-tunnel") : this was the default mode for versions
1276 1.0 to 1.5-dev21 : only the first request and response are processed, and
1277 everything else is forwarded with no analysis at all. This mode should not
1278 be used as it creates lots of trouble with logging and HTTP processing.
1279
1280 - PCL: passive close ("option httpclose") : exactly the same as tunnel mode,
1281 but with "Connection: close" appended in both directions to try to make
1282 both ends close after the first request/response exchange.
1283
1284 - SCL: server close ("option http-server-close") : the server-facing
1285 connection is closed after the end of the response is received, but the
1286 client-facing connection remains open.
1287
1288 - FCL: forced close ("option forceclose") : the connection is actively closed
1289 after the end of the response.
1290
1291The effective mode that will be applied to a connection passing through a
1292frontend and a backend can be determined by both proxy modes according to the
1293following matrix, but in short, the modes are symmetric, keep-alive is the
1294weakest option and force close is the strongest.
1295
1296 Backend mode
1297
1298 | KAL | TUN | PCL | SCL | FCL
1299 ----+-----+-----+-----+-----+----
1300 KAL | KAL | TUN | PCL | SCL | FCL
1301 ----+-----+-----+-----+-----+----
1302 TUN | TUN | TUN | PCL | SCL | FCL
1303 Frontend ----+-----+-----+-----+-----+----
1304 mode PCL | PCL | PCL | PCL | FCL | FCL
1305 ----+-----+-----+-----+-----+----
1306 SCL | SCL | SCL | FCL | SCL | FCL
1307 ----+-----+-----+-----+-----+----
1308 FCL | FCL | FCL | FCL | FCL | FCL
1309
Willy Tarreau0ba27502007-12-24 16:55:16 +01001310
Willy Tarreau70dffda2014-01-30 03:07:23 +01001311
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013124.1. Proxy keywords matrix
1313--------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001314
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001315The following list of keywords is supported. Most of them may only be used in a
1316limited set of section types. Some of them are marked as "deprecated" because
1317they are inherited from an old syntax which may be confusing or functionally
1318limited, and there are new recommended keywords to replace them. Keywords
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001319marked with "(*)" can be optionally inverted using the "no" prefix, eg. "no
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001320option contstats". This makes sense when the option has been enabled by default
Willy Tarreau3842f002009-06-14 11:39:52 +02001321and must be disabled for a specific instance. Such options may also be prefixed
1322with "default" in order to restore default settings regardless of what has been
1323specified in a previous "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001324
Willy Tarreau6a06a402007-07-15 20:15:28 +02001325
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001326 keyword defaults frontend listen backend
1327------------------------------------+----------+----------+---------+---------
1328acl - X X X
1329appsession - - X X
1330backlog X X X -
1331balance X - X X
1332bind - X X -
1333bind-process X X X X
1334block - X X X
1335capture cookie - X X -
1336capture request header - X X -
1337capture response header - X X -
1338clitimeout (deprecated) X X X -
William Lallemand82fe75c2012-10-23 10:25:10 +02001339compression X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001340contimeout (deprecated) X - X X
1341cookie X - X X
1342default-server X - X X
1343default_backend X X X -
1344description - X X X
1345disabled X X X X
1346dispatch - - X X
1347enabled X X X X
1348errorfile X X X X
1349errorloc X X X X
1350errorloc302 X X X X
1351-- keyword -------------------------- defaults - frontend - listen -- backend -
1352errorloc303 X X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001353force-persist - X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001354fullconn X - X X
1355grace X X X X
1356hash-type X - X X
1357http-check disable-on-404 X - X X
Willy Tarreaubd741542010-03-16 18:46:54 +01001358http-check expect - - X X
Willy Tarreau7ab6aff2010-10-12 06:30:16 +02001359http-check send-state X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001360http-request - X X X
Willy Tarreaue365c0b2013-06-11 16:06:12 +02001361http-response - X X X
Baptiste Assmann2c42ef52013-10-09 21:57:02 +02001362http-send-name-header - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001363id - X X X
Cyril Bonté0d4bf012010-04-25 23:21:46 +02001364ignore-persist - X X X
William Lallemand0f99e342011-10-12 17:50:54 +02001365log (*) X X X X
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01001366log-format X X X -
Willy Tarreau094af4e2015-01-07 15:03:42 +01001367log-tag X X X X
Willy Tarreauc35362a2014-04-25 13:58:37 +02001368max-keep-alive-queue X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001369maxconn X X X -
1370mode X X X X
1371monitor fail - X X -
1372monitor-net X X X -
1373monitor-uri X X X -
1374option abortonclose (*) X - X X
1375option accept-invalid-http-request (*) X X X -
1376option accept-invalid-http-response (*) X - X X
1377option allbackups (*) X - X X
1378option checkcache (*) X - X X
1379option clitcpka (*) X X X -
1380option contstats (*) X X X -
1381option dontlog-normal (*) X X X -
1382option dontlognull (*) X X X -
1383option forceclose (*) X X X X
1384-- keyword -------------------------- defaults - frontend - listen -- backend -
1385option forwardfor X X X X
Willy Tarreau16bfb022010-01-16 19:48:41 +01001386option http-keep-alive (*) X X X X
Willy Tarreau96e31212011-05-30 18:10:30 +02001387option http-no-delay (*) X X X X
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02001388option http-pretend-keepalive (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001389option http-server-close (*) X X X X
Willy Tarreau02bce8b2014-01-30 00:15:28 +01001390option http-tunnel (*) X X X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001391option http-use-proxy-header (*) X X X -
1392option httpchk X - X X
1393option httpclose (*) X X X X
1394option httplog X X X X
1395option http_proxy (*) X X X X
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001396option independent-streams (*) X X X X
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02001397option ldap-check X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001398option external-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001399option log-health-checks (*) X - X X
1400option log-separate-errors (*) X X X -
1401option logasap (*) X X X -
1402option mysql-check X - X X
Rauf Kuliyev38b41562011-01-04 15:14:13 +01001403option pgsql-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001404option nolinger (*) X X X X
1405option originalto X X X X
1406option persist (*) X - X X
1407option redispatch (*) X - X X
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02001408option redis-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001409option smtpchk X - X X
1410option socket-stats (*) X X X -
1411option splice-auto (*) X X X X
1412option splice-request (*) X X X X
1413option splice-response (*) X X X X
1414option srvtcpka (*) X - X X
1415option ssl-hello-chk X - X X
1416-- keyword -------------------------- defaults - frontend - listen -- backend -
Willy Tarreaued179852013-12-16 01:07:00 +01001417option tcp-check X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001418option tcp-smart-accept (*) X X X -
1419option tcp-smart-connect (*) X - X X
1420option tcpka X X X X
1421option tcplog X X X X
1422option transparent (*) X - X X
Simon Horman98637e52014-06-20 12:30:16 +09001423external-check command X - X X
1424external-check path X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001425persist rdp-cookie X - X X
1426rate-limit sessions X X X -
1427redirect - X X X
1428redisp (deprecated) X - X X
1429redispatch (deprecated) X - X X
1430reqadd - X X X
1431reqallow - X X X
1432reqdel - X X X
1433reqdeny - X X X
1434reqiallow - X X X
1435reqidel - X X X
1436reqideny - X X X
1437reqipass - X X X
1438reqirep - X X X
1439reqisetbe - X X X
1440reqitarpit - X X X
1441reqpass - X X X
1442reqrep - X X X
1443-- keyword -------------------------- defaults - frontend - listen -- backend -
1444reqsetbe - X X X
1445reqtarpit - X X X
1446retries X - X X
1447rspadd - X X X
1448rspdel - X X X
1449rspdeny - X X X
1450rspidel - X X X
1451rspideny - X X X
1452rspirep - X X X
1453rsprep - X X X
1454server - - X X
1455source X - X X
1456srvtimeout (deprecated) X - X X
Cyril Bonté66c327d2010-10-12 00:14:37 +02001457stats admin - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001458stats auth X - X X
1459stats enable X - X X
1460stats hide-version X - X X
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02001461stats http-request - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001462stats realm X - X X
1463stats refresh X - X X
1464stats scope X - X X
1465stats show-desc X - X X
1466stats show-legends X - X X
1467stats show-node X - X X
1468stats uri X - X X
1469-- keyword -------------------------- defaults - frontend - listen -- backend -
1470stick match - - X X
1471stick on - - X X
1472stick store-request - - X X
Willy Tarreaud8dc99f2011-07-01 11:33:25 +02001473stick store-response - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001474stick-table - - X X
Willy Tarreau938c7fe2014-04-25 14:21:39 +02001475tcp-check connect - - X X
1476tcp-check expect - - X X
1477tcp-check send - - X X
1478tcp-check send-binary - - X X
Willy Tarreaue9656522010-08-17 15:40:09 +02001479tcp-request connection - X X -
1480tcp-request content - X X X
Willy Tarreaua56235c2010-09-14 11:31:36 +02001481tcp-request inspect-delay - X X X
Emeric Brun0a3b67f2010-09-24 15:34:53 +02001482tcp-response content - - X X
1483tcp-response inspect-delay - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001484timeout check X - X X
1485timeout client X X X -
Willy Tarreau05cdd962014-05-10 14:30:07 +02001486timeout client-fin X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001487timeout clitimeout (deprecated) X X X -
1488timeout connect X - X X
1489timeout contimeout (deprecated) X - X X
1490timeout http-keep-alive X X X X
1491timeout http-request X X X X
1492timeout queue X - X X
1493timeout server X - X X
Willy Tarreau05cdd962014-05-10 14:30:07 +02001494timeout server-fin X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001495timeout srvtimeout (deprecated) X - X X
1496timeout tarpit X X X X
Willy Tarreauce887fd2012-05-12 12:50:00 +02001497timeout tunnel X - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001498transparent (deprecated) X - X X
William Lallemanda73203e2012-03-12 12:48:57 +01001499unique-id-format X X X -
1500unique-id-header X X X -
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001501use_backend - X X -
Willy Tarreau4a5cade2012-04-05 21:09:48 +02001502use-server - - X X
Willy Tarreau5c6f7b32010-02-26 13:34:29 +01001503------------------------------------+----------+----------+---------+---------
1504 keyword defaults frontend listen backend
Willy Tarreau6a06a402007-07-15 20:15:28 +02001505
Willy Tarreau0ba27502007-12-24 16:55:16 +01001506
Willy Tarreauc57f0e22009-05-10 13:12:33 +020015074.2. Alphabetically sorted keywords reference
1508---------------------------------------------
Willy Tarreau0ba27502007-12-24 16:55:16 +01001509
1510This section provides a description of each keyword and its usage.
1511
1512
1513acl <aclname> <criterion> [flags] [operator] <value> ...
1514 Declare or complete an access list.
1515 May be used in sections : defaults | frontend | listen | backend
1516 no | yes | yes | yes
1517 Example:
1518 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
1519 acl invalid_src src_port 0:1023
1520 acl local_dst hdr(host) -i localhost
1521
Willy Tarreauc57f0e22009-05-10 13:12:33 +02001522 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001523
1524
Cyril Bontéb21570a2009-11-29 20:04:48 +01001525appsession <cookie> len <length> timeout <holdtime>
1526 [request-learn] [prefix] [mode <path-parameters|query-string>]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001527 Define session stickiness on an existing application cookie.
1528 May be used in sections : defaults | frontend | listen | backend
1529 no | no | yes | yes
1530 Arguments :
1531 <cookie> this is the name of the cookie used by the application and which
1532 HAProxy will have to learn for each new session.
1533
Cyril Bontéb21570a2009-11-29 20:04:48 +01001534 <length> this is the max number of characters that will be memorized and
Willy Tarreau0ba27502007-12-24 16:55:16 +01001535 checked in each cookie value.
1536
1537 <holdtime> this is the time after which the cookie will be removed from
1538 memory if unused. If no unit is specified, this time is in
1539 milliseconds.
1540
Cyril Bontébf47aeb2009-10-15 00:15:40 +02001541 request-learn
1542 If this option is specified, then haproxy will be able to learn
1543 the cookie found in the request in case the server does not
1544 specify any in response. This is typically what happens with
1545 PHPSESSID cookies, or when haproxy's session expires before
1546 the application's session and the correct server is selected.
1547 It is recommended to specify this option to improve reliability.
1548
Cyril Bontéb21570a2009-11-29 20:04:48 +01001549 prefix When this option is specified, haproxy will match on the cookie
1550 prefix (or URL parameter prefix). The appsession value is the
1551 data following this prefix.
1552
1553 Example :
1554 appsession ASPSESSIONID len 64 timeout 3h prefix
1555
1556 This will match the cookie ASPSESSIONIDXXXX=XXXXX,
1557 the appsession value will be XXXX=XXXXX.
1558
1559 mode This option allows to change the URL parser mode.
1560 2 modes are currently supported :
1561 - path-parameters :
1562 The parser looks for the appsession in the path parameters
1563 part (each parameter is separated by a semi-colon), which is
1564 convenient for JSESSIONID for example.
1565 This is the default mode if the option is not set.
1566 - query-string :
1567 In this mode, the parser will look for the appsession in the
1568 query string.
1569
Willy Tarreau0ba27502007-12-24 16:55:16 +01001570 When an application cookie is defined in a backend, HAProxy will check when
1571 the server sets such a cookie, and will store its value in a table, and
1572 associate it with the server's identifier. Up to <length> characters from
1573 the value will be retained. On each connection, haproxy will look for this
Cyril Bontéb21570a2009-11-29 20:04:48 +01001574 cookie both in the "Cookie:" headers, and as a URL parameter (depending on
1575 the mode used). If a known value is found, the client will be directed to the
1576 server associated with this value. Otherwise, the load balancing algorithm is
Willy Tarreau0ba27502007-12-24 16:55:16 +01001577 applied. Cookies are automatically removed from memory when they have been
1578 unused for a duration longer than <holdtime>.
1579
1580 The definition of an application cookie is limited to one per backend.
1581
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001582 Note : Consider not using this feature in multi-process mode (nbproc > 1)
1583 unless you know what you do : memory is not shared between the
1584 processes, which can result in random behaviours.
1585
Willy Tarreau0ba27502007-12-24 16:55:16 +01001586 Example :
1587 appsession JSESSIONID len 52 timeout 3h
1588
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01001589 See also : "cookie", "capture cookie", "balance", "stick", "stick-table",
1590 "ignore-persist", "nbproc" and "bind-process".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001591
1592
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001593backlog <conns>
1594 Give hints to the system about the approximate listen backlog desired size
1595 May be used in sections : defaults | frontend | listen | backend
1596 yes | yes | yes | no
1597 Arguments :
1598 <conns> is the number of pending connections. Depending on the operating
1599 system, it may represent the number of already acknowledged
Cyril Bontédc4d9032012-04-08 21:57:39 +02001600 connections, of non-acknowledged ones, or both.
Willy Tarreauc73ce2b2008-01-06 10:55:10 +01001601
1602 In order to protect against SYN flood attacks, one solution is to increase
1603 the system's SYN backlog size. Depending on the system, sometimes it is just
1604 tunable via a system parameter, sometimes it is not adjustable at all, and
1605 sometimes the system relies on hints given by the application at the time of
1606 the listen() syscall. By default, HAProxy passes the frontend's maxconn value
1607 to the listen() syscall. On systems which can make use of this value, it can
1608 sometimes be useful to be able to specify a different value, hence this
1609 backlog parameter.
1610
1611 On Linux 2.4, the parameter is ignored by the system. On Linux 2.6, it is
1612 used as a hint and the system accepts up to the smallest greater power of
1613 two, and never more than some limits (usually 32768).
1614
1615 See also : "maxconn" and the target operating system's tuning guide.
1616
1617
Willy Tarreau0ba27502007-12-24 16:55:16 +01001618balance <algorithm> [ <arguments> ]
Willy Tarreau226071e2014-04-10 11:55:45 +02001619balance url_param <param> [check_post]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001620 Define the load balancing algorithm to be used in a backend.
1621 May be used in sections : defaults | frontend | listen | backend
1622 yes | no | yes | yes
1623 Arguments :
1624 <algorithm> is the algorithm used to select a server when doing load
1625 balancing. This only applies when no persistence information
1626 is available, or when a connection is redispatched to another
1627 server. <algorithm> may be one of the following :
1628
1629 roundrobin Each server is used in turns, according to their weights.
1630 This is the smoothest and fairest algorithm when the server's
1631 processing time remains equally distributed. This algorithm
1632 is dynamic, which means that server weights may be adjusted
Willy Tarreau9757a382009-10-03 12:56:50 +02001633 on the fly for slow starts for instance. It is limited by
Godbacha34bdc02013-07-22 07:44:53 +08001634 design to 4095 active servers per backend. Note that in some
Willy Tarreau9757a382009-10-03 12:56:50 +02001635 large farms, when a server becomes up after having been down
1636 for a very short time, it may sometimes take a few hundreds
1637 requests for it to be re-integrated into the farm and start
1638 receiving traffic. This is normal, though very rare. It is
1639 indicated here in case you would have the chance to observe
1640 it, so that you don't worry.
1641
1642 static-rr Each server is used in turns, according to their weights.
1643 This algorithm is as similar to roundrobin except that it is
1644 static, which means that changing a server's weight on the
1645 fly will have no effect. On the other hand, it has no design
1646 limitation on the number of servers, and when a server goes
1647 up, it is always immediately reintroduced into the farm, once
1648 the full map is recomputed. It also uses slightly less CPU to
1649 run (around -1%).
Willy Tarreau0ba27502007-12-24 16:55:16 +01001650
Willy Tarreau2d2a7f82008-03-17 12:07:56 +01001651 leastconn The server with the lowest number of connections receives the
1652 connection. Round-robin is performed within groups of servers
1653 of the same load to ensure that all servers will be used. Use
1654 of this algorithm is recommended where very long sessions are
1655 expected, such as LDAP, SQL, TSE, etc... but is not very well
1656 suited for protocols using short sessions such as HTTP. This
1657 algorithm is dynamic, which means that server weights may be
1658 adjusted on the fly for slow starts for instance.
1659
Willy Tarreauf09c6602012-02-13 17:12:08 +01001660 first The first server with available connection slots receives the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03001661 connection. The servers are chosen from the lowest numeric
Willy Tarreauf09c6602012-02-13 17:12:08 +01001662 identifier to the highest (see server parameter "id"), which
1663 defaults to the server's position in the farm. Once a server
Willy Tarreau64559c52012-04-07 09:08:45 +02001664 reaches its maxconn value, the next server is used. It does
Willy Tarreauf09c6602012-02-13 17:12:08 +01001665 not make sense to use this algorithm without setting maxconn.
1666 The purpose of this algorithm is to always use the smallest
1667 number of servers so that extra servers can be powered off
1668 during non-intensive hours. This algorithm ignores the server
1669 weight, and brings more benefit to long session such as RDP
Willy Tarreau64559c52012-04-07 09:08:45 +02001670 or IMAP than HTTP, though it can be useful there too. In
1671 order to use this algorithm efficiently, it is recommended
1672 that a cloud controller regularly checks server usage to turn
1673 them off when unused, and regularly checks backend queue to
1674 turn new servers on when the queue inflates. Alternatively,
1675 using "http-check send-state" may inform servers on the load.
Willy Tarreauf09c6602012-02-13 17:12:08 +01001676
Willy Tarreau0ba27502007-12-24 16:55:16 +01001677 source The source IP address is hashed and divided by the total
1678 weight of the running servers to designate which server will
1679 receive the request. This ensures that the same client IP
1680 address will always reach the same server as long as no
1681 server goes down or up. If the hash result changes due to the
1682 number of running servers changing, many clients will be
1683 directed to a different server. This algorithm is generally
1684 used in TCP mode where no cookie may be inserted. It may also
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001685 be used on the Internet to provide a best-effort stickiness
Willy Tarreau0ba27502007-12-24 16:55:16 +01001686 to clients which refuse session cookies. This algorithm is
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001687 static by default, which means that changing a server's
1688 weight on the fly will have no effect, but this can be
1689 changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001690
Oskar Stolc8dc41842012-05-19 10:19:54 +01001691 uri This algorithm hashes either the left part of the URI (before
1692 the question mark) or the whole URI (if the "whole" parameter
1693 is present) and divides the hash value by the total weight of
1694 the running servers. The result designates which server will
1695 receive the request. This ensures that the same URI will
1696 always be directed to the same server as long as no server
1697 goes up or down. This is used with proxy caches and
1698 anti-virus proxies in order to maximize the cache hit rate.
1699 Note that this algorithm may only be used in an HTTP backend.
1700 This algorithm is static by default, which means that
1701 changing a server's weight on the fly will have no effect,
1702 but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001703
Oskar Stolc8dc41842012-05-19 10:19:54 +01001704 This algorithm supports two optional parameters "len" and
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001705 "depth", both followed by a positive integer number. These
1706 options may be helpful when it is needed to balance servers
1707 based on the beginning of the URI only. The "len" parameter
1708 indicates that the algorithm should only consider that many
1709 characters at the beginning of the URI to compute the hash.
1710 Note that having "len" set to 1 rarely makes sense since most
1711 URIs start with a leading "/".
1712
1713 The "depth" parameter indicates the maximum directory depth
1714 to be used to compute the hash. One level is counted for each
1715 slash in the request. If both parameters are specified, the
1716 evaluation stops when either is reached.
1717
Willy Tarreau0ba27502007-12-24 16:55:16 +01001718 url_param The URL parameter specified in argument will be looked up in
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001719 the query string of each HTTP GET request.
1720
1721 If the modifier "check_post" is used, then an HTTP POST
Cyril Bontédc4d9032012-04-08 21:57:39 +02001722 request entity will be searched for the parameter argument,
1723 when it is not found in a query string after a question mark
Willy Tarreau226071e2014-04-10 11:55:45 +02001724 ('?') in the URL. The message body will only start to be
1725 analyzed once either the advertised amount of data has been
1726 received or the request buffer is full. In the unlikely event
1727 that chunked encoding is used, only the first chunk is
Cyril Bontédc4d9032012-04-08 21:57:39 +02001728 scanned. Parameter values separated by a chunk boundary, may
Willy Tarreau226071e2014-04-10 11:55:45 +02001729 be randomly balanced if at all. This keyword used to support
1730 an optional <max_wait> parameter which is now ignored.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001731
1732 If the parameter is found followed by an equal sign ('=') and
1733 a value, then the value is hashed and divided by the total
1734 weight of the running servers. The result designates which
1735 server will receive the request.
1736
1737 This is used to track user identifiers in requests and ensure
1738 that a same user ID will always be sent to the same server as
1739 long as no server goes up or down. If no value is found or if
1740 the parameter is not found, then a round robin algorithm is
1741 applied. Note that this algorithm may only be used in an HTTP
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001742 backend. This algorithm is static by default, which means
1743 that changing a server's weight on the fly will have no
1744 effect, but this can be changed using "hash-type".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001745
Cyril Bontédc4d9032012-04-08 21:57:39 +02001746 hdr(<name>) The HTTP header <name> will be looked up in each HTTP
1747 request. Just as with the equivalent ACL 'hdr()' function,
1748 the header name in parenthesis is not case sensitive. If the
1749 header is absent or if it does not contain any value, the
1750 roundrobin algorithm is applied instead.
Benoitaffb4812009-03-25 13:02:10 +01001751
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001752 An optional 'use_domain_only' parameter is available, for
Benoitaffb4812009-03-25 13:02:10 +01001753 reducing the hash algorithm to the main domain part with some
1754 specific headers such as 'Host'. For instance, in the Host
1755 value "haproxy.1wt.eu", only "1wt" will be considered.
1756
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001757 This algorithm is static by default, which means that
1758 changing a server's weight on the fly will have no effect,
1759 but this can be changed using "hash-type".
1760
Emeric Brun736aa232009-06-30 17:56:00 +02001761 rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02001762 rdp-cookie(<name>)
Emeric Brun736aa232009-06-30 17:56:00 +02001763 The RDP cookie <name> (or "mstshash" if omitted) will be
1764 looked up and hashed for each incoming TCP request. Just as
1765 with the equivalent ACL 'req_rdp_cookie()' function, the name
1766 is not case-sensitive. This mechanism is useful as a degraded
1767 persistence mode, as it makes it possible to always send the
1768 same user (or the same session ID) to the same server. If the
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001769 cookie is not found, the normal roundrobin algorithm is
Emeric Brun736aa232009-06-30 17:56:00 +02001770 used instead.
1771
1772 Note that for this to work, the frontend must ensure that an
1773 RDP cookie is already present in the request buffer. For this
1774 you must use 'tcp-request content accept' rule combined with
1775 a 'req_rdp_cookie_cnt' ACL.
1776
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001777 This algorithm is static by default, which means that
1778 changing a server's weight on the fly will have no effect,
1779 but this can be changed using "hash-type".
1780
Cyril Bontédc4d9032012-04-08 21:57:39 +02001781 See also the rdp_cookie pattern fetch function.
Simon Hormanab814e02011-06-24 14:50:20 +09001782
Willy Tarreau0ba27502007-12-24 16:55:16 +01001783 <arguments> is an optional list of arguments which may be needed by some
Marek Majkowski9c30fc12008-04-27 23:25:55 +02001784 algorithms. Right now, only "url_param" and "uri" support an
1785 optional argument.
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001786
Willy Tarreau3cd9af22009-03-15 14:06:41 +01001787 The load balancing algorithm of a backend is set to roundrobin when no other
1788 algorithm, mode nor option have been set. The algorithm may only be set once
1789 for each backend.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001790
1791 Examples :
1792 balance roundrobin
1793 balance url_param userid
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001794 balance url_param session_id check_post 64
Benoitaffb4812009-03-25 13:02:10 +01001795 balance hdr(User-Agent)
1796 balance hdr(host)
1797 balance hdr(Host) use_domain_only
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001798
1799 Note: the following caveats and limitations on using the "check_post"
1800 extension with "url_param" must be considered :
1801
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001802 - all POST requests are eligible for consideration, because there is no way
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001803 to determine if the parameters will be found in the body or entity which
1804 may contain binary data. Therefore another method may be required to
1805 restrict consideration of POST requests that have no URL parameters in
1806 the body. (see acl reqideny http_end)
1807
1808 - using a <max_wait> value larger than the request buffer size does not
1809 make sense and is useless. The buffer size is set at build time, and
1810 defaults to 16 kB.
1811
1812 - Content-Encoding is not supported, the parameter search will probably
1813 fail; and load balancing will fall back to Round Robin.
1814
1815 - Expect: 100-continue is not supported, load balancing will fall back to
1816 Round Robin.
1817
1818 - Transfer-Encoding (RFC2616 3.6.1) is only supported in the first chunk.
1819 If the entire parameter value is not present in the first chunk, the
1820 selection of server is undefined (actually, defined by how little
1821 actually appeared in the first chunk).
1822
1823 - This feature does not support generation of a 100, 411 or 501 response.
1824
1825 - In some cases, requesting "check_post" MAY attempt to scan the entire
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01001826 contents of a message body. Scanning normally terminates when linear
matt.farnsworth@nokia.com1c2ab962008-04-14 20:47:37 +02001827 white space or control characters are found, indicating the end of what
1828 might be a URL parameter list. This is probably not a concern with SGML
1829 type message bodies.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001830
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02001831 See also : "dispatch", "cookie", "appsession", "transparent", "hash-type" and
1832 "http_proxy".
Willy Tarreau0ba27502007-12-24 16:55:16 +01001833
1834
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001835bind [<address>]:<port_range> [, ...] [param*]
1836bind /<path> [, ...] [param*]
Willy Tarreau0ba27502007-12-24 16:55:16 +01001837 Define one or several listening addresses and/or ports in a frontend.
1838 May be used in sections : defaults | frontend | listen | backend
1839 no | yes | yes | no
1840 Arguments :
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001841 <address> is optional and can be a host name, an IPv4 address, an IPv6
1842 address, or '*'. It designates the address the frontend will
1843 listen on. If unset, all IPv4 addresses of the system will be
1844 listened on. The same will apply for '*' or the system's
David du Colombier9c938da2011-03-17 10:40:27 +01001845 special address "0.0.0.0". The IPv6 equivalent is '::'.
Willy Tarreau24709282013-03-10 21:32:12 +01001846 Optionally, an address family prefix may be used before the
1847 address to force the family regardless of the address format,
1848 which can be useful to specify a path to a unix socket with
1849 no slash ('/'). Currently supported prefixes are :
1850 - 'ipv4@' -> address is always IPv4
1851 - 'ipv6@' -> address is always IPv6
1852 - 'unix@' -> address is a path to a local unix socket
Willy Tarreau70f72e02014-07-08 00:37:50 +02001853 - 'abns@' -> address is in abstract namespace (Linux only).
1854 Note: since abstract sockets are not "rebindable", they
1855 do not cope well with multi-process mode during
1856 soft-restart, so it is better to avoid them if
1857 nbproc is greater than 1. The effect is that if the
1858 new process fails to start, only one of the old ones
1859 will be able to rebind to the socket.
Willy Tarreau40aa0702013-03-10 23:51:38 +01001860 - 'fd@<n>' -> use file descriptor <n> inherited from the
1861 parent. The fd must be bound and may or may not already
1862 be listening.
Willy Tarreaudad36a32013-03-11 01:20:04 +01001863 Any part of the address string may reference any number of
1864 environment variables by preceding their name with a dollar
1865 sign ('$') and optionally enclosing them with braces ('{}'),
1866 similarly to what is done in Bourne shell.
Willy Tarreaub1e52e82008-01-13 14:49:51 +01001867
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001868 <port_range> is either a unique TCP port, or a port range for which the
1869 proxy will accept connections for the IP address specified
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001870 above. The port is mandatory for TCP listeners. Note that in
1871 the case of an IPv6 address, the port is always the number
1872 after the last colon (':'). A range can either be :
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001873 - a numerical port (ex: '80')
1874 - a dash-delimited ports range explicitly stating the lower
1875 and upper bounds (ex: '2000-2100') which are included in
1876 the range.
1877
1878 Particular care must be taken against port ranges, because
1879 every <address:port> couple consumes one socket (= a file
1880 descriptor), so it's easy to consume lots of descriptors
1881 with a simple range, and to run out of sockets. Also, each
1882 <address:port> couple must be used only once among all
1883 instances running on a same system. Please note that binding
1884 to ports lower than 1024 generally require particular
Jamie Gloudon801a0a32012-08-25 00:18:33 -04001885 privileges to start the program, which are independent of
Willy Tarreauc5011ca2010-03-22 11:53:56 +01001886 the 'uid' parameter.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001887
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001888 <path> is a UNIX socket path beginning with a slash ('/'). This is
1889 alternative to the TCP listening port. Haproxy will then
1890 receive UNIX connections on the socket located at this place.
1891 The path must begin with a slash and by default is absolute.
1892 It can be relative to the prefix defined by "unix-bind" in
1893 the global section. Note that the total length of the prefix
1894 followed by the socket path cannot exceed some system limits
1895 for UNIX sockets, which commonly are set to 107 characters.
1896
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001897 <param*> is a list of parameters common to all sockets declared on the
1898 same line. These numerous parameters depend on OS and build
1899 options and have a complete section dedicated to them. Please
1900 refer to section 5 to for more details.
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001901
Willy Tarreau0ba27502007-12-24 16:55:16 +01001902 It is possible to specify a list of address:port combinations delimited by
1903 commas. The frontend will then listen on all of these addresses. There is no
1904 fixed limit to the number of addresses and ports which can be listened on in
1905 a frontend, as well as there is no limit to the number of "bind" statements
1906 in a frontend.
1907
1908 Example :
1909 listen http_proxy
1910 bind :80,:443
1911 bind 10.0.0.1:10080,10.0.0.1:10443
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001912 bind /var/run/ssl-frontend.sock user root mode 600 accept-proxy
Willy Tarreau0ba27502007-12-24 16:55:16 +01001913
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001914 listen http_https_proxy
1915 bind :80
Cyril Bonté0d44fc62012-10-09 22:45:33 +02001916 bind :443 ssl crt /etc/haproxy/site.pem
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02001917
Willy Tarreau24709282013-03-10 21:32:12 +01001918 listen http_https_proxy_explicit
1919 bind ipv6@:80
1920 bind ipv4@public_ssl:443 ssl crt /etc/haproxy/site.pem
1921 bind unix@ssl-frontend.sock user root mode 600 accept-proxy
1922
Willy Tarreaudad36a32013-03-11 01:20:04 +01001923 listen external_bind_app1
1924 bind fd@${FD_APP1}
1925
Willy Tarreauceb24bc2010-11-09 12:46:41 +01001926 See also : "source", "option forwardfor", "unix-bind" and the PROXY protocol
Willy Tarreaub6205fd2012-09-24 12:27:33 +02001927 documentation, and section 5 about bind options.
Willy Tarreau0ba27502007-12-24 16:55:16 +01001928
1929
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001930bind-process [ all | odd | even | <number 1-64>[-<number 1-64>] ] ...
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001931 Limit visibility of an instance to a certain set of processes numbers.
1932 May be used in sections : defaults | frontend | listen | backend
1933 yes | yes | yes | yes
1934 Arguments :
1935 all All process will see this instance. This is the default. It
1936 may be used to override a default value.
1937
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001938 odd This instance will be enabled on processes 1,3,5,...63. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001939 option may be combined with other numbers.
1940
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001941 even This instance will be enabled on processes 2,4,6,...64. This
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001942 option may be combined with other numbers. Do not use it
1943 with less than 2 processes otherwise some instances might be
1944 missing from all processes.
1945
Willy Tarreau110ecc12012-11-15 17:50:01 +01001946 number The instance will be enabled on this process number or range,
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001947 whose values must all be between 1 and 32 or 64 depending on
Willy Tarreau102df612014-05-07 23:56:38 +02001948 the machine's word size. If a proxy is bound to process
1949 numbers greater than the configured global.nbproc, it will
1950 either be forced to process #1 if a single process was
1951 specified, or to all processes otherwise.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001952
1953 This keyword limits binding of certain instances to certain processes. This
1954 is useful in order not to have too many processes listening to the same
1955 ports. For instance, on a dual-core machine, it might make sense to set
1956 'nbproc 2' in the global section, then distributes the listeners among 'odd'
1957 and 'even' instances.
1958
Willy Tarreaua9db57e2013-01-18 11:29:29 +01001959 At the moment, it is not possible to reference more than 32 or 64 processes
1960 using this keyword, but this should be more than enough for most setups.
1961 Please note that 'all' really means all processes regardless of the machine's
1962 word size, and is not limited to the first 32 or 64.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001963
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02001964 Each "bind" line may further be limited to a subset of the proxy's processes,
1965 please consult the "process" bind keyword in section 5.1.
1966
Willy Tarreaub369a042014-09-16 13:21:03 +02001967 When a frontend has no explicit "bind-process" line, it tries to bind to all
1968 the processes referenced by its "bind" lines. That means that frontends can
1969 easily adapt to their listeners' processes.
1970
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001971 If some backends are referenced by frontends bound to other processes, the
1972 backend automatically inherits the frontend's processes.
1973
1974 Example :
1975 listen app_ip1
1976 bind 10.0.0.1:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001977 bind-process odd
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001978
1979 listen app_ip2
1980 bind 10.0.0.2:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001981 bind-process even
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001982
1983 listen management
1984 bind 10.0.0.3:80
Willy Tarreaubfcd3112010-10-23 11:22:08 +02001985 bind-process 1 2 3 4
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001986
Willy Tarreau110ecc12012-11-15 17:50:01 +01001987 listen management
1988 bind 10.0.0.4:80
1989 bind-process 1-4
1990
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02001991 See also : "nbproc" in global section, and "process" in section 5.1.
Willy Tarreau0b9c02c2009-02-04 22:05:05 +01001992
1993
Willy Tarreau0ba27502007-12-24 16:55:16 +01001994block { if | unless } <condition>
1995 Block a layer 7 request if/unless a condition is matched
1996 May be used in sections : defaults | frontend | listen | backend
1997 no | yes | yes | yes
1998
1999 The HTTP request will be blocked very early in the layer 7 processing
2000 if/unless <condition> is matched. A 403 error will be returned if the request
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002001 is blocked. The condition has to reference ACLs (see section 7). This is
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02002002 typically used to deny access to certain sensitive resources if some
Willy Tarreau0ba27502007-12-24 16:55:16 +01002003 conditions are met or not met. There is no fixed limit to the number of
2004 "block" statements per instance.
2005
2006 Example:
2007 acl invalid_src src 0.0.0.0/7 224.0.0.0/3
2008 acl invalid_src src_port 0:1023
2009 acl local_dst hdr(host) -i localhost
2010 block if invalid_src || local_dst
2011
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002012 See section 7 about ACL usage.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002013
2014
2015capture cookie <name> len <length>
2016 Capture and log a cookie in the request and in the response.
2017 May be used in sections : defaults | frontend | listen | backend
2018 no | yes | yes | no
2019 Arguments :
2020 <name> is the beginning of the name of the cookie to capture. In order
2021 to match the exact name, simply suffix the name with an equal
2022 sign ('='). The full name will appear in the logs, which is
2023 useful with application servers which adjust both the cookie name
2024 and value (eg: ASPSESSIONXXXXX).
2025
2026 <length> is the maximum number of characters to report in the logs, which
2027 include the cookie name, the equal sign and the value, all in the
2028 standard "name=value" form. The string will be truncated on the
2029 right if it exceeds <length>.
2030
2031 Only the first cookie is captured. Both the "cookie" request headers and the
2032 "set-cookie" response headers are monitored. This is particularly useful to
2033 check for application bugs causing session crossing or stealing between
2034 users, because generally the user's cookies can only change on a login page.
2035
2036 When the cookie was not presented by the client, the associated log column
2037 will report "-". When a request does not cause a cookie to be assigned by the
2038 server, a "-" is reported in the response column.
2039
2040 The capture is performed in the frontend only because it is necessary that
2041 the log format does not change for a given frontend depending on the
2042 backends. This may change in the future. Note that there can be only one
Willy Tarreau193b8c62012-11-22 00:17:38 +01002043 "capture cookie" statement in a frontend. The maximum capture length is set
2044 by the global "tune.http.cookielen" setting and defaults to 63 characters. It
2045 is not possible to specify a capture in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002046
2047 Example:
2048 capture cookie ASPSESSION len 32
2049
2050 See also : "capture request header", "capture response header" as well as
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002051 section 8 about logging.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002052
2053
2054capture request header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002055 Capture and log the last occurrence of the specified request header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002056 May be used in sections : defaults | frontend | listen | backend
2057 no | yes | yes | no
2058 Arguments :
2059 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002060 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002061 appear in the requests, with the first letter of each word in
2062 upper case. The header name will not appear in the logs, only the
2063 value is reported, but the position in the logs is respected.
2064
2065 <length> is the maximum number of characters to extract from the value and
2066 report in the logs. The string will be truncated on the right if
2067 it exceeds <length>.
2068
Willy Tarreau4460d032012-11-21 23:37:37 +01002069 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002070 value will be added to the logs between braces ('{}'). If multiple headers
2071 are captured, they will be delimited by a vertical bar ('|') and will appear
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002072 in the same order they were declared in the configuration. Non-existent
2073 headers will be logged just as an empty string. Common uses for request
2074 header captures include the "Host" field in virtual hosting environments, the
2075 "Content-length" when uploads are supported, "User-agent" to quickly
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002076 differentiate between real users and robots, and "X-Forwarded-For" in proxied
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002077 environments to find where the request came from.
2078
2079 Note that when capturing headers such as "User-agent", some spaces may be
2080 logged, making the log analysis more difficult. Thus be careful about what
2081 you log if you know your log parser is not smart enough to rely on the
2082 braces.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002083
Willy Tarreau0900abb2012-11-22 00:21:46 +01002084 There is no limit to the number of captured request headers nor to their
2085 length, though it is wise to keep them low to limit memory usage per session.
2086 In order to keep log format consistent for a same frontend, header captures
2087 can only be declared in a frontend. It is not possible to specify a capture
2088 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002089
2090 Example:
2091 capture request header Host len 15
2092 capture request header X-Forwarded-For len 15
2093 capture request header Referrer len 15
2094
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002095 See also : "capture cookie", "capture response header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002096 about logging.
2097
2098
2099capture response header <name> len <length>
Willy Tarreau4460d032012-11-21 23:37:37 +01002100 Capture and log the last occurrence of the specified response header.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002101 May be used in sections : defaults | frontend | listen | backend
2102 no | yes | yes | no
2103 Arguments :
2104 <name> is the name of the header to capture. The header names are not
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002105 case-sensitive, but it is a common practice to write them as they
Willy Tarreau0ba27502007-12-24 16:55:16 +01002106 appear in the response, with the first letter of each word in
2107 upper case. The header name will not appear in the logs, only the
2108 value is reported, but the position in the logs is respected.
2109
2110 <length> is the maximum number of characters to extract from the value and
2111 report in the logs. The string will be truncated on the right if
2112 it exceeds <length>.
2113
Willy Tarreau4460d032012-11-21 23:37:37 +01002114 The complete value of the last occurrence of the header is captured. The
Willy Tarreau0ba27502007-12-24 16:55:16 +01002115 result will be added to the logs between braces ('{}') after the captured
2116 request headers. If multiple headers are captured, they will be delimited by
2117 a vertical bar ('|') and will appear in the same order they were declared in
Willy Tarreaucc6c8912009-02-22 10:53:55 +01002118 the configuration. Non-existent headers will be logged just as an empty
2119 string. Common uses for response header captures include the "Content-length"
2120 header which indicates how many bytes are expected to be returned, the
2121 "Location" header to track redirections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002122
Willy Tarreau0900abb2012-11-22 00:21:46 +01002123 There is no limit to the number of captured response headers nor to their
2124 length, though it is wise to keep them low to limit memory usage per session.
2125 In order to keep log format consistent for a same frontend, header captures
2126 can only be declared in a frontend. It is not possible to specify a capture
2127 in a "defaults" section.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002128
2129 Example:
2130 capture response header Content-length len 9
2131 capture response header Location len 15
2132
Willy Tarreauc57f0e22009-05-10 13:12:33 +02002133 See also : "capture cookie", "capture request header" as well as section 8
Willy Tarreau0ba27502007-12-24 16:55:16 +01002134 about logging.
2135
2136
Cyril Bontéf0c60612010-02-06 14:44:47 +01002137clitimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002138 Set the maximum inactivity time on the client side.
2139 May be used in sections : defaults | frontend | listen | backend
2140 yes | yes | yes | no
2141 Arguments :
2142 <timeout> is the timeout value is specified in milliseconds by default, but
2143 can be in any other unit if the number is suffixed by the unit,
2144 as explained at the top of this document.
2145
2146 The inactivity timeout applies when the client is expected to acknowledge or
2147 send data. In HTTP mode, this timeout is particularly important to consider
2148 during the first phase, when the client sends the request, and during the
2149 response while it is reading data sent by the server. The value is specified
2150 in milliseconds by default, but can be in any other unit if the number is
2151 suffixed by the unit, as specified at the top of this document. In TCP mode
2152 (and to a lesser extent, in HTTP mode), it is highly recommended that the
2153 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002154 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01002155 losses by specifying timeouts that are slightly above multiples of 3 seconds
2156 (eg: 4 or 5 seconds).
2157
2158 This parameter is specific to frontends, but can be specified once for all in
2159 "defaults" sections. This is in fact one of the easiest solutions not to
2160 forget about it. An unspecified timeout results in an infinite timeout, which
2161 is not recommended. Such a usage is accepted and works but reports a warning
2162 during startup because it may results in accumulation of expired sessions in
2163 the system if the system's timeouts are not configured either.
2164
2165 This parameter is provided for compatibility but is currently deprecated.
2166 Please use "timeout client" instead.
2167
Willy Tarreau036fae02008-01-06 13:24:40 +01002168 See also : "timeout client", "timeout http-request", "timeout server", and
2169 "srvtimeout".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002170
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002171compression algo <algorithm> ...
2172compression type <mime type> ...
Willy Tarreau70737d12012-10-27 00:34:28 +02002173compression offload
William Lallemand82fe75c2012-10-23 10:25:10 +02002174 Enable HTTP compression.
2175 May be used in sections : defaults | frontend | listen | backend
2176 yes | yes | yes | yes
2177 Arguments :
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002178 algo is followed by the list of supported compression algorithms.
2179 type is followed by the list of MIME types that will be compressed.
2180 offload makes haproxy work as a compression offloader only (see notes).
2181
2182 The currently supported algorithms are :
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002183 identity this is mostly for debugging, and it was useful for developing
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002184 the compression feature. Identity does not apply any change on
2185 data.
2186
2187 gzip applies gzip compression. This setting is only available when
2188 support for zlib was built in.
2189
2190 deflate same as gzip, but with deflate algorithm and zlib format.
2191 Note that this algorithm has ambiguous support on many browsers
2192 and no support at all from recent ones. It is strongly
2193 recommended not to use it for anything else than experimentation.
2194 This setting is only available when support for zlib was built
2195 in.
2196
Dmitry Sivachenko87c208b2012-11-22 20:03:26 +04002197 Compression will be activated depending on the Accept-Encoding request
Cyril Bonté316a8cf2012-11-11 13:38:27 +01002198 header. With identity, it does not take care of that header.
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002199 If backend servers support HTTP compression, these directives
2200 will be no-op: haproxy will see the compressed response and will not
2201 compress again. If backend servers do not support HTTP compression and
2202 there is Accept-Encoding header in request, haproxy will compress the
2203 matching response.
Willy Tarreau70737d12012-10-27 00:34:28 +02002204
2205 The "offload" setting makes haproxy remove the Accept-Encoding header to
2206 prevent backend servers from compressing responses. It is strongly
2207 recommended not to do this because this means that all the compression work
2208 will be done on the single point where haproxy is located. However in some
2209 deployment scenarios, haproxy may be installed in front of a buggy gateway
Dmitry Sivachenkoc9f3b452012-11-28 17:47:11 +04002210 with broken HTTP compression implementation which can't be turned off.
2211 In that case haproxy can be used to prevent that gateway from emitting
2212 invalid payloads. In this case, simply removing the header in the
2213 configuration does not work because it applies before the header is parsed,
2214 so that prevents haproxy from compressing. The "offload" setting should
Willy Tarreauffea9fd2014-07-12 16:37:02 +02002215 then be used for such scenarios. Note: for now, the "offload" setting is
2216 ignored when set in a defaults section.
William Lallemand82fe75c2012-10-23 10:25:10 +02002217
William Lallemand05097442012-11-20 12:14:28 +01002218 Compression is disabled when:
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002219 * the request does not advertise a supported compression algorithm in the
2220 "Accept-Encoding" header
2221 * the response message is not HTTP/1.1
William Lallemandd3002612012-11-26 14:34:47 +01002222 * HTTP status code is not 200
William Lallemand8bb4e342013-12-10 17:28:48 +01002223 * response header "Transfer-Encoding" contains "chunked" (Temporary
2224 Workaround)
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002225 * response contain neither a "Content-Length" header nor a
2226 "Transfer-Encoding" whose last value is "chunked"
2227 * response contains a "Content-Type" header whose first value starts with
2228 "multipart"
2229 * the response contains the "no-transform" value in the "Cache-control"
2230 header
2231 * User-Agent matches "Mozilla/4" unless it is MSIE 6 with XP SP2, or MSIE 7
2232 and later
2233 * The response contains a "Content-Encoding" header, indicating that the
2234 response is already compressed (see compression offload)
William Lallemand05097442012-11-20 12:14:28 +01002235
Baptiste Assmann650d53d2013-01-05 15:44:44 +01002236 Note: The compression does not rewrite Etag headers, and does not emit the
2237 Warning header.
William Lallemand05097442012-11-20 12:14:28 +01002238
William Lallemand82fe75c2012-10-23 10:25:10 +02002239 Examples :
2240 compression algo gzip
2241 compression type text/html text/plain
Willy Tarreau0ba27502007-12-24 16:55:16 +01002242
Cyril Bontéf0c60612010-02-06 14:44:47 +01002243contimeout <timeout> (deprecated)
Willy Tarreau0ba27502007-12-24 16:55:16 +01002244 Set the maximum time to wait for a connection attempt to a server to succeed.
2245 May be used in sections : defaults | frontend | listen | backend
2246 yes | no | yes | yes
2247 Arguments :
2248 <timeout> is the timeout value is specified in milliseconds by default, but
2249 can be in any other unit if the number is suffixed by the unit,
2250 as explained at the top of this document.
2251
2252 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002253 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01002254 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01002255 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
2256 connect timeout also presets the queue timeout to the same value if this one
2257 has not been specified. Historically, the contimeout was also used to set the
2258 tarpit timeout in a listen section, which is not possible in a pure frontend.
2259
2260 This parameter is specific to backends, but can be specified once for all in
2261 "defaults" sections. This is in fact one of the easiest solutions not to
2262 forget about it. An unspecified timeout results in an infinite timeout, which
2263 is not recommended. Such a usage is accepted and works but reports a warning
2264 during startup because it may results in accumulation of failed sessions in
2265 the system if the system's timeouts are not configured either.
2266
2267 This parameter is provided for backwards compatibility but is currently
2268 deprecated. Please use "timeout connect", "timeout queue" or "timeout tarpit"
2269 instead.
2270
2271 See also : "timeout connect", "timeout queue", "timeout tarpit",
2272 "timeout server", "contimeout".
2273
2274
Willy Tarreau55165fe2009-05-10 12:02:55 +02002275cookie <name> [ rewrite | insert | prefix ] [ indirect ] [ nocache ]
Willy Tarreau4992dd22012-05-31 21:02:17 +02002276 [ postonly ] [ preserve ] [ httponly ] [ secure ]
2277 [ domain <domain> ]* [ maxidle <idle> ] [ maxlife <life> ]
Willy Tarreau0ba27502007-12-24 16:55:16 +01002278 Enable cookie-based persistence in a backend.
2279 May be used in sections : defaults | frontend | listen | backend
2280 yes | no | yes | yes
2281 Arguments :
2282 <name> is the name of the cookie which will be monitored, modified or
2283 inserted in order to bring persistence. This cookie is sent to
2284 the client via a "Set-Cookie" header in the response, and is
2285 brought back by the client in a "Cookie" header in all requests.
2286 Special care should be taken to choose a name which does not
2287 conflict with any likely application cookie. Also, if the same
2288 backends are subject to be used by the same clients (eg:
2289 HTTP/HTTPS), care should be taken to use different cookie names
2290 between all backends if persistence between them is not desired.
2291
2292 rewrite This keyword indicates that the cookie will be provided by the
2293 server and that haproxy will have to modify its value to set the
2294 server's identifier in it. This mode is handy when the management
2295 of complex combinations of "Set-cookie" and "Cache-control"
2296 headers is left to the application. The application can then
2297 decide whether or not it is appropriate to emit a persistence
2298 cookie. Since all responses should be monitored, this mode only
2299 works in HTTP close mode. Unless the application behaviour is
2300 very complex and/or broken, it is advised not to start with this
2301 mode for new deployments. This keyword is incompatible with
2302 "insert" and "prefix".
2303
2304 insert This keyword indicates that the persistence cookie will have to
Willy Tarreaua79094d2010-08-31 22:54:15 +02002305 be inserted by haproxy in server responses if the client did not
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002306
Willy Tarreaua79094d2010-08-31 22:54:15 +02002307 already have a cookie that would have permitted it to access this
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002308 server. When used without the "preserve" option, if the server
2309 emits a cookie with the same name, it will be remove before
2310 processing. For this reason, this mode can be used to upgrade
2311 existing configurations running in the "rewrite" mode. The cookie
2312 will only be a session cookie and will not be stored on the
2313 client's disk. By default, unless the "indirect" option is added,
2314 the server will see the cookies emitted by the client. Due to
2315 caching effects, it is generally wise to add the "nocache" or
2316 "postonly" keywords (see below). The "insert" keyword is not
2317 compatible with "rewrite" and "prefix".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002318
2319 prefix This keyword indicates that instead of relying on a dedicated
2320 cookie for the persistence, an existing one will be completed.
2321 This may be needed in some specific environments where the client
2322 does not support more than one single cookie and the application
2323 already needs it. In this case, whenever the server sets a cookie
2324 named <name>, it will be prefixed with the server's identifier
2325 and a delimiter. The prefix will be removed from all client
2326 requests so that the server still finds the cookie it emitted.
2327 Since all requests and responses are subject to being modified,
2328 this mode requires the HTTP close mode. The "prefix" keyword is
Willy Tarreau37229df2011-10-17 12:24:55 +02002329 not compatible with "rewrite" and "insert". Note: it is highly
2330 recommended not to use "indirect" with "prefix", otherwise server
2331 cookie updates would not be sent to clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002332
Willy Tarreaua79094d2010-08-31 22:54:15 +02002333 indirect When this option is specified, no cookie will be emitted to a
2334 client which already has a valid one for the server which has
2335 processed the request. If the server sets such a cookie itself,
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002336 it will be removed, unless the "preserve" option is also set. In
2337 "insert" mode, this will additionally remove cookies from the
2338 requests transmitted to the server, making the persistence
2339 mechanism totally transparent from an application point of view.
Willy Tarreau37229df2011-10-17 12:24:55 +02002340 Note: it is highly recommended not to use "indirect" with
2341 "prefix", otherwise server cookie updates would not be sent to
2342 clients.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002343
2344 nocache This option is recommended in conjunction with the insert mode
2345 when there is a cache between the client and HAProxy, as it
2346 ensures that a cacheable response will be tagged non-cacheable if
2347 a cookie needs to be inserted. This is important because if all
2348 persistence cookies are added on a cacheable home page for
2349 instance, then all customers will then fetch the page from an
2350 outer cache and will all share the same persistence cookie,
2351 leading to one server receiving much more traffic than others.
2352 See also the "insert" and "postonly" options.
2353
2354 postonly This option ensures that cookie insertion will only be performed
2355 on responses to POST requests. It is an alternative to the
2356 "nocache" option, because POST responses are not cacheable, so
2357 this ensures that the persistence cookie will never get cached.
2358 Since most sites do not need any sort of persistence before the
2359 first POST which generally is a login request, this is a very
2360 efficient method to optimize caching without risking to find a
2361 persistence cookie in the cache.
2362 See also the "insert" and "nocache" options.
2363
Willy Tarreauba4c5be2010-10-23 12:46:42 +02002364 preserve This option may only be used with "insert" and/or "indirect". It
2365 allows the server to emit the persistence cookie itself. In this
2366 case, if a cookie is found in the response, haproxy will leave it
2367 untouched. This is useful in order to end persistence after a
2368 logout request for instance. For this, the server just has to
2369 emit a cookie with an invalid value (eg: empty) or with a date in
2370 the past. By combining this mechanism with the "disable-on-404"
2371 check option, it is possible to perform a completely graceful
2372 shutdown because users will definitely leave the server after
2373 they logout.
2374
Willy Tarreau4992dd22012-05-31 21:02:17 +02002375 httponly This option tells haproxy to add an "HttpOnly" cookie attribute
2376 when a cookie is inserted. This attribute is used so that a
2377 user agent doesn't share the cookie with non-HTTP components.
2378 Please check RFC6265 for more information on this attribute.
2379
2380 secure This option tells haproxy to add a "Secure" cookie attribute when
2381 a cookie is inserted. This attribute is used so that a user agent
2382 never emits this cookie over non-secure channels, which means
2383 that a cookie learned with this flag will be presented only over
2384 SSL/TLS connections. Please check RFC6265 for more information on
2385 this attribute.
2386
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002387 domain This option allows to specify the domain at which a cookie is
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002388 inserted. It requires exactly one parameter: a valid domain
Willy Tarreau68a897b2009-12-03 23:28:34 +01002389 name. If the domain begins with a dot, the browser is allowed to
2390 use it for any host ending with that name. It is also possible to
2391 specify several domain names by invoking this option multiple
2392 times. Some browsers might have small limits on the number of
2393 domains, so be careful when doing that. For the record, sending
2394 10 domains to MSIE 6 or Firefox 2 works as expected.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +02002395
Willy Tarreau996a92c2010-10-13 19:30:47 +02002396 maxidle This option allows inserted cookies to be ignored after some idle
2397 time. It only works with insert-mode cookies. When a cookie is
2398 sent to the client, the date this cookie was emitted is sent too.
2399 Upon further presentations of this cookie, if the date is older
2400 than the delay indicated by the parameter (in seconds), it will
2401 be ignored. Otherwise, it will be refreshed if needed when the
2402 response is sent to the client. This is particularly useful to
2403 prevent users who never close their browsers from remaining for
2404 too long on the same server (eg: after a farm size change). When
2405 this option is set and a cookie has no date, it is always
2406 accepted, but gets refreshed in the response. This maintains the
2407 ability for admins to access their sites. Cookies that have a
2408 date in the future further than 24 hours are ignored. Doing so
2409 lets admins fix timezone issues without risking kicking users off
2410 the site.
2411
2412 maxlife This option allows inserted cookies to be ignored after some life
2413 time, whether they're in use or not. It only works with insert
2414 mode cookies. When a cookie is first sent to the client, the date
2415 this cookie was emitted is sent too. Upon further presentations
2416 of this cookie, if the date is older than the delay indicated by
2417 the parameter (in seconds), it will be ignored. If the cookie in
2418 the request has no date, it is accepted and a date will be set.
2419 Cookies that have a date in the future further than 24 hours are
2420 ignored. Doing so lets admins fix timezone issues without risking
2421 kicking users off the site. Contrary to maxidle, this value is
2422 not refreshed, only the first visit date counts. Both maxidle and
2423 maxlife may be used at the time. This is particularly useful to
2424 prevent users who never close their browsers from remaining for
2425 too long on the same server (eg: after a farm size change). This
2426 is stronger than the maxidle method in that it forces a
2427 redispatch after some absolute delay.
2428
Willy Tarreau0ba27502007-12-24 16:55:16 +01002429 There can be only one persistence cookie per HTTP backend, and it can be
2430 declared in a defaults section. The value of the cookie will be the value
2431 indicated after the "cookie" keyword in a "server" statement. If no cookie
2432 is declared for a given server, the cookie is not set.
Willy Tarreau6a06a402007-07-15 20:15:28 +02002433
Willy Tarreau0ba27502007-12-24 16:55:16 +01002434 Examples :
2435 cookie JSESSIONID prefix
2436 cookie SRV insert indirect nocache
2437 cookie SRV insert postonly indirect
Willy Tarreau996a92c2010-10-13 19:30:47 +02002438 cookie SRV insert indirect nocache maxidle 30m maxlife 8h
Willy Tarreau0ba27502007-12-24 16:55:16 +01002439
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002440 See also : "appsession", "balance source", "capture cookie", "server"
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002441 and "ignore-persist".
Willy Tarreau0ba27502007-12-24 16:55:16 +01002442
Willy Tarreau983e01e2010-01-11 18:42:06 +01002443
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002444default-server [param*]
2445 Change default options for a server in a backend
2446 May be used in sections : defaults | frontend | listen | backend
2447 yes | no | yes | yes
2448 Arguments:
Willy Tarreau983e01e2010-01-11 18:42:06 +01002449 <param*> is a list of parameters for this server. The "default-server"
2450 keyword accepts an important number of options and has a complete
2451 section dedicated to it. Please refer to section 5 for more
2452 details.
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002453
Willy Tarreau983e01e2010-01-11 18:42:06 +01002454 Example :
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01002455 default-server inter 1000 weight 13
2456
2457 See also: "server" and section 5 about server options
Willy Tarreau0ba27502007-12-24 16:55:16 +01002458
Willy Tarreau983e01e2010-01-11 18:42:06 +01002459
Willy Tarreau0ba27502007-12-24 16:55:16 +01002460default_backend <backend>
2461 Specify the backend to use when no "use_backend" rule has been matched.
2462 May be used in sections : defaults | frontend | listen | backend
2463 yes | yes | yes | no
2464 Arguments :
2465 <backend> is the name of the backend to use.
2466
2467 When doing content-switching between frontend and backends using the
2468 "use_backend" keyword, it is often useful to indicate which backend will be
2469 used when no rule has matched. It generally is the dynamic backend which
2470 will catch all undetermined requests.
2471
Willy Tarreau0ba27502007-12-24 16:55:16 +01002472 Example :
2473
2474 use_backend dynamic if url_dyn
2475 use_backend static if url_css url_img extension_img
2476 default_backend dynamic
2477
Willy Tarreau2769aa02007-12-27 18:26:09 +01002478 See also : "use_backend", "reqsetbe", "reqisetbe"
2479
Willy Tarreau0ba27502007-12-24 16:55:16 +01002480
Baptiste Assmann27f51342013-10-09 06:51:49 +02002481description <string>
2482 Describe a listen, frontend or backend.
2483 May be used in sections : defaults | frontend | listen | backend
2484 no | yes | yes | yes
2485 Arguments : string
2486
2487 Allows to add a sentence to describe the related object in the HAProxy HTML
2488 stats page. The description will be printed on the right of the object name
2489 it describes.
2490 No need to backslash spaces in the <string> arguments.
2491
2492
Willy Tarreau0ba27502007-12-24 16:55:16 +01002493disabled
2494 Disable a proxy, frontend or backend.
2495 May be used in sections : defaults | frontend | listen | backend
2496 yes | yes | yes | yes
2497 Arguments : none
2498
2499 The "disabled" keyword is used to disable an instance, mainly in order to
2500 liberate a listening port or to temporarily disable a service. The instance
2501 will still be created and its configuration will be checked, but it will be
2502 created in the "stopped" state and will appear as such in the statistics. It
2503 will not receive any traffic nor will it send any health-checks or logs. It
2504 is possible to disable many instances at once by adding the "disabled"
2505 keyword in a "defaults" section.
2506
2507 See also : "enabled"
2508
2509
Willy Tarreau5ce94572010-06-07 14:35:41 +02002510dispatch <address>:<port>
2511 Set a default server address
2512 May be used in sections : defaults | frontend | listen | backend
2513 no | no | yes | yes
Cyril Bonté108cf6e2012-04-21 23:30:29 +02002514 Arguments :
Willy Tarreau5ce94572010-06-07 14:35:41 +02002515
2516 <address> is the IPv4 address of the default server. Alternatively, a
2517 resolvable hostname is supported, but this name will be resolved
2518 during start-up.
2519
2520 <ports> is a mandatory port specification. All connections will be sent
2521 to this port, and it is not permitted to use port offsets as is
2522 possible with normal servers.
2523
Willy Tarreau787aed52011-04-15 06:45:37 +02002524 The "dispatch" keyword designates a default server for use when no other
Willy Tarreau5ce94572010-06-07 14:35:41 +02002525 server can take the connection. In the past it was used to forward non
2526 persistent connections to an auxiliary load balancer. Due to its simple
2527 syntax, it has also been used for simple TCP relays. It is recommended not to
2528 use it for more clarity, and to use the "server" directive instead.
2529
2530 See also : "server"
2531
2532
Willy Tarreau0ba27502007-12-24 16:55:16 +01002533enabled
2534 Enable a proxy, frontend or backend.
2535 May be used in sections : defaults | frontend | listen | backend
2536 yes | yes | yes | yes
2537 Arguments : none
2538
2539 The "enabled" keyword is used to explicitly enable an instance, when the
2540 defaults has been set to "disabled". This is very rarely used.
2541
2542 See also : "disabled"
2543
2544
2545errorfile <code> <file>
2546 Return a file contents instead of errors generated by HAProxy
2547 May be used in sections : defaults | frontend | listen | backend
2548 yes | yes | yes | yes
2549 Arguments :
2550 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002551 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002552
2553 <file> designates a file containing the full HTTP response. It is
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002554 recommended to follow the common practice of appending ".http" to
Willy Tarreau0ba27502007-12-24 16:55:16 +01002555 the filename so that people do not confuse the response with HTML
Willy Tarreau59140a22009-02-22 12:02:30 +01002556 error pages, and to use absolute paths, since files are read
2557 before any chroot is performed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01002558
2559 It is important to understand that this keyword is not meant to rewrite
2560 errors returned by the server, but errors detected and returned by HAProxy.
2561 This is why the list of supported errors is limited to a small set.
2562
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002563 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2564
Willy Tarreau0ba27502007-12-24 16:55:16 +01002565 The files are returned verbatim on the TCP socket. This allows any trick such
2566 as redirections to another URL or site, as well as tricks to clean cookies,
2567 force enable or disable caching, etc... The package provides default error
2568 files returning the same contents as default errors.
2569
Willy Tarreau59140a22009-02-22 12:02:30 +01002570 The files should not exceed the configured buffer size (BUFSIZE), which
2571 generally is 8 or 16 kB, otherwise they will be truncated. It is also wise
2572 not to put any reference to local contents (eg: images) in order to avoid
2573 loops between the client and HAProxy when all servers are down, causing an
2574 error to be returned instead of an image. For better HTTP compliance, it is
2575 recommended that all header lines end with CR-LF and not LF alone.
2576
Willy Tarreau0ba27502007-12-24 16:55:16 +01002577 The files are read at the same time as the configuration and kept in memory.
2578 For this reason, the errors continue to be returned even when the process is
2579 chrooted, and no file change is considered while the process is running. A
Willy Tarreauc27debf2008-01-06 08:57:02 +01002580 simple method for developing those files consists in associating them to the
Willy Tarreau0ba27502007-12-24 16:55:16 +01002581 403 status code and interrogating a blocked URL.
2582
2583 See also : "errorloc", "errorloc302", "errorloc303"
2584
Willy Tarreau59140a22009-02-22 12:02:30 +01002585 Example :
2586 errorfile 400 /etc/haproxy/errorfiles/400badreq.http
Willy Tarreau2705a612014-05-23 17:38:34 +02002587 errorfile 408 /dev/null # workaround Chrome pre-connect bug
Willy Tarreau59140a22009-02-22 12:02:30 +01002588 errorfile 403 /etc/haproxy/errorfiles/403forbid.http
2589 errorfile 503 /etc/haproxy/errorfiles/503sorry.http
2590
Willy Tarreau2769aa02007-12-27 18:26:09 +01002591
2592errorloc <code> <url>
2593errorloc302 <code> <url>
2594 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2595 May be used in sections : defaults | frontend | listen | backend
2596 yes | yes | yes | yes
2597 Arguments :
2598 <code> is the HTTP status code. Currently, HAProxy is capable of
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002599 generating codes 200, 400, 403, 408, 500, 502, 503, and 504.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002600
2601 <url> it is the exact contents of the "Location" header. It may contain
2602 either a relative URI to an error page hosted on the same site,
2603 or an absolute URI designating an error page on another site.
2604 Special care should be given to relative URIs to avoid redirect
2605 loops if the URI itself may generate the same error (eg: 500).
2606
2607 It is important to understand that this keyword is not meant to rewrite
2608 errors returned by the server, but errors detected and returned by HAProxy.
2609 This is why the list of supported errors is limited to a small set.
2610
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002611 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2612
Willy Tarreau2769aa02007-12-27 18:26:09 +01002613 Note that both keyword return the HTTP 302 status code, which tells the
2614 client to fetch the designated URL using the same HTTP method. This can be
2615 quite problematic in case of non-GET methods such as POST, because the URL
2616 sent to the client might not be allowed for something other than GET. To
2617 workaround this problem, please use "errorloc303" which send the HTTP 303
2618 status code, indicating to the client that the URL must be fetched with a GET
2619 request.
2620
2621 See also : "errorfile", "errorloc303"
2622
2623
2624errorloc303 <code> <url>
2625 Return an HTTP redirection to a URL instead of errors generated by HAProxy
2626 May be used in sections : defaults | frontend | listen | backend
2627 yes | yes | yes | yes
2628 Arguments :
2629 <code> is the HTTP status code. Currently, HAProxy is capable of
2630 generating codes 400, 403, 408, 500, 502, 503, and 504.
2631
2632 <url> it is the exact contents of the "Location" header. It may contain
2633 either a relative URI to an error page hosted on the same site,
2634 or an absolute URI designating an error page on another site.
2635 Special care should be given to relative URIs to avoid redirect
2636 loops if the URI itself may generate the same error (eg: 500).
2637
2638 It is important to understand that this keyword is not meant to rewrite
2639 errors returned by the server, but errors detected and returned by HAProxy.
2640 This is why the list of supported errors is limited to a small set.
2641
Willy Tarreauae94d4d2011-05-11 16:28:49 +02002642 Code 200 is emitted in response to requests matching a "monitor-uri" rule.
2643
Willy Tarreau2769aa02007-12-27 18:26:09 +01002644 Note that both keyword return the HTTP 303 status code, which tells the
2645 client to fetch the designated URL using the same HTTP GET method. This
2646 solves the usual problems associated with "errorloc" and the 302 code. It is
2647 possible that some very old browsers designed before HTTP/1.1 do not support
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01002648 it, but no such problem has been reported till now.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002649
2650 See also : "errorfile", "errorloc", "errorloc302"
2651
2652
Willy Tarreau4de91492010-01-22 19:10:05 +01002653force-persist { if | unless } <condition>
2654 Declare a condition to force persistence on down servers
2655 May be used in sections: defaults | frontend | listen | backend
2656 no | yes | yes | yes
2657
2658 By default, requests are not dispatched to down servers. It is possible to
2659 force this using "option persist", but it is unconditional and redispatches
2660 to a valid server if "option redispatch" is set. That leaves with very little
2661 possibilities to force some requests to reach a server which is artificially
2662 marked down for maintenance operations.
2663
2664 The "force-persist" statement allows one to declare various ACL-based
2665 conditions which, when met, will cause a request to ignore the down status of
2666 a server and still try to connect to it. That makes it possible to start a
2667 server, still replying an error to the health checks, and run a specially
2668 configured browser to test the service. Among the handy methods, one could
2669 use a specific source IP address, or a specific cookie. The cookie also has
2670 the advantage that it can easily be added/removed on the browser from a test
2671 page. Once the service is validated, it is then possible to open the service
2672 to the world by returning a valid response to health checks.
2673
2674 The forced persistence is enabled when an "if" condition is met, or unless an
2675 "unless" condition is met. The final redispatch is always disabled when this
2676 is used.
2677
Cyril Bonté0d4bf012010-04-25 23:21:46 +02002678 See also : "option redispatch", "ignore-persist", "persist",
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +02002679 and section 7 about ACL usage.
Willy Tarreau4de91492010-01-22 19:10:05 +01002680
2681
Willy Tarreau2769aa02007-12-27 18:26:09 +01002682fullconn <conns>
2683 Specify at what backend load the servers will reach their maxconn
2684 May be used in sections : defaults | frontend | listen | backend
2685 yes | no | yes | yes
2686 Arguments :
2687 <conns> is the number of connections on the backend which will make the
2688 servers use the maximal number of connections.
2689
Willy Tarreau198a7442008-01-17 12:05:32 +01002690 When a server has a "maxconn" parameter specified, it means that its number
Willy Tarreau2769aa02007-12-27 18:26:09 +01002691 of concurrent connections will never go higher. Additionally, if it has a
Willy Tarreau198a7442008-01-17 12:05:32 +01002692 "minconn" parameter, it indicates a dynamic limit following the backend's
Willy Tarreau2769aa02007-12-27 18:26:09 +01002693 load. The server will then always accept at least <minconn> connections,
2694 never more than <maxconn>, and the limit will be on the ramp between both
2695 values when the backend has less than <conns> concurrent connections. This
2696 makes it possible to limit the load on the servers during normal loads, but
2697 push it further for important loads without overloading the servers during
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002698 exceptional loads.
Willy Tarreau2769aa02007-12-27 18:26:09 +01002699
Willy Tarreaufbb78422011-06-05 15:38:35 +02002700 Since it's hard to get this value right, haproxy automatically sets it to
2701 10% of the sum of the maxconns of all frontends that may branch to this
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01002702 backend (based on "use_backend" and "default_backend" rules). That way it's
2703 safe to leave it unset. However, "use_backend" involving dynamic names are
2704 not counted since there is no way to know if they could match or not.
Willy Tarreaufbb78422011-06-05 15:38:35 +02002705
Willy Tarreau2769aa02007-12-27 18:26:09 +01002706 Example :
2707 # The servers will accept between 100 and 1000 concurrent connections each
2708 # and the maximum of 1000 will be reached when the backend reaches 10000
2709 # connections.
2710 backend dynamic
2711 fullconn 10000
2712 server srv1 dyn1:80 minconn 100 maxconn 1000
2713 server srv2 dyn2:80 minconn 100 maxconn 1000
2714
2715 See also : "maxconn", "server"
2716
2717
2718grace <time>
2719 Maintain a proxy operational for some time after a soft stop
2720 May be used in sections : defaults | frontend | listen | backend
Cyril Bonté99ed3272010-01-24 23:29:44 +01002721 yes | yes | yes | yes
Willy Tarreau2769aa02007-12-27 18:26:09 +01002722 Arguments :
2723 <time> is the time (by default in milliseconds) for which the instance
2724 will remain operational with the frontend sockets still listening
2725 when a soft-stop is received via the SIGUSR1 signal.
2726
2727 This may be used to ensure that the services disappear in a certain order.
2728 This was designed so that frontends which are dedicated to monitoring by an
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01002729 external equipment fail immediately while other ones remain up for the time
Willy Tarreau2769aa02007-12-27 18:26:09 +01002730 needed by the equipment to detect the failure.
2731
2732 Note that currently, there is very little benefit in using this parameter,
2733 and it may in fact complicate the soft-reconfiguration process more than
2734 simplify it.
2735
Willy Tarreau0ba27502007-12-24 16:55:16 +01002736
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002737hash-type <method> <function> <modifier>
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002738 Specify a method to use for mapping hashes to servers
2739 May be used in sections : defaults | frontend | listen | backend
2740 yes | no | yes | yes
2741 Arguments :
Bhaskar98634f02013-10-29 23:30:51 -04002742 <method> is the method used to select a server from the hash computed by
2743 the <function> :
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002744
Bhaskar98634f02013-10-29 23:30:51 -04002745 map-based the hash table is a static array containing all alive servers.
2746 The hashes will be very smooth, will consider weights, but
2747 will be static in that weight changes while a server is up
2748 will be ignored. This means that there will be no slow start.
2749 Also, since a server is selected by its position in the array,
2750 most mappings are changed when the server count changes. This
2751 means that when a server goes up or down, or when a server is
2752 added to a farm, most connections will be redistributed to
2753 different servers. This can be inconvenient with caches for
2754 instance.
Willy Tarreau798a39c2010-11-24 15:04:29 +01002755
Bhaskar98634f02013-10-29 23:30:51 -04002756 consistent the hash table is a tree filled with many occurrences of each
2757 server. The hash key is looked up in the tree and the closest
2758 server is chosen. This hash is dynamic, it supports changing
2759 weights while the servers are up, so it is compatible with the
2760 slow start feature. It has the advantage that when a server
2761 goes up or down, only its associations are moved. When a
2762 server is added to the farm, only a few part of the mappings
2763 are redistributed, making it an ideal method for caches.
2764 However, due to its principle, the distribution will never be
2765 very smooth and it may sometimes be necessary to adjust a
2766 server's weight or its ID to get a more balanced distribution.
2767 In order to get the same distribution on multiple load
2768 balancers, it is important that all servers have the exact
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002769 same IDs. Note: consistent hash uses sdbm and avalanche if no
2770 hash function is specified.
Bhaskar98634f02013-10-29 23:30:51 -04002771
2772 <function> is the hash function to be used :
2773
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03002774 sdbm this function was created initially for sdbm (a public-domain
Bhaskar98634f02013-10-29 23:30:51 -04002775 reimplementation of ndbm) database library. It was found to do
2776 well in scrambling bits, causing better distribution of the keys
2777 and fewer splits. It also happens to be a good general hashing
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002778 function with good distribution, unless the total server weight
2779 is a multiple of 64, in which case applying the avalanche
2780 modifier may help.
Bhaskar98634f02013-10-29 23:30:51 -04002781
2782 djb2 this function was first proposed by Dan Bernstein many years ago
2783 on comp.lang.c. Studies have shown that for certain workload this
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002784 function provides a better distribution than sdbm. It generally
2785 works well with text-based inputs though it can perform extremely
2786 poorly with numeric-only input or when the total server weight is
2787 a multiple of 33, unless the avalanche modifier is also used.
2788
Willy Tarreaua0f42712013-11-14 14:30:35 +01002789 wt6 this function was designed for haproxy while testing other
2790 functions in the past. It is not as smooth as the other ones, but
2791 is much less sensible to the input data set or to the number of
2792 servers. It can make sense as an alternative to sdbm+avalanche or
2793 djb2+avalanche for consistent hashing or when hashing on numeric
2794 data such as a source IP address or a visitor identifier in a URL
2795 parameter.
2796
Willy Tarreau324f07f2015-01-20 19:44:50 +01002797 crc32 this is the most common CRC32 implementation as used in Ethernet,
2798 gzip, PNG, etc. It is slower than the other ones but may provide
2799 a better distribution or less predictable results especially when
2800 used on strings.
2801
Bhaskar Maddalab6c0ac92013-11-05 11:54:02 -05002802 <modifier> indicates an optional method applied after hashing the key :
2803
2804 avalanche This directive indicates that the result from the hash
2805 function above should not be used in its raw form but that
2806 a 4-byte full avalanche hash must be applied first. The
2807 purpose of this step is to mix the resulting bits from the
2808 previous hash in order to avoid any undesired effect when
2809 the input contains some limited values or when the number of
2810 servers is a multiple of one of the hash's components (64
2811 for SDBM, 33 for DJB2). Enabling avalanche tends to make the
2812 result less predictable, but it's also not as smooth as when
2813 using the original function. Some testing might be needed
2814 with some workloads. This hash is one of the many proposed
2815 by Bob Jenkins.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002816
Bhaskar98634f02013-10-29 23:30:51 -04002817 The default hash type is "map-based" and is recommended for most usages. The
2818 default function is "sdbm", the selection of a function should be based on
2819 the range of the values being hashed.
Willy Tarreau6b2e11b2009-10-01 07:52:15 +02002820
2821 See also : "balance", "server"
2822
2823
Willy Tarreau0ba27502007-12-24 16:55:16 +01002824http-check disable-on-404
2825 Enable a maintenance mode upon HTTP/404 response to health-checks
2826 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau2769aa02007-12-27 18:26:09 +01002827 yes | no | yes | yes
Willy Tarreau0ba27502007-12-24 16:55:16 +01002828 Arguments : none
2829
2830 When this option is set, a server which returns an HTTP code 404 will be
2831 excluded from further load-balancing, but will still receive persistent
2832 connections. This provides a very convenient method for Web administrators
2833 to perform a graceful shutdown of their servers. It is also important to note
2834 that a server which is detected as failed while it was in this mode will not
2835 generate an alert, just a notice. If the server responds 2xx or 3xx again, it
2836 will immediately be reinserted into the farm. The status on the stats page
2837 reports "NOLB" for a server in this mode. It is important to note that this
Willy Tarreaubd741542010-03-16 18:46:54 +01002838 option only works in conjunction with the "httpchk" option. If this option
2839 is used with "http-check expect", then it has precedence over it so that 404
2840 responses will still be considered as soft-stop.
2841
2842 See also : "option httpchk", "http-check expect"
2843
2844
2845http-check expect [!] <match> <pattern>
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002846 Make HTTP health checks consider response contents or specific status codes
Willy Tarreaubd741542010-03-16 18:46:54 +01002847 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau1ee51a62011-08-19 20:04:17 +02002848 yes | no | yes | yes
Willy Tarreaubd741542010-03-16 18:46:54 +01002849 Arguments :
2850 <match> is a keyword indicating how to look for a specific pattern in the
2851 response. The keyword may be one of "status", "rstatus",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002852 "string", or "rstring". The keyword may be preceded by an
Willy Tarreaubd741542010-03-16 18:46:54 +01002853 exclamation mark ("!") to negate the match. Spaces are allowed
2854 between the exclamation mark and the keyword. See below for more
2855 details on the supported keywords.
2856
2857 <pattern> is the pattern to look for. It may be a string or a regular
2858 expression. If the pattern contains spaces, they must be escaped
2859 with the usual backslash ('\').
2860
2861 By default, "option httpchk" considers that response statuses 2xx and 3xx
2862 are valid, and that others are invalid. When "http-check expect" is used,
2863 it defines what is considered valid or invalid. Only one "http-check"
2864 statement is supported in a backend. If a server fails to respond or times
2865 out, the check obviously fails. The available matches are :
2866
2867 status <string> : test the exact string match for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002868 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002869 response's status code is exactly this string. If the
2870 "status" keyword is prefixed with "!", then the response
2871 will be considered invalid if the status code matches.
2872
2873 rstatus <regex> : test a regular expression for the HTTP status code.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002874 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002875 response's status code matches the expression. If the
2876 "rstatus" keyword is prefixed with "!", then the response
2877 will be considered invalid if the status code matches.
2878 This is mostly used to check for multiple codes.
2879
2880 string <string> : test the exact string match in the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002881 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002882 response's body contains this exact string. If the
2883 "string" keyword is prefixed with "!", then the response
2884 will be considered invalid if the body contains this
2885 string. This can be used to look for a mandatory word at
2886 the end of a dynamic page, or to detect a failure when a
2887 specific error appears on the check page (eg: a stack
2888 trace).
2889
2890 rstring <regex> : test a regular expression on the HTTP response body.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04002891 A health check response will be considered valid if the
Willy Tarreaubd741542010-03-16 18:46:54 +01002892 response's body matches this expression. If the "rstring"
2893 keyword is prefixed with "!", then the response will be
2894 considered invalid if the body matches the expression.
2895 This can be used to look for a mandatory word at the end
2896 of a dynamic page, or to detect a failure when a specific
2897 error appears on the check page (eg: a stack trace).
2898
2899 It is important to note that the responses will be limited to a certain size
2900 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
2901 Thus, too large responses may not contain the mandatory pattern when using
2902 "string" or "rstring". If a large response is absolutely required, it is
2903 possible to change the default max size by setting the global variable.
2904 However, it is worth keeping in mind that parsing very large responses can
2905 waste some CPU cycles, especially when regular expressions are used, and that
2906 it is always better to focus the checks on smaller resources.
2907
Cyril Bonté32602d22015-01-30 00:07:07 +01002908 Also "http-check expect" doesn't support HTTP keep-alive. Keep in mind that it
2909 will automatically append a "Connection: close" header, meaning that this
2910 header should not be present in the request provided by "option httpchk".
2911
Willy Tarreaubd741542010-03-16 18:46:54 +01002912 Last, if "http-check expect" is combined with "http-check disable-on-404",
2913 then this last one has precedence when the server responds with 404.
2914
2915 Examples :
2916 # only accept status 200 as valid
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002917 http-check expect status 200
Willy Tarreaubd741542010-03-16 18:46:54 +01002918
2919 # consider SQL errors as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002920 http-check expect ! string SQL\ Error
Willy Tarreaubd741542010-03-16 18:46:54 +01002921
2922 # consider status 5xx only as errors
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002923 http-check expect ! rstatus ^5
Willy Tarreaubd741542010-03-16 18:46:54 +01002924
2925 # check that we have a correct hexadecimal tag before /html
Willy Tarreau8f2a1e72011-01-06 16:36:10 +01002926 http-check expect rstring <!--tag:[0-9a-f]*</html>
Willy Tarreau0ba27502007-12-24 16:55:16 +01002927
Willy Tarreaubd741542010-03-16 18:46:54 +01002928 See also : "option httpchk", "http-check disable-on-404"
Willy Tarreau2769aa02007-12-27 18:26:09 +01002929
2930
Willy Tarreauef781042010-01-27 11:53:01 +01002931http-check send-state
2932 Enable emission of a state header with HTTP health checks
2933 May be used in sections : defaults | frontend | listen | backend
2934 yes | no | yes | yes
2935 Arguments : none
2936
2937 When this option is set, haproxy will systematically send a special header
2938 "X-Haproxy-Server-State" with a list of parameters indicating to each server
2939 how they are seen by haproxy. This can be used for instance when a server is
2940 manipulated without access to haproxy and the operator needs to know whether
2941 haproxy still sees it up or not, or if the server is the last one in a farm.
2942
2943 The header is composed of fields delimited by semi-colons, the first of which
2944 is a word ("UP", "DOWN", "NOLB"), possibly followed by a number of valid
2945 checks on the total number before transition, just as appears in the stats
2946 interface. Next headers are in the form "<variable>=<value>", indicating in
2947 no specific order some values available in the stats interface :
2948 - a variable "name", containing the name of the backend followed by a slash
2949 ("/") then the name of the server. This can be used when a server is
2950 checked in multiple backends.
2951
2952 - a variable "node" containing the name of the haproxy node, as set in the
2953 global "node" variable, otherwise the system's hostname if unspecified.
2954
2955 - a variable "weight" indicating the weight of the server, a slash ("/")
2956 and the total weight of the farm (just counting usable servers). This
2957 helps to know if other servers are available to handle the load when this
2958 one fails.
2959
2960 - a variable "scur" indicating the current number of concurrent connections
2961 on the server, followed by a slash ("/") then the total number of
2962 connections on all servers of the same backend.
2963
2964 - a variable "qcur" indicating the current number of requests in the
2965 server's queue.
2966
2967 Example of a header received by the application server :
2968 >>> X-Haproxy-Server-State: UP 2/3; name=bck/srv2; node=lb1; weight=1/2; \
2969 scur=13/22; qcur=0
2970
2971 See also : "option httpchk", "http-check disable-on-404"
2972
Willy Tarreauccbcc372012-12-27 12:37:57 +01002973http-request { allow | deny | tarpit | auth [realm <realm>] | redirect <rule> |
Willy Tarreauf4c43c12013-06-11 17:01:13 +02002974 add-header <name> <fmt> | set-header <name> <fmt> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02002975 del-header <name> | set-nice <nice> | set-log-level <level> |
Sasha Pachev218f0642014-06-16 12:05:59 -06002976 replace-header <name> <match-regex> <replace-fmt> |
2977 replace-value <name> <match-regex> <replace-fmt> |
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01002978 set-method <fmt> | set-path <fmt> | set-query <fmt> |
2979 set-uri <fmt> | set-tos <tos> | set-mark <mark> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02002980 add-acl(<file name>) <key fmt> |
2981 del-acl(<file name>) <key fmt> |
2982 del-map(<file name>) <key fmt> |
Baptiste Assmannbb7e86a2014-09-03 18:29:47 +02002983 set-map(<file name>) <key fmt> <value fmt> |
2984 { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02002985 }
Cyril Bontéf0c60612010-02-06 14:44:47 +01002986 [ { if | unless } <condition> ]
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002987 Access control for Layer 7 requests
2988
2989 May be used in sections: defaults | frontend | listen | backend
2990 no | yes | yes | yes
2991
Willy Tarreau20b0de52012-12-24 15:45:22 +01002992 The http-request statement defines a set of rules which apply to layer 7
2993 processing. The rules are evaluated in their declaration order when they are
2994 met in a frontend, listen or backend section. Any rule may optionally be
2995 followed by an ACL-based condition, in which case it will only be evaluated
2996 if the condition is true.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01002997
Willy Tarreau20b0de52012-12-24 15:45:22 +01002998 The first keyword is the rule's action. Currently supported actions include :
2999 - "allow" : this stops the evaluation of the rules and lets the request
3000 pass the check. No further "http-request" rules are evaluated.
3001
3002 - "deny" : this stops the evaluation of the rules and immediately rejects
3003 the request and emits an HTTP 403 error. No further "http-request" rules
3004 are evaluated.
3005
Willy Tarreauccbcc372012-12-27 12:37:57 +01003006 - "tarpit" : this stops the evaluation of the rules and immediately blocks
3007 the request without responding for a delay specified by "timeout tarpit"
3008 or "timeout connect" if the former is not set. After that delay, if the
3009 client is still connected, an HTTP error 500 is returned so that the
3010 client does not suspect it has been tarpitted. Logs will report the flags
3011 "PT". The goal of the tarpit rule is to slow down robots during an attack
3012 when they're limited on the number of concurrent requests. It can be very
3013 efficient against very dumb robots, and will significantly reduce the
3014 load on firewalls compared to a "deny" rule. But when facing "correctly"
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003015 developed robots, it can make things worse by forcing haproxy and the
Willy Tarreauccbcc372012-12-27 12:37:57 +01003016 front firewall to support insane number of concurrent connections.
3017
Willy Tarreau20b0de52012-12-24 15:45:22 +01003018 - "auth" : this stops the evaluation of the rules and immediately responds
3019 with an HTTP 401 or 407 error code to invite the user to present a valid
3020 user name and password. No further "http-request" rules are evaluated. An
3021 optional "realm" parameter is supported, it sets the authentication realm
3022 that is returned with the response (typically the application's name).
3023
Willy Tarreau81499eb2012-12-27 12:19:02 +01003024 - "redirect" : this performs an HTTP redirection based on a redirect rule.
3025 This is exactly the same as the "redirect" statement except that it
3026 inserts a redirect rule which can be processed in the middle of other
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01003027 "http-request" rules and that these rules use the "log-format" strings.
3028 See the "redirect" keyword for the rule's syntax.
Willy Tarreau81499eb2012-12-27 12:19:02 +01003029
Willy Tarreau20b0de52012-12-24 15:45:22 +01003030 - "add-header" appends an HTTP header field whose name is specified in
3031 <name> and whose value is defined by <fmt> which follows the log-format
3032 rules (see Custom Log Format in section 8.2.4). This is particularly
3033 useful to pass connection-specific information to the server (eg: the
3034 client's SSL certificate), or to combine several headers into one. This
3035 rule is not final, so it is possible to add other similar rules. Note
3036 that header addition is performed immediately, so one rule might reuse
3037 the resulting header from a previous rule.
3038
3039 - "set-header" does the same as "add-header" except that the header name
3040 is first removed if it existed. This is useful when passing security
3041 information to the server, where the header must not be manipulated by
Willy Tarreau85603282015-01-21 20:39:27 +01003042 external users. Note that the new value is computed before the removal so
3043 it is possible to concatenate a value to an existing header.
Willy Tarreau20b0de52012-12-24 15:45:22 +01003044
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003045 - "del-header" removes all HTTP header fields whose name is specified in
3046 <name>.
3047
Sasha Pachev218f0642014-06-16 12:05:59 -06003048 - "replace-header" matches the regular expression in all occurrences of
3049 header field <name> according to <match-regex>, and replaces them with
3050 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3051 and work like in <fmt> arguments in "add-header". The match is only
3052 case-sensitive. It is important to understand that this action only
3053 considers whole header lines, regardless of the number of values they
3054 may contain. This usage is suited to headers naturally containing commas
3055 in their value, such as If-Modified-Since and so on.
3056
3057 Example:
3058
3059 http-request replace-header Cookie foo=([^;]*);(.*) foo=\1;ip=%bi;\2
3060
3061 applied to:
3062
3063 Cookie: foo=foobar; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3064
3065 outputs:
3066
3067 Cookie: foo=foobar;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT;
3068
3069 assuming the backend IP is 192.168.1.20
3070
3071 - "replace-value" works like "replace-header" except that it matches the
3072 regex against every comma-delimited value of the header field <name>
3073 instead of the entire header. This is suited for all headers which are
3074 allowed to carry more than one value. An example could be the Accept
3075 header.
3076
3077 Example:
3078
3079 http-request replace-value X-Forwarded-For ^192\.168\.(.*)$ 172.16.\1
3080
3081 applied to:
3082
3083 X-Forwarded-For: 192.168.10.1, 192.168.13.24, 10.0.0.37
3084
3085 outputs:
3086
3087 X-Forwarded-For: 172.16.10.1, 172.16.13.24, 10.0.0.37
3088
Willy Tarreaua0dc23f2015-01-22 20:46:11 +01003089 - "set-method" rewrites the request method with the result of the
3090 evaluation of format string <fmt>. There should be very few valid reasons
3091 for having to do so as this is more likely to break something than to fix
3092 it.
3093
3094 - "set-path" rewrites the request path with the result of the evaluation of
3095 format string <fmt>. The query string, if any, is left intact. If a
3096 scheme and authority is found before the path, they are left intact as
3097 well. If the request doesn't have a path ("*"), this one is replaced with
3098 the format. This can be used to prepend a directory component in front of
3099 a path for example. See also "set-query" and "set-uri".
3100
3101 Example :
3102 # prepend the host name before the path
3103 http-request set-path /%[hdr(host)]%[path]
3104
3105 - "set-query" rewrites the request's query string which appears after the
3106 first question mark ("?") with the result of the evaluation of format
3107 string <fmt>. The part prior to the question mark is left intact. If the
3108 request doesn't contain a question mark and the new value is not empty,
3109 then one is added at the end of the URI, followed by the new value. If
3110 a question mark was present, it will never be removed even if the value
3111 is empty. This can be used to add or remove parameters from the query
3112 string. See also "set-query" and "set-uri".
3113
3114 Example :
3115 # replace "%3D" with "=" in the query string
3116 http-request set-query %[query,regsub(%3D,=,g)]
3117
3118 - "set-uri" rewrites the request URI with the result of the evaluation of
3119 format string <fmt>. The scheme, authority, path and query string are all
3120 replaced at once. This can be used to rewrite hosts in front of proxies,
3121 or to perform complex modifications to the URI such as moving parts
3122 between the path and the query string. See also "set-path" and
3123 "set-query".
3124
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003125 - "set-nice" sets the "nice" factor of the current request being processed.
3126 It only has effect against the other requests being processed at the same
3127 time. The default value is 0, unless altered by the "nice" setting on the
3128 "bind" line. The accepted range is -1024..1024. The higher the value, the
3129 nicest the request will be. Lower values will make the request more
3130 important than other ones. This can be useful to improve the speed of
3131 some requests, or lower the priority of non-important requests. Using
3132 this setting without prior experimentation can cause some major slowdown.
3133
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003134 - "set-log-level" is used to change the log level of the current request
3135 when a certain condition is met. Valid levels are the 8 syslog levels
3136 (see the "log" keyword) plus the special level "silent" which disables
3137 logging for this request. This rule is not final so the last matching
3138 rule wins. This rule can be useful to disable health checks coming from
3139 another equipment.
3140
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003141 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3142 the client to the value passed in <tos> on platforms which support this.
3143 This value represents the whole 8 bits of the IP TOS field, and can be
3144 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3145 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3146 bits are always 0. This can be used to adjust some routing behaviour on
3147 border routers based on some information from the request. See RFC 2474,
3148 2597, 3260 and 4594 for more information.
3149
Willy Tarreau51347ed2013-06-11 19:34:13 +02003150 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3151 client to the value passed in <mark> on platforms which support it. This
3152 value is an unsigned 32 bit value which can be matched by netfilter and
3153 by the routing table. It can be expressed both in decimal or hexadecimal
3154 format (prefixed by "0x"). This can be useful to force certain packets to
3155 take a different route (for example a cheaper network path for bulk
3156 downloads). This works on Linux kernels 2.6.32 and above and requires
3157 admin privileges.
3158
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003159 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3160 from a file (even a dummy empty file). The file name of the ACL to be
3161 updated is passed between parentheses. It takes one argument: <key fmt>,
3162 which follows log-format rules, to collect content of the new entry. It
3163 performs a lookup in the ACL before insertion, to avoid duplicated (or
3164 more) values. This lookup is done by a linear search and can be expensive
3165 with large lists! It is the equivalent of the "add acl" command from the
3166 stats socket, but can be triggered by an HTTP request.
3167
3168 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3169 from a file (even a dummy empty file). The file name of the ACL to be
3170 updated is passed between parentheses. It takes one argument: <key fmt>,
3171 which follows log-format rules, to collect content of the entry to delete.
3172 It is the equivalent of the "del acl" command from the stats socket, but
3173 can be triggered by an HTTP request.
3174
3175 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3176 from a file (even a dummy empty file). The file name of the MAP to be
3177 updated is passed between parentheses. It takes one argument: <key fmt>,
3178 which follows log-format rules, to collect content of the entry to delete.
3179 It takes one argument: "file name" It is the equivalent of the "del map"
3180 command from the stats socket, but can be triggered by an HTTP request.
3181
3182 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3183 from a file (even a dummy empty file). The file name of the MAP to be
3184 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3185 which follows log-format rules, used to collect MAP key, and <value fmt>,
3186 which follows log-format rules, used to collect content for the new entry.
3187 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3188 more) values. This lookup is done by a linear search and can be expensive
3189 with large lists! It is the equivalent of the "set map" command from the
3190 stats socket, but can be triggered by an HTTP request.
3191
Willy Tarreau09448f72014-06-25 18:12:15 +02003192 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
3193 enables tracking of sticky counters from current request. These rules
3194 do not stop evaluation and do not change default action. Three sets of
3195 counters may be simultaneously tracked by the same connection. The first
3196 "track-sc0" rule executed enables tracking of the counters of the
3197 specified table as the first set. The first "track-sc1" rule executed
3198 enables tracking of the counters of the specified table as the second
3199 set. The first "track-sc2" rule executed enables tracking of the
3200 counters of the specified table as the third set. It is a recommended
3201 practice to use the first set of counters for the per-frontend counters
3202 and the second set for the per-backend ones. But this is just a
3203 guideline, all may be used everywhere.
3204
3205 These actions take one or two arguments :
3206 <key> is mandatory, and is a sample expression rule as described
3207 in section 7.3. It describes what elements of the incoming
3208 request or connection will be analysed, extracted, combined,
3209 and used to select which table entry to update the counters.
3210
3211 <table> is an optional table to be used instead of the default one,
3212 which is the stick-table declared in the current proxy. All
3213 the counters for the matches and updates for the key will
3214 then be performed in that table until the session ends.
3215
3216 Once a "track-sc*" rule is executed, the key is looked up in the table
3217 and if it is not found, an entry is allocated for it. Then a pointer to
3218 that entry is kept during all the session's life, and this entry's
3219 counters are updated as often as possible, every time the session's
3220 counters are updated, and also systematically when the session ends.
3221 Counters are only updated for events that happen after the tracking has
3222 been started. As an exception, connection counters and request counters
3223 are systematically updated so that they reflect useful information.
3224
3225 If the entry tracks concurrent connection counters, one connection is
3226 counted for as long as the entry is tracked, and the entry will not
3227 expire during that time. Tracking counters also provides a performance
3228 advantage over just checking the keys, because only one table lookup is
3229 performed for all ACL checks that make use of it.
3230
Willy Tarreau20b0de52012-12-24 15:45:22 +01003231 There is no limit to the number of http-request statements per instance.
3232
3233 It is important to know that http-request rules are processed very early in
3234 the HTTP processing, just after "block" rules and before "reqdel" or "reqrep"
3235 rules. That way, headers added by "add-header"/"set-header" are visible by
3236 almost all further ACL rules.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003237
3238 Example:
Cyril Bonté78caf842010-03-10 22:41:43 +01003239 acl nagios src 192.168.129.3
3240 acl local_net src 192.168.0.0/16
3241 acl auth_ok http_auth(L1)
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003242
Cyril Bonté78caf842010-03-10 22:41:43 +01003243 http-request allow if nagios
3244 http-request allow if local_net auth_ok
3245 http-request auth realm Gimme if local_net auth_ok
3246 http-request deny
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003247
Cyril Bonté78caf842010-03-10 22:41:43 +01003248 Example:
3249 acl auth_ok http_auth_group(L1) G1
Cyril Bonté78caf842010-03-10 22:41:43 +01003250 http-request auth unless auth_ok
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +01003251
Willy Tarreau20b0de52012-12-24 15:45:22 +01003252 Example:
3253 http-request set-header X-Haproxy-Current-Date %T
3254 http-request set-header X-SSL %[ssl_fc]
3255 http-request set-header X-SSL-Session_ID %[ssl_fc_session_id]
3256 http-request set-header X-SSL-Client-Verify %[ssl_c_verify]
3257 http-request set-header X-SSL-Client-DN %{+Q}[ssl_c_s_dn]
3258 http-request set-header X-SSL-Client-CN %{+Q}[ssl_c_s_dn(cn)]
3259 http-request set-header X-SSL-Issuer %{+Q}[ssl_c_i_dn]
3260 http-request set-header X-SSL-Client-NotBefore %{+Q}[ssl_c_notbefore]
3261 http-request set-header X-SSL-Client-NotAfter %{+Q}[ssl_c_notafter]
3262
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003263 Example:
3264 acl key req.hdr(X-Add-Acl-Key) -m found
3265 acl add path /addacl
3266 acl del path /delacl
3267
3268 acl myhost hdr(Host) -f myhost.lst
3269
3270 http-request add-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key add
3271 http-request del-acl(myhost.lst) %[req.hdr(X-Add-Acl-Key)] if key del
3272
3273 Example:
3274 acl value req.hdr(X-Value) -m found
3275 acl setmap path /setmap
3276 acl delmap path /delmap
3277
3278 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3279
3280 http-request set-map(map.lst) %[src] %[req.hdr(X-Value)] if setmap value
3281 http-request del-map(map.lst) %[src] if delmap
3282
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02003283 See also : "stats http-request", section 3.4 about userlists and section 7
3284 about ACL usage.
Willy Tarreauef781042010-01-27 11:53:01 +01003285
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003286http-response { allow | deny | add-header <name> <fmt> | set-nice <nice> |
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003287 set-header <name> <fmt> | del-header <name> |
Sasha Pachev218f0642014-06-16 12:05:59 -06003288 replace-header <name> <regex-match> <replace-fmt> |
3289 replace-value <name> <regex-match> <replace-fmt> |
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003290 set-log-level <level> | set-mark <mark> | set-tos <tos> |
3291 add-acl(<file name>) <key fmt> |
3292 del-acl(<file name>) <key fmt> |
3293 del-map(<file name>) <key fmt> |
3294 set-map(<file name>) <key fmt> <value fmt>
3295 }
Lukas Tribus2dd1d1a2013-06-19 23:34:41 +02003296 [ { if | unless } <condition> ]
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003297 Access control for Layer 7 responses
3298
3299 May be used in sections: defaults | frontend | listen | backend
3300 no | yes | yes | yes
3301
3302 The http-response statement defines a set of rules which apply to layer 7
3303 processing. The rules are evaluated in their declaration order when they are
3304 met in a frontend, listen or backend section. Any rule may optionally be
3305 followed by an ACL-based condition, in which case it will only be evaluated
3306 if the condition is true. Since these rules apply on responses, the backend
3307 rules are applied first, followed by the frontend's rules.
3308
3309 The first keyword is the rule's action. Currently supported actions include :
3310 - "allow" : this stops the evaluation of the rules and lets the response
3311 pass the check. No further "http-response" rules are evaluated for the
3312 current section.
3313
3314 - "deny" : this stops the evaluation of the rules and immediately rejects
3315 the response and emits an HTTP 502 error. No further "http-response"
3316 rules are evaluated.
3317
3318 - "add-header" appends an HTTP header field whose name is specified in
3319 <name> and whose value is defined by <fmt> which follows the log-format
3320 rules (see Custom Log Format in section 8.2.4). This may be used to send
3321 a cookie to a client for example, or to pass some internal information.
3322 This rule is not final, so it is possible to add other similar rules.
3323 Note that header addition is performed immediately, so one rule might
3324 reuse the resulting header from a previous rule.
3325
3326 - "set-header" does the same as "add-header" except that the header name
3327 is first removed if it existed. This is useful when passing security
3328 information to the server, where the header must not be manipulated by
3329 external users.
3330
Thierry FOURNIERdad3d1d2014-04-22 18:07:25 +02003331 - "del-header" removes all HTTP header fields whose name is specified in
3332 <name>.
3333
Sasha Pachev218f0642014-06-16 12:05:59 -06003334 - "replace-header" matches the regular expression in all occurrences of
3335 header field <name> according to <match-regex>, and replaces them with
3336 the <replace-fmt> argument. Format characters are allowed in replace-fmt
3337 and work like in <fmt> arguments in "add-header". The match is only
3338 case-sensitive. It is important to understand that this action only
3339 considers whole header lines, regardless of the number of values they
3340 may contain. This usage is suited to headers naturally containing commas
3341 in their value, such as Set-Cookie, Expires and so on.
3342
3343 Example:
3344
3345 http-response replace-header Set-Cookie (C=[^;]*);(.*) \1;ip=%bi;\2
3346
3347 applied to:
3348
3349 Set-Cookie: C=1; expires=Tue, 14-Jun-2016 01:40:45 GMT
3350
3351 outputs:
3352
3353 Set-Cookie: C=1;ip=192.168.1.20; expires=Tue, 14-Jun-2016 01:40:45 GMT
3354
3355 assuming the backend IP is 192.168.1.20.
3356
3357 - "replace-value" works like "replace-header" except that it matches the
3358 regex against every comma-delimited value of the header field <name>
3359 instead of the entire header. This is suited for all headers which are
3360 allowed to carry more than one value. An example could be the Accept
3361 header.
3362
3363 Example:
3364
3365 http-response replace-value Cache-control ^public$ private
3366
3367 applied to:
3368
3369 Cache-Control: max-age=3600, public
3370
3371 outputs:
3372
3373 Cache-Control: max-age=3600, private
3374
Willy Tarreauf4c43c12013-06-11 17:01:13 +02003375 - "set-nice" sets the "nice" factor of the current request being processed.
3376 It only has effect against the other requests being processed at the same
3377 time. The default value is 0, unless altered by the "nice" setting on the
3378 "bind" line. The accepted range is -1024..1024. The higher the value, the
3379 nicest the request will be. Lower values will make the request more
3380 important than other ones. This can be useful to improve the speed of
3381 some requests, or lower the priority of non-important requests. Using
3382 this setting without prior experimentation can cause some major slowdown.
3383
Willy Tarreau9a355ec2013-06-11 17:45:46 +02003384 - "set-log-level" is used to change the log level of the current request
3385 when a certain condition is met. Valid levels are the 8 syslog levels
3386 (see the "log" keyword) plus the special level "silent" which disables
3387 logging for this request. This rule is not final so the last matching
3388 rule wins. This rule can be useful to disable health checks coming from
3389 another equipment.
3390
Willy Tarreau42cf39e2013-06-11 18:51:32 +02003391 - "set-tos" is used to set the TOS or DSCP field value of packets sent to
3392 the client to the value passed in <tos> on platforms which support this.
3393 This value represents the whole 8 bits of the IP TOS field, and can be
3394 expressed both in decimal or hexadecimal format (prefixed by "0x"). Note
3395 that only the 6 higher bits are used in DSCP or TOS, and the two lower
3396 bits are always 0. This can be used to adjust some routing behaviour on
3397 border routers based on some information from the request. See RFC 2474,
3398 2597, 3260 and 4594 for more information.
3399
Willy Tarreau51347ed2013-06-11 19:34:13 +02003400 - "set-mark" is used to set the Netfilter MARK on all packets sent to the
3401 client to the value passed in <mark> on platforms which support it. This
3402 value is an unsigned 32 bit value which can be matched by netfilter and
3403 by the routing table. It can be expressed both in decimal or hexadecimal
3404 format (prefixed by "0x"). This can be useful to force certain packets to
3405 take a different route (for example a cheaper network path for bulk
3406 downloads). This works on Linux kernels 2.6.32 and above and requires
3407 admin privileges.
3408
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003409 - "add-acl" is used to add a new entry into an ACL. The ACL must be loaded
3410 from a file (even a dummy empty file). The file name of the ACL to be
3411 updated is passed between parentheses. It takes one argument: <key fmt>,
3412 which follows log-format rules, to collect content of the new entry. It
3413 performs a lookup in the ACL before insertion, to avoid duplicated (or
3414 more) values. This lookup is done by a linear search and can be expensive
3415 with large lists! It is the equivalent of the "add acl" command from the
3416 stats socket, but can be triggered by an HTTP response.
3417
3418 - "del-acl" is used to delete an entry from an ACL. The ACL must be loaded
3419 from a file (even a dummy empty file). The file name of the ACL to be
3420 updated is passed between parentheses. It takes one argument: <key fmt>,
3421 which follows log-format rules, to collect content of the entry to delete.
3422 It is the equivalent of the "del acl" command from the stats socket, but
3423 can be triggered by an HTTP response.
3424
3425 - "del-map" is used to delete an entry from a MAP. The MAP must be loaded
3426 from a file (even a dummy empty file). The file name of the MAP to be
3427 updated is passed between parentheses. It takes one argument: <key fmt>,
3428 which follows log-format rules, to collect content of the entry to delete.
3429 It takes one argument: "file name" It is the equivalent of the "del map"
3430 command from the stats socket, but can be triggered by an HTTP response.
3431
3432 - "set-map" is used to add a new entry into a MAP. The MAP must be loaded
3433 from a file (even a dummy empty file). The file name of the MAP to be
3434 updated is passed between parentheses. It takes 2 arguments: <key fmt>,
3435 which follows log-format rules, used to collect MAP key, and <value fmt>,
3436 which follows log-format rules, used to collect content for the new entry.
3437 It performs a lookup in the MAP before insertion, to avoid duplicated (or
3438 more) values. This lookup is done by a linear search and can be expensive
3439 with large lists! It is the equivalent of the "set map" command from the
3440 stats socket, but can be triggered by an HTTP response.
3441
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003442 There is no limit to the number of http-response statements per instance.
3443
Godbach09250262013-07-02 01:19:15 +08003444 It is important to know that http-response rules are processed very early in
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003445 the HTTP processing, before "reqdel" or "reqrep" rules. That way, headers
3446 added by "add-header"/"set-header" are visible by almost all further ACL
3447 rules.
3448
Baptiste Assmannfabcbe02014-04-24 22:16:59 +02003449 Example:
3450 acl key_acl res.hdr(X-Acl-Key) -m found
3451
3452 acl myhost hdr(Host) -f myhost.lst
3453
3454 http-response add-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
3455 http-response del-acl(myhost.lst) %[res.hdr(X-Acl-Key)] if key_acl
3456
3457 Example:
3458 acl value res.hdr(X-Value) -m found
3459
3460 use_backend bk_appli if { hdr(Host),map_str(map.lst) -m found }
3461
3462 http-response set-map(map.lst) %[src] %[res.hdr(X-Value)] if value
3463 http-response del-map(map.lst) %[src] if ! value
3464
Willy Tarreaue365c0b2013-06-11 16:06:12 +02003465 See also : "http-request", section 3.4 about userlists and section 7 about
3466 ACL usage.
3467
Baptiste Assmann5ecb77f2013-10-06 23:24:13 +02003468
Mark Lamourinec2247f02012-01-04 13:02:01 -05003469http-send-name-header [<header>]
3470 Add the server name to a request. Use the header string given by <header>
3471
3472 May be used in sections: defaults | frontend | listen | backend
3473 yes | no | yes | yes
3474
3475 Arguments :
3476
3477 <header> The header string to use to send the server name
3478
3479 The "http-send-name-header" statement causes the name of the target
3480 server to be added to the headers of an HTTP request. The name
3481 is added with the header string proved.
3482
3483 See also : "server"
3484
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003485id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02003486 Set a persistent ID to a proxy.
3487 May be used in sections : defaults | frontend | listen | backend
3488 no | yes | yes | yes
3489 Arguments : none
3490
3491 Set a persistent ID for the proxy. This ID must be unique and positive.
3492 An unused ID will automatically be assigned if unset. The first assigned
3493 value will be 1. This ID is currently only returned in statistics.
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +01003494
3495
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003496ignore-persist { if | unless } <condition>
3497 Declare a condition to ignore persistence
3498 May be used in sections: defaults | frontend | listen | backend
3499 no | yes | yes | yes
3500
3501 By default, when cookie persistence is enabled, every requests containing
3502 the cookie are unconditionally persistent (assuming the target server is up
3503 and running).
3504
3505 The "ignore-persist" statement allows one to declare various ACL-based
3506 conditions which, when met, will cause a request to ignore persistence.
3507 This is sometimes useful to load balance requests for static files, which
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03003508 often don't require persistence. This can also be used to fully disable
Cyril Bonté0d4bf012010-04-25 23:21:46 +02003509 persistence for a specific User-Agent (for example, some web crawler bots).
3510
3511 Combined with "appsession", it can also help reduce HAProxy memory usage, as
3512 the appsession table won't grow if persistence is ignored.
3513
3514 The persistence is ignored when an "if" condition is met, or unless an
3515 "unless" condition is met.
3516
3517 See also : "force-persist", "cookie", and section 7 about ACL usage.
3518
3519
Willy Tarreau2769aa02007-12-27 18:26:09 +01003520log global
Willy Tarreau18324f52014-06-27 18:10:07 +02003521log <address> [len <length>] <facility> [<level> [<minlevel>]]
William Lallemand0f99e342011-10-12 17:50:54 +02003522no log
Willy Tarreau2769aa02007-12-27 18:26:09 +01003523 Enable per-instance logging of events and traffic.
3524 May be used in sections : defaults | frontend | listen | backend
3525 yes | yes | yes | yes
William Lallemand0f99e342011-10-12 17:50:54 +02003526
3527 Prefix :
3528 no should be used when the logger list must be flushed. For example,
3529 if you don't want to inherit from the default logger list. This
3530 prefix does not allow arguments.
3531
Willy Tarreau2769aa02007-12-27 18:26:09 +01003532 Arguments :
3533 global should be used when the instance's logging parameters are the
3534 same as the global ones. This is the most common usage. "global"
3535 replaces <address>, <facility> and <level> with those of the log
3536 entries found in the "global" section. Only one "log global"
3537 statement may be used per instance, and this form takes no other
3538 parameter.
3539
3540 <address> indicates where to send the logs. It takes the same format as
3541 for the "global" section's logs, and can be one of :
3542
3543 - An IPv4 address optionally followed by a colon (':') and a UDP
3544 port. If no port is specified, 514 is used by default (the
3545 standard syslog port).
3546
David du Colombier24bb5f52011-03-17 10:40:23 +01003547 - An IPv6 address followed by a colon (':') and optionally a UDP
3548 port. If no port is specified, 514 is used by default (the
3549 standard syslog port).
3550
Willy Tarreau2769aa02007-12-27 18:26:09 +01003551 - A filesystem path to a UNIX domain socket, keeping in mind
3552 considerations for chroot (be sure the path is accessible
3553 inside the chroot) and uid/gid (be sure the path is
3554 appropriately writeable).
3555
Willy Tarreaudad36a32013-03-11 01:20:04 +01003556 Any part of the address string may reference any number of
3557 environment variables by preceding their name with a dollar
3558 sign ('$') and optionally enclosing them with braces ('{}'),
3559 similarly to what is done in Bourne shell.
3560
Willy Tarreau18324f52014-06-27 18:10:07 +02003561 <length> is an optional maximum line length. Log lines larger than this
3562 value will be truncated before being sent. The reason is that
3563 syslog servers act differently on log line length. All servers
3564 support the default value of 1024, but some servers simply drop
3565 larger lines while others do log them. If a server supports long
3566 lines, it may make sense to set this value here in order to avoid
3567 truncating long lines. Similarly, if a server drops long lines,
3568 it is preferable to truncate them before sending them. Accepted
3569 values are 80 to 65535 inclusive. The default value of 1024 is
3570 generally fine for all standard usages. Some specific cases of
3571 long captures or JSON-formated logs may require larger values.
3572
Willy Tarreau2769aa02007-12-27 18:26:09 +01003573 <facility> must be one of the 24 standard syslog facilities :
3574
3575 kern user mail daemon auth syslog lpr news
3576 uucp cron auth2 ftp ntp audit alert cron2
3577 local0 local1 local2 local3 local4 local5 local6 local7
3578
3579 <level> is optional and can be specified to filter outgoing messages. By
3580 default, all messages are sent. If a level is specified, only
3581 messages with a severity at least as important as this level
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003582 will be sent. An optional minimum level can be specified. If it
3583 is set, logs emitted with a more severe level than this one will
3584 be capped to this level. This is used to avoid sending "emerg"
3585 messages on all terminals on some default syslog configurations.
3586 Eight levels are known :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003587
3588 emerg alert crit err warning notice info debug
3589
William Lallemand0f99e342011-10-12 17:50:54 +02003590 It is important to keep in mind that it is the frontend which decides what to
3591 log from a connection, and that in case of content switching, the log entries
3592 from the backend will be ignored. Connections are logged at level "info".
Willy Tarreaucc6c8912009-02-22 10:53:55 +01003593
3594 However, backend log declaration define how and where servers status changes
3595 will be logged. Level "notice" will be used to indicate a server going up,
3596 "warning" will be used for termination signals and definitive service
3597 termination, and "alert" will be used for when a server goes down.
3598
3599 Note : According to RFC3164, messages are truncated to 1024 bytes before
3600 being emitted.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003601
3602 Example :
3603 log global
Willy Tarreauf7edefa2009-05-10 17:20:05 +02003604 log 127.0.0.1:514 local0 notice # only send important events
3605 log 127.0.0.1:514 local0 notice notice # same but limit output level
Willy Tarreaudad36a32013-03-11 01:20:04 +01003606 log ${LOCAL_SYSLOG}:514 local0 notice # send to local server
3607
Willy Tarreau2769aa02007-12-27 18:26:09 +01003608
William Lallemand48940402012-01-30 16:47:22 +01003609log-format <string>
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01003610 Specifies the log format string to use for traffic logs
3611 May be used in sections: defaults | frontend | listen | backend
3612 yes | yes | yes | no
William Lallemand48940402012-01-30 16:47:22 +01003613
Willy Tarreaufb4e7ea2015-01-07 14:55:17 +01003614 This directive specifies the log format string that will be used for all logs
3615 resulting from traffic passing through the frontend using this line. If the
3616 directive is used in a defaults section, all subsequent frontends will use
3617 the same log format. Please see section 8.2.4 which covers the log format
3618 string in depth.
William Lallemand48940402012-01-30 16:47:22 +01003619
Willy Tarreau094af4e2015-01-07 15:03:42 +01003620log-tag <string>
3621 Specifies the log tag to use for all outgoing logs
3622 May be used in sections: defaults | frontend | listen | backend
3623 yes | yes | yes | yes
3624
3625 Sets the tag field in the syslog header to this string. It defaults to the
3626 log-tag set in the global section, otherwise the program name as launched
3627 from the command line, which usually is "haproxy". Sometimes it can be useful
3628 to differentiate between multiple processes running on the same host, or to
3629 differentiate customer instances running in the same process. In the backend,
3630 logs about servers up/down will use this tag. As a hint, it can be convenient
3631 to set a log-tag related to a hosted customer in a defaults section then put
3632 all the frontends and backends for that customer, then start another customer
3633 in a new defaults section. See also the global "log-tag" directive.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003634
Willy Tarreauc35362a2014-04-25 13:58:37 +02003635max-keep-alive-queue <value>
3636 Set the maximum server queue size for maintaining keep-alive connections
3637 May be used in sections: defaults | frontend | listen | backend
3638 yes | no | yes | yes
3639
3640 HTTP keep-alive tries to reuse the same server connection whenever possible,
3641 but sometimes it can be counter-productive, for example if a server has a lot
3642 of connections while other ones are idle. This is especially true for static
3643 servers.
3644
3645 The purpose of this setting is to set a threshold on the number of queued
3646 connections at which haproxy stops trying to reuse the same server and prefers
3647 to find another one. The default value, -1, means there is no limit. A value
3648 of zero means that keep-alive requests will never be queued. For very close
3649 servers which can be reached with a low latency and which are not sensible to
3650 breaking keep-alive, a low value is recommended (eg: local static server can
3651 use a value of 10 or less). For remote servers suffering from a high latency,
3652 higher values might be needed to cover for the latency and/or the cost of
3653 picking a different server.
3654
3655 Note that this has no impact on responses which are maintained to the same
3656 server consecutively to a 401 response. They will still go to the same server
3657 even if they have to be queued.
3658
3659 See also : "option http-server-close", "option prefer-last-server", server
3660 "maxconn" and cookie persistence.
3661
3662
Willy Tarreau2769aa02007-12-27 18:26:09 +01003663maxconn <conns>
3664 Fix the maximum number of concurrent connections on a frontend
3665 May be used in sections : defaults | frontend | listen | backend
3666 yes | yes | yes | no
3667 Arguments :
3668 <conns> is the maximum number of concurrent connections the frontend will
3669 accept to serve. Excess connections will be queued by the system
3670 in the socket's listen queue and will be served once a connection
3671 closes.
3672
3673 If the system supports it, it can be useful on big sites to raise this limit
3674 very high so that haproxy manages connection queues, instead of leaving the
3675 clients with unanswered connection attempts. This value should not exceed the
3676 global maxconn. Also, keep in mind that a connection contains two buffers
3677 of 8kB each, as well as some other data resulting in about 17 kB of RAM being
3678 consumed per established connection. That means that a medium system equipped
3679 with 1GB of RAM can withstand around 40000-50000 concurrent connections if
3680 properly tuned.
3681
3682 Also, when <conns> is set to large values, it is possible that the servers
3683 are not sized to accept such loads, and for this reason it is generally wise
3684 to assign them some reasonable connection limits.
3685
Vincent Bernat6341be52012-06-27 17:18:30 +02003686 By default, this value is set to 2000.
3687
Willy Tarreau2769aa02007-12-27 18:26:09 +01003688 See also : "server", global section's "maxconn", "fullconn"
3689
3690
3691mode { tcp|http|health }
3692 Set the running mode or protocol of the instance
3693 May be used in sections : defaults | frontend | listen | backend
3694 yes | yes | yes | yes
3695 Arguments :
3696 tcp The instance will work in pure TCP mode. A full-duplex connection
3697 will be established between clients and servers, and no layer 7
3698 examination will be performed. This is the default mode. It
3699 should be used for SSL, SSH, SMTP, ...
3700
3701 http The instance will work in HTTP mode. The client request will be
3702 analyzed in depth before connecting to any server. Any request
3703 which is not RFC-compliant will be rejected. Layer 7 filtering,
3704 processing and switching will be possible. This is the mode which
3705 brings HAProxy most of its value.
3706
3707 health The instance will work in "health" mode. It will just reply "OK"
Willy Tarreau82569f92012-09-27 23:48:56 +02003708 to incoming connections and close the connection. Alternatively,
3709 If the "httpchk" option is set, "HTTP/1.0 200 OK" will be sent
3710 instead. Nothing will be logged in either case. This mode is used
3711 to reply to external components health checks. This mode is
3712 deprecated and should not be used anymore as it is possible to do
3713 the same and even better by combining TCP or HTTP modes with the
3714 "monitor" keyword.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003715
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003716 When doing content switching, it is mandatory that the frontend and the
3717 backend are in the same mode (generally HTTP), otherwise the configuration
3718 will be refused.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003719
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003720 Example :
Willy Tarreau2769aa02007-12-27 18:26:09 +01003721 defaults http_instances
3722 mode http
3723
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003724 See also : "monitor", "monitor-net"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003725
Willy Tarreau0ba27502007-12-24 16:55:16 +01003726
Cyril Bontéf0c60612010-02-06 14:44:47 +01003727monitor fail { if | unless } <condition>
Willy Tarreau2769aa02007-12-27 18:26:09 +01003728 Add a condition to report a failure to a monitor HTTP request.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003729 May be used in sections : defaults | frontend | listen | backend
3730 no | yes | yes | no
Willy Tarreau0ba27502007-12-24 16:55:16 +01003731 Arguments :
3732 if <cond> the monitor request will fail if the condition is satisfied,
3733 and will succeed otherwise. The condition should describe a
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003734 combined test which must induce a failure if all conditions
Willy Tarreau0ba27502007-12-24 16:55:16 +01003735 are met, for instance a low number of servers both in a
3736 backend and its backup.
3737
3738 unless <cond> the monitor request will succeed only if the condition is
3739 satisfied, and will fail otherwise. Such a condition may be
3740 based on a test on the presence of a minimum number of active
3741 servers in a list of backends.
3742
3743 This statement adds a condition which can force the response to a monitor
3744 request to report a failure. By default, when an external component queries
3745 the URI dedicated to monitoring, a 200 response is returned. When one of the
3746 conditions above is met, haproxy will return 503 instead of 200. This is
3747 very useful to report a site failure to an external component which may base
3748 routing advertisements between multiple sites on the availability reported by
3749 haproxy. In this case, one would rely on an ACL involving the "nbsrv"
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003750 criterion. Note that "monitor fail" only works in HTTP mode. Both status
3751 messages may be tweaked using "errorfile" or "errorloc" if needed.
Willy Tarreau0ba27502007-12-24 16:55:16 +01003752
3753 Example:
3754 frontend www
Willy Tarreau2769aa02007-12-27 18:26:09 +01003755 mode http
Willy Tarreau0ba27502007-12-24 16:55:16 +01003756 acl site_dead nbsrv(dynamic) lt 2
3757 acl site_dead nbsrv(static) lt 2
3758 monitor-uri /site_alive
3759 monitor fail if site_dead
3760
Willy Tarreauae94d4d2011-05-11 16:28:49 +02003761 See also : "monitor-net", "monitor-uri", "errorfile", "errorloc"
Willy Tarreau2769aa02007-12-27 18:26:09 +01003762
3763
3764monitor-net <source>
3765 Declare a source network which is limited to monitor requests
3766 May be used in sections : defaults | frontend | listen | backend
3767 yes | yes | yes | no
3768 Arguments :
3769 <source> is the source IPv4 address or network which will only be able to
3770 get monitor responses to any request. It can be either an IPv4
3771 address, a host name, or an address followed by a slash ('/')
3772 followed by a mask.
3773
3774 In TCP mode, any connection coming from a source matching <source> will cause
3775 the connection to be immediately closed without any log. This allows another
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003776 equipment to probe the port and verify that it is still listening, without
Willy Tarreau2769aa02007-12-27 18:26:09 +01003777 forwarding the connection to a remote server.
3778
3779 In HTTP mode, a connection coming from a source matching <source> will be
3780 accepted, the following response will be sent without waiting for a request,
3781 then the connection will be closed : "HTTP/1.0 200 OK". This is normally
3782 enough for any front-end HTTP probe to detect that the service is UP and
Willy Tarreau82569f92012-09-27 23:48:56 +02003783 running without forwarding the request to a backend server. Note that this
3784 response is sent in raw format, without any transformation. This is important
3785 as it means that it will not be SSL-encrypted on SSL listeners.
Willy Tarreau2769aa02007-12-27 18:26:09 +01003786
Willy Tarreau82569f92012-09-27 23:48:56 +02003787 Monitor requests are processed very early, just after tcp-request connection
3788 ACLs which are the only ones able to block them. These connections are short
3789 lived and never wait for any data from the client. They cannot be logged, and
3790 it is the intended purpose. They are only used to report HAProxy's health to
3791 an upper component, nothing more. Please note that "monitor fail" rules do
3792 not apply to connections intercepted by "monitor-net".
Willy Tarreau2769aa02007-12-27 18:26:09 +01003793
Willy Tarreau95cd2832010-03-04 23:36:33 +01003794 Last, please note that only one "monitor-net" statement can be specified in
3795 a frontend. If more than one is found, only the last one will be considered.
Cyril Bonté108cf6e2012-04-21 23:30:29 +02003796
Willy Tarreau2769aa02007-12-27 18:26:09 +01003797 Example :
3798 # addresses .252 and .253 are just probing us.
3799 frontend www
3800 monitor-net 192.168.0.252/31
3801
3802 See also : "monitor fail", "monitor-uri"
3803
3804
3805monitor-uri <uri>
3806 Intercept a URI used by external components' monitor requests
3807 May be used in sections : defaults | frontend | listen | backend
3808 yes | yes | yes | no
3809 Arguments :
3810 <uri> is the exact URI which we want to intercept to return HAProxy's
3811 health status instead of forwarding the request.
3812
3813 When an HTTP request referencing <uri> will be received on a frontend,
3814 HAProxy will not forward it nor log it, but instead will return either
3815 "HTTP/1.0 200 OK" or "HTTP/1.0 503 Service unavailable", depending on failure
3816 conditions defined with "monitor fail". This is normally enough for any
3817 front-end HTTP probe to detect that the service is UP and running without
3818 forwarding the request to a backend server. Note that the HTTP method, the
3819 version and all headers are ignored, but the request must at least be valid
3820 at the HTTP level. This keyword may only be used with an HTTP-mode frontend.
3821
3822 Monitor requests are processed very early. It is not possible to block nor
3823 divert them using ACLs. They cannot be logged either, and it is the intended
3824 purpose. They are only used to report HAProxy's health to an upper component,
3825 nothing more. However, it is possible to add any number of conditions using
3826 "monitor fail" and ACLs so that the result can be adjusted to whatever check
3827 can be imagined (most often the number of available servers in a backend).
3828
3829 Example :
3830 # Use /haproxy_test to report haproxy's status
3831 frontend www
3832 mode http
3833 monitor-uri /haproxy_test
3834
3835 See also : "monitor fail", "monitor-net"
3836
Willy Tarreau0ba27502007-12-24 16:55:16 +01003837
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003838option abortonclose
3839no option abortonclose
3840 Enable or disable early dropping of aborted requests pending in queues.
3841 May be used in sections : defaults | frontend | listen | backend
3842 yes | no | yes | yes
3843 Arguments : none
3844
3845 In presence of very high loads, the servers will take some time to respond.
3846 The per-instance connection queue will inflate, and the response time will
3847 increase respective to the size of the queue times the average per-session
3848 response time. When clients will wait for more than a few seconds, they will
Willy Tarreau198a7442008-01-17 12:05:32 +01003849 often hit the "STOP" button on their browser, leaving a useless request in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003850 the queue, and slowing down other users, and the servers as well, because the
3851 request will eventually be served, then aborted at the first error
3852 encountered while delivering the response.
3853
3854 As there is no way to distinguish between a full STOP and a simple output
3855 close on the client side, HTTP agents should be conservative and consider
3856 that the client might only have closed its output channel while waiting for
3857 the response. However, this introduces risks of congestion when lots of users
3858 do the same, and is completely useless nowadays because probably no client at
3859 all will close the session while waiting for the response. Some HTTP agents
Willy Tarreaud72758d2010-01-12 10:42:19 +01003860 support this behaviour (Squid, Apache, HAProxy), and others do not (TUX, most
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003861 hardware-based load balancers). So the probability for a closed input channel
Willy Tarreau198a7442008-01-17 12:05:32 +01003862 to represent a user hitting the "STOP" button is close to 100%, and the risk
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003863 of being the single component to break rare but valid traffic is extremely
3864 low, which adds to the temptation to be able to abort a session early while
3865 still not served and not pollute the servers.
3866
3867 In HAProxy, the user can choose the desired behaviour using the option
3868 "abortonclose". By default (without the option) the behaviour is HTTP
3869 compliant and aborted requests will be served. But when the option is
3870 specified, a session with an incoming channel closed will be aborted while
3871 it is still possible, either pending in the queue for a connection slot, or
3872 during the connection establishment if the server has not yet acknowledged
3873 the connection request. This considerably reduces the queue size and the load
3874 on saturated servers when users are tempted to click on STOP, which in turn
Willy Tarreaud72758d2010-01-12 10:42:19 +01003875 reduces the response time for other users.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003876
3877 If this option has been enabled in a "defaults" section, it can be disabled
3878 in a specific instance by prepending the "no" keyword before it.
3879
3880 See also : "timeout queue" and server's "maxconn" and "maxqueue" parameters
3881
3882
Willy Tarreau4076a152009-04-02 15:18:36 +02003883option accept-invalid-http-request
3884no option accept-invalid-http-request
3885 Enable or disable relaxing of HTTP request parsing
3886 May be used in sections : defaults | frontend | listen | backend
3887 yes | yes | yes | no
3888 Arguments : none
3889
3890 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3891 means that invalid characters in header names are not permitted and cause an
3892 error to be returned to the client. This is the desired behaviour as such
3893 forbidden characters are essentially used to build attacks exploiting server
3894 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3895 server will emit invalid header names for whatever reason (configuration,
3896 implementation) and the issue will not be immediately fixed. In such a case,
3897 it is possible to relax HAProxy's header name parser to accept any character
Willy Tarreau422246e2012-01-07 23:54:13 +01003898 even if that does not make sense, by specifying this option. Similarly, the
3899 list of characters allowed to appear in a URI is well defined by RFC3986, and
3900 chars 0-31, 32 (space), 34 ('"'), 60 ('<'), 62 ('>'), 92 ('\'), 94 ('^'), 96
3901 ('`'), 123 ('{'), 124 ('|'), 125 ('}'), 127 (delete) and anything above are
3902 not allowed at all. Haproxy always blocks a number of them (0..32, 127). The
3903 remaining ones are blocked by default unless this option is enabled.
Willy Tarreau4076a152009-04-02 15:18:36 +02003904
3905 This option should never be enabled by default as it hides application bugs
3906 and open security breaches. It should only be deployed after a problem has
3907 been confirmed.
3908
3909 When this option is enabled, erroneous header names will still be accepted in
3910 requests, but the complete request will be captured in order to permit later
Willy Tarreau422246e2012-01-07 23:54:13 +01003911 analysis using the "show errors" request on the UNIX stats socket. Similarly,
3912 requests containing invalid chars in the URI part will be logged. Doing this
Willy Tarreau4076a152009-04-02 15:18:36 +02003913 also helps confirming that the issue has been solved.
3914
3915 If this option has been enabled in a "defaults" section, it can be disabled
3916 in a specific instance by prepending the "no" keyword before it.
3917
3918 See also : "option accept-invalid-http-response" and "show errors" on the
3919 stats socket.
3920
3921
3922option accept-invalid-http-response
3923no option accept-invalid-http-response
3924 Enable or disable relaxing of HTTP response parsing
3925 May be used in sections : defaults | frontend | listen | backend
3926 yes | no | yes | yes
3927 Arguments : none
3928
3929 By default, HAProxy complies with RFC2616 in terms of message parsing. This
3930 means that invalid characters in header names are not permitted and cause an
3931 error to be returned to the client. This is the desired behaviour as such
3932 forbidden characters are essentially used to build attacks exploiting server
3933 weaknesses, and bypass security filtering. Sometimes, a buggy browser or
3934 server will emit invalid header names for whatever reason (configuration,
3935 implementation) and the issue will not be immediately fixed. In such a case,
3936 it is possible to relax HAProxy's header name parser to accept any character
3937 even if that does not make sense, by specifying this option.
3938
3939 This option should never be enabled by default as it hides application bugs
3940 and open security breaches. It should only be deployed after a problem has
3941 been confirmed.
3942
3943 When this option is enabled, erroneous header names will still be accepted in
3944 responses, but the complete response will be captured in order to permit
3945 later analysis using the "show errors" request on the UNIX stats socket.
3946 Doing this also helps confirming that the issue has been solved.
3947
3948 If this option has been enabled in a "defaults" section, it can be disabled
3949 in a specific instance by prepending the "no" keyword before it.
3950
3951 See also : "option accept-invalid-http-request" and "show errors" on the
3952 stats socket.
3953
3954
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003955option allbackups
3956no option allbackups
3957 Use either all backup servers at a time or only the first one
3958 May be used in sections : defaults | frontend | listen | backend
3959 yes | no | yes | yes
3960 Arguments : none
3961
3962 By default, the first operational backup server gets all traffic when normal
3963 servers are all down. Sometimes, it may be preferred to use multiple backups
3964 at once, because one will not be enough. When "option allbackups" is enabled,
3965 the load balancing will be performed among all backup servers when all normal
3966 ones are unavailable. The same load balancing algorithm will be used and the
3967 servers' weights will be respected. Thus, there will not be any priority
3968 order between the backup servers anymore.
3969
3970 This option is mostly used with static server farms dedicated to return a
3971 "sorry" page when an application is completely offline.
3972
3973 If this option has been enabled in a "defaults" section, it can be disabled
3974 in a specific instance by prepending the "no" keyword before it.
3975
3976
3977option checkcache
3978no option checkcache
Godbach7056a352013-12-11 20:01:07 +08003979 Analyze all server responses and block responses with cacheable cookies
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003980 May be used in sections : defaults | frontend | listen | backend
3981 yes | no | yes | yes
3982 Arguments : none
3983
3984 Some high-level frameworks set application cookies everywhere and do not
3985 always let enough control to the developer to manage how the responses should
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003986 be cached. When a session cookie is returned on a cacheable object, there is a
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003987 high risk of session crossing or stealing between users traversing the same
3988 caches. In some situations, it is better to block the response than to let
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02003989 some sensitive session information go in the wild.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003990
3991 The option "checkcache" enables deep inspection of all server responses for
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01003992 strict compliance with HTTP specification in terms of cacheability. It
Willy Tarreau198a7442008-01-17 12:05:32 +01003993 carefully checks "Cache-control", "Pragma" and "Set-cookie" headers in server
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003994 response to check if there's a risk of caching a cookie on a client-side
3995 proxy. When this option is enabled, the only responses which can be delivered
Willy Tarreau198a7442008-01-17 12:05:32 +01003996 to the client are :
3997 - all those without "Set-Cookie" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01003998 - all those with a return code other than 200, 203, 206, 300, 301, 410,
Willy Tarreau198a7442008-01-17 12:05:32 +01003999 provided that the server has not set a "Cache-control: public" header ;
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004000 - all those that come from a POST request, provided that the server has not
4001 set a 'Cache-Control: public' header ;
4002 - those with a 'Pragma: no-cache' header
4003 - those with a 'Cache-control: private' header
4004 - those with a 'Cache-control: no-store' header
4005 - those with a 'Cache-control: max-age=0' header
4006 - those with a 'Cache-control: s-maxage=0' header
4007 - those with a 'Cache-control: no-cache' header
4008 - those with a 'Cache-control: no-cache="set-cookie"' header
4009 - those with a 'Cache-control: no-cache="set-cookie,' header
4010 (allowing other fields after set-cookie)
4011
4012 If a response doesn't respect these requirements, then it will be blocked
Willy Tarreau198a7442008-01-17 12:05:32 +01004013 just as if it was from an "rspdeny" filter, with an "HTTP 502 bad gateway".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004014 The session state shows "PH--" meaning that the proxy blocked the response
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004015 during headers processing. Additionally, an alert will be sent in the logs so
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004016 that admins are informed that there's something to be fixed.
4017
4018 Due to the high impact on the application, the application should be tested
4019 in depth with the option enabled before going to production. It is also a
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004020 good practice to always activate it during tests, even if it is not used in
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004021 production, as it will report potentially dangerous application behaviours.
4022
4023 If this option has been enabled in a "defaults" section, it can be disabled
4024 in a specific instance by prepending the "no" keyword before it.
4025
4026
4027option clitcpka
4028no option clitcpka
4029 Enable or disable the sending of TCP keepalive packets on the client side
4030 May be used in sections : defaults | frontend | listen | backend
4031 yes | yes | yes | no
4032 Arguments : none
4033
4034 When there is a firewall or any session-aware component between a client and
4035 a server, and when the protocol involves very long sessions with long idle
4036 periods (eg: remote desktops), there is a risk that one of the intermediate
4037 components decides to expire a session which has remained idle for too long.
4038
4039 Enabling socket-level TCP keep-alives makes the system regularly send packets
4040 to the other end of the connection, leaving it active. The delay between
4041 keep-alive probes is controlled by the system only and depends both on the
4042 operating system and its tuning parameters.
4043
4044 It is important to understand that keep-alive packets are neither emitted nor
4045 received at the application level. It is only the network stacks which sees
4046 them. For this reason, even if one side of the proxy already uses keep-alives
4047 to maintain its connection alive, those keep-alive packets will not be
4048 forwarded to the other side of the proxy.
4049
4050 Please note that this has nothing to do with HTTP keep-alive.
4051
4052 Using option "clitcpka" enables the emission of TCP keep-alive probes on the
4053 client side of a connection, which should help when session expirations are
4054 noticed between HAProxy and a client.
4055
4056 If this option has been enabled in a "defaults" section, it can be disabled
4057 in a specific instance by prepending the "no" keyword before it.
4058
4059 See also : "option srvtcpka", "option tcpka"
4060
4061
Willy Tarreau0ba27502007-12-24 16:55:16 +01004062option contstats
4063 Enable continuous traffic statistics updates
4064 May be used in sections : defaults | frontend | listen | backend
4065 yes | yes | yes | no
4066 Arguments : none
4067
4068 By default, counters used for statistics calculation are incremented
4069 only when a session finishes. It works quite well when serving small
4070 objects, but with big ones (for example large images or archives) or
4071 with A/V streaming, a graph generated from haproxy counters looks like
4072 a hedgehog. With this option enabled counters get incremented continuously,
4073 during a whole session. Recounting touches a hotpath directly so
4074 it is not enabled by default, as it has small performance impact (~0.5%).
4075
4076
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004077option dontlog-normal
4078no option dontlog-normal
4079 Enable or disable logging of normal, successful connections
4080 May be used in sections : defaults | frontend | listen | backend
4081 yes | yes | yes | no
4082 Arguments : none
4083
4084 There are large sites dealing with several thousand connections per second
4085 and for which logging is a major pain. Some of them are even forced to turn
4086 logs off and cannot debug production issues. Setting this option ensures that
4087 normal connections, those which experience no error, no timeout, no retry nor
4088 redispatch, will not be logged. This leaves disk space for anomalies. In HTTP
4089 mode, the response status code is checked and return codes 5xx will still be
4090 logged.
4091
4092 It is strongly discouraged to use this option as most of the time, the key to
4093 complex issues is in the normal logs which will not be logged here. If you
4094 need to separate logs, see the "log-separate-errors" option instead.
4095
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004096 See also : "log", "dontlognull", "log-separate-errors" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004097 logging.
4098
4099
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004100option dontlognull
4101no option dontlognull
4102 Enable or disable logging of null connections
4103 May be used in sections : defaults | frontend | listen | backend
4104 yes | yes | yes | no
4105 Arguments : none
4106
4107 In certain environments, there are components which will regularly connect to
4108 various systems to ensure that they are still alive. It can be the case from
4109 another load balancer as well as from monitoring systems. By default, even a
4110 simple port probe or scan will produce a log. If those connections pollute
4111 the logs too much, it is possible to enable option "dontlognull" to indicate
4112 that a connection on which no data has been transferred will not be logged,
4113 which typically corresponds to those probes.
4114
4115 It is generally recommended not to use this option in uncontrolled
4116 environments (eg: internet), otherwise scans and other malicious activities
4117 would not be logged.
4118
4119 If this option has been enabled in a "defaults" section, it can be disabled
4120 in a specific instance by prepending the "no" keyword before it.
4121
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004122 See also : "log", "monitor-net", "monitor-uri" and section 8 about logging.
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004123
4124
4125option forceclose
4126no option forceclose
4127 Enable or disable active connection closing after response is transferred.
4128 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaua31e5df2009-12-30 01:10:35 +01004129 yes | yes | yes | yes
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004130 Arguments : none
4131
4132 Some HTTP servers do not necessarily close the connections when they receive
4133 the "Connection: close" set by "option httpclose", and if the client does not
4134 close either, then the connection remains open till the timeout expires. This
4135 causes high number of simultaneous connections on the servers and shows high
4136 global session times in the logs.
4137
4138 When this happens, it is possible to use "option forceclose". It will
Willy Tarreau82eeaf22009-12-29 12:09:05 +01004139 actively close the outgoing server channel as soon as the server has finished
Cyril Bonté653dcd62014-02-20 00:13:15 +01004140 to respond and release some resources earlier than with "option httpclose".
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004141
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004142 This option may also be combined with "option http-pretend-keepalive", which
4143 will disable sending of the "Connection: close" header, but will still cause
4144 the connection to be closed once the whole response is received.
4145
Cyril Bonté653dcd62014-02-20 00:13:15 +01004146 This option disables and replaces any previous "option httpclose", "option
4147 http-server-close", "option http-keep-alive", or "option http-tunnel".
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004148
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004149 If this option has been enabled in a "defaults" section, it can be disabled
4150 in a specific instance by prepending the "no" keyword before it.
4151
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004152 See also : "option httpclose" and "option http-pretend-keepalive"
Willy Tarreaubf1f8162007-12-28 17:42:56 +01004153
4154
Willy Tarreau87cf5142011-08-19 22:57:24 +02004155option forwardfor [ except <network> ] [ header <name> ] [ if-none ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004156 Enable insertion of the X-Forwarded-For header to requests sent to servers
4157 May be used in sections : defaults | frontend | listen | backend
4158 yes | yes | yes | yes
4159 Arguments :
4160 <network> is an optional argument used to disable this option for sources
4161 matching <network>
Ross Westaf72a1d2008-08-03 10:51:45 +02004162 <name> an optional argument to specify a different "X-Forwarded-For"
Willy Tarreaud72758d2010-01-12 10:42:19 +01004163 header name.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004164
4165 Since HAProxy works in reverse-proxy mode, the servers see its IP address as
4166 their client address. This is sometimes annoying when the client's IP address
4167 is expected in server logs. To solve this problem, the well-known HTTP header
4168 "X-Forwarded-For" may be added by HAProxy to all requests sent to the server.
4169 This header contains a value representing the client's IP address. Since this
4170 header is always appended at the end of the existing header list, the server
4171 must be configured to always use the last occurrence of this header only. See
Ross Westaf72a1d2008-08-03 10:51:45 +02004172 the server's manual to find how to enable use of this standard header. Note
4173 that only the last occurrence of the header must be used, since it is really
4174 possible that the client has already brought one.
4175
Willy Tarreaud72758d2010-01-12 10:42:19 +01004176 The keyword "header" may be used to supply a different header name to replace
Ross Westaf72a1d2008-08-03 10:51:45 +02004177 the default "X-Forwarded-For". This can be useful where you might already
Willy Tarreaud72758d2010-01-12 10:42:19 +01004178 have a "X-Forwarded-For" header from a different application (eg: stunnel),
4179 and you need preserve it. Also if your backend server doesn't use the
Ross Westaf72a1d2008-08-03 10:51:45 +02004180 "X-Forwarded-For" header and requires different one (eg: Zeus Web Servers
4181 require "X-Cluster-Client-IP").
Willy Tarreauc27debf2008-01-06 08:57:02 +01004182
4183 Sometimes, a same HAProxy instance may be shared between a direct client
4184 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4185 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4186 header for a known source address or network by adding the "except" keyword
4187 followed by the network address. In this case, any source IP matching the
4188 network will not cause an addition of this header. Most common uses are with
4189 private networks or 127.0.0.1.
4190
Willy Tarreau87cf5142011-08-19 22:57:24 +02004191 Alternatively, the keyword "if-none" states that the header will only be
4192 added if it is not present. This should only be used in perfectly trusted
4193 environment, as this might cause a security issue if headers reaching haproxy
4194 are under the control of the end-user.
4195
Willy Tarreauc27debf2008-01-06 08:57:02 +01004196 This option may be specified either in the frontend or in the backend. If at
Ross Westaf72a1d2008-08-03 10:51:45 +02004197 least one of them uses it, the header will be added. Note that the backend's
4198 setting of the header subargument takes precedence over the frontend's if
Willy Tarreau87cf5142011-08-19 22:57:24 +02004199 both are defined. In the case of the "if-none" argument, if at least one of
4200 the frontend or the backend does not specify it, it wants the addition to be
4201 mandatory, so it wins.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004202
Ross Westaf72a1d2008-08-03 10:51:45 +02004203 Examples :
Willy Tarreauc27debf2008-01-06 08:57:02 +01004204 # Public HTTP address also used by stunnel on the same machine
4205 frontend www
4206 mode http
4207 option forwardfor except 127.0.0.1 # stunnel already adds the header
4208
Ross Westaf72a1d2008-08-03 10:51:45 +02004209 # Those servers want the IP Address in X-Client
4210 backend www
4211 mode http
4212 option forwardfor header X-Client
4213
Willy Tarreau87cf5142011-08-19 22:57:24 +02004214 See also : "option httpclose", "option http-server-close",
Willy Tarreau70dffda2014-01-30 03:07:23 +01004215 "option forceclose", "option http-keep-alive"
Willy Tarreauc27debf2008-01-06 08:57:02 +01004216
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004217
Willy Tarreau16bfb022010-01-16 19:48:41 +01004218option http-keep-alive
4219no option http-keep-alive
4220 Enable or disable HTTP keep-alive from client to server
4221 May be used in sections : defaults | frontend | listen | backend
4222 yes | yes | yes | yes
4223 Arguments : none
4224
Willy Tarreau70dffda2014-01-30 03:07:23 +01004225 By default HAProxy operates in keep-alive mode with regards to persistent
4226 connections: for each connection it processes each request and response, and
4227 leaves the connection idle on both sides between the end of a response and the
4228 start of a new request. This mode may be changed by several options such as
4229 "option http-server-close", "option forceclose", "option httpclose" or
4230 "option http-tunnel". This option allows to set back the keep-alive mode,
4231 which can be useful when another mode was used in a defaults section.
4232
4233 Setting "option http-keep-alive" enables HTTP keep-alive mode on the client-
4234 and server- sides. This provides the lowest latency on the client side (slow
Willy Tarreau16bfb022010-01-16 19:48:41 +01004235 network) and the fastest session reuse on the server side at the expense
4236 of maintaining idle connections to the servers. In general, it is possible
4237 with this option to achieve approximately twice the request rate that the
4238 "http-server-close" option achieves on small objects. There are mainly two
4239 situations where this option may be useful :
4240
4241 - when the server is non-HTTP compliant and authenticates the connection
4242 instead of requests (eg: NTLM authentication)
4243
4244 - when the cost of establishing the connection to the server is significant
4245 compared to the cost of retrieving the associated object from the server.
4246
4247 This last case can happen when the server is a fast static server of cache.
4248 In this case, the server will need to be properly tuned to support high enough
4249 connection counts because connections will last until the client sends another
4250 request.
4251
4252 If the client request has to go to another backend or another server due to
4253 content switching or the load balancing algorithm, the idle connection will
Willy Tarreau9420b122013-12-15 18:58:25 +01004254 immediately be closed and a new one re-opened. Option "prefer-last-server" is
4255 available to try optimize server selection so that if the server currently
4256 attached to an idle connection is usable, it will be used.
Willy Tarreau16bfb022010-01-16 19:48:41 +01004257
4258 In general it is preferred to use "option http-server-close" with application
4259 servers, and some static servers might benefit from "option http-keep-alive".
4260
4261 At the moment, logs will not indicate whether requests came from the same
4262 session or not. The accept date reported in the logs corresponds to the end
4263 of the previous request, and the request time corresponds to the time spent
4264 waiting for a new request. The keep-alive request time is still bound to the
4265 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
4266 not set.
4267
Cyril Bonté653dcd62014-02-20 00:13:15 +01004268 This option disables and replaces any previous "option httpclose", "option
4269 http-server-close", "option forceclose" or "option http-tunnel". When backend
Willy Tarreau70dffda2014-01-30 03:07:23 +01004270 and frontend options differ, all of these 4 options have precedence over
Cyril Bonté653dcd62014-02-20 00:13:15 +01004271 "option http-keep-alive".
Willy Tarreau16bfb022010-01-16 19:48:41 +01004272
4273 See also : "option forceclose", "option http-server-close",
Willy Tarreau9420b122013-12-15 18:58:25 +01004274 "option prefer-last-server", "option http-pretend-keepalive",
4275 "option httpclose", and "1.1. The HTTP transaction model".
Willy Tarreau16bfb022010-01-16 19:48:41 +01004276
4277
Willy Tarreau96e31212011-05-30 18:10:30 +02004278option http-no-delay
4279no option http-no-delay
4280 Instruct the system to favor low interactive delays over performance in HTTP
4281 May be used in sections : defaults | frontend | listen | backend
4282 yes | yes | yes | yes
4283 Arguments : none
4284
4285 In HTTP, each payload is unidirectional and has no notion of interactivity.
4286 Any agent is expected to queue data somewhat for a reasonably low delay.
4287 There are some very rare server-to-server applications that abuse the HTTP
4288 protocol and expect the payload phase to be highly interactive, with many
4289 interleaved data chunks in both directions within a single request. This is
4290 absolutely not supported by the HTTP specification and will not work across
4291 most proxies or servers. When such applications attempt to do this through
4292 haproxy, it works but they will experience high delays due to the network
4293 optimizations which favor performance by instructing the system to wait for
4294 enough data to be available in order to only send full packets. Typical
4295 delays are around 200 ms per round trip. Note that this only happens with
4296 abnormal uses. Normal uses such as CONNECT requests nor WebSockets are not
4297 affected.
4298
4299 When "option http-no-delay" is present in either the frontend or the backend
4300 used by a connection, all such optimizations will be disabled in order to
4301 make the exchanges as fast as possible. Of course this offers no guarantee on
4302 the functionality, as it may break at any other place. But if it works via
4303 HAProxy, it will work as fast as possible. This option should never be used
4304 by default, and should never be used at all unless such a buggy application
4305 is discovered. The impact of using this option is an increase of bandwidth
4306 usage and CPU usage, which may significantly lower performance in high
4307 latency environments.
4308
4309
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004310option http-pretend-keepalive
4311no option http-pretend-keepalive
4312 Define whether haproxy will announce keepalive to the server or not
4313 May be used in sections : defaults | frontend | listen | backend
4314 yes | yes | yes | yes
4315 Arguments : none
4316
4317 When running with "option http-server-close" or "option forceclose", haproxy
4318 adds a "Connection: close" header to the request forwarded to the server.
4319 Unfortunately, when some servers see this header, they automatically refrain
4320 from using the chunked encoding for responses of unknown length, while this
4321 is totally unrelated. The immediate effect is that this prevents haproxy from
4322 maintaining the client connection alive. A second effect is that a client or
4323 a cache could receive an incomplete response without being aware of it, and
4324 consider the response complete.
4325
4326 By setting "option http-pretend-keepalive", haproxy will make the server
4327 believe it will keep the connection alive. The server will then not fall back
4328 to the abnormal undesired above. When haproxy gets the whole response, it
4329 will close the connection with the server just as it would do with the
4330 "forceclose" option. That way the client gets a normal response and the
4331 connection is correctly closed on the server side.
4332
4333 It is recommended not to enable this option by default, because most servers
4334 will more efficiently close the connection themselves after the last packet,
4335 and release its buffers slightly earlier. Also, the added packet on the
4336 network could slightly reduce the overall peak performance. However it is
4337 worth noting that when this option is enabled, haproxy will have slightly
4338 less work to do. So if haproxy is the bottleneck on the whole architecture,
4339 enabling this option might save a few CPU cycles.
4340
4341 This option may be set both in a frontend and in a backend. It is enabled if
4342 at least one of the frontend or backend holding a connection has it enabled.
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004343 This option may be combined with "option httpclose", which will cause
Willy Tarreau22a95342010-09-29 14:31:41 +02004344 keepalive to be announced to the server and close to be announced to the
4345 client. This practice is discouraged though.
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004346
4347 If this option has been enabled in a "defaults" section, it can be disabled
4348 in a specific instance by prepending the "no" keyword before it.
4349
Willy Tarreau16bfb022010-01-16 19:48:41 +01004350 See also : "option forceclose", "option http-server-close", and
4351 "option http-keep-alive"
Willy Tarreau8a8e1d92010-04-05 16:15:16 +02004352
Willy Tarreauc27debf2008-01-06 08:57:02 +01004353
Willy Tarreaub608feb2010-01-02 22:47:18 +01004354option http-server-close
4355no option http-server-close
4356 Enable or disable HTTP connection closing on the server side
4357 May be used in sections : defaults | frontend | listen | backend
4358 yes | yes | yes | yes
4359 Arguments : none
4360
Willy Tarreau70dffda2014-01-30 03:07:23 +01004361 By default HAProxy operates in keep-alive mode with regards to persistent
4362 connections: for each connection it processes each request and response, and
4363 leaves the connection idle on both sides between the end of a response and
4364 the start of a new request. This mode may be changed by several options such
4365 as "option http-server-close", "option forceclose", "option httpclose" or
4366 "option http-tunnel". Setting "option http-server-close" enables HTTP
4367 connection-close mode on the server side while keeping the ability to support
4368 HTTP keep-alive and pipelining on the client side. This provides the lowest
4369 latency on the client side (slow network) and the fastest session reuse on
4370 the server side to save server resources, similarly to "option forceclose".
4371 It also permits non-keepalive capable servers to be served in keep-alive mode
4372 to the clients if they conform to the requirements of RFC2616. Please note
4373 that some servers do not always conform to those requirements when they see
4374 "Connection: close" in the request. The effect will be that keep-alive will
4375 never be used. A workaround consists in enabling "option
4376 http-pretend-keepalive".
Willy Tarreaub608feb2010-01-02 22:47:18 +01004377
4378 At the moment, logs will not indicate whether requests came from the same
4379 session or not. The accept date reported in the logs corresponds to the end
4380 of the previous request, and the request time corresponds to the time spent
4381 waiting for a new request. The keep-alive request time is still bound to the
Willy Tarreaub16a5742010-01-10 14:46:16 +01004382 timeout defined by "timeout http-keep-alive" or "timeout http-request" if
4383 not set.
Willy Tarreaub608feb2010-01-02 22:47:18 +01004384
4385 This option may be set both in a frontend and in a backend. It is enabled if
4386 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01004387 It disables and replaces any previous "option httpclose", "option forceclose",
4388 "option http-tunnel" or "option http-keep-alive". Please check section 4
Willy Tarreau70dffda2014-01-30 03:07:23 +01004389 ("Proxies") to see how this option combines with others when frontend and
4390 backend options differ.
Willy Tarreaub608feb2010-01-02 22:47:18 +01004391
4392 If this option has been enabled in a "defaults" section, it can be disabled
4393 in a specific instance by prepending the "no" keyword before it.
4394
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004395 See also : "option forceclose", "option http-pretend-keepalive",
Willy Tarreau16bfb022010-01-16 19:48:41 +01004396 "option httpclose", "option http-keep-alive", and
4397 "1.1. The HTTP transaction model".
Willy Tarreaub608feb2010-01-02 22:47:18 +01004398
4399
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004400option http-tunnel
4401no option http-tunnel
4402 Disable or enable HTTP connection processing after first transaction
4403 May be used in sections : defaults | frontend | listen | backend
4404 yes | yes | yes | yes
4405 Arguments : none
4406
Willy Tarreau70dffda2014-01-30 03:07:23 +01004407 By default HAProxy operates in keep-alive mode with regards to persistent
4408 connections: for each connection it processes each request and response, and
4409 leaves the connection idle on both sides between the end of a response and
4410 the start of a new request. This mode may be changed by several options such
4411 as "option http-server-close", "option forceclose", "option httpclose" or
4412 "option http-tunnel".
4413
4414 Option "http-tunnel" disables any HTTP processing past the first request and
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004415 the first response. This is the mode which was used by default in versions
Willy Tarreau70dffda2014-01-30 03:07:23 +01004416 1.0 to 1.5-dev21. It is the mode with the lowest processing overhead, which
4417 is normally not needed anymore unless in very specific cases such as when
4418 using an in-house protocol that looks like HTTP but is not compatible, or
4419 just to log one request per client in order to reduce log size. Note that
4420 everything which works at the HTTP level, including header parsing/addition,
4421 cookie processing or content switching will only work for the first request
4422 and will be ignored after the first response.
Willy Tarreau02bce8b2014-01-30 00:15:28 +01004423
4424 If this option has been enabled in a "defaults" section, it can be disabled
4425 in a specific instance by prepending the "no" keyword before it.
4426
4427 See also : "option forceclose", "option http-server-close",
4428 "option httpclose", "option http-keep-alive", and
4429 "1.1. The HTTP transaction model".
4430
4431
Willy Tarreau88d349d2010-01-25 12:15:43 +01004432option http-use-proxy-header
Cyril Bontéf0c60612010-02-06 14:44:47 +01004433no option http-use-proxy-header
Willy Tarreau88d349d2010-01-25 12:15:43 +01004434 Make use of non-standard Proxy-Connection header instead of Connection
4435 May be used in sections : defaults | frontend | listen | backend
4436 yes | yes | yes | no
4437 Arguments : none
4438
4439 While RFC2616 explicitly states that HTTP/1.1 agents must use the
4440 Connection header to indicate their wish of persistent or non-persistent
4441 connections, both browsers and proxies ignore this header for proxied
4442 connections and make use of the undocumented, non-standard Proxy-Connection
4443 header instead. The issue begins when trying to put a load balancer between
4444 browsers and such proxies, because there will be a difference between what
4445 haproxy understands and what the client and the proxy agree on.
4446
4447 By setting this option in a frontend, haproxy can automatically switch to use
4448 that non-standard header if it sees proxied requests. A proxied request is
4449 defined here as one where the URI begins with neither a '/' nor a '*'. The
4450 choice of header only affects requests passing through proxies making use of
4451 one of the "httpclose", "forceclose" and "http-server-close" options. Note
4452 that this option can only be specified in a frontend and will affect the
4453 request along its whole life.
4454
Willy Tarreau844a7e72010-01-31 21:46:18 +01004455 Also, when this option is set, a request which requires authentication will
4456 automatically switch to use proxy authentication headers if it is itself a
4457 proxied request. That makes it possible to check or enforce authentication in
4458 front of an existing proxy.
4459
Willy Tarreau88d349d2010-01-25 12:15:43 +01004460 This option should normally never be used, except in front of a proxy.
4461
4462 See also : "option httpclose", "option forceclose" and "option
4463 http-server-close".
4464
4465
Willy Tarreaud63335a2010-02-26 12:56:52 +01004466option httpchk
4467option httpchk <uri>
4468option httpchk <method> <uri>
4469option httpchk <method> <uri> <version>
4470 Enable HTTP protocol to check on the servers health
4471 May be used in sections : defaults | frontend | listen | backend
4472 yes | no | yes | yes
4473 Arguments :
4474 <method> is the optional HTTP method used with the requests. When not set,
4475 the "OPTIONS" method is used, as it generally requires low server
4476 processing and is easy to filter out from the logs. Any method
4477 may be used, though it is not recommended to invent non-standard
4478 ones.
4479
4480 <uri> is the URI referenced in the HTTP requests. It defaults to " / "
4481 which is accessible by default on almost any server, but may be
4482 changed to any other URI. Query strings are permitted.
4483
4484 <version> is the optional HTTP version string. It defaults to "HTTP/1.0"
4485 but some servers might behave incorrectly in HTTP 1.0, so turning
4486 it to HTTP/1.1 may sometimes help. Note that the Host field is
4487 mandatory in HTTP/1.1, and as a trick, it is possible to pass it
4488 after "\r\n" following the version string.
4489
4490 By default, server health checks only consist in trying to establish a TCP
4491 connection. When "option httpchk" is specified, a complete HTTP request is
4492 sent once the TCP connection is established, and responses 2xx and 3xx are
4493 considered valid, while all other ones indicate a server failure, including
4494 the lack of any response.
4495
4496 The port and interval are specified in the server configuration.
4497
4498 This option does not necessarily require an HTTP backend, it also works with
4499 plain TCP backends. This is particularly useful to check simple scripts bound
4500 to some dedicated ports using the inetd daemon.
4501
4502 Examples :
4503 # Relay HTTPS traffic to Apache instance and check service availability
4504 # using HTTP request "OPTIONS * HTTP/1.1" on port 80.
4505 backend https_relay
4506 mode tcp
4507 option httpchk OPTIONS * HTTP/1.1\r\nHost:\ www
4508 server apache1 192.168.1.1:443 check port 80
4509
Simon Hormanafc47ee2013-11-25 10:46:35 +09004510 See also : "option ssl-hello-chk", "option smtpchk", "option mysql-check",
4511 "option pgsql-check", "http-check" and the "check", "port" and
4512 "inter" server options.
Willy Tarreaud63335a2010-02-26 12:56:52 +01004513
4514
Willy Tarreauc27debf2008-01-06 08:57:02 +01004515option httpclose
4516no option httpclose
4517 Enable or disable passive HTTP connection closing
4518 May be used in sections : defaults | frontend | listen | backend
4519 yes | yes | yes | yes
4520 Arguments : none
4521
Willy Tarreau70dffda2014-01-30 03:07:23 +01004522 By default HAProxy operates in keep-alive mode with regards to persistent
4523 connections: for each connection it processes each request and response, and
4524 leaves the connection idle on both sides between the end of a response and
4525 the start of a new request. This mode may be changed by several options such
Cyril Bonté653dcd62014-02-20 00:13:15 +01004526 as "option http-server-close", "option forceclose", "option httpclose" or
Willy Tarreau70dffda2014-01-30 03:07:23 +01004527 "option http-tunnel".
4528
4529 If "option httpclose" is set, HAProxy will work in HTTP tunnel mode and check
4530 if a "Connection: close" header is already set in each direction, and will
4531 add one if missing. Each end should react to this by actively closing the TCP
4532 connection after each transfer, thus resulting in a switch to the HTTP close
4533 mode. Any "Connection" header different from "close" will also be removed.
4534 Note that this option is deprecated since what it does is very cheap but not
4535 reliable. Using "option http-server-close" or "option forceclose" is strongly
4536 recommended instead.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004537
4538 It seldom happens that some servers incorrectly ignore this header and do not
Jamie Gloudonaaa21002012-08-25 00:18:33 -04004539 close the connection even though they reply "Connection: close". For this
Willy Tarreau0dfdf192010-01-05 11:33:11 +01004540 reason, they are not compatible with older HTTP 1.0 browsers. If this happens
4541 it is possible to use the "option forceclose" which actively closes the
4542 request connection once the server responds. Option "forceclose" also
4543 releases the server connection earlier because it does not have to wait for
4544 the client to acknowledge it.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004545
4546 This option may be set both in a frontend and in a backend. It is enabled if
4547 at least one of the frontend or backend holding a connection has it enabled.
Cyril Bonté653dcd62014-02-20 00:13:15 +01004548 It disables and replaces any previous "option http-server-close",
4549 "option forceclose", "option http-keep-alive" or "option http-tunnel". Please
Willy Tarreau70dffda2014-01-30 03:07:23 +01004550 check section 4 ("Proxies") to see how this option combines with others when
4551 frontend and backend options differ.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004552
4553 If this option has been enabled in a "defaults" section, it can be disabled
4554 in a specific instance by prepending the "no" keyword before it.
4555
Patrick Mezard9ec2ec42010-06-12 17:02:45 +02004556 See also : "option forceclose", "option http-server-close" and
4557 "1.1. The HTTP transaction model".
Willy Tarreauc27debf2008-01-06 08:57:02 +01004558
4559
Emeric Brun3a058f32009-06-30 18:26:00 +02004560option httplog [ clf ]
Willy Tarreauc27debf2008-01-06 08:57:02 +01004561 Enable logging of HTTP request, session state and timers
4562 May be used in sections : defaults | frontend | listen | backend
4563 yes | yes | yes | yes
Emeric Brun3a058f32009-06-30 18:26:00 +02004564 Arguments :
4565 clf if the "clf" argument is added, then the output format will be
4566 the CLF format instead of HAProxy's default HTTP format. You can
4567 use this when you need to feed HAProxy's logs through a specific
4568 log analyser which only support the CLF format and which is not
4569 extensible.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004570
4571 By default, the log output format is very poor, as it only contains the
4572 source and destination addresses, and the instance name. By specifying
4573 "option httplog", each log line turns into a much richer format including,
4574 but not limited to, the HTTP request, the connection timers, the session
4575 status, the connections numbers, the captured headers and cookies, the
4576 frontend, backend and server name, and of course the source address and
4577 ports.
4578
4579 This option may be set either in the frontend or the backend.
4580
PiBa-NLbd556bf2014-12-11 21:31:54 +01004581 Specifying only "option httplog" will automatically clear the 'clf' mode
4582 if it was set by default.
Emeric Brun3a058f32009-06-30 18:26:00 +02004583
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004584 See also : section 8 about logging.
Willy Tarreauc27debf2008-01-06 08:57:02 +01004585
Willy Tarreau55165fe2009-05-10 12:02:55 +02004586
4587option http_proxy
4588no option http_proxy
4589 Enable or disable plain HTTP proxy mode
4590 May be used in sections : defaults | frontend | listen | backend
4591 yes | yes | yes | yes
4592 Arguments : none
4593
4594 It sometimes happens that people need a pure HTTP proxy which understands
4595 basic proxy requests without caching nor any fancy feature. In this case,
4596 it may be worth setting up an HAProxy instance with the "option http_proxy"
4597 set. In this mode, no server is declared, and the connection is forwarded to
4598 the IP address and port found in the URL after the "http://" scheme.
4599
4600 No host address resolution is performed, so this only works when pure IP
4601 addresses are passed. Since this option's usage perimeter is rather limited,
4602 it will probably be used only by experts who know they need exactly it. Last,
4603 if the clients are susceptible of sending keep-alive requests, it will be
Cyril Bonté2409e682010-12-14 22:47:51 +01004604 needed to add "option httpclose" to ensure that all requests will correctly
Willy Tarreau55165fe2009-05-10 12:02:55 +02004605 be analyzed.
4606
4607 If this option has been enabled in a "defaults" section, it can be disabled
4608 in a specific instance by prepending the "no" keyword before it.
4609
4610 Example :
4611 # this backend understands HTTP proxy requests and forwards them directly.
4612 backend direct_forward
4613 option httpclose
4614 option http_proxy
4615
4616 See also : "option httpclose"
4617
Willy Tarreau211ad242009-10-03 21:45:07 +02004618
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004619option independent-streams
4620no option independent-streams
4621 Enable or disable independent timeout processing for both directions
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02004622 May be used in sections : defaults | frontend | listen | backend
4623 yes | yes | yes | yes
4624 Arguments : none
4625
4626 By default, when data is sent over a socket, both the write timeout and the
4627 read timeout for that socket are refreshed, because we consider that there is
4628 activity on that socket, and we have no other means of guessing if we should
4629 receive data or not.
4630
4631 While this default behaviour is desirable for almost all applications, there
4632 exists a situation where it is desirable to disable it, and only refresh the
4633 read timeout if there are incoming data. This happens on sessions with large
4634 timeouts and low amounts of exchanged data such as telnet session. If the
4635 server suddenly disappears, the output data accumulates in the system's
4636 socket buffers, both timeouts are correctly refreshed, and there is no way
4637 to know the server does not receive them, so we don't timeout. However, when
4638 the underlying protocol always echoes sent data, it would be enough by itself
4639 to detect the issue using the read timeout. Note that this problem does not
4640 happen with more verbose protocols because data won't accumulate long in the
4641 socket buffers.
4642
4643 When this option is set on the frontend, it will disable read timeout updates
4644 on data sent to the client. There probably is little use of this case. When
4645 the option is set on the backend, it will disable read timeout updates on
4646 data sent to the server. Doing so will typically break large HTTP posts from
4647 slow lines, so use it with caution.
4648
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03004649 Note: older versions used to call this setting "option independent-streams"
Jamie Gloudon801a0a32012-08-25 00:18:33 -04004650 with a spelling mistake. This spelling is still supported but
4651 deprecated.
4652
Willy Tarreauce887fd2012-05-12 12:50:00 +02004653 See also : "timeout client", "timeout server" and "timeout tunnel"
Willy Tarreauf27b5ea2009-10-03 22:01:18 +02004654
4655
Gabor Lekenyb4c81e42010-09-29 18:17:05 +02004656option ldap-check
4657 Use LDAPv3 health checks for server testing
4658 May be used in sections : defaults | frontend | listen | backend
4659 yes | no | yes | yes
4660 Arguments : none
4661
4662 It is possible to test that the server correctly talks LDAPv3 instead of just
4663 testing that it accepts the TCP connection. When this option is set, an
4664 LDAPv3 anonymous simple bind message is sent to the server, and the response
4665 is analyzed to find an LDAPv3 bind response message.
4666
4667 The server is considered valid only when the LDAP response contains success
4668 resultCode (http://tools.ietf.org/html/rfc4511#section-4.1.9).
4669
4670 Logging of bind requests is server dependent see your documentation how to
4671 configure it.
4672
4673 Example :
4674 option ldap-check
4675
4676 See also : "option httpchk"
4677
4678
Simon Horman98637e52014-06-20 12:30:16 +09004679option external-check
4680 Use external processes for server health checks
4681 May be used in sections : defaults | frontend | listen | backend
4682 yes | no | yes | yes
4683
4684 It is possible to test the health of a server using an external command.
4685 This is achieved by running the executable set using "external-check
4686 command".
4687
4688 Requires the "external-check" global to be set.
4689
4690 See also : "external-check", "external-check command", "external-check path"
4691
4692
Willy Tarreau211ad242009-10-03 21:45:07 +02004693option log-health-checks
4694no option log-health-checks
Willy Tarreaubef1b322014-05-13 21:01:39 +02004695 Enable or disable logging of health checks status updates
Willy Tarreau211ad242009-10-03 21:45:07 +02004696 May be used in sections : defaults | frontend | listen | backend
4697 yes | no | yes | yes
4698 Arguments : none
4699
Willy Tarreaubef1b322014-05-13 21:01:39 +02004700 By default, failed health check are logged if server is UP and successful
4701 health checks are logged if server is DOWN, so the amount of additional
4702 information is limited.
Willy Tarreau211ad242009-10-03 21:45:07 +02004703
Willy Tarreaubef1b322014-05-13 21:01:39 +02004704 When this option is enabled, any change of the health check status or to
4705 the server's health will be logged, so that it becomes possible to know
4706 that a server was failing occasional checks before crashing, or exactly when
4707 it failed to respond a valid HTTP status, then when the port started to
4708 reject connections, then when the server stopped responding at all.
4709
4710 Note that status changes not caused by health checks (eg: enable/disable on
4711 the CLI) are intentionally not logged by this option.
Willy Tarreau211ad242009-10-03 21:45:07 +02004712
Willy Tarreaubef1b322014-05-13 21:01:39 +02004713 See also: "option httpchk", "option ldap-check", "option mysql-check",
4714 "option pgsql-check", "option redis-check", "option smtpchk",
4715 "option tcp-check", "log" and section 8 about logging.
Willy Tarreau211ad242009-10-03 21:45:07 +02004716
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004717
4718option log-separate-errors
4719no option log-separate-errors
4720 Change log level for non-completely successful connections
4721 May be used in sections : defaults | frontend | listen | backend
4722 yes | yes | yes | no
4723 Arguments : none
4724
4725 Sometimes looking for errors in logs is not easy. This option makes haproxy
4726 raise the level of logs containing potentially interesting information such
4727 as errors, timeouts, retries, redispatches, or HTTP status codes 5xx. The
4728 level changes from "info" to "err". This makes it possible to log them
4729 separately to a different file with most syslog daemons. Be careful not to
4730 remove them from the original file, otherwise you would lose ordering which
4731 provides very important information.
4732
4733 Using this option, large sites dealing with several thousand connections per
4734 second may log normal traffic to a rotating buffer and only archive smaller
4735 error logs.
4736
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004737 See also : "log", "dontlognull", "dontlog-normal" and section 8 about
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +02004738 logging.
4739
Willy Tarreauc27debf2008-01-06 08:57:02 +01004740
4741option logasap
4742no option logasap
4743 Enable or disable early logging of HTTP requests
4744 May be used in sections : defaults | frontend | listen | backend
4745 yes | yes | yes | no
4746 Arguments : none
4747
4748 By default, HTTP requests are logged upon termination so that the total
4749 transfer time and the number of bytes appear in the logs. When large objects
4750 are being transferred, it may take a while before the request appears in the
4751 logs. Using "option logasap", the request gets logged as soon as the server
4752 sends the complete headers. The only missing information in the logs will be
4753 the total number of bytes which will indicate everything except the amount
4754 of data transferred, and the total time which will not take the transfer
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01004755 time into account. In such a situation, it's a good practice to capture the
Willy Tarreauc27debf2008-01-06 08:57:02 +01004756 "Content-Length" response header so that the logs at least indicate how many
4757 bytes are expected to be transferred.
4758
Willy Tarreaucc6c8912009-02-22 10:53:55 +01004759 Examples :
4760 listen http_proxy 0.0.0.0:80
4761 mode http
4762 option httplog
4763 option logasap
4764 log 192.168.2.200 local3
4765
4766 >>> Feb 6 12:14:14 localhost \
4767 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
4768 static/srv1 9/10/7/14/+30 200 +243 - - ---- 3/1/1/1/0 1/0 \
4769 "GET /image.iso HTTP/1.0"
4770
Willy Tarreauc57f0e22009-05-10 13:12:33 +02004771 See also : "option httplog", "capture response header", and section 8 about
Willy Tarreauc27debf2008-01-06 08:57:02 +01004772 logging.
4773
4774
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02004775option mysql-check [ user <username> [ post-41 ] ]
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004776 Use MySQL health checks for server testing
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004777 May be used in sections : defaults | frontend | listen | backend
4778 yes | no | yes | yes
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004779 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02004780 <username> This is the username which will be used when connecting to MySQL
4781 server.
Nenad Merdanovic6639a7c2014-05-30 14:26:32 +02004782 post-41 Send post v4.1 client compatible checks
Hervé COMMOWICK8776f1b2010-10-18 15:58:36 +02004783
4784 If you specify a username, the check consists of sending two MySQL packet,
4785 one Client Authentication packet, and one QUIT packet, to correctly close
4786 MySQL session. We then parse the MySQL Handshake Initialisation packet and/or
4787 Error packet. It is a basic but useful test which does not produce error nor
4788 aborted connect on the server. However, it requires adding an authorization
4789 in the MySQL table, like this :
4790
4791 USE mysql;
4792 INSERT INTO user (Host,User) values ('<ip_of_haproxy>','<username>');
4793 FLUSH PRIVILEGES;
4794
4795 If you don't specify a username (it is deprecated and not recommended), the
4796 check only consists in parsing the Mysql Handshake Initialisation packet or
4797 Error packet, we don't send anything in this mode. It was reported that it
4798 can generate lockout if check is too frequent and/or if there is not enough
4799 traffic. In fact, you need in this case to check MySQL "max_connect_errors"
4800 value as if a connection is established successfully within fewer than MySQL
4801 "max_connect_errors" attempts after a previous connection was interrupted,
4802 the error count for the host is cleared to zero. If HAProxy's server get
4803 blocked, the "FLUSH HOSTS" statement is the only way to unblock it.
4804
4805 Remember that this does not check database presence nor database consistency.
4806 To do this, you can use an external check with xinetd for example.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004807
Hervé COMMOWICK212f7782011-06-10 14:05:59 +02004808 The check requires MySQL >=3.22, for older version, please use TCP check.
Hervé COMMOWICK698ae002010-01-12 09:25:13 +01004809
4810 Most often, an incoming MySQL server needs to see the client's IP address for
4811 various purposes, including IP privilege matching and connection logging.
4812 When possible, it is often wise to masquerade the client's IP address when
4813 connecting to the server using the "usesrc" argument of the "source" keyword,
4814 which requires the cttproxy feature to be compiled in, and the MySQL server
4815 to route the client via the machine hosting haproxy.
4816
4817 See also: "option httpchk"
4818
4819
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004820option nolinger
4821no option nolinger
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01004822 Enable or disable immediate session resource cleaning after close
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004823 May be used in sections: defaults | frontend | listen | backend
4824 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004825 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004826
4827 When clients or servers abort connections in a dirty way (eg: they are
4828 physically disconnected), the session timeouts triggers and the session is
4829 closed. But it will remain in FIN_WAIT1 state for some time in the system,
4830 using some resources and possibly limiting the ability to establish newer
4831 connections.
4832
4833 When this happens, it is possible to activate "option nolinger" which forces
4834 the system to immediately remove any socket's pending data on close. Thus,
4835 the session is instantly purged from the system's tables. This usually has
4836 side effects such as increased number of TCP resets due to old retransmits
4837 getting immediately rejected. Some firewalls may sometimes complain about
4838 this too.
4839
4840 For this reason, it is not recommended to use this option when not absolutely
4841 needed. You know that you need it when you have thousands of FIN_WAIT1
4842 sessions on your system (TIME_WAIT ones do not count).
4843
4844 This option may be used both on frontends and backends, depending on the side
4845 where it is required. Use it on the frontend for clients, and on the backend
4846 for servers.
4847
4848 If this option has been enabled in a "defaults" section, it can be disabled
4849 in a specific instance by prepending the "no" keyword before it.
4850
4851
Willy Tarreau55165fe2009-05-10 12:02:55 +02004852option originalto [ except <network> ] [ header <name> ]
4853 Enable insertion of the X-Original-To header to requests sent to servers
4854 May be used in sections : defaults | frontend | listen | backend
4855 yes | yes | yes | yes
4856 Arguments :
4857 <network> is an optional argument used to disable this option for sources
4858 matching <network>
4859 <name> an optional argument to specify a different "X-Original-To"
4860 header name.
4861
4862 Since HAProxy can work in transparent mode, every request from a client can
4863 be redirected to the proxy and HAProxy itself can proxy every request to a
4864 complex SQUID environment and the destination host from SO_ORIGINAL_DST will
4865 be lost. This is annoying when you want access rules based on destination ip
4866 addresses. To solve this problem, a new HTTP header "X-Original-To" may be
4867 added by HAProxy to all requests sent to the server. This header contains a
4868 value representing the original destination IP address. Since this must be
4869 configured to always use the last occurrence of this header only. Note that
4870 only the last occurrence of the header must be used, since it is really
4871 possible that the client has already brought one.
4872
4873 The keyword "header" may be used to supply a different header name to replace
4874 the default "X-Original-To". This can be useful where you might already
4875 have a "X-Original-To" header from a different application, and you need
4876 preserve it. Also if your backend server doesn't use the "X-Original-To"
4877 header and requires different one.
4878
4879 Sometimes, a same HAProxy instance may be shared between a direct client
4880 access and a reverse-proxy access (for instance when an SSL reverse-proxy is
4881 used to decrypt HTTPS traffic). It is possible to disable the addition of the
4882 header for a known source address or network by adding the "except" keyword
4883 followed by the network address. In this case, any source IP matching the
4884 network will not cause an addition of this header. Most common uses are with
4885 private networks or 127.0.0.1.
4886
4887 This option may be specified either in the frontend or in the backend. If at
4888 least one of them uses it, the header will be added. Note that the backend's
4889 setting of the header subargument takes precedence over the frontend's if
4890 both are defined.
4891
Willy Tarreau55165fe2009-05-10 12:02:55 +02004892 Examples :
4893 # Original Destination address
4894 frontend www
4895 mode http
4896 option originalto except 127.0.0.1
4897
4898 # Those servers want the IP Address in X-Client-Dst
4899 backend www
4900 mode http
4901 option originalto header X-Client-Dst
4902
Willy Tarreau87cf5142011-08-19 22:57:24 +02004903 See also : "option httpclose", "option http-server-close",
4904 "option forceclose"
Willy Tarreau55165fe2009-05-10 12:02:55 +02004905
4906
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004907option persist
4908no option persist
4909 Enable or disable forced persistence on down servers
4910 May be used in sections: defaults | frontend | listen | backend
4911 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004912 Arguments : none
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004913
4914 When an HTTP request reaches a backend with a cookie which references a dead
4915 server, by default it is redispatched to another server. It is possible to
4916 force the request to be sent to the dead server first using "option persist"
4917 if absolutely needed. A common use case is when servers are under extreme
4918 load and spend their time flapping. In this case, the users would still be
4919 directed to the server they opened the session on, in the hope they would be
4920 correctly served. It is recommended to use "option redispatch" in conjunction
4921 with this option so that in the event it would not be possible to connect to
4922 the server at all (server definitely dead), the client would finally be
4923 redirected to another valid server.
4924
4925 If this option has been enabled in a "defaults" section, it can be disabled
4926 in a specific instance by prepending the "no" keyword before it.
4927
Willy Tarreau4de91492010-01-22 19:10:05 +01004928 See also : "option redispatch", "retries", "force-persist"
Willy Tarreaua453bdd2008-01-08 19:50:52 +01004929
4930
Willy Tarreau0c122822013-12-15 18:49:01 +01004931option pgsql-check [ user <username> ]
4932 Use PostgreSQL health checks for server testing
4933 May be used in sections : defaults | frontend | listen | backend
4934 yes | no | yes | yes
4935 Arguments :
4936 <username> This is the username which will be used when connecting to
4937 PostgreSQL server.
4938
4939 The check sends a PostgreSQL StartupMessage and waits for either
4940 Authentication request or ErrorResponse message. It is a basic but useful
4941 test which does not produce error nor aborted connect on the server.
4942 This check is identical with the "mysql-check".
4943
4944 See also: "option httpchk"
4945
4946
Willy Tarreau9420b122013-12-15 18:58:25 +01004947option prefer-last-server
4948no option prefer-last-server
4949 Allow multiple load balanced requests to remain on the same server
4950 May be used in sections: defaults | frontend | listen | backend
4951 yes | no | yes | yes
4952 Arguments : none
4953
4954 When the load balancing algorithm in use is not deterministic, and a previous
4955 request was sent to a server to which haproxy still holds a connection, it is
4956 sometimes desirable that subsequent requests on a same session go to the same
4957 server as much as possible. Note that this is different from persistence, as
4958 we only indicate a preference which haproxy tries to apply without any form
4959 of warranty. The real use is for keep-alive connections sent to servers. When
4960 this option is used, haproxy will try to reuse the same connection that is
4961 attached to the server instead of rebalancing to another server, causing a
4962 close of the connection. This can make sense for static file servers. It does
Willy Tarreau068621e2013-12-23 15:11:25 +01004963 not make much sense to use this in combination with hashing algorithms. Note,
4964 haproxy already automatically tries to stick to a server which sends a 401 or
4965 to a proxy which sends a 407 (authentication required). This is mandatory for
4966 use with the broken NTLM authentication challenge, and significantly helps in
4967 troubleshooting some faulty applications. Option prefer-last-server might be
4968 desirable in these environments as well, to avoid redistributing the traffic
4969 after every other response.
Willy Tarreau9420b122013-12-15 18:58:25 +01004970
4971 If this option has been enabled in a "defaults" section, it can be disabled
4972 in a specific instance by prepending the "no" keyword before it.
4973
4974 See also: "option http-keep-alive"
4975
4976
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004977option redispatch
4978no option redispatch
4979 Enable or disable session redistribution in case of connection failure
4980 May be used in sections: defaults | frontend | listen | backend
4981 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01004982 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004983
4984 In HTTP mode, if a server designated by a cookie is down, clients may
4985 definitely stick to it because they cannot flush the cookie, so they will not
4986 be able to access the service anymore.
4987
4988 Specifying "option redispatch" will allow the proxy to break their
4989 persistence and redistribute them to a working server.
4990
4991 It also allows to retry last connection to another server in case of multiple
4992 connection failures. Of course, it requires having "retries" set to a nonzero
4993 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01004994
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01004995 This form is the preferred form, which replaces both the "redispatch" and
4996 "redisp" keywords.
4997
4998 If this option has been enabled in a "defaults" section, it can be disabled
4999 in a specific instance by prepending the "no" keyword before it.
5000
Willy Tarreau4de91492010-01-22 19:10:05 +01005001 See also : "redispatch", "retries", "force-persist"
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005002
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005003
Hervé COMMOWICKec032d62011-08-05 16:23:48 +02005004option redis-check
5005 Use redis health checks for server testing
5006 May be used in sections : defaults | frontend | listen | backend
5007 yes | no | yes | yes
5008 Arguments : none
5009
5010 It is possible to test that the server correctly talks REDIS protocol instead
5011 of just testing that it accepts the TCP connection. When this option is set,
5012 a PING redis command is sent to the server, and the response is analyzed to
5013 find the "+PONG" response message.
5014
5015 Example :
5016 option redis-check
5017
5018 See also : "option httpchk"
5019
5020
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005021option smtpchk
5022option smtpchk <hello> <domain>
5023 Use SMTP health checks for server testing
5024 May be used in sections : defaults | frontend | listen | backend
5025 yes | no | yes | yes
Willy Tarreaud72758d2010-01-12 10:42:19 +01005026 Arguments :
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005027 <hello> is an optional argument. It is the "hello" command to use. It can
5028 be either "HELO" (for SMTP) or "EHLO" (for ESTMP). All other
5029 values will be turned into the default command ("HELO").
5030
5031 <domain> is the domain name to present to the server. It may only be
5032 specified (and is mandatory) if the hello command has been
5033 specified. By default, "localhost" is used.
5034
5035 When "option smtpchk" is set, the health checks will consist in TCP
5036 connections followed by an SMTP command. By default, this command is
5037 "HELO localhost". The server's return code is analyzed and only return codes
5038 starting with a "2" will be considered as valid. All other responses,
5039 including a lack of response will constitute an error and will indicate a
5040 dead server.
5041
5042 This test is meant to be used with SMTP servers or relays. Depending on the
5043 request, it is possible that some servers do not log each connection attempt,
5044 so you may want to experiment to improve the behaviour. Using telnet on port
5045 25 is often easier than adjusting the configuration.
5046
5047 Most often, an incoming SMTP server needs to see the client's IP address for
5048 various purposes, including spam filtering, anti-spoofing and logging. When
5049 possible, it is often wise to masquerade the client's IP address when
5050 connecting to the server using the "usesrc" argument of the "source" keyword,
5051 which requires the cttproxy feature to be compiled in.
5052
5053 Example :
5054 option smtpchk HELO mydomain.org
5055
5056 See also : "option httpchk", "source"
5057
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005058
Krzysztof Piotr Oledzkiaeebf9b2009-10-04 15:43:17 +02005059option socket-stats
5060no option socket-stats
5061
5062 Enable or disable collecting & providing separate statistics for each socket.
5063 May be used in sections : defaults | frontend | listen | backend
5064 yes | yes | yes | no
5065
5066 Arguments : none
5067
5068
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005069option splice-auto
5070no option splice-auto
5071 Enable or disable automatic kernel acceleration on sockets in both directions
5072 May be used in sections : defaults | frontend | listen | backend
5073 yes | yes | yes | yes
5074 Arguments : none
5075
5076 When this option is enabled either on a frontend or on a backend, haproxy
5077 will automatically evaluate the opportunity to use kernel tcp splicing to
5078 forward data between the client and the server, in either direction. Haproxy
5079 uses heuristics to estimate if kernel splicing might improve performance or
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005080 not. Both directions are handled independently. Note that the heuristics used
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005081 are not much aggressive in order to limit excessive use of splicing. This
5082 option requires splicing to be enabled at compile time, and may be globally
5083 disabled with the global option "nosplice". Since splice uses pipes, using it
5084 requires that there are enough spare pipes.
5085
5086 Important note: kernel-based TCP splicing is a Linux-specific feature which
5087 first appeared in kernel 2.6.25. It offers kernel-based acceleration to
5088 transfer data between sockets without copying these data to user-space, thus
5089 providing noticeable performance gains and CPU cycles savings. Since many
5090 early implementations are buggy, corrupt data and/or are inefficient, this
5091 feature is not enabled by default, and it should be used with extreme care.
5092 While it is not possible to detect the correctness of an implementation,
5093 2.6.29 is the first version offering a properly working implementation. In
5094 case of doubt, splicing may be globally disabled using the global "nosplice"
5095 keyword.
5096
5097 Example :
5098 option splice-auto
5099
5100 If this option has been enabled in a "defaults" section, it can be disabled
5101 in a specific instance by prepending the "no" keyword before it.
5102
5103 See also : "option splice-request", "option splice-response", and global
5104 options "nosplice" and "maxpipes"
5105
5106
5107option splice-request
5108no option splice-request
5109 Enable or disable automatic kernel acceleration on sockets for requests
5110 May be used in sections : defaults | frontend | listen | backend
5111 yes | yes | yes | yes
5112 Arguments : none
5113
5114 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005115 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005116 the client to the server. It might still use the recv/send scheme if there
5117 are no spare pipes left. This option requires splicing to be enabled at
5118 compile time, and may be globally disabled with the global option "nosplice".
5119 Since splice uses pipes, using it requires that there are enough spare pipes.
5120
5121 Important note: see "option splice-auto" for usage limitations.
5122
5123 Example :
5124 option splice-request
5125
5126 If this option has been enabled in a "defaults" section, it can be disabled
5127 in a specific instance by prepending the "no" keyword before it.
5128
5129 See also : "option splice-auto", "option splice-response", and global options
5130 "nosplice" and "maxpipes"
5131
5132
5133option splice-response
5134no option splice-response
5135 Enable or disable automatic kernel acceleration on sockets for responses
5136 May be used in sections : defaults | frontend | listen | backend
5137 yes | yes | yes | yes
5138 Arguments : none
5139
5140 When this option is enabled either on a frontend or on a backend, haproxy
Jamie Gloudonaaa21002012-08-25 00:18:33 -04005141 will use kernel tcp splicing whenever possible to forward data going from
Willy Tarreauff4f82d2009-02-06 11:28:13 +01005142 the server to the client. It might still use the recv/send scheme if there
5143 are no spare pipes left. This option requires splicing to be enabled at
5144 compile time, and may be globally disabled with the global option "nosplice".
5145 Since splice uses pipes, using it requires that there are enough spare pipes.
5146
5147 Important note: see "option splice-auto" for usage limitations.
5148
5149 Example :
5150 option splice-response
5151
5152 If this option has been enabled in a "defaults" section, it can be disabled
5153 in a specific instance by prepending the "no" keyword before it.
5154
5155 See also : "option splice-auto", "option splice-request", and global options
5156 "nosplice" and "maxpipes"
5157
5158
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005159option srvtcpka
5160no option srvtcpka
5161 Enable or disable the sending of TCP keepalive packets on the server side
5162 May be used in sections : defaults | frontend | listen | backend
5163 yes | no | yes | yes
5164 Arguments : none
5165
5166 When there is a firewall or any session-aware component between a client and
5167 a server, and when the protocol involves very long sessions with long idle
5168 periods (eg: remote desktops), there is a risk that one of the intermediate
5169 components decides to expire a session which has remained idle for too long.
5170
5171 Enabling socket-level TCP keep-alives makes the system regularly send packets
5172 to the other end of the connection, leaving it active. The delay between
5173 keep-alive probes is controlled by the system only and depends both on the
5174 operating system and its tuning parameters.
5175
5176 It is important to understand that keep-alive packets are neither emitted nor
5177 received at the application level. It is only the network stacks which sees
5178 them. For this reason, even if one side of the proxy already uses keep-alives
5179 to maintain its connection alive, those keep-alive packets will not be
5180 forwarded to the other side of the proxy.
5181
5182 Please note that this has nothing to do with HTTP keep-alive.
5183
5184 Using option "srvtcpka" enables the emission of TCP keep-alive probes on the
5185 server side of a connection, which should help when session expirations are
5186 noticed between HAProxy and a server.
5187
5188 If this option has been enabled in a "defaults" section, it can be disabled
5189 in a specific instance by prepending the "no" keyword before it.
5190
5191 See also : "option clitcpka", "option tcpka"
5192
5193
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005194option ssl-hello-chk
5195 Use SSLv3 client hello health checks for server testing
5196 May be used in sections : defaults | frontend | listen | backend
5197 yes | no | yes | yes
5198 Arguments : none
5199
5200 When some SSL-based protocols are relayed in TCP mode through HAProxy, it is
5201 possible to test that the server correctly talks SSL instead of just testing
5202 that it accepts the TCP connection. When "option ssl-hello-chk" is set, pure
5203 SSLv3 client hello messages are sent once the connection is established to
5204 the server, and the response is analyzed to find an SSL server hello message.
5205 The server is considered valid only when the response contains this server
5206 hello message.
5207
5208 All servers tested till there correctly reply to SSLv3 client hello messages,
5209 and most servers tested do not even log the requests containing only hello
5210 messages, which is appreciable.
5211
Willy Tarreau763a95b2012-10-04 23:15:39 +02005212 Note that this check works even when SSL support was not built into haproxy
5213 because it forges the SSL message. When SSL support is available, it is best
5214 to use native SSL health checks instead of this one.
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005215
Willy Tarreau763a95b2012-10-04 23:15:39 +02005216 See also: "option httpchk", "check-ssl"
5217
Willy Tarreaua453bdd2008-01-08 19:50:52 +01005218
Willy Tarreaued179852013-12-16 01:07:00 +01005219option tcp-check
5220 Perform health checks using tcp-check send/expect sequences
5221 May be used in sections: defaults | frontend | listen | backend
5222 yes | no | yes | yes
5223
5224 This health check method is intended to be combined with "tcp-check" command
5225 lists in order to support send/expect types of health check sequences.
5226
5227 TCP checks currently support 4 modes of operations :
5228 - no "tcp-check" directive : the health check only consists in a connection
5229 attempt, which remains the default mode.
5230
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005231 - "tcp-check send" or "tcp-check send-binary" only is mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01005232 used to send a string along with a connection opening. With some
5233 protocols, it helps sending a "QUIT" message for example that prevents
5234 the server from logging a connection error for each health check. The
5235 check result will still be based on the ability to open the connection
5236 only.
5237
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005238 - "tcp-check expect" only is mentioned : this is used to test a banner.
Willy Tarreaued179852013-12-16 01:07:00 +01005239 The connection is opened and haproxy waits for the server to present some
5240 contents which must validate some rules. The check result will be based
5241 on the matching between the contents and the rules. This is suited for
5242 POP, IMAP, SMTP, FTP, SSH, TELNET.
5243
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005244 - both "tcp-check send" and "tcp-check expect" are mentioned : this is
Willy Tarreaued179852013-12-16 01:07:00 +01005245 used to test a hello-type protocol. Haproxy sends a message, the server
5246 responds and its response is analysed. the check result will be based on
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005247 the matching between the response contents and the rules. This is often
Willy Tarreaued179852013-12-16 01:07:00 +01005248 suited for protocols which require a binding or a request/response model.
5249 LDAP, MySQL, Redis and SSL are example of such protocols, though they
5250 already all have their dedicated checks with a deeper understanding of
5251 the respective protocols.
5252 In this mode, many questions may be sent and many answers may be
5253 analysed.
5254
5255 Examples :
5256 # perform a POP check (analyse only server's banner)
5257 option tcp-check
5258 tcp-check expect string +OK\ POP3\ ready
5259
5260 # perform an IMAP check (analyse only server's banner)
5261 option tcp-check
5262 tcp-check expect string *\ OK\ IMAP4\ ready
5263
5264 # look for the redis master server after ensuring it speaks well
5265 # redis protocol, then it exits properly.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005266 # (send a command then analyse the response 3 times)
Willy Tarreaued179852013-12-16 01:07:00 +01005267 option tcp-check
5268 tcp-check send PING\r\n
5269 tcp-check expect +PONG
5270 tcp-check send info\ replication\r\n
5271 tcp-check expect string role:master
5272 tcp-check send QUIT\r\n
5273 tcp-check expect string +OK
5274
5275 forge a HTTP request, then analyse the response
5276 (send many headers before analyzing)
5277 option tcp-check
5278 tcp-check send HEAD\ /\ HTTP/1.1\r\n
5279 tcp-check send Host:\ www.mydomain.com\r\n
5280 tcp-check send User-Agent:\ HAProxy\ tcpcheck\r\n
5281 tcp-check send \r\n
5282 tcp-check expect rstring HTTP/1\..\ (2..|3..)
5283
5284
5285 See also : "tcp-check expect", "tcp-check send"
5286
5287
Willy Tarreau9ea05a72009-06-14 12:07:01 +02005288option tcp-smart-accept
5289no option tcp-smart-accept
5290 Enable or disable the saving of one ACK packet during the accept sequence
5291 May be used in sections : defaults | frontend | listen | backend
5292 yes | yes | yes | no
5293 Arguments : none
5294
5295 When an HTTP connection request comes in, the system acknowledges it on
5296 behalf of HAProxy, then the client immediately sends its request, and the
5297 system acknowledges it too while it is notifying HAProxy about the new
5298 connection. HAProxy then reads the request and responds. This means that we
5299 have one TCP ACK sent by the system for nothing, because the request could
5300 very well be acknowledged by HAProxy when it sends its response.
5301
5302 For this reason, in HTTP mode, HAProxy automatically asks the system to avoid
5303 sending this useless ACK on platforms which support it (currently at least
5304 Linux). It must not cause any problem, because the system will send it anyway
5305 after 40 ms if the response takes more time than expected to come.
5306
5307 During complex network debugging sessions, it may be desirable to disable
5308 this optimization because delayed ACKs can make troubleshooting more complex
5309 when trying to identify where packets are delayed. It is then possible to
5310 fall back to normal behaviour by specifying "no option tcp-smart-accept".
5311
5312 It is also possible to force it for non-HTTP proxies by simply specifying
5313 "option tcp-smart-accept". For instance, it can make sense with some services
5314 such as SMTP where the server speaks first.
5315
5316 It is recommended to avoid forcing this option in a defaults section. In case
5317 of doubt, consider setting it back to automatic values by prepending the
5318 "default" keyword before it, or disabling it using the "no" keyword.
5319
Willy Tarreaud88edf22009-06-14 15:48:17 +02005320 See also : "option tcp-smart-connect"
5321
5322
5323option tcp-smart-connect
5324no option tcp-smart-connect
5325 Enable or disable the saving of one ACK packet during the connect sequence
5326 May be used in sections : defaults | frontend | listen | backend
5327 yes | no | yes | yes
5328 Arguments : none
5329
5330 On certain systems (at least Linux), HAProxy can ask the kernel not to
5331 immediately send an empty ACK upon a connection request, but to directly
5332 send the buffer request instead. This saves one packet on the network and
5333 thus boosts performance. It can also be useful for some servers, because they
5334 immediately get the request along with the incoming connection.
5335
5336 This feature is enabled when "option tcp-smart-connect" is set in a backend.
5337 It is not enabled by default because it makes network troubleshooting more
5338 complex.
5339
5340 It only makes sense to enable it with protocols where the client speaks first
5341 such as HTTP. In other situations, if there is no data to send in place of
5342 the ACK, a normal ACK is sent.
5343
5344 If this option has been enabled in a "defaults" section, it can be disabled
5345 in a specific instance by prepending the "no" keyword before it.
5346
5347 See also : "option tcp-smart-accept"
5348
Willy Tarreau9ea05a72009-06-14 12:07:01 +02005349
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005350option tcpka
5351 Enable or disable the sending of TCP keepalive packets on both sides
5352 May be used in sections : defaults | frontend | listen | backend
5353 yes | yes | yes | yes
5354 Arguments : none
5355
5356 When there is a firewall or any session-aware component between a client and
5357 a server, and when the protocol involves very long sessions with long idle
5358 periods (eg: remote desktops), there is a risk that one of the intermediate
5359 components decides to expire a session which has remained idle for too long.
5360
5361 Enabling socket-level TCP keep-alives makes the system regularly send packets
5362 to the other end of the connection, leaving it active. The delay between
5363 keep-alive probes is controlled by the system only and depends both on the
5364 operating system and its tuning parameters.
5365
5366 It is important to understand that keep-alive packets are neither emitted nor
5367 received at the application level. It is only the network stacks which sees
5368 them. For this reason, even if one side of the proxy already uses keep-alives
5369 to maintain its connection alive, those keep-alive packets will not be
5370 forwarded to the other side of the proxy.
5371
5372 Please note that this has nothing to do with HTTP keep-alive.
5373
5374 Using option "tcpka" enables the emission of TCP keep-alive probes on both
5375 the client and server sides of a connection. Note that this is meaningful
5376 only in "defaults" or "listen" sections. If this option is used in a
5377 frontend, only the client side will get keep-alives, and if this option is
5378 used in a backend, only the server side will get keep-alives. For this
5379 reason, it is strongly recommended to explicitly use "option clitcpka" and
5380 "option srvtcpka" when the configuration is split between frontends and
5381 backends.
5382
5383 See also : "option clitcpka", "option srvtcpka"
5384
Willy Tarreau844e3c52008-01-11 16:28:18 +01005385
5386option tcplog
5387 Enable advanced logging of TCP connections with session state and timers
5388 May be used in sections : defaults | frontend | listen | backend
5389 yes | yes | yes | yes
5390 Arguments : none
5391
5392 By default, the log output format is very poor, as it only contains the
5393 source and destination addresses, and the instance name. By specifying
5394 "option tcplog", each log line turns into a much richer format including, but
5395 not limited to, the connection timers, the session status, the connections
5396 numbers, the frontend, backend and server name, and of course the source
5397 address and ports. This option is useful for pure TCP proxies in order to
5398 find which of the client or server disconnects or times out. For normal HTTP
5399 proxies, it's better to use "option httplog" which is even more complete.
5400
5401 This option may be set either in the frontend or the backend.
5402
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005403 See also : "option httplog", and section 8 about logging.
Willy Tarreau844e3c52008-01-11 16:28:18 +01005404
5405
Willy Tarreau844e3c52008-01-11 16:28:18 +01005406option transparent
5407no option transparent
5408 Enable client-side transparent proxying
5409 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01005410 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01005411 Arguments : none
5412
5413 This option was introduced in order to provide layer 7 persistence to layer 3
5414 load balancers. The idea is to use the OS's ability to redirect an incoming
5415 connection for a remote address to a local process (here HAProxy), and let
5416 this process know what address was initially requested. When this option is
5417 used, sessions without cookies will be forwarded to the original destination
5418 IP address of the incoming request (which should match that of another
5419 equipment), while requests with cookies will still be forwarded to the
5420 appropriate server.
5421
5422 Note that contrary to a common belief, this option does NOT make HAProxy
5423 present the client's IP to the server when establishing the connection.
5424
Willy Tarreaua1146052011-03-01 09:51:54 +01005425 See also: the "usesrc" argument of the "source" keyword, and the
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005426 "transparent" option of the "bind" keyword.
Willy Tarreau844e3c52008-01-11 16:28:18 +01005427
Willy Tarreaubf1f8162007-12-28 17:42:56 +01005428
Simon Horman98637e52014-06-20 12:30:16 +09005429external-check command <command>
5430 Executable to run when performing an external-check
5431 May be used in sections : defaults | frontend | listen | backend
5432 yes | no | yes | yes
5433
5434 Arguments :
5435 <command> is the external command to run
5436
Simon Horman98637e52014-06-20 12:30:16 +09005437 The arguments passed to the to the command are:
5438
Cyril Bonté777be862014-12-02 21:21:35 +01005439 <proxy_address> <proxy_port> <server_address> <server_port>
Simon Horman98637e52014-06-20 12:30:16 +09005440
Cyril Bonté777be862014-12-02 21:21:35 +01005441 The <proxy_address> and <proxy_port> are derived from the first listener
5442 that is either IPv4, IPv6 or a UNIX socket. In the case of a UNIX socket
5443 listener the proxy_address will be the path of the socket and the
5444 <proxy_port> will be the string "NOT_USED". In a backend section, it's not
5445 possible to determine a listener, and both <proxy_address> and <proxy_port>
5446 will have the string value "NOT_USED".
Simon Horman98637e52014-06-20 12:30:16 +09005447
Cyril Bonté72cda2a2014-12-27 22:28:39 +01005448 Some values are also provided through environment variables.
5449
5450 Environment variables :
5451 HAPROXY_PROXY_ADDR The first bind address if available (or empty if not
5452 applicable, for example in a "backend" section).
5453
5454 HAPROXY_PROXY_ID The backend id.
5455
5456 HAPROXY_PROXY_NAME The backend name.
5457
5458 HAPROXY_PROXY_PORT The first bind port if available (or empty if not
5459 applicable, for example in a "backend" section or
5460 for a UNIX socket).
5461
5462 HAPROXY_SERVER_ADDR The server address.
5463
5464 HAPROXY_SERVER_CURCONN The current number of connections on the server.
5465
5466 HAPROXY_SERVER_ID The server id.
5467
5468 HAPROXY_SERVER_MAXCONN The server max connections.
5469
5470 HAPROXY_SERVER_NAME The server name.
5471
5472 HAPROXY_SERVER_PORT The server port if available (or empty for a UNIX
5473 socket).
5474
5475 PATH The PATH environment variable used when executing
5476 the command may be set using "external-check path".
5477
Simon Horman98637e52014-06-20 12:30:16 +09005478 If the command executed and exits with a zero status then the check is
5479 considered to have passed, otherwise the check is considered to have
5480 failed.
5481
5482 Example :
5483 external-check command /bin/true
5484
5485 See also : "external-check", "option external-check", "external-check path"
5486
5487
5488external-check path <path>
5489 The value of the PATH environment variable used when running an external-check
5490 May be used in sections : defaults | frontend | listen | backend
5491 yes | no | yes | yes
5492
5493 Arguments :
5494 <path> is the path used when executing external command to run
5495
5496 The default path is "".
5497
5498 Example :
5499 external-check path "/usr/bin:/bin"
5500
5501 See also : "external-check", "option external-check",
5502 "external-check command"
5503
5504
Emeric Brun647caf12009-06-30 17:57:00 +02005505persist rdp-cookie
Hervé COMMOWICKa3eb39c2011-08-05 18:48:51 +02005506persist rdp-cookie(<name>)
Emeric Brun647caf12009-06-30 17:57:00 +02005507 Enable RDP cookie-based persistence
5508 May be used in sections : defaults | frontend | listen | backend
5509 yes | no | yes | yes
5510 Arguments :
5511 <name> is the optional name of the RDP cookie to check. If omitted, the
Willy Tarreau61e28f22010-05-16 22:31:05 +02005512 default cookie name "msts" will be used. There currently is no
5513 valid reason to change this name.
Emeric Brun647caf12009-06-30 17:57:00 +02005514
5515 This statement enables persistence based on an RDP cookie. The RDP cookie
5516 contains all information required to find the server in the list of known
5517 servers. So when this option is set in the backend, the request is analysed
5518 and if an RDP cookie is found, it is decoded. If it matches a known server
5519 which is still UP (or if "option persist" is set), then the connection is
5520 forwarded to this server.
5521
5522 Note that this only makes sense in a TCP backend, but for this to work, the
5523 frontend must have waited long enough to ensure that an RDP cookie is present
5524 in the request buffer. This is the same requirement as with the "rdp-cookie"
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01005525 load-balancing method. Thus it is highly recommended to put all statements in
Emeric Brun647caf12009-06-30 17:57:00 +02005526 a single "listen" section.
5527
Willy Tarreau61e28f22010-05-16 22:31:05 +02005528 Also, it is important to understand that the terminal server will emit this
5529 RDP cookie only if it is configured for "token redirection mode", which means
5530 that the "IP address redirection" option is disabled.
5531
Emeric Brun647caf12009-06-30 17:57:00 +02005532 Example :
5533 listen tse-farm
5534 bind :3389
5535 # wait up to 5s for an RDP cookie in the request
5536 tcp-request inspect-delay 5s
5537 tcp-request content accept if RDP_COOKIE
5538 # apply RDP cookie persistence
5539 persist rdp-cookie
5540 # if server is unknown, let's balance on the same cookie.
Cyril Bontédc4d9032012-04-08 21:57:39 +02005541 # alternatively, "balance leastconn" may be useful too.
Emeric Brun647caf12009-06-30 17:57:00 +02005542 balance rdp-cookie
5543 server srv1 1.1.1.1:3389
5544 server srv2 1.1.1.2:3389
5545
Simon Hormanab814e02011-06-24 14:50:20 +09005546 See also : "balance rdp-cookie", "tcp-request", the "req_rdp_cookie" ACL and
5547 the rdp_cookie pattern fetch function.
Emeric Brun647caf12009-06-30 17:57:00 +02005548
5549
Willy Tarreau3a7d2072009-03-05 23:48:25 +01005550rate-limit sessions <rate>
5551 Set a limit on the number of new sessions accepted per second on a frontend
5552 May be used in sections : defaults | frontend | listen | backend
5553 yes | yes | yes | no
5554 Arguments :
5555 <rate> The <rate> parameter is an integer designating the maximum number
5556 of new sessions per second to accept on the frontend.
5557
5558 When the frontend reaches the specified number of new sessions per second, it
5559 stops accepting new connections until the rate drops below the limit again.
5560 During this time, the pending sessions will be kept in the socket's backlog
5561 (in system buffers) and haproxy will not even be aware that sessions are
5562 pending. When applying very low limit on a highly loaded service, it may make
5563 sense to increase the socket's backlog using the "backlog" keyword.
5564
5565 This feature is particularly efficient at blocking connection-based attacks
5566 or service abuse on fragile servers. Since the session rate is measured every
5567 millisecond, it is extremely accurate. Also, the limit applies immediately,
5568 no delay is needed at all to detect the threshold.
5569
5570 Example : limit the connection rate on SMTP to 10 per second max
5571 listen smtp
5572 mode tcp
5573 bind :25
5574 rate-limit sessions 10
5575 server 127.0.0.1:1025
5576
Willy Tarreaua17c2d92011-07-25 08:16:20 +02005577 Note : when the maximum rate is reached, the frontend's status is not changed
5578 but its sockets appear as "WAITING" in the statistics if the
5579 "socket-stats" option is enabled.
Willy Tarreau3a7d2072009-03-05 23:48:25 +01005580
5581 See also : the "backlog" keyword and the "fe_sess_rate" ACL criterion.
5582
5583
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005584redirect location <loc> [code <code>] <option> [{if | unless} <condition>]
5585redirect prefix <pfx> [code <code>] <option> [{if | unless} <condition>]
5586redirect scheme <sch> [code <code>] <option> [{if | unless} <condition>]
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005587 Return an HTTP redirection if/unless a condition is matched
5588 May be used in sections : defaults | frontend | listen | backend
5589 no | yes | yes | yes
5590
5591 If/unless the condition is matched, the HTTP request will lead to a redirect
Willy Tarreauf285f542010-01-03 20:03:03 +01005592 response. If no condition is specified, the redirect applies unconditionally.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005593
Willy Tarreau0140f252008-11-19 21:07:09 +01005594 Arguments :
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005595 <loc> With "redirect location", the exact value in <loc> is placed into
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005596 the HTTP "Location" header. When used in an "http-request" rule,
5597 <loc> value follows the log-format rules and can include some
5598 dynamic values (see Custom Log Format in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005599
5600 <pfx> With "redirect prefix", the "Location" header is built from the
5601 concatenation of <pfx> and the complete URI path, including the
5602 query string, unless the "drop-query" option is specified (see
5603 below). As a special case, if <pfx> equals exactly "/", then
5604 nothing is inserted before the original URI. It allows one to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005605 redirect to the same URL (for instance, to insert a cookie). When
5606 used in an "http-request" rule, <pfx> value follows the log-format
5607 rules and can include some dynamic values (see Custom Log Format
5608 in section 8.2.4).
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005609
5610 <sch> With "redirect scheme", then the "Location" header is built by
5611 concatenating <sch> with "://" then the first occurrence of the
5612 "Host" header, and then the URI path, including the query string
5613 unless the "drop-query" option is specified (see below). If no
5614 path is found or if the path is "*", then "/" is used instead. If
5615 no "Host" header is found, then an empty host component will be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03005616 returned, which most recent browsers interpret as redirecting to
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005617 the same host. This directive is mostly used to redirect HTTP to
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005618 HTTPS. When used in an "http-request" rule, <sch> value follows
5619 the log-format rules and can include some dynamic values (see
5620 Custom Log Format in section 8.2.4).
Willy Tarreau0140f252008-11-19 21:07:09 +01005621
5622 <code> The code is optional. It indicates which type of HTTP redirection
Willy Tarreaub67fdc42013-03-29 19:28:11 +01005623 is desired. Only codes 301, 302, 303, 307 and 308 are supported,
5624 with 302 used by default if no code is specified. 301 means
5625 "Moved permanently", and a browser may cache the Location. 302
5626 means "Moved permanently" and means that the browser should not
5627 cache the redirection. 303 is equivalent to 302 except that the
5628 browser will fetch the location with a GET method. 307 is just
5629 like 302 but makes it clear that the same method must be reused.
5630 Likewise, 308 replaces 301 if the same method must be used.
Willy Tarreau0140f252008-11-19 21:07:09 +01005631
5632 <option> There are several options which can be specified to adjust the
5633 expected behaviour of a redirection :
5634
5635 - "drop-query"
5636 When this keyword is used in a prefix-based redirection, then the
5637 location will be set without any possible query-string, which is useful
5638 for directing users to a non-secure page for instance. It has no effect
5639 with a location-type redirect.
5640
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01005641 - "append-slash"
5642 This keyword may be used in conjunction with "drop-query" to redirect
5643 users who use a URL not ending with a '/' to the same one with the '/'.
5644 It can be useful to ensure that search engines will only see one URL.
5645 For this, a return code 301 is preferred.
5646
Willy Tarreau0140f252008-11-19 21:07:09 +01005647 - "set-cookie NAME[=value]"
5648 A "Set-Cookie" header will be added with NAME (and optionally "=value")
5649 to the response. This is sometimes used to indicate that a user has
5650 been seen, for instance to protect against some types of DoS. No other
5651 cookie option is added, so the cookie will be a session cookie. Note
5652 that for a browser, a sole cookie name without an equal sign is
5653 different from a cookie with an equal sign.
5654
5655 - "clear-cookie NAME[=]"
5656 A "Set-Cookie" header will be added with NAME (and optionally "="), but
5657 with the "Max-Age" attribute set to zero. This will tell the browser to
5658 delete this cookie. It is useful for instance on logout pages. It is
5659 important to note that clearing the cookie "NAME" will not remove a
5660 cookie set with "NAME=value". You have to clear the cookie "NAME=" for
5661 that, because the browser makes the difference.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005662
5663 Example: move the login URL only to HTTPS.
5664 acl clear dst_port 80
5665 acl secure dst_port 8080
5666 acl login_page url_beg /login
Willy Tarreau0140f252008-11-19 21:07:09 +01005667 acl logout url_beg /logout
Willy Tarreau79da4692008-11-19 20:03:04 +01005668 acl uid_given url_reg /login?userid=[^&]+
Willy Tarreau0140f252008-11-19 21:07:09 +01005669 acl cookie_set hdr_sub(cookie) SEEN=1
5670
5671 redirect prefix https://mysite.com set-cookie SEEN=1 if !cookie_set
Willy Tarreau79da4692008-11-19 20:03:04 +01005672 redirect prefix https://mysite.com if login_page !secure
5673 redirect prefix http://mysite.com drop-query if login_page !uid_given
5674 redirect location http://mysite.com/ if !login_page secure
Willy Tarreau0140f252008-11-19 21:07:09 +01005675 redirect location / clear-cookie USERID= if logout
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005676
Willy Tarreau81e3b4f2010-01-10 00:42:19 +01005677 Example: send redirects for request for articles without a '/'.
5678 acl missing_slash path_reg ^/article/[^/]*$
5679 redirect code 301 prefix / drop-query append-slash if missing_slash
5680
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005681 Example: redirect all HTTP traffic to HTTPS when SSL is handled by haproxy.
David BERARDe7153042012-11-03 00:11:31 +01005682 redirect scheme https if !{ ssl_fc }
Willy Tarreau2e1dca82012-09-12 08:43:15 +02005683
Thierry FOURNIERd18cd0f2013-11-29 12:15:45 +01005684 Example: append 'www.' prefix in front of all hosts not having it
5685 http-request redirect code 301 location www.%[hdr(host)]%[req.uri] \
5686 unless { hdr_beg(host) -i www }
5687
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005688 See section 7 about ACL usage.
Willy Tarreaub463dfb2008-06-07 23:08:56 +02005689
5690
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005691redisp (deprecated)
5692redispatch (deprecated)
5693 Enable or disable session redistribution in case of connection failure
5694 May be used in sections: defaults | frontend | listen | backend
5695 yes | no | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005696 Arguments : none
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005697
5698 In HTTP mode, if a server designated by a cookie is down, clients may
5699 definitely stick to it because they cannot flush the cookie, so they will not
5700 be able to access the service anymore.
5701
5702 Specifying "redispatch" will allow the proxy to break their persistence and
5703 redistribute them to a working server.
5704
5705 It also allows to retry last connection to another server in case of multiple
5706 connection failures. Of course, it requires having "retries" set to a nonzero
5707 value.
Willy Tarreaud72758d2010-01-12 10:42:19 +01005708
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01005709 This form is deprecated, do not use it in any new configuration, use the new
5710 "option redispatch" instead.
5711
5712 See also : "option redispatch"
5713
Willy Tarreaueabeafa2008-01-16 16:17:06 +01005714
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005715reqadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01005716 Add a header at the end of the HTTP request
5717 May be used in sections : defaults | frontend | listen | backend
5718 no | yes | yes | yes
5719 Arguments :
5720 <string> is the complete line to be added. Any space or known delimiter
5721 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005722 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005723
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005724 <cond> is an optional matching condition built from ACLs. It makes it
5725 possible to ignore this rule when other conditions are not met.
5726
Willy Tarreau303c0352008-01-17 19:01:39 +01005727 A new line consisting in <string> followed by a line feed will be added after
5728 the last header of an HTTP request.
5729
5730 Header transformations only apply to traffic which passes through HAProxy,
5731 and not to traffic generated by HAProxy, such as health-checks or error
5732 responses.
5733
Willy Tarreau8abd4cd2010-01-31 14:30:44 +01005734 Example : add "X-Proto: SSL" to requests coming via port 81
5735 acl is-ssl dst_port 81
5736 reqadd X-Proto:\ SSL if is-ssl
5737
5738 See also: "rspadd", section 6 about HTTP header manipulation, and section 7
5739 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005740
5741
Willy Tarreau5321c422010-01-28 20:35:13 +01005742reqallow <search> [{if | unless} <cond>]
5743reqiallow <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005744 Definitely allow an HTTP request if a line matches a regular expression
5745 May be used in sections : defaults | frontend | listen | backend
5746 no | yes | yes | yes
5747 Arguments :
5748 <search> is the regular expression applied to HTTP headers and to the
5749 request line. This is an extended regular expression. Parenthesis
5750 grouping is supported and no preliminary backslash is required.
5751 Any space or known delimiter must be escaped using a backslash
5752 ('\'). The pattern applies to a full line at a time. The
5753 "reqallow" keyword strictly matches case while "reqiallow"
5754 ignores case.
5755
Willy Tarreau5321c422010-01-28 20:35:13 +01005756 <cond> is an optional matching condition built from ACLs. It makes it
5757 possible to ignore this rule when other conditions are not met.
5758
Willy Tarreau303c0352008-01-17 19:01:39 +01005759 A request containing any line which matches extended regular expression
5760 <search> will mark the request as allowed, even if any later test would
5761 result in a deny. The test applies both to the request line and to request
5762 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01005763 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01005764
5765 It is easier, faster and more powerful to use ACLs to write access policies.
5766 Reqdeny, reqallow and reqpass should be avoided in new designs.
5767
5768 Example :
5769 # allow www.* but refuse *.local
5770 reqiallow ^Host:\ www\.
5771 reqideny ^Host:\ .*\.local
5772
Willy Tarreau5321c422010-01-28 20:35:13 +01005773 See also: "reqdeny", "block", section 6 about HTTP header manipulation, and
5774 section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005775
5776
Willy Tarreau5321c422010-01-28 20:35:13 +01005777reqdel <search> [{if | unless} <cond>]
5778reqidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005779 Delete all headers matching a regular expression in an HTTP request
5780 May be used in sections : defaults | frontend | listen | backend
5781 no | yes | yes | yes
5782 Arguments :
5783 <search> is the regular expression applied to HTTP headers and to the
5784 request line. This is an extended regular expression. Parenthesis
5785 grouping is supported and no preliminary backslash is required.
5786 Any space or known delimiter must be escaped using a backslash
5787 ('\'). The pattern applies to a full line at a time. The "reqdel"
5788 keyword strictly matches case while "reqidel" ignores case.
5789
Willy Tarreau5321c422010-01-28 20:35:13 +01005790 <cond> is an optional matching condition built from ACLs. It makes it
5791 possible to ignore this rule when other conditions are not met.
5792
Willy Tarreau303c0352008-01-17 19:01:39 +01005793 Any header line matching extended regular expression <search> in the request
5794 will be completely deleted. Most common use of this is to remove unwanted
5795 and/or dangerous headers or cookies from a request before passing it to the
5796 next servers.
5797
5798 Header transformations only apply to traffic which passes through HAProxy,
5799 and not to traffic generated by HAProxy, such as health-checks or error
5800 responses. Keep in mind that header names are not case-sensitive.
5801
5802 Example :
5803 # remove X-Forwarded-For header and SERVER cookie
5804 reqidel ^X-Forwarded-For:.*
5805 reqidel ^Cookie:.*SERVER=
5806
Willy Tarreau5321c422010-01-28 20:35:13 +01005807 See also: "reqadd", "reqrep", "rspdel", section 6 about HTTP header
5808 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005809
5810
Willy Tarreau5321c422010-01-28 20:35:13 +01005811reqdeny <search> [{if | unless} <cond>]
5812reqideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005813 Deny an HTTP request if a line matches a regular expression
5814 May be used in sections : defaults | frontend | listen | backend
5815 no | yes | yes | yes
5816 Arguments :
5817 <search> is the regular expression applied to HTTP headers and to the
5818 request line. This is an extended regular expression. Parenthesis
5819 grouping is supported and no preliminary backslash is required.
5820 Any space or known delimiter must be escaped using a backslash
5821 ('\'). The pattern applies to a full line at a time. The
5822 "reqdeny" keyword strictly matches case while "reqideny" ignores
5823 case.
5824
Willy Tarreau5321c422010-01-28 20:35:13 +01005825 <cond> is an optional matching condition built from ACLs. It makes it
5826 possible to ignore this rule when other conditions are not met.
5827
Willy Tarreau303c0352008-01-17 19:01:39 +01005828 A request containing any line which matches extended regular expression
5829 <search> will mark the request as denied, even if any later test would
5830 result in an allow. The test applies both to the request line and to request
5831 headers. Keep in mind that URLs in request line are case-sensitive while
Willy Tarreaud72758d2010-01-12 10:42:19 +01005832 header names are not.
Willy Tarreau303c0352008-01-17 19:01:39 +01005833
Willy Tarreauced27012008-01-17 20:35:34 +01005834 A denied request will generate an "HTTP 403 forbidden" response once the
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01005835 complete request has been parsed. This is consistent with what is practiced
Willy Tarreaud72758d2010-01-12 10:42:19 +01005836 using ACLs.
Willy Tarreauced27012008-01-17 20:35:34 +01005837
Willy Tarreau303c0352008-01-17 19:01:39 +01005838 It is easier, faster and more powerful to use ACLs to write access policies.
5839 Reqdeny, reqallow and reqpass should be avoided in new designs.
5840
5841 Example :
5842 # refuse *.local, then allow www.*
5843 reqideny ^Host:\ .*\.local
5844 reqiallow ^Host:\ www\.
5845
Willy Tarreau5321c422010-01-28 20:35:13 +01005846 See also: "reqallow", "rspdeny", "block", section 6 about HTTP header
5847 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005848
5849
Willy Tarreau5321c422010-01-28 20:35:13 +01005850reqpass <search> [{if | unless} <cond>]
5851reqipass <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005852 Ignore any HTTP request line matching a regular expression in next rules
5853 May be used in sections : defaults | frontend | listen | backend
5854 no | yes | yes | yes
5855 Arguments :
5856 <search> is the regular expression applied to HTTP headers and to the
5857 request line. This is an extended regular expression. Parenthesis
5858 grouping is supported and no preliminary backslash is required.
5859 Any space or known delimiter must be escaped using a backslash
5860 ('\'). The pattern applies to a full line at a time. The
5861 "reqpass" keyword strictly matches case while "reqipass" ignores
5862 case.
5863
Willy Tarreau5321c422010-01-28 20:35:13 +01005864 <cond> is an optional matching condition built from ACLs. It makes it
5865 possible to ignore this rule when other conditions are not met.
5866
Willy Tarreau303c0352008-01-17 19:01:39 +01005867 A request containing any line which matches extended regular expression
5868 <search> will skip next rules, without assigning any deny or allow verdict.
5869 The test applies both to the request line and to request headers. Keep in
5870 mind that URLs in request line are case-sensitive while header names are not.
5871
5872 It is easier, faster and more powerful to use ACLs to write access policies.
5873 Reqdeny, reqallow and reqpass should be avoided in new designs.
5874
5875 Example :
5876 # refuse *.local, then allow www.*, but ignore "www.private.local"
5877 reqipass ^Host:\ www.private\.local
5878 reqideny ^Host:\ .*\.local
5879 reqiallow ^Host:\ www\.
5880
Willy Tarreau5321c422010-01-28 20:35:13 +01005881 See also: "reqallow", "reqdeny", "block", section 6 about HTTP header
5882 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005883
5884
Willy Tarreau5321c422010-01-28 20:35:13 +01005885reqrep <search> <string> [{if | unless} <cond>]
5886reqirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005887 Replace a regular expression with a string in an HTTP request line
5888 May be used in sections : defaults | frontend | listen | backend
5889 no | yes | yes | yes
5890 Arguments :
5891 <search> is the regular expression applied to HTTP headers and to the
5892 request line. This is an extended regular expression. Parenthesis
5893 grouping is supported and no preliminary backslash is required.
5894 Any space or known delimiter must be escaped using a backslash
5895 ('\'). The pattern applies to a full line at a time. The "reqrep"
5896 keyword strictly matches case while "reqirep" ignores case.
5897
5898 <string> is the complete line to be added. Any space or known delimiter
5899 must be escaped using a backslash ('\'). References to matched
5900 pattern groups are possible using the common \N form, with N
5901 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02005902 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01005903
Willy Tarreau5321c422010-01-28 20:35:13 +01005904 <cond> is an optional matching condition built from ACLs. It makes it
5905 possible to ignore this rule when other conditions are not met.
5906
Willy Tarreau303c0352008-01-17 19:01:39 +01005907 Any line matching extended regular expression <search> in the request (both
5908 the request line and header lines) will be completely replaced with <string>.
5909 Most common use of this is to rewrite URLs or domain names in "Host" headers.
5910
5911 Header transformations only apply to traffic which passes through HAProxy,
5912 and not to traffic generated by HAProxy, such as health-checks or error
5913 responses. Note that for increased readability, it is suggested to add enough
5914 spaces between the request and the response. Keep in mind that URLs in
5915 request line are case-sensitive while header names are not.
5916
5917 Example :
5918 # replace "/static/" with "/" at the beginning of any request path.
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04005919 reqrep ^([^\ :]*)\ /static/(.*) \1\ /\2
Willy Tarreau303c0352008-01-17 19:01:39 +01005920 # replace "www.mydomain.com" with "www" in the host name.
5921 reqirep ^Host:\ www.mydomain.com Host:\ www
5922
Dmitry Sivachenkof6f4f7b2012-10-21 18:10:25 +04005923 See also: "reqadd", "reqdel", "rsprep", "tune.bufsize", section 6 about
5924 HTTP header manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005925
5926
Willy Tarreau5321c422010-01-28 20:35:13 +01005927reqtarpit <search> [{if | unless} <cond>]
5928reqitarpit <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01005929 Tarpit an HTTP request containing a line matching a regular expression
5930 May be used in sections : defaults | frontend | listen | backend
5931 no | yes | yes | yes
5932 Arguments :
5933 <search> is the regular expression applied to HTTP headers and to the
5934 request line. This is an extended regular expression. Parenthesis
5935 grouping is supported and no preliminary backslash is required.
5936 Any space or known delimiter must be escaped using a backslash
5937 ('\'). The pattern applies to a full line at a time. The
5938 "reqtarpit" keyword strictly matches case while "reqitarpit"
5939 ignores case.
5940
Willy Tarreau5321c422010-01-28 20:35:13 +01005941 <cond> is an optional matching condition built from ACLs. It makes it
5942 possible to ignore this rule when other conditions are not met.
5943
Willy Tarreau303c0352008-01-17 19:01:39 +01005944 A request containing any line which matches extended regular expression
5945 <search> will be tarpitted, which means that it will connect to nowhere, will
Willy Tarreauced27012008-01-17 20:35:34 +01005946 be kept open for a pre-defined time, then will return an HTTP error 500 so
5947 that the attacker does not suspect it has been tarpitted. The status 500 will
5948 be reported in the logs, but the completion flags will indicate "PT". The
Willy Tarreau303c0352008-01-17 19:01:39 +01005949 delay is defined by "timeout tarpit", or "timeout connect" if the former is
5950 not set.
5951
5952 The goal of the tarpit is to slow down robots attacking servers with
5953 identifiable requests. Many robots limit their outgoing number of connections
5954 and stay connected waiting for a reply which can take several minutes to
5955 come. Depending on the environment and attack, it may be particularly
5956 efficient at reducing the load on the network and firewalls.
5957
Willy Tarreau5321c422010-01-28 20:35:13 +01005958 Examples :
Willy Tarreau303c0352008-01-17 19:01:39 +01005959 # ignore user-agents reporting any flavour of "Mozilla" or "MSIE", but
5960 # block all others.
5961 reqipass ^User-Agent:\.*(Mozilla|MSIE)
5962 reqitarpit ^User-Agent:
5963
Willy Tarreau5321c422010-01-28 20:35:13 +01005964 # block bad guys
5965 acl badguys src 10.1.0.3 172.16.13.20/28
5966 reqitarpit . if badguys
5967
5968 See also: "reqallow", "reqdeny", "reqpass", section 6 about HTTP header
5969 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01005970
5971
Willy Tarreaue5c5ce92008-06-20 17:27:19 +02005972retries <value>
5973 Set the number of retries to perform on a server after a connection failure
5974 May be used in sections: defaults | frontend | listen | backend
5975 yes | no | yes | yes
5976 Arguments :
5977 <value> is the number of times a connection attempt should be retried on
5978 a server when a connection either is refused or times out. The
5979 default value is 3.
5980
5981 It is important to understand that this value applies to the number of
5982 connection attempts, not full requests. When a connection has effectively
5983 been established to a server, there will be no more retry.
5984
5985 In order to avoid immediate reconnections to a server which is restarting,
5986 a turn-around timer of 1 second is applied before a retry occurs.
5987
5988 When "option redispatch" is set, the last retry may be performed on another
5989 server even if a cookie references a different server.
5990
5991 See also : "option redispatch"
5992
5993
Willy Tarreaufdb563c2010-01-31 15:43:27 +01005994rspadd <string> [{if | unless} <cond>]
Willy Tarreau303c0352008-01-17 19:01:39 +01005995 Add a header at the end of the HTTP response
5996 May be used in sections : defaults | frontend | listen | backend
5997 no | yes | yes | yes
5998 Arguments :
5999 <string> is the complete line to be added. Any space or known delimiter
6000 must be escaped using a backslash ('\'). Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006001 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006002
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006003 <cond> is an optional matching condition built from ACLs. It makes it
6004 possible to ignore this rule when other conditions are not met.
6005
Willy Tarreau303c0352008-01-17 19:01:39 +01006006 A new line consisting in <string> followed by a line feed will be added after
6007 the last header of an HTTP response.
6008
6009 Header transformations only apply to traffic which passes through HAProxy,
6010 and not to traffic generated by HAProxy, such as health-checks or error
6011 responses.
6012
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006013 See also: "reqadd", section 6 about HTTP header manipulation, and section 7
6014 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006015
6016
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006017rspdel <search> [{if | unless} <cond>]
6018rspidel <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006019 Delete all headers matching a regular expression in an HTTP response
6020 May be used in sections : defaults | frontend | listen | backend
6021 no | yes | yes | yes
6022 Arguments :
6023 <search> is the regular expression applied to HTTP headers and to the
6024 response line. This is an extended regular expression, so
6025 parenthesis grouping is supported and no preliminary backslash
6026 is required. Any space or known delimiter must be escaped using
6027 a backslash ('\'). The pattern applies to a full line at a time.
6028 The "rspdel" keyword strictly matches case while "rspidel"
6029 ignores case.
6030
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006031 <cond> is an optional matching condition built from ACLs. It makes it
6032 possible to ignore this rule when other conditions are not met.
6033
Willy Tarreau303c0352008-01-17 19:01:39 +01006034 Any header line matching extended regular expression <search> in the response
6035 will be completely deleted. Most common use of this is to remove unwanted
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006036 and/or sensitive headers or cookies from a response before passing it to the
Willy Tarreau303c0352008-01-17 19:01:39 +01006037 client.
6038
6039 Header transformations only apply to traffic which passes through HAProxy,
6040 and not to traffic generated by HAProxy, such as health-checks or error
6041 responses. Keep in mind that header names are not case-sensitive.
6042
6043 Example :
6044 # remove the Server header from responses
Willy Tarreau5e80e022013-05-25 08:31:25 +02006045 rspidel ^Server:.*
Willy Tarreau303c0352008-01-17 19:01:39 +01006046
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006047 See also: "rspadd", "rsprep", "reqdel", section 6 about HTTP header
6048 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006049
6050
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006051rspdeny <search> [{if | unless} <cond>]
6052rspideny <search> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006053 Block an HTTP response if a line matches a regular expression
6054 May be used in sections : defaults | frontend | listen | backend
6055 no | yes | yes | yes
6056 Arguments :
6057 <search> is the regular expression applied to HTTP headers and to the
6058 response line. This is an extended regular expression, so
6059 parenthesis grouping is supported and no preliminary backslash
6060 is required. Any space or known delimiter must be escaped using
6061 a backslash ('\'). The pattern applies to a full line at a time.
6062 The "rspdeny" keyword strictly matches case while "rspideny"
6063 ignores case.
6064
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006065 <cond> is an optional matching condition built from ACLs. It makes it
6066 possible to ignore this rule when other conditions are not met.
6067
Willy Tarreau303c0352008-01-17 19:01:39 +01006068 A response containing any line which matches extended regular expression
6069 <search> will mark the request as denied. The test applies both to the
6070 response line and to response headers. Keep in mind that header names are not
6071 case-sensitive.
6072
6073 Main use of this keyword is to prevent sensitive information leak and to
Willy Tarreauced27012008-01-17 20:35:34 +01006074 block the response before it reaches the client. If a response is denied, it
6075 will be replaced with an HTTP 502 error so that the client never retrieves
6076 any sensitive data.
Willy Tarreau303c0352008-01-17 19:01:39 +01006077
6078 It is easier, faster and more powerful to use ACLs to write access policies.
6079 Rspdeny should be avoided in new designs.
6080
6081 Example :
6082 # Ensure that no content type matching ms-word will leak
6083 rspideny ^Content-type:\.*/ms-word
6084
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006085 See also: "reqdeny", "acl", "block", section 6 about HTTP header manipulation
6086 and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006087
6088
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006089rsprep <search> <string> [{if | unless} <cond>]
6090rspirep <search> <string> [{if | unless} <cond>] (ignore case)
Willy Tarreau303c0352008-01-17 19:01:39 +01006091 Replace a regular expression with a string in an HTTP response line
6092 May be used in sections : defaults | frontend | listen | backend
6093 no | yes | yes | yes
6094 Arguments :
6095 <search> is the regular expression applied to HTTP headers and to the
6096 response line. This is an extended regular expression, so
6097 parenthesis grouping is supported and no preliminary backslash
6098 is required. Any space or known delimiter must be escaped using
6099 a backslash ('\'). The pattern applies to a full line at a time.
6100 The "rsprep" keyword strictly matches case while "rspirep"
6101 ignores case.
6102
6103 <string> is the complete line to be added. Any space or known delimiter
6104 must be escaped using a backslash ('\'). References to matched
6105 pattern groups are possible using the common \N form, with N
6106 being a single digit between 0 and 9. Please refer to section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006107 6 about HTTP header manipulation for more information.
Willy Tarreau303c0352008-01-17 19:01:39 +01006108
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006109 <cond> is an optional matching condition built from ACLs. It makes it
6110 possible to ignore this rule when other conditions are not met.
6111
Willy Tarreau303c0352008-01-17 19:01:39 +01006112 Any line matching extended regular expression <search> in the response (both
6113 the response line and header lines) will be completely replaced with
6114 <string>. Most common use of this is to rewrite Location headers.
6115
6116 Header transformations only apply to traffic which passes through HAProxy,
6117 and not to traffic generated by HAProxy, such as health-checks or error
6118 responses. Note that for increased readability, it is suggested to add enough
6119 spaces between the request and the response. Keep in mind that header names
6120 are not case-sensitive.
6121
6122 Example :
6123 # replace "Location: 127.0.0.1:8080" with "Location: www.mydomain.com"
6124 rspirep ^Location:\ 127.0.0.1:8080 Location:\ www.mydomain.com
6125
Willy Tarreaufdb563c2010-01-31 15:43:27 +01006126 See also: "rspadd", "rspdel", "reqrep", section 6 about HTTP header
6127 manipulation, and section 7 about ACLs.
Willy Tarreau303c0352008-01-17 19:01:39 +01006128
6129
David du Colombier486df472011-03-17 10:40:26 +01006130server <name> <address>[:[port]] [param*]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006131 Declare a server in a backend
6132 May be used in sections : defaults | frontend | listen | backend
6133 no | no | yes | yes
6134 Arguments :
6135 <name> is the internal name assigned to this server. This name will
Cyril Bonté941a0c62012-10-15 19:44:24 +02006136 appear in logs and alerts. If "http-send-name-header" is
Mark Lamourinec2247f02012-01-04 13:02:01 -05006137 set, it will be added to the request header sent to the server.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006138
David du Colombier486df472011-03-17 10:40:26 +01006139 <address> is the IPv4 or IPv6 address of the server. Alternatively, a
6140 resolvable hostname is supported, but this name will be resolved
6141 during start-up. Address "0.0.0.0" or "*" has a special meaning.
6142 It indicates that the connection will be forwarded to the same IP
Willy Tarreaud669a4f2010-07-13 14:49:50 +02006143 address as the one from the client connection. This is useful in
6144 transparent proxy architectures where the client's connection is
6145 intercepted and haproxy must forward to the original destination
6146 address. This is more or less what the "transparent" keyword does
6147 except that with a server it's possible to limit concurrency and
Willy Tarreau24709282013-03-10 21:32:12 +01006148 to report statistics. Optionally, an address family prefix may be
6149 used before the address to force the family regardless of the
6150 address format, which can be useful to specify a path to a unix
6151 socket with no slash ('/'). Currently supported prefixes are :
6152 - 'ipv4@' -> address is always IPv4
6153 - 'ipv6@' -> address is always IPv6
6154 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02006155 - 'abns@' -> address is in abstract namespace (Linux only)
Willy Tarreaudad36a32013-03-11 01:20:04 +01006156 Any part of the address string may reference any number of
6157 environment variables by preceding their name with a dollar
6158 sign ('$') and optionally enclosing them with braces ('{}'),
6159 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006160
Willy Tarreaub6205fd2012-09-24 12:27:33 +02006161 <port> is an optional port specification. If set, all connections will
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006162 be sent to this port. If unset, the same port the client
6163 connected to will be used. The port may also be prefixed by a "+"
6164 or a "-". In this case, the server's port will be determined by
6165 adding this value to the client's port.
6166
6167 <param*> is a list of parameters for this server. The "server" keywords
6168 accepts an important number of options and has a complete section
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006169 dedicated to it. Please refer to section 5 for more details.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006170
6171 Examples :
6172 server first 10.1.1.1:1080 cookie first check inter 1000
6173 server second 10.1.1.2:1080 cookie second check inter 1000
Willy Tarreau24709282013-03-10 21:32:12 +01006174 server transp ipv4@
Willy Tarreaudad36a32013-03-11 01:20:04 +01006175 server backup ${SRV_BACKUP}:1080 backup
6176 server www1_dc1 ${LAN_DC1}.101:80
6177 server www1_dc2 ${LAN_DC2}.101:80
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006178
Mark Lamourinec2247f02012-01-04 13:02:01 -05006179 See also: "default-server", "http-send-name-header" and section 5 about
6180 server options
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006181
6182
6183source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02006184source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreaud53f96b2009-02-04 18:46:54 +01006185source <addr>[:<port>] [interface <name>]
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006186 Set the source address for outgoing connections
6187 May be used in sections : defaults | frontend | listen | backend
6188 yes | no | yes | yes
6189 Arguments :
6190 <addr> is the IPv4 address HAProxy will bind to before connecting to a
6191 server. This address is also used as a source for health checks.
Willy Tarreau24709282013-03-10 21:32:12 +01006192
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006193 The default value of 0.0.0.0 means that the system will select
Willy Tarreau24709282013-03-10 21:32:12 +01006194 the most appropriate address to reach its destination. Optionally
6195 an address family prefix may be used before the address to force
6196 the family regardless of the address format, which can be useful
6197 to specify a path to a unix socket with no slash ('/'). Currently
6198 supported prefixes are :
6199 - 'ipv4@' -> address is always IPv4
6200 - 'ipv6@' -> address is always IPv6
6201 - 'unix@' -> address is a path to a local unix socket
Willy Tarreauccfccef2014-05-10 01:49:15 +02006202 - 'abns@' -> address is in abstract namespace (Linux only)
Willy Tarreaudad36a32013-03-11 01:20:04 +01006203 Any part of the address string may reference any number of
6204 environment variables by preceding their name with a dollar
6205 sign ('$') and optionally enclosing them with braces ('{}'),
6206 similarly to what is done in Bourne shell.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006207
6208 <port> is an optional port. It is normally not needed but may be useful
6209 in some very specific contexts. The default value of zero means
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02006210 the system will select a free port. Note that port ranges are not
6211 supported in the backend. If you want to force port ranges, you
6212 have to specify them on each "server" line.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006213
6214 <addr2> is the IP address to present to the server when connections are
6215 forwarded in full transparent proxy mode. This is currently only
6216 supported on some patched Linux kernels. When this address is
6217 specified, clients connecting to the server will be presented
6218 with this address, while health checks will still use the address
6219 <addr>.
6220
6221 <port2> is the optional port to present to the server when connections
6222 are forwarded in full transparent proxy mode (see <addr2> above).
6223 The default value of zero means the system will select a free
6224 port.
6225
Willy Tarreaubce70882009-09-07 11:51:47 +02006226 <hdr> is the name of a HTTP header in which to fetch the IP to bind to.
6227 This is the name of a comma-separated header list which can
6228 contain multiple IP addresses. By default, the last occurrence is
6229 used. This is designed to work with the X-Forwarded-For header
Baptiste Assmannea3e73b2013-02-02 23:47:49 +01006230 and to automatically bind to the client's IP address as seen
Willy Tarreaubce70882009-09-07 11:51:47 +02006231 by previous proxy, typically Stunnel. In order to use another
6232 occurrence from the last one, please see the <occ> parameter
6233 below. When the header (or occurrence) is not found, no binding
6234 is performed so that the proxy's default IP address is used. Also
6235 keep in mind that the header name is case insensitive, as for any
6236 HTTP header.
6237
6238 <occ> is the occurrence number of a value to be used in a multi-value
6239 header. This is to be used in conjunction with "hdr_ip(<hdr>)",
Jamie Gloudonaaa21002012-08-25 00:18:33 -04006240 in order to specify which occurrence to use for the source IP
Willy Tarreaubce70882009-09-07 11:51:47 +02006241 address. Positive values indicate a position from the first
6242 occurrence, 1 being the first one. Negative values indicate
6243 positions relative to the last one, -1 being the last one. This
6244 is helpful for situations where an X-Forwarded-For header is set
6245 at the entry point of an infrastructure and must be used several
6246 proxy layers away. When this value is not specified, -1 is
6247 assumed. Passing a zero here disables the feature.
6248
Willy Tarreaud53f96b2009-02-04 18:46:54 +01006249 <name> is an optional interface name to which to bind to for outgoing
6250 traffic. On systems supporting this features (currently, only
6251 Linux), this allows one to bind all traffic to the server to
6252 this interface even if it is not the one the system would select
6253 based on routing tables. This should be used with extreme care.
6254 Note that using this option requires root privileges.
6255
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006256 The "source" keyword is useful in complex environments where a specific
6257 address only is allowed to connect to the servers. It may be needed when a
6258 private address must be used through a public gateway for instance, and it is
6259 known that the system cannot determine the adequate source address by itself.
6260
6261 An extension which is available on certain patched Linux kernels may be used
6262 through the "usesrc" optional keyword. It makes it possible to connect to the
6263 servers with an IP address which does not belong to the system itself. This
6264 is called "full transparent proxy mode". For this to work, the destination
6265 servers have to route their traffic back to this address through the machine
6266 running HAProxy, and IP forwarding must generally be enabled on this machine.
6267
6268 In this "full transparent proxy" mode, it is possible to force a specific IP
6269 address to be presented to the servers. This is not much used in fact. A more
6270 common use is to tell HAProxy to present the client's IP address. For this,
6271 there are two methods :
6272
6273 - present the client's IP and port addresses. This is the most transparent
6274 mode, but it can cause problems when IP connection tracking is enabled on
6275 the machine, because a same connection may be seen twice with different
6276 states. However, this solution presents the huge advantage of not
6277 limiting the system to the 64k outgoing address+port couples, because all
6278 of the client ranges may be used.
6279
6280 - present only the client's IP address and select a spare port. This
6281 solution is still quite elegant but slightly less transparent (downstream
6282 firewalls logs will not match upstream's). It also presents the downside
6283 of limiting the number of concurrent connections to the usual 64k ports.
6284 However, since the upstream and downstream ports are different, local IP
6285 connection tracking on the machine will not be upset by the reuse of the
6286 same session.
6287
6288 Note that depending on the transparent proxy technology used, it may be
6289 required to force the source address. In fact, cttproxy version 2 requires an
6290 IP address in <addr> above, and does not support setting of "0.0.0.0" as the
6291 IP address because it creates NAT entries which much match the exact outgoing
6292 address. Tproxy version 4 and some other kernel patches which work in pure
6293 forwarding mode generally will not have this limitation.
6294
6295 This option sets the default source for all servers in the backend. It may
6296 also be specified in a "defaults" section. Finer source address specification
6297 is possible at the server level using the "source" server option. Refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006298 section 5 for more information.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006299
6300 Examples :
6301 backend private
6302 # Connect to the servers using our 192.168.1.200 source address
6303 source 192.168.1.200
6304
6305 backend transparent_ssl1
6306 # Connect to the SSL farm from the client's source address
6307 source 192.168.1.200 usesrc clientip
6308
6309 backend transparent_ssl2
6310 # Connect to the SSL farm from the client's source address and port
6311 # not recommended if IP conntrack is present on the local machine.
6312 source 192.168.1.200 usesrc client
6313
6314 backend transparent_ssl3
6315 # Connect to the SSL farm from the client's source address. It
6316 # is more conntrack-friendly.
6317 source 192.168.1.200 usesrc clientip
6318
6319 backend transparent_smtp
6320 # Connect to the SMTP farm from the client's source address/port
6321 # with Tproxy version 4.
6322 source 0.0.0.0 usesrc clientip
6323
Willy Tarreaubce70882009-09-07 11:51:47 +02006324 backend transparent_http
6325 # Connect to the servers using the client's IP as seen by previous
6326 # proxy.
6327 source 0.0.0.0 usesrc hdr_ip(x-forwarded-for,-1)
6328
Willy Tarreauc57f0e22009-05-10 13:12:33 +02006329 See also : the "source" server option in section 5, the Tproxy patches for
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006330 the Linux kernel on www.balabit.com, the "bind" keyword.
6331
Krzysztof Piotr Oledzki25b501a2008-01-06 16:36:16 +01006332
Willy Tarreau844e3c52008-01-11 16:28:18 +01006333srvtimeout <timeout> (deprecated)
6334 Set the maximum inactivity time on the server side.
6335 May be used in sections : defaults | frontend | listen | backend
6336 yes | no | yes | yes
6337 Arguments :
6338 <timeout> is the timeout value specified in milliseconds by default, but
6339 can be in any other unit if the number is suffixed by the unit,
6340 as explained at the top of this document.
6341
6342 The inactivity timeout applies when the server is expected to acknowledge or
6343 send data. In HTTP mode, this timeout is particularly important to consider
6344 during the first phase of the server's response, when it has to send the
6345 headers, as it directly represents the server's processing time for the
6346 request. To find out what value to put there, it's often good to start with
6347 what would be considered as unacceptable response times, then check the logs
6348 to observe the response time distribution, and adjust the value accordingly.
6349
6350 The value is specified in milliseconds by default, but can be in any other
6351 unit if the number is suffixed by the unit, as specified at the top of this
6352 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
6353 recommended that the client timeout remains equal to the server timeout in
6354 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01006355 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01006356 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreaud72758d2010-01-12 10:42:19 +01006357 seconds (eg: 4 or 5 seconds minimum).
Willy Tarreau844e3c52008-01-11 16:28:18 +01006358
6359 This parameter is specific to backends, but can be specified once for all in
6360 "defaults" sections. This is in fact one of the easiest solutions not to
6361 forget about it. An unspecified timeout results in an infinite timeout, which
6362 is not recommended. Such a usage is accepted and works but reports a warning
6363 during startup because it may results in accumulation of expired sessions in
6364 the system if the system's timeouts are not configured either.
6365
6366 This parameter is provided for compatibility but is currently deprecated.
6367 Please use "timeout server" instead.
6368
Willy Tarreauce887fd2012-05-12 12:50:00 +02006369 See also : "timeout server", "timeout tunnel", "timeout client" and
6370 "clitimeout".
Willy Tarreau844e3c52008-01-11 16:28:18 +01006371
6372
Cyril Bonté66c327d2010-10-12 00:14:37 +02006373stats admin { if | unless } <cond>
6374 Enable statistics admin level if/unless a condition is matched
6375 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006376 no | yes | yes | yes
Cyril Bonté66c327d2010-10-12 00:14:37 +02006377
6378 This statement enables the statistics admin level if/unless a condition is
6379 matched.
6380
6381 The admin level allows to enable/disable servers from the web interface. By
6382 default, statistics page is read-only for security reasons.
6383
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006384 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6385 unless you know what you do : memory is not shared between the
6386 processes, which can result in random behaviours.
6387
Cyril Bonté23b39d92011-02-10 22:54:44 +01006388 Currently, the POST request is limited to the buffer size minus the reserved
6389 buffer space, which means that if the list of servers is too long, the
6390 request won't be processed. It is recommended to alter few servers at a
6391 time.
Cyril Bonté66c327d2010-10-12 00:14:37 +02006392
6393 Example :
6394 # statistics admin level only for localhost
6395 backend stats_localhost
6396 stats enable
6397 stats admin if LOCALHOST
6398
6399 Example :
6400 # statistics admin level always enabled because of the authentication
6401 backend stats_auth
6402 stats enable
6403 stats auth admin:AdMiN123
6404 stats admin if TRUE
6405
6406 Example :
6407 # statistics admin level depends on the authenticated user
6408 userlist stats-auth
6409 group admin users admin
6410 user admin insecure-password AdMiN123
6411 group readonly users haproxy
6412 user haproxy insecure-password haproxy
6413
6414 backend stats_auth
6415 stats enable
6416 acl AUTH http_auth(stats-auth)
6417 acl AUTH_ADMIN http_auth_group(stats-auth) admin
6418 stats http-request auth unless AUTH
6419 stats admin if AUTH_ADMIN
6420
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006421 See also : "stats enable", "stats auth", "stats http-request", "nbproc",
6422 "bind-process", section 3.4 about userlists and section 7 about
6423 ACL usage.
Cyril Bonté66c327d2010-10-12 00:14:37 +02006424
6425
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006426stats auth <user>:<passwd>
6427 Enable statistics with authentication and grant access to an account
6428 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006429 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006430 Arguments :
6431 <user> is a user name to grant access to
6432
6433 <passwd> is the cleartext password associated to this user
6434
6435 This statement enables statistics with default settings, and restricts access
6436 to declared users only. It may be repeated as many times as necessary to
6437 allow as many users as desired. When a user tries to access the statistics
6438 without a valid account, a "401 Forbidden" response will be returned so that
6439 the browser asks the user to provide a valid user and password. The real
6440 which will be returned to the browser is configurable using "stats realm".
6441
6442 Since the authentication method is HTTP Basic Authentication, the passwords
6443 circulate in cleartext on the network. Thus, it was decided that the
6444 configuration file would also use cleartext passwords to remind the users
Willy Tarreau3c92c5f2011-08-28 09:45:47 +02006445 that those ones should not be sensitive and not shared with any other account.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006446
6447 It is also possible to reduce the scope of the proxies which appear in the
6448 report using "stats scope".
6449
6450 Though this statement alone is enough to enable statistics reporting, it is
6451 recommended to set all other settings in order to avoid relying on default
6452 unobvious parameters.
6453
6454 Example :
6455 # public access (limited to this backend only)
6456 backend public_www
6457 server srv1 192.168.0.1:80
6458 stats enable
6459 stats hide-version
6460 stats scope .
6461 stats uri /admin?stats
6462 stats realm Haproxy\ Statistics
6463 stats auth admin1:AdMiN123
6464 stats auth admin2:AdMiN321
6465
6466 # internal monitoring access (unlimited)
6467 backend private_monitoring
6468 stats enable
6469 stats uri /admin?stats
6470 stats refresh 5s
6471
6472 See also : "stats enable", "stats realm", "stats scope", "stats uri"
6473
6474
6475stats enable
6476 Enable statistics reporting with default settings
6477 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006478 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006479 Arguments : none
6480
6481 This statement enables statistics reporting with default settings defined
6482 at build time. Unless stated otherwise, these settings are used :
6483 - stats uri : /haproxy?stats
6484 - stats realm : "HAProxy Statistics"
6485 - stats auth : no authentication
6486 - stats scope : no restriction
6487
6488 Though this statement alone is enough to enable statistics reporting, it is
6489 recommended to set all other settings in order to avoid relying on default
6490 unobvious parameters.
6491
6492 Example :
6493 # public access (limited to this backend only)
6494 backend public_www
6495 server srv1 192.168.0.1:80
6496 stats enable
6497 stats hide-version
6498 stats scope .
6499 stats uri /admin?stats
6500 stats realm Haproxy\ Statistics
6501 stats auth admin1:AdMiN123
6502 stats auth admin2:AdMiN321
6503
6504 # internal monitoring access (unlimited)
6505 backend private_monitoring
6506 stats enable
6507 stats uri /admin?stats
6508 stats refresh 5s
6509
6510 See also : "stats auth", "stats realm", "stats uri"
6511
6512
Willy Tarreaud63335a2010-02-26 12:56:52 +01006513stats hide-version
6514 Enable statistics and hide HAProxy version reporting
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006515 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006516 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01006517 Arguments : none
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006518
Willy Tarreaud63335a2010-02-26 12:56:52 +01006519 By default, the stats page reports some useful status information along with
6520 the statistics. Among them is HAProxy's version. However, it is generally
6521 considered dangerous to report precise version to anyone, as it can help them
6522 target known weaknesses with specific attacks. The "stats hide-version"
6523 statement removes the version from the statistics report. This is recommended
6524 for public sites or any site with a weak login/password.
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006525
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02006526 Though this statement alone is enough to enable statistics reporting, it is
6527 recommended to set all other settings in order to avoid relying on default
6528 unobvious parameters.
6529
Willy Tarreaud63335a2010-02-26 12:56:52 +01006530 Example :
6531 # public access (limited to this backend only)
6532 backend public_www
6533 server srv1 192.168.0.1:80
Krzysztof Piotr Oledzki48cb2ae2009-10-02 22:51:14 +02006534 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01006535 stats hide-version
6536 stats scope .
6537 stats uri /admin?stats
6538 stats realm Haproxy\ Statistics
6539 stats auth admin1:AdMiN123
6540 stats auth admin2:AdMiN321
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006541
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006542 # internal monitoring access (unlimited)
6543 backend private_monitoring
6544 stats enable
Willy Tarreaud63335a2010-02-26 12:56:52 +01006545 stats uri /admin?stats
6546 stats refresh 5s
Krzysztof Piotr Oledzki15514c22010-01-04 16:03:09 +01006547
Willy Tarreaud63335a2010-02-26 12:56:52 +01006548 See also : "stats auth", "stats enable", "stats realm", "stats uri"
Willy Tarreau1d45b7c2009-08-16 10:29:18 +02006549
Willy Tarreau983e01e2010-01-11 18:42:06 +01006550
Cyril Bonté2be1b3f2010-09-30 23:46:30 +02006551stats http-request { allow | deny | auth [realm <realm>] }
6552 [ { if | unless } <condition> ]
6553 Access control for statistics
6554
6555 May be used in sections: defaults | frontend | listen | backend
6556 no | no | yes | yes
6557
6558 As "http-request", these set of options allow to fine control access to
6559 statistics. Each option may be followed by if/unless and acl.
6560 First option with matched condition (or option without condition) is final.
6561 For "deny" a 403 error will be returned, for "allow" normal processing is
6562 performed, for "auth" a 401/407 error code is returned so the client
6563 should be asked to enter a username and password.
6564
6565 There is no fixed limit to the number of http-request statements per
6566 instance.
6567
6568 See also : "http-request", section 3.4 about userlists and section 7
6569 about ACL usage.
6570
6571
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006572stats realm <realm>
6573 Enable statistics and set authentication realm
6574 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006575 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006576 Arguments :
6577 <realm> is the name of the HTTP Basic Authentication realm reported to
6578 the browser. The browser uses it to display it in the pop-up
6579 inviting the user to enter a valid username and password.
6580
6581 The realm is read as a single word, so any spaces in it should be escaped
6582 using a backslash ('\').
6583
6584 This statement is useful only in conjunction with "stats auth" since it is
6585 only related to authentication.
6586
6587 Though this statement alone is enough to enable statistics reporting, it is
6588 recommended to set all other settings in order to avoid relying on default
6589 unobvious parameters.
6590
6591 Example :
6592 # public access (limited to this backend only)
6593 backend public_www
6594 server srv1 192.168.0.1:80
6595 stats enable
6596 stats hide-version
6597 stats scope .
6598 stats uri /admin?stats
6599 stats realm Haproxy\ Statistics
6600 stats auth admin1:AdMiN123
6601 stats auth admin2:AdMiN321
6602
6603 # internal monitoring access (unlimited)
6604 backend private_monitoring
6605 stats enable
6606 stats uri /admin?stats
6607 stats refresh 5s
6608
6609 See also : "stats auth", "stats enable", "stats uri"
6610
6611
6612stats refresh <delay>
6613 Enable statistics with automatic refresh
6614 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006615 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006616 Arguments :
6617 <delay> is the suggested refresh delay, specified in seconds, which will
6618 be returned to the browser consulting the report page. While the
6619 browser is free to apply any delay, it will generally respect it
6620 and refresh the page this every seconds. The refresh interval may
6621 be specified in any other non-default time unit, by suffixing the
6622 unit after the value, as explained at the top of this document.
6623
6624 This statement is useful on monitoring displays with a permanent page
6625 reporting the load balancer's activity. When set, the HTML report page will
6626 include a link "refresh"/"stop refresh" so that the user can select whether
6627 he wants automatic refresh of the page or not.
6628
6629 Though this statement alone is enough to enable statistics reporting, it is
6630 recommended to set all other settings in order to avoid relying on default
6631 unobvious parameters.
6632
6633 Example :
6634 # public access (limited to this backend only)
6635 backend public_www
6636 server srv1 192.168.0.1:80
6637 stats enable
6638 stats hide-version
6639 stats scope .
6640 stats uri /admin?stats
6641 stats realm Haproxy\ Statistics
6642 stats auth admin1:AdMiN123
6643 stats auth admin2:AdMiN321
6644
6645 # internal monitoring access (unlimited)
6646 backend private_monitoring
6647 stats enable
6648 stats uri /admin?stats
6649 stats refresh 5s
6650
6651 See also : "stats auth", "stats enable", "stats realm", "stats uri"
6652
6653
6654stats scope { <name> | "." }
6655 Enable statistics and limit access scope
6656 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006657 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006658 Arguments :
6659 <name> is the name of a listen, frontend or backend section to be
6660 reported. The special name "." (a single dot) designates the
6661 section in which the statement appears.
6662
6663 When this statement is specified, only the sections enumerated with this
6664 statement will appear in the report. All other ones will be hidden. This
6665 statement may appear as many times as needed if multiple sections need to be
6666 reported. Please note that the name checking is performed as simple string
6667 comparisons, and that it is never checked that a give section name really
6668 exists.
6669
6670 Though this statement alone is enough to enable statistics reporting, it is
6671 recommended to set all other settings in order to avoid relying on default
6672 unobvious parameters.
6673
6674 Example :
6675 # public access (limited to this backend only)
6676 backend public_www
6677 server srv1 192.168.0.1:80
6678 stats enable
6679 stats hide-version
6680 stats scope .
6681 stats uri /admin?stats
6682 stats realm Haproxy\ Statistics
6683 stats auth admin1:AdMiN123
6684 stats auth admin2:AdMiN321
6685
6686 # internal monitoring access (unlimited)
6687 backend private_monitoring
6688 stats enable
6689 stats uri /admin?stats
6690 stats refresh 5s
6691
6692 See also : "stats auth", "stats enable", "stats realm", "stats uri"
6693
Willy Tarreaud63335a2010-02-26 12:56:52 +01006694
Willy Tarreauc9705a12010-07-27 20:05:50 +02006695stats show-desc [ <desc> ]
Willy Tarreaud63335a2010-02-26 12:56:52 +01006696 Enable reporting of a description on the statistics page.
6697 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006698 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01006699
Willy Tarreauc9705a12010-07-27 20:05:50 +02006700 <desc> is an optional description to be reported. If unspecified, the
Willy Tarreaud63335a2010-02-26 12:56:52 +01006701 description from global section is automatically used instead.
6702
6703 This statement is useful for users that offer shared services to their
6704 customers, where node or description should be different for each customer.
6705
6706 Though this statement alone is enough to enable statistics reporting, it is
6707 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006708 unobvious parameters. By default description is not shown.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006709
6710 Example :
6711 # internal monitoring access (unlimited)
6712 backend private_monitoring
6713 stats enable
6714 stats show-desc Master node for Europe, Asia, Africa
6715 stats uri /admin?stats
6716 stats refresh 5s
6717
6718 See also: "show-node", "stats enable", "stats uri" and "description" in
6719 global section.
6720
6721
6722stats show-legends
Willy Tarreaued2119c2014-04-24 22:10:39 +02006723 Enable reporting additional information on the statistics page
6724 May be used in sections : defaults | frontend | listen | backend
6725 yes | yes | yes | yes
6726 Arguments : none
6727
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03006728 Enable reporting additional information on the statistics page :
Willy Tarreaud63335a2010-02-26 12:56:52 +01006729 - cap: capabilities (proxy)
6730 - mode: one of tcp, http or health (proxy)
6731 - id: SNMP ID (proxy, socket, server)
6732 - IP (socket, server)
6733 - cookie (backend, server)
6734
6735 Though this statement alone is enough to enable statistics reporting, it is
6736 recommended to set all other settings in order to avoid relying on default
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006737 unobvious parameters. Default behaviour is not to show this information.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006738
6739 See also: "stats enable", "stats uri".
6740
6741
6742stats show-node [ <name> ]
6743 Enable reporting of a host name on the statistics page.
6744 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006745 yes | yes | yes | yes
Willy Tarreaud63335a2010-02-26 12:56:52 +01006746 Arguments:
6747 <name> is an optional name to be reported. If unspecified, the
6748 node name from global section is automatically used instead.
6749
6750 This statement is useful for users that offer shared services to their
6751 customers, where node or description might be different on a stats page
Dmitry Sivachenko7823de32012-05-16 14:00:26 +04006752 provided for each customer. Default behaviour is not to show host name.
Willy Tarreaud63335a2010-02-26 12:56:52 +01006753
6754 Though this statement alone is enough to enable statistics reporting, it is
6755 recommended to set all other settings in order to avoid relying on default
6756 unobvious parameters.
6757
6758 Example:
6759 # internal monitoring access (unlimited)
6760 backend private_monitoring
6761 stats enable
6762 stats show-node Europe-1
6763 stats uri /admin?stats
6764 stats refresh 5s
6765
6766 See also: "show-desc", "stats enable", "stats uri", and "node" in global
6767 section.
6768
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006769
6770stats uri <prefix>
6771 Enable statistics and define the URI prefix to access them
6772 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaued2119c2014-04-24 22:10:39 +02006773 yes | yes | yes | yes
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006774 Arguments :
6775 <prefix> is the prefix of any URI which will be redirected to stats. This
6776 prefix may contain a question mark ('?') to indicate part of a
6777 query string.
6778
6779 The statistics URI is intercepted on the relayed traffic, so it appears as a
6780 page within the normal application. It is strongly advised to ensure that the
6781 selected URI will never appear in the application, otherwise it will never be
6782 possible to reach it in the application.
6783
6784 The default URI compiled in haproxy is "/haproxy?stats", but this may be
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01006785 changed at build time, so it's better to always explicitly specify it here.
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006786 It is generally a good idea to include a question mark in the URI so that
6787 intermediate proxies refrain from caching the results. Also, since any string
6788 beginning with the prefix will be accepted as a stats request, the question
6789 mark helps ensuring that no valid URI will begin with the same words.
6790
6791 It is sometimes very convenient to use "/" as the URI prefix, and put that
6792 statement in a "listen" instance of its own. That makes it easy to dedicate
6793 an address or a port to statistics only.
6794
6795 Though this statement alone is enough to enable statistics reporting, it is
6796 recommended to set all other settings in order to avoid relying on default
6797 unobvious parameters.
6798
6799 Example :
6800 # public access (limited to this backend only)
6801 backend public_www
6802 server srv1 192.168.0.1:80
6803 stats enable
6804 stats hide-version
6805 stats scope .
6806 stats uri /admin?stats
6807 stats realm Haproxy\ Statistics
6808 stats auth admin1:AdMiN123
6809 stats auth admin2:AdMiN321
6810
6811 # internal monitoring access (unlimited)
6812 backend private_monitoring
6813 stats enable
6814 stats uri /admin?stats
6815 stats refresh 5s
6816
6817 See also : "stats auth", "stats enable", "stats realm"
6818
6819
Willy Tarreaud63335a2010-02-26 12:56:52 +01006820stick match <pattern> [table <table>] [{if | unless} <cond>]
6821 Define a request pattern matching condition to stick a user to a server
Willy Tarreaueabeafa2008-01-16 16:17:06 +01006822 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaud63335a2010-02-26 12:56:52 +01006823 no | no | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006824
6825 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02006826 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006827 describes what elements of the incoming request or connection
6828 will be analysed in the hope to find a matching entry in a
6829 stickiness table. This rule is mandatory.
6830
6831 <table> is an optional stickiness table name. If unspecified, the same
6832 backend's table is used. A stickiness table is declared using
6833 the "stick-table" statement.
6834
6835 <cond> is an optional matching condition. It makes it possible to match
6836 on a certain criterion only when other conditions are met (or
6837 not met). For instance, it could be used to match on a source IP
6838 address except when a request passes through a known proxy, in
6839 which case we'd match on a header containing that IP address.
6840
6841 Some protocols or applications require complex stickiness rules and cannot
6842 always simply rely on cookies nor hashing. The "stick match" statement
6843 describes a rule to extract the stickiness criterion from an incoming request
6844 or connection. See section 7 for a complete list of possible patterns and
6845 transformation rules.
6846
6847 The table has to be declared using the "stick-table" statement. It must be of
6848 a type compatible with the pattern. By default it is the one which is present
6849 in the same backend. It is possible to share a table with other backends by
6850 referencing it using the "table" keyword. If another table is referenced,
6851 the server's ID inside the backends are used. By default, all server IDs
6852 start at 1 in each backend, so the server ordering is enough. But in case of
6853 doubt, it is highly recommended to force server IDs using their "id" setting.
6854
6855 It is possible to restrict the conditions where a "stick match" statement
6856 will apply, using "if" or "unless" followed by a condition. See section 7 for
6857 ACL based conditions.
6858
6859 There is no limit on the number of "stick match" statements. The first that
6860 applies and matches will cause the request to be directed to the same server
6861 as was used for the request which created the entry. That way, multiple
6862 matches can be used as fallbacks.
6863
6864 The stick rules are checked after the persistence cookies, so they will not
6865 affect stickiness if a cookie has already been used to select a server. That
6866 way, it becomes very easy to insert cookies and match on IP addresses in
6867 order to maintain stickiness between HTTP and HTTPS.
6868
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006869 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6870 unless you know what you do : memory is not shared between the
6871 processes, which can result in random behaviours.
6872
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006873 Example :
6874 # forward SMTP users to the same server they just used for POP in the
6875 # last 30 minutes
6876 backend pop
6877 mode tcp
6878 balance roundrobin
6879 stick store-request src
6880 stick-table type ip size 200k expire 30m
6881 server s1 192.168.1.1:110
6882 server s2 192.168.1.1:110
6883
6884 backend smtp
6885 mode tcp
6886 balance roundrobin
6887 stick match src table pop
6888 server s1 192.168.1.1:25
6889 server s2 192.168.1.1:25
6890
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006891 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02006892 about ACLs and samples fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006893
6894
6895stick on <pattern> [table <table>] [{if | unless} <condition>]
6896 Define a request pattern to associate a user to a server
6897 May be used in sections : defaults | frontend | listen | backend
6898 no | no | yes | yes
6899
6900 Note : This form is exactly equivalent to "stick match" followed by
6901 "stick store-request", all with the same arguments. Please refer
6902 to both keywords for details. It is only provided as a convenience
6903 for writing more maintainable configurations.
6904
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006905 Note : Consider not using this feature in multi-process mode (nbproc > 1)
6906 unless you know what you do : memory is not shared between the
6907 processes, which can result in random behaviours.
6908
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006909 Examples :
6910 # The following form ...
Willy Tarreauec579d82010-02-26 19:15:04 +01006911 stick on src table pop if !localhost
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006912
6913 # ...is strictly equivalent to this one :
6914 stick match src table pop if !localhost
6915 stick store-request src table pop if !localhost
6916
6917
6918 # Use cookie persistence for HTTP, and stick on source address for HTTPS as
6919 # well as HTTP without cookie. Share the same table between both accesses.
6920 backend http
6921 mode http
6922 balance roundrobin
6923 stick on src table https
6924 cookie SRV insert indirect nocache
6925 server s1 192.168.1.1:80 cookie s1
6926 server s2 192.168.1.1:80 cookie s2
6927
6928 backend https
6929 mode tcp
6930 balance roundrobin
6931 stick-table type ip size 200k expire 30m
6932 stick on src
6933 server s1 192.168.1.1:443
6934 server s2 192.168.1.1:443
6935
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01006936 See also : "stick match", "stick store-request", "nbproc" and "bind-process".
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006937
6938
6939stick store-request <pattern> [table <table>] [{if | unless} <condition>]
6940 Define a request pattern used to create an entry in a stickiness table
6941 May be used in sections : defaults | frontend | listen | backend
6942 no | no | yes | yes
6943
6944 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02006945 <pattern> is a sample expression rule as described in section 7.3. It
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006946 describes what elements of the incoming request or connection
6947 will be analysed, extracted and stored in the table once a
6948 server is selected.
6949
6950 <table> is an optional stickiness table name. If unspecified, the same
6951 backend's table is used. A stickiness table is declared using
6952 the "stick-table" statement.
6953
6954 <cond> is an optional storage condition. It makes it possible to store
6955 certain criteria only when some conditions are met (or not met).
6956 For instance, it could be used to store the source IP address
6957 except when the request passes through a known proxy, in which
6958 case we'd store a converted form of a header containing that IP
6959 address.
6960
6961 Some protocols or applications require complex stickiness rules and cannot
6962 always simply rely on cookies nor hashing. The "stick store-request" statement
6963 describes a rule to decide what to extract from the request and when to do
6964 it, in order to store it into a stickiness table for further requests to
6965 match it using the "stick match" statement. Obviously the extracted part must
6966 make sense and have a chance to be matched in a further request. Storing a
6967 client's IP address for instance often makes sense. Storing an ID found in a
6968 URL parameter also makes sense. Storing a source port will almost never make
6969 any sense because it will be randomly matched. See section 7 for a complete
6970 list of possible patterns and transformation rules.
6971
6972 The table has to be declared using the "stick-table" statement. It must be of
6973 a type compatible with the pattern. By default it is the one which is present
6974 in the same backend. It is possible to share a table with other backends by
6975 referencing it using the "table" keyword. If another table is referenced,
6976 the server's ID inside the backends are used. By default, all server IDs
6977 start at 1 in each backend, so the server ordering is enough. But in case of
6978 doubt, it is highly recommended to force server IDs using their "id" setting.
6979
6980 It is possible to restrict the conditions where a "stick store-request"
6981 statement will apply, using "if" or "unless" followed by a condition. This
6982 condition will be evaluated while parsing the request, so any criteria can be
6983 used. See section 7 for ACL based conditions.
6984
6985 There is no limit on the number of "stick store-request" statements, but
6986 there is a limit of 8 simultaneous stores per request or response. This
6987 makes it possible to store up to 8 criteria, all extracted from either the
6988 request or the response, regardless of the number of rules. Only the 8 first
6989 ones which match will be kept. Using this, it is possible to feed multiple
6990 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01006991 another protocol or access method. Using multiple store-request rules with
6992 the same table is possible and may be used to find the best criterion to rely
6993 on, by arranging the rules by decreasing preference order. Only the first
6994 extracted criterion for a given table will be stored. All subsequent store-
6995 request rules referencing the same table will be skipped and their ACLs will
6996 not be evaluated.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01006997
6998 The "store-request" rules are evaluated once the server connection has been
6999 established, so that the table will contain the real server that processed
7000 the request.
7001
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007002 Note : Consider not using this feature in multi-process mode (nbproc > 1)
7003 unless you know what you do : memory is not shared between the
7004 processes, which can result in random behaviours.
7005
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007006 Example :
7007 # forward SMTP users to the same server they just used for POP in the
7008 # last 30 minutes
7009 backend pop
7010 mode tcp
7011 balance roundrobin
7012 stick store-request src
7013 stick-table type ip size 200k expire 30m
7014 server s1 192.168.1.1:110
7015 server s2 192.168.1.1:110
7016
7017 backend smtp
7018 mode tcp
7019 balance roundrobin
7020 stick match src table pop
7021 server s1 192.168.1.1:25
7022 server s2 192.168.1.1:25
7023
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007024 See also : "stick-table", "stick on", "nbproc", "bind-process" and section 7
Willy Tarreaube722a22014-06-13 16:31:59 +02007025 about ACLs and sample fetching.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007026
7027
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007028stick-table type {ip | integer | string [len <length>] | binary [len <length>]}
Emeric Brunf099e792010-09-27 12:05:28 +02007029 size <size> [expire <expire>] [nopurge] [peers <peersect>]
7030 [store <data_type>]*
Godbach64cef792013-12-04 16:08:22 +08007031 Configure the stickiness table for the current section
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007032 May be used in sections : defaults | frontend | listen | backend
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007033 no | yes | yes | yes
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007034
7035 Arguments :
7036 ip a table declared with "type ip" will only store IPv4 addresses.
7037 This form is very compact (about 50 bytes per entry) and allows
7038 very fast entry lookup and stores with almost no overhead. This
7039 is mainly used to store client source IP addresses.
7040
David du Colombier9a6d3c92011-03-17 10:40:24 +01007041 ipv6 a table declared with "type ipv6" will only store IPv6 addresses.
7042 This form is very compact (about 60 bytes per entry) and allows
7043 very fast entry lookup and stores with almost no overhead. This
7044 is mainly used to store client source IP addresses.
7045
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007046 integer a table declared with "type integer" will store 32bit integers
7047 which can represent a client identifier found in a request for
7048 instance.
7049
7050 string a table declared with "type string" will store substrings of up
7051 to <len> characters. If the string provided by the pattern
7052 extractor is larger than <len>, it will be truncated before
7053 being stored. During matching, at most <len> characters will be
7054 compared between the string in the table and the extracted
7055 pattern. When not specified, the string is automatically limited
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007056 to 32 characters.
7057
7058 binary a table declared with "type binary" will store binary blocks
7059 of <len> bytes. If the block provided by the pattern
7060 extractor is larger than <len>, it will be truncated before
Willy Tarreaube722a22014-06-13 16:31:59 +02007061 being stored. If the block provided by the sample expression
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007062 is shorter than <len>, it will be padded by 0. When not
7063 specified, the block is automatically limited to 32 bytes.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007064
7065 <length> is the maximum number of characters that will be stored in a
Emeric Brun7c6b82e2010-09-24 16:34:28 +02007066 "string" type table (See type "string" above). Or the number
7067 of bytes of the block in "binary" type table. Be careful when
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007068 changing this parameter as memory usage will proportionally
7069 increase.
7070
7071 <size> is the maximum number of entries that can fit in the table. This
Cyril Bonté78caf842010-03-10 22:41:43 +01007072 value directly impacts memory usage. Count approximately
7073 50 bytes per entry, plus the size of a string if any. The size
7074 supports suffixes "k", "m", "g" for 2^10, 2^20 and 2^30 factors.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007075
7076 [nopurge] indicates that we refuse to purge older entries when the table
7077 is full. When not specified and the table is full when haproxy
7078 wants to store an entry in it, it will flush a few of the oldest
7079 entries in order to release some space for the new ones. This is
7080 most often the desired behaviour. In some specific cases, it
7081 be desirable to refuse new entries instead of purging the older
7082 ones. That may be the case when the amount of data to store is
7083 far above the hardware limits and we prefer not to offer access
7084 to new clients than to reject the ones already connected. When
7085 using this parameter, be sure to properly set the "expire"
7086 parameter (see below).
7087
Emeric Brunf099e792010-09-27 12:05:28 +02007088 <peersect> is the name of the peers section to use for replication. Entries
7089 which associate keys to server IDs are kept synchronized with
7090 the remote peers declared in this section. All entries are also
7091 automatically learned from the local peer (old process) during a
7092 soft restart.
7093
Cyril Bonté02ff8ef2010-12-14 22:48:49 +01007094 NOTE : peers can't be used in multi-process mode.
7095
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007096 <expire> defines the maximum duration of an entry in the table since it
7097 was last created, refreshed or matched. The expiration delay is
7098 defined using the standard time format, similarly as the various
7099 timeouts. The maximum duration is slightly above 24 days. See
7100 section 2.2 for more information. If this delay is not specified,
Cyril Bontédc4d9032012-04-08 21:57:39 +02007101 the session won't automatically expire, but older entries will
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007102 be removed once full. Be sure not to use the "nopurge" parameter
7103 if not expiration delay is specified.
7104
Willy Tarreau08d5f982010-06-06 13:34:54 +02007105 <data_type> is used to store additional information in the stick-table. This
7106 may be used by ACLs in order to control various criteria related
7107 to the activity of the client matching the stick-table. For each
7108 item specified here, the size of each entry will be inflated so
Willy Tarreauc9705a12010-07-27 20:05:50 +02007109 that the additional data can fit. Several data types may be
7110 stored with an entry. Multiple data types may be specified after
7111 the "store" keyword, as a comma-separated list. Alternatively,
7112 it is possible to repeat the "store" keyword followed by one or
7113 several data types. Except for the "server_id" type which is
7114 automatically detected and enabled, all data types must be
7115 explicitly declared to be stored. If an ACL references a data
7116 type which is not stored, the ACL will simply not match. Some
7117 data types require an argument which must be passed just after
7118 the type between parenthesis. See below for the supported data
7119 types and their arguments.
7120
7121 The data types that can be stored with an entry are the following :
7122 - server_id : this is an integer which holds the numeric ID of the server a
7123 request was assigned to. It is used by the "stick match", "stick store",
7124 and "stick on" rules. It is automatically enabled when referenced.
7125
7126 - gpc0 : first General Purpose Counter. It is a positive 32-bit integer
7127 integer which may be used for anything. Most of the time it will be used
7128 to put a special tag on some entries, for instance to note that a
7129 specific behaviour was detected and must be known for future matches.
7130
Willy Tarreauba2ffd12013-05-29 15:54:14 +02007131 - gpc0_rate(<period>) : increment rate of the first General Purpose Counter
7132 over a period. It is a positive 32-bit integer integer which may be used
7133 for anything. Just like <gpc0>, it counts events, but instead of keeping
7134 a cumulative count, it maintains the rate at which the counter is
7135 incremented. Most of the time it will be used to measure the frequency of
7136 occurrence of certain events (eg: requests to a specific URL).
7137
Willy Tarreauc9705a12010-07-27 20:05:50 +02007138 - conn_cnt : Connection Count. It is a positive 32-bit integer which counts
7139 the absolute number of connections received from clients which matched
7140 this entry. It does not mean the connections were accepted, just that
7141 they were received.
7142
7143 - conn_cur : Current Connections. It is a positive 32-bit integer which
7144 stores the concurrent connection counts for the entry. It is incremented
7145 once an incoming connection matches the entry, and decremented once the
7146 connection leaves. That way it is possible to know at any time the exact
7147 number of concurrent connections for an entry.
7148
7149 - conn_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7150 integer parameter <period> which indicates in milliseconds the length
7151 of the period over which the average is measured. It reports the average
7152 incoming connection rate over that period, in connections per period. The
7153 result is an integer which can be matched using ACLs.
7154
7155 - sess_cnt : Session Count. It is a positive 32-bit integer which counts
7156 the absolute number of sessions received from clients which matched this
7157 entry. A session is a connection that was accepted by the layer 4 rules.
7158
7159 - sess_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7160 integer parameter <period> which indicates in milliseconds the length
7161 of the period over which the average is measured. It reports the average
7162 incoming session rate over that period, in sessions per period. The
7163 result is an integer which can be matched using ACLs.
7164
7165 - http_req_cnt : HTTP request Count. It is a positive 32-bit integer which
7166 counts the absolute number of HTTP requests received from clients which
7167 matched this entry. It does not matter whether they are valid requests or
7168 not. Note that this is different from sessions when keep-alive is used on
7169 the client side.
7170
7171 - http_req_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7172 integer parameter <period> which indicates in milliseconds the length
7173 of the period over which the average is measured. It reports the average
7174 HTTP request rate over that period, in requests per period. The result is
7175 an integer which can be matched using ACLs. It does not matter whether
7176 they are valid requests or not. Note that this is different from sessions
7177 when keep-alive is used on the client side.
7178
7179 - http_err_cnt : HTTP Error Count. It is a positive 32-bit integer which
7180 counts the absolute number of HTTP requests errors induced by clients
7181 which matched this entry. Errors are counted on invalid and truncated
7182 requests, as well as on denied or tarpitted requests, and on failed
7183 authentications. If the server responds with 4xx, then the request is
7184 also counted as an error since it's an error triggered by the client
7185 (eg: vulnerability scan).
7186
7187 - http_err_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7188 integer parameter <period> which indicates in milliseconds the length
7189 of the period over which the average is measured. It reports the average
7190 HTTP request error rate over that period, in requests per period (see
7191 http_err_cnt above for what is accounted as an error). The result is an
7192 integer which can be matched using ACLs.
7193
7194 - bytes_in_cnt : client to server byte count. It is a positive 64-bit
7195 integer which counts the cumulated amount of bytes received from clients
7196 which matched this entry. Headers are included in the count. This may be
7197 used to limit abuse of upload features on photo or video servers.
7198
7199 - bytes_in_rate(<period>) : frequency counter (takes 12 bytes). It takes an
7200 integer parameter <period> which indicates in milliseconds the length
7201 of the period over which the average is measured. It reports the average
7202 incoming bytes rate over that period, in bytes per period. It may be used
7203 to detect users which upload too much and too fast. Warning: with large
7204 uploads, it is possible that the amount of uploaded data will be counted
7205 once upon termination, thus causing spikes in the average transfer speed
7206 instead of having a smooth one. This may partially be smoothed with
7207 "option contstats" though this is not perfect yet. Use of byte_in_cnt is
7208 recommended for better fairness.
7209
7210 - bytes_out_cnt : server to client byte count. It is a positive 64-bit
7211 integer which counts the cumulated amount of bytes sent to clients which
7212 matched this entry. Headers are included in the count. This may be used
7213 to limit abuse of bots sucking the whole site.
7214
7215 - bytes_out_rate(<period>) : frequency counter (takes 12 bytes). It takes
7216 an integer parameter <period> which indicates in milliseconds the length
7217 of the period over which the average is measured. It reports the average
7218 outgoing bytes rate over that period, in bytes per period. It may be used
7219 to detect users which download too much and too fast. Warning: with large
7220 transfers, it is possible that the amount of transferred data will be
7221 counted once upon termination, thus causing spikes in the average
7222 transfer speed instead of having a smooth one. This may partially be
7223 smoothed with "option contstats" though this is not perfect yet. Use of
7224 byte_out_cnt is recommended for better fairness.
Willy Tarreau08d5f982010-06-06 13:34:54 +02007225
Willy Tarreauc00cdc22010-06-06 16:48:26 +02007226 There is only one stick-table per proxy. At the moment of writing this doc,
7227 it does not seem useful to have multiple tables per proxy. If this happens
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007228 to be required, simply create a dummy backend with a stick-table in it and
7229 reference it.
7230
7231 It is important to understand that stickiness based on learning information
7232 has some limitations, including the fact that all learned associations are
7233 lost upon restart. In general it can be good as a complement but not always
7234 as an exclusive stickiness.
7235
Willy Tarreauc9705a12010-07-27 20:05:50 +02007236 Last, memory requirements may be important when storing many data types.
7237 Indeed, storing all indicators above at once in each entry requires 116 bytes
7238 per entry, or 116 MB for a 1-million entries table. This is definitely not
7239 something that can be ignored.
7240
7241 Example:
7242 # Keep track of counters of up to 1 million IP addresses over 5 minutes
7243 # and store a general purpose counter and the average connection rate
7244 # computed over a sliding window of 30 seconds.
7245 stick-table type ip size 1m expire 5m store gpc0,conn_rate(30s)
7246
7247 See also : "stick match", "stick on", "stick store-request", section 2.2
David du Colombiera13d1b92011-03-17 10:40:22 +01007248 about time format and section 7 about ACLs.
Willy Tarreaub937b7e2010-01-12 15:27:54 +01007249
7250
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007251stick store-response <pattern> [table <table>] [{if | unless} <condition>]
7252 Define a request pattern used to create an entry in a stickiness table
7253 May be used in sections : defaults | frontend | listen | backend
7254 no | no | yes | yes
7255
7256 Arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007257 <pattern> is a sample expression rule as described in section 7.3. It
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007258 describes what elements of the response or connection will
7259 be analysed, extracted and stored in the table once a
7260 server is selected.
7261
7262 <table> is an optional stickiness table name. If unspecified, the same
7263 backend's table is used. A stickiness table is declared using
7264 the "stick-table" statement.
7265
7266 <cond> is an optional storage condition. It makes it possible to store
7267 certain criteria only when some conditions are met (or not met).
7268 For instance, it could be used to store the SSL session ID only
7269 when the response is a SSL server hello.
7270
7271 Some protocols or applications require complex stickiness rules and cannot
7272 always simply rely on cookies nor hashing. The "stick store-response"
7273 statement describes a rule to decide what to extract from the response and
7274 when to do it, in order to store it into a stickiness table for further
7275 requests to match it using the "stick match" statement. Obviously the
7276 extracted part must make sense and have a chance to be matched in a further
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007277 request. Storing an ID found in a header of a response makes sense.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007278 See section 7 for a complete list of possible patterns and transformation
7279 rules.
7280
7281 The table has to be declared using the "stick-table" statement. It must be of
7282 a type compatible with the pattern. By default it is the one which is present
7283 in the same backend. It is possible to share a table with other backends by
7284 referencing it using the "table" keyword. If another table is referenced,
7285 the server's ID inside the backends are used. By default, all server IDs
7286 start at 1 in each backend, so the server ordering is enough. But in case of
7287 doubt, it is highly recommended to force server IDs using their "id" setting.
7288
7289 It is possible to restrict the conditions where a "stick store-response"
7290 statement will apply, using "if" or "unless" followed by a condition. This
7291 condition will be evaluated while parsing the response, so any criteria can
7292 be used. See section 7 for ACL based conditions.
7293
7294 There is no limit on the number of "stick store-response" statements, but
7295 there is a limit of 8 simultaneous stores per request or response. This
7296 makes it possible to store up to 8 criteria, all extracted from either the
7297 request or the response, regardless of the number of rules. Only the 8 first
7298 ones which match will be kept. Using this, it is possible to feed multiple
7299 tables at once in the hope to increase the chance to recognize a user on
Willy Tarreau9667a802013-12-09 12:52:13 +01007300 another protocol or access method. Using multiple store-response rules with
7301 the same table is possible and may be used to find the best criterion to rely
7302 on, by arranging the rules by decreasing preference order. Only the first
7303 extracted criterion for a given table will be stored. All subsequent store-
7304 response rules referencing the same table will be skipped and their ACLs will
7305 not be evaluated. However, even if a store-request rule references a table, a
7306 store-response rule may also use the same table. This means that each table
7307 may learn exactly one element from the request and one element from the
7308 response at once.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007309
7310 The table will contain the real server that processed the request.
7311
7312 Example :
7313 # Learn SSL session ID from both request and response and create affinity.
7314 backend https
7315 mode tcp
7316 balance roundrobin
Cyril Bontédc4d9032012-04-08 21:57:39 +02007317 # maximum SSL session ID length is 32 bytes.
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007318 stick-table type binary len 32 size 30k expire 30m
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007319
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007320 acl clienthello req_ssl_hello_type 1
7321 acl serverhello rep_ssl_hello_type 2
7322
7323 # use tcp content accepts to detects ssl client and server hello.
7324 tcp-request inspect-delay 5s
7325 tcp-request content accept if clienthello
7326
7327 # no timeout on response inspect delay by default.
7328 tcp-response content accept if serverhello
Cyril Bonté108cf6e2012-04-21 23:30:29 +02007329
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007330 # SSL session ID (SSLID) may be present on a client or server hello.
7331 # Its length is coded on 1 byte at offset 43 and its value starts
7332 # at offset 44.
7333
7334 # Match and learn on request if client hello.
7335 stick on payload_lv(43,1) if clienthello
7336
7337 # Learn on response if server hello.
7338 stick store-response payload_lv(43,1) if serverhello
Cyril Bontédc4d9032012-04-08 21:57:39 +02007339
Emeric Brun6a1cefa2010-09-24 18:15:17 +02007340 server s1 192.168.1.1:443
7341 server s2 192.168.1.1:443
7342
7343 See also : "stick-table", "stick on", and section 7 about ACLs and pattern
7344 extraction.
7345
7346
Willy Tarreau938c7fe2014-04-25 14:21:39 +02007347tcp-check connect [params*]
7348 Opens a new connection
7349 May be used in sections: defaults | frontend | listen | backend
7350 no | no | yes | yes
7351
7352 When an application lies on more than a single TCP port or when HAProxy
7353 load-balance many services in a single backend, it makes sense to probe all
7354 the services individually before considering a server as operational.
7355
7356 When there are no TCP port configured on the server line neither server port
7357 directive, then the 'tcp-check connect port <port>' must be the first step
7358 of the sequence.
7359
7360 In a tcp-check ruleset a 'connect' is required, it is also mandatory to start
7361 the ruleset with a 'connect' rule. Purpose is to ensure admin know what they
7362 do.
7363
7364 Parameters :
7365 They are optional and can be used to describe how HAProxy should open and
7366 use the TCP connection.
7367
7368 port if not set, check port or server port is used.
7369 It tells HAProxy where to open the connection to.
7370 <port> must be a valid TCP port source integer, from 1 to 65535.
7371
7372 send-proxy send a PROXY protocol string
7373
7374 ssl opens a ciphered connection
7375
7376 Examples:
7377 # check HTTP and HTTPs services on a server.
7378 # first open port 80 thanks to server line port directive, then
7379 # tcp-check opens port 443, ciphered and run a request on it:
7380 option tcp-check
7381 tcp-check connect
7382 tcp-check send GET\ /\ HTTP/1.0\r\n
7383 tcp-check send Host:\ haproxy.1wt.eu\r\n
7384 tcp-check send \r\n
7385 tcp-check expect rstring (2..|3..)
7386 tcp-check connect port 443 ssl
7387 tcp-check send GET\ /\ HTTP/1.0\r\n
7388 tcp-check send Host:\ haproxy.1wt.eu\r\n
7389 tcp-check send \r\n
7390 tcp-check expect rstring (2..|3..)
7391 server www 10.0.0.1 check port 80
7392
7393 # check both POP and IMAP from a single server:
7394 option tcp-check
7395 tcp-check connect port 110
7396 tcp-check expect string +OK\ POP3\ ready
7397 tcp-check connect port 143
7398 tcp-check expect string *\ OK\ IMAP4\ ready
7399 server mail 10.0.0.1 check
7400
7401 See also : "option tcp-check", "tcp-check send", "tcp-check expect"
7402
7403
7404tcp-check expect [!] <match> <pattern>
7405 Specify data to be collected and analysed during a generic health check
7406 May be used in sections: defaults | frontend | listen | backend
7407 no | no | yes | yes
7408
7409 Arguments :
7410 <match> is a keyword indicating how to look for a specific pattern in the
7411 response. The keyword may be one of "string", "rstring" or
7412 binary.
7413 The keyword may be preceded by an exclamation mark ("!") to negate
7414 the match. Spaces are allowed between the exclamation mark and the
7415 keyword. See below for more details on the supported keywords.
7416
7417 <pattern> is the pattern to look for. It may be a string or a regular
7418 expression. If the pattern contains spaces, they must be escaped
7419 with the usual backslash ('\').
7420 If the match is set to binary, then the pattern must be passed as
7421 a serie of hexadecimal digits in an even number. Each sequence of
7422 two digits will represent a byte. The hexadecimal digits may be
7423 used upper or lower case.
7424
7425
7426 The available matches are intentionally similar to their http-check cousins :
7427
7428 string <string> : test the exact string matches in the response buffer.
7429 A health check response will be considered valid if the
7430 response's buffer contains this exact string. If the
7431 "string" keyword is prefixed with "!", then the response
7432 will be considered invalid if the body contains this
7433 string. This can be used to look for a mandatory pattern
7434 in a protocol response, or to detect a failure when a
7435 specific error appears in a protocol banner.
7436
7437 rstring <regex> : test a regular expression on the response buffer.
7438 A health check response will be considered valid if the
7439 response's buffer matches this expression. If the
7440 "rstring" keyword is prefixed with "!", then the response
7441 will be considered invalid if the body matches the
7442 expression.
7443
7444 binary <hexstring> : test the exact string in its hexadecimal form matches
7445 in the response buffer. A health check response will
7446 be considered valid if the response's buffer contains
7447 this exact hexadecimal string.
7448 Purpose is to match data on binary protocols.
7449
7450 It is important to note that the responses will be limited to a certain size
7451 defined by the global "tune.chksize" option, which defaults to 16384 bytes.
7452 Thus, too large responses may not contain the mandatory pattern when using
7453 "string", "rstring" or binary. If a large response is absolutely required, it
7454 is possible to change the default max size by setting the global variable.
7455 However, it is worth keeping in mind that parsing very large responses can
7456 waste some CPU cycles, especially when regular expressions are used, and that
7457 it is always better to focus the checks on smaller resources. Also, in its
7458 current state, the check will not find any string nor regex past a null
7459 character in the response. Similarly it is not possible to request matching
7460 the null character.
7461
7462 Examples :
7463 # perform a POP check
7464 option tcp-check
7465 tcp-check expect string +OK\ POP3\ ready
7466
7467 # perform an IMAP check
7468 option tcp-check
7469 tcp-check expect string *\ OK\ IMAP4\ ready
7470
7471 # look for the redis master server
7472 option tcp-check
7473 tcp-check send PING\r\n
7474 tcp-check expect +PONG
7475 tcp-check send info\ replication\r\n
7476 tcp-check expect string role:master
7477 tcp-check send QUIT\r\n
7478 tcp-check expect string +OK
7479
7480
7481 See also : "option tcp-check", "tcp-check connect", "tcp-check send",
7482 "tcp-check send-binary", "http-check expect", tune.chksize
7483
7484
7485tcp-check send <data>
7486 Specify a string to be sent as a question during a generic health check
7487 May be used in sections: defaults | frontend | listen | backend
7488 no | no | yes | yes
7489
7490 <data> : the data to be sent as a question during a generic health check
7491 session. For now, <data> must be a string.
7492
7493 Examples :
7494 # look for the redis master server
7495 option tcp-check
7496 tcp-check send info\ replication\r\n
7497 tcp-check expect string role:master
7498
7499 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
7500 "tcp-check send-binary", tune.chksize
7501
7502
7503tcp-check send-binary <hexastring>
7504 Specify an hexa digits string to be sent as a binary question during a raw
7505 tcp health check
7506 May be used in sections: defaults | frontend | listen | backend
7507 no | no | yes | yes
7508
7509 <data> : the data to be sent as a question during a generic health check
7510 session. For now, <data> must be a string.
7511 <hexastring> : test the exact string in its hexadecimal form matches in the
7512 response buffer. A health check response will be considered
7513 valid if the response's buffer contains this exact
7514 hexadecimal string.
7515 Purpose is to send binary data to ask on binary protocols.
7516
7517 Examples :
7518 # redis check in binary
7519 option tcp-check
7520 tcp-check send-binary 50494e470d0a # PING\r\n
7521 tcp-check expect binary 2b504F4e47 # +PONG
7522
7523
7524 See also : "option tcp-check", "tcp-check connect", "tcp-check expect",
7525 "tcp-check send", tune.chksize
7526
7527
Willy Tarreaue9656522010-08-17 15:40:09 +02007528tcp-request connection <action> [{if | unless} <condition>]
7529 Perform an action on an incoming connection depending on a layer 4 condition
Willy Tarreau1a687942010-05-23 22:40:30 +02007530 May be used in sections : defaults | frontend | listen | backend
7531 no | yes | yes | no
Willy Tarreaue9656522010-08-17 15:40:09 +02007532 Arguments :
7533 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007534 actions include : "accept", "reject", "track-sc0", "track-sc1",
7535 "track-sc2", and "expect-proxy". See below for more details.
Willy Tarreau1a687942010-05-23 22:40:30 +02007536
Willy Tarreaue9656522010-08-17 15:40:09 +02007537 <condition> is a standard layer4-only ACL-based condition (see section 7).
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007538
7539 Immediately after acceptance of a new incoming connection, it is possible to
7540 evaluate some conditions to decide whether this connection must be accepted
Willy Tarreaue9656522010-08-17 15:40:09 +02007541 or dropped or have its counters tracked. Those conditions cannot make use of
7542 any data contents because the connection has not been read from yet, and the
7543 buffers are not yet allocated. This is used to selectively and very quickly
7544 accept or drop connections from various sources with a very low overhead. If
7545 some contents need to be inspected in order to take the decision, the
7546 "tcp-request content" statements must be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007547
Willy Tarreaue9656522010-08-17 15:40:09 +02007548 The "tcp-request connection" rules are evaluated in their exact declaration
7549 order. If no rule matches or if there is no rule, the default action is to
7550 accept the incoming connection. There is no specific limit to the number of
7551 rules which may be inserted.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007552
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007553 Five types of actions are supported :
Willy Tarreaue9656522010-08-17 15:40:09 +02007554 - accept :
7555 accepts the connection if the condition is true (when used with "if")
7556 or false (when used with "unless"). The first such rule executed ends
7557 the rules evaluation.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007558
Willy Tarreaue9656522010-08-17 15:40:09 +02007559 - reject :
7560 rejects the connection if the condition is true (when used with "if")
7561 or false (when used with "unless"). The first such rule executed ends
7562 the rules evaluation. Rejected connections do not even become a
7563 session, which is why they are accounted separately for in the stats,
7564 as "denied connections". They are not considered for the session
7565 rate-limit and are not logged either. The reason is that these rules
7566 should only be used to filter extremely high connection rates such as
7567 the ones encountered during a massive DDoS attack. Under these extreme
7568 conditions, the simple action of logging each event would make the
7569 system collapse and would considerably lower the filtering capacity. If
7570 logging is absolutely desired, then "tcp-request content" rules should
7571 be used instead.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007572
Willy Tarreau4f0d9192013-06-11 20:40:55 +02007573 - expect-proxy layer4 :
7574 configures the client-facing connection to receive a PROXY protocol
7575 header before any byte is read from the socket. This is equivalent to
7576 having the "accept-proxy" keyword on the "bind" line, except that using
7577 the TCP rule allows the PROXY protocol to be accepted only for certain
7578 IP address ranges using an ACL. This is convenient when multiple layers
7579 of load balancers are passed through by traffic coming from public
7580 hosts.
7581
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007582 - capture <sample> len <length> :
7583 This only applies to "tcp-request content" rules. It captures sample
7584 expression <sample> from the request buffer, and converts it to a
7585 string of at most <len> characters. The resulting string is stored into
7586 the next request "capture" slot, so it will possibly appear next to
7587 some captured HTTP headers. It will then automatically appear in the
7588 logs, and it will be possible to extract it using sample fetch rules to
7589 feed it into headers or anything. The length should be limited given
7590 that this size will be allocated for each capture during the whole
7591 session life. Since it applies to Please check section 7.3 (Fetching
7592 samples) and "capture request header" for more information.
7593
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007594 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>] :
Willy Tarreaue9656522010-08-17 15:40:09 +02007595 enables tracking of sticky counters from current connection. These
Willy Tarreau09448f72014-06-25 18:12:15 +02007596 rules do not stop evaluation and do not change default action. 3 sets
Willy Tarreaue9656522010-08-17 15:40:09 +02007597 of counters may be simultaneously tracked by the same connection. The
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007598 first "track-sc0" rule executed enables tracking of the counters of the
7599 specified table as the first set. The first "track-sc1" rule executed
Willy Tarreaue9656522010-08-17 15:40:09 +02007600 enables tracking of the counters of the specified table as the second
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007601 set. The first "track-sc2" rule executed enables tracking of the
7602 counters of the specified table as the third set. It is a recommended
7603 practice to use the first set of counters for the per-frontend counters
7604 and the second set for the per-backend ones. But this is just a
7605 guideline, all may be used everywhere.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007606
Willy Tarreaue9656522010-08-17 15:40:09 +02007607 These actions take one or two arguments :
Willy Tarreaube722a22014-06-13 16:31:59 +02007608 <key> is mandatory, and is a sample expression rule as described
Willy Tarreau74ca5042013-06-11 23:12:07 +02007609 in section 7.3. It describes what elements of the incoming
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007610 request or connection will be analysed, extracted, combined,
7611 and used to select which table entry to update the counters.
7612 Note that "tcp-request connection" cannot use content-based
7613 fetches.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007614
Willy Tarreaue9656522010-08-17 15:40:09 +02007615 <table> is an optional table to be used instead of the default one,
7616 which is the stick-table declared in the current proxy. All
7617 the counters for the matches and updates for the key will
7618 then be performed in that table until the session ends.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007619
Willy Tarreaue9656522010-08-17 15:40:09 +02007620 Once a "track-sc*" rule is executed, the key is looked up in the table
7621 and if it is not found, an entry is allocated for it. Then a pointer to
7622 that entry is kept during all the session's life, and this entry's
7623 counters are updated as often as possible, every time the session's
7624 counters are updated, and also systematically when the session ends.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007625 Counters are only updated for events that happen after the tracking has
7626 been started. For example, connection counters will not be updated when
7627 tracking layer 7 information, since the connection event happens before
7628 layer7 information is extracted.
7629
Willy Tarreaue9656522010-08-17 15:40:09 +02007630 If the entry tracks concurrent connection counters, one connection is
7631 counted for as long as the entry is tracked, and the entry will not
7632 expire during that time. Tracking counters also provides a performance
7633 advantage over just checking the keys, because only one table lookup is
7634 performed for all ACL checks that make use of it.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007635
Willy Tarreaue9656522010-08-17 15:40:09 +02007636 Note that the "if/unless" condition is optional. If no condition is set on
7637 the action, it is simply performed unconditionally. That can be useful for
7638 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007639
Willy Tarreaue9656522010-08-17 15:40:09 +02007640 Example: accept all connections from white-listed hosts, reject too fast
7641 connection without counting them, and track accepted connections.
7642 This results in connection rate being capped from abusive sources.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007643
Willy Tarreaue9656522010-08-17 15:40:09 +02007644 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007645 tcp-request connection reject if { src_conn_rate gt 10 }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007646 tcp-request connection track-sc0 src
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007647
Willy Tarreaue9656522010-08-17 15:40:09 +02007648 Example: accept all connections from white-listed hosts, count all other
7649 connections and reject too fast ones. This results in abusive ones
7650 being blocked as long as they don't slow down.
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007651
Willy Tarreaue9656522010-08-17 15:40:09 +02007652 tcp-request connection accept if { src -f /etc/haproxy/whitelist.lst }
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007653 tcp-request connection track-sc0 src
7654 tcp-request connection reject if { sc0_conn_rate gt 10 }
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007655
Willy Tarreau4f0d9192013-06-11 20:40:55 +02007656 Example: enable the PROXY protocol for traffic coming from all known proxies.
7657
7658 tcp-request connection expect-proxy layer4 if { src -f proxies.lst }
7659
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007660 See section 7 about ACL usage.
7661
Willy Tarreaue9656522010-08-17 15:40:09 +02007662 See also : "tcp-request content", "stick-table"
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007663
7664
Willy Tarreaue9656522010-08-17 15:40:09 +02007665tcp-request content <action> [{if | unless} <condition>]
7666 Perform an action on a new session depending on a layer 4-7 condition
Willy Tarreau62644772008-07-16 18:36:06 +02007667 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02007668 no | yes | yes | yes
Willy Tarreaue9656522010-08-17 15:40:09 +02007669 Arguments :
7670 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007671 actions include : "accept", "reject", "track-sc0", "track-sc1",
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007672 "track-sc2" and "capture". See "tcp-request connection" above
7673 for their signification.
Willy Tarreau62644772008-07-16 18:36:06 +02007674
Willy Tarreaue9656522010-08-17 15:40:09 +02007675 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
Willy Tarreau62644772008-07-16 18:36:06 +02007676
Willy Tarreaue9656522010-08-17 15:40:09 +02007677 A request's contents can be analysed at an early stage of request processing
7678 called "TCP content inspection". During this stage, ACL-based rules are
7679 evaluated every time the request contents are updated, until either an
7680 "accept" or a "reject" rule matches, or the TCP request inspection delay
7681 expires with no matching rule.
Willy Tarreau62644772008-07-16 18:36:06 +02007682
Willy Tarreaue9656522010-08-17 15:40:09 +02007683 The first difference between these rules and "tcp-request connection" rules
7684 is that "tcp-request content" rules can make use of contents to take a
7685 decision. Most often, these decisions will consider a protocol recognition or
7686 validity. The second difference is that content-based rules can be used in
Willy Tarreauf3338342014-01-28 21:40:28 +01007687 both frontends and backends. In case of HTTP keep-alive with the client, all
7688 tcp-request content rules are evaluated again, so haproxy keeps a record of
7689 what sticky counters were assigned by a "tcp-request connection" versus a
7690 "tcp-request content" rule, and flushes all the content-related ones after
7691 processing an HTTP request, so that they may be evaluated again by the rules
7692 being evaluated again for the next request. This is of particular importance
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007693 when the rule tracks some L7 information or when it is conditioned by an
Willy Tarreauf3338342014-01-28 21:40:28 +01007694 L7-based ACL, since tracking may change between requests.
Willy Tarreau62644772008-07-16 18:36:06 +02007695
Willy Tarreaue9656522010-08-17 15:40:09 +02007696 Content-based rules are evaluated in their exact declaration order. If no
7697 rule matches or if there is no rule, the default action is to accept the
7698 contents. There is no specific limit to the number of rules which may be
7699 inserted.
Willy Tarreau62644772008-07-16 18:36:06 +02007700
Willy Tarreau18bf01e2014-06-13 16:18:52 +02007701 Four types of actions are supported :
7702 - accept : the request is accepted
7703 - reject : the request is rejected and the connection is closed
7704 - capture : the specified sample expression is captured
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007705 - { track-sc0 | track-sc1 | track-sc2 } <key> [table <table>]
Willy Tarreau62644772008-07-16 18:36:06 +02007706
Willy Tarreaue9656522010-08-17 15:40:09 +02007707 They have the same meaning as their counter-parts in "tcp-request connection"
7708 so please refer to that section for a complete description.
Willy Tarreau62644772008-07-16 18:36:06 +02007709
Willy Tarreauf3338342014-01-28 21:40:28 +01007710 While there is nothing mandatory about it, it is recommended to use the
7711 track-sc0 in "tcp-request connection" rules, track-sc1 for "tcp-request
7712 content" rules in the frontend, and track-sc2 for "tcp-request content"
7713 rules in the backend, because that makes the configuration more readable
7714 and easier to troubleshoot, but this is just a guideline and all counters
7715 may be used everywhere.
Willy Tarreau62644772008-07-16 18:36:06 +02007716
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007717 Note that the "if/unless" condition is optional. If no condition is set on
Willy Tarreaue9656522010-08-17 15:40:09 +02007718 the action, it is simply performed unconditionally. That can be useful for
7719 "track-sc*" actions as well as for changing the default action to a reject.
Willy Tarreau62644772008-07-16 18:36:06 +02007720
Willy Tarreaue9656522010-08-17 15:40:09 +02007721 It is perfectly possible to match layer 7 contents with "tcp-request content"
Willy Tarreauc0239e02012-04-16 14:42:55 +02007722 rules, since HTTP-specific ACL matches are able to preliminarily parse the
7723 contents of a buffer before extracting the required data. If the buffered
7724 contents do not parse as a valid HTTP message, then the ACL does not match.
7725 The parser which is involved there is exactly the same as for all other HTTP
Willy Tarreauf3338342014-01-28 21:40:28 +01007726 processing, so there is no risk of parsing something differently. In an HTTP
7727 backend connected to from an HTTP frontend, it is guaranteed that HTTP
7728 contents will always be immediately present when the rule is evaluated first.
Willy Tarreau62644772008-07-16 18:36:06 +02007729
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007730 Tracking layer7 information is also possible provided that the information
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02007731 are present when the rule is processed. The rule processing engine is able to
7732 wait until the inspect delay expires when the data to be tracked is not yet
7733 available.
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007734
Willy Tarreau62644772008-07-16 18:36:06 +02007735 Example:
Willy Tarreaue9656522010-08-17 15:40:09 +02007736 # Accept HTTP requests containing a Host header saying "example.com"
7737 # and reject everything else.
7738 acl is_host_com hdr(Host) -i example.com
7739 tcp-request inspect-delay 30s
Willy Tarreauc0239e02012-04-16 14:42:55 +02007740 tcp-request content accept if is_host_com
Willy Tarreaue9656522010-08-17 15:40:09 +02007741 tcp-request content reject
7742
7743 Example:
Willy Tarreau62644772008-07-16 18:36:06 +02007744 # reject SMTP connection if client speaks first
7745 tcp-request inspect-delay 30s
7746 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007747 tcp-request content reject if content_present
Willy Tarreau62644772008-07-16 18:36:06 +02007748
7749 # Forward HTTPS connection only if client speaks
7750 tcp-request inspect-delay 30s
7751 acl content_present req_len gt 0
Willy Tarreau68c03ab2010-08-06 15:08:45 +02007752 tcp-request content accept if content_present
Willy Tarreaue9656522010-08-17 15:40:09 +02007753 tcp-request content reject
7754
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007755 Example:
7756 # Track the last IP from X-Forwarded-For
7757 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02007758 tcp-request content track-sc0 hdr(x-forwarded-for,-1)
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007759
7760 Example:
7761 # track request counts per "base" (concatenation of Host+URL)
7762 tcp-request inspect-delay 10s
Willy Tarreau4d54c7c2014-09-16 15:48:15 +02007763 tcp-request content track-sc0 base table req-rate
Willy Tarreau5d5b5d82012-12-09 12:00:04 +01007764
Willy Tarreaue9656522010-08-17 15:40:09 +02007765 Example: track per-frontend and per-backend counters, block abusers at the
7766 frontend when the backend detects abuse.
7767
7768 frontend http
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007769 # Use General Purpose Couter 0 in SC0 as a global abuse counter
Willy Tarreaue9656522010-08-17 15:40:09 +02007770 # protecting all our sites
7771 stick-table type ip size 1m expire 5m store gpc0
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007772 tcp-request connection track-sc0 src
7773 tcp-request connection reject if { sc0_get_gpc0 gt 0 }
Willy Tarreaue9656522010-08-17 15:40:09 +02007774 ...
7775 use_backend http_dynamic if { path_end .php }
7776
7777 backend http_dynamic
7778 # if a source makes too fast requests to this dynamic site (tracked
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007779 # by SC1), block it globally in the frontend.
Willy Tarreaue9656522010-08-17 15:40:09 +02007780 stick-table type ip size 1m expire 5m store http_req_rate(10s)
Willy Tarreaube4a3ef2013-06-17 15:04:07 +02007781 acl click_too_fast sc1_http_req_rate gt 10
7782 acl mark_as_abuser sc0_inc_gpc0 gt 0
7783 tcp-request content track-sc1 src
Willy Tarreaue9656522010-08-17 15:40:09 +02007784 tcp-request content reject if click_too_fast mark_as_abuser
Willy Tarreau62644772008-07-16 18:36:06 +02007785
Willy Tarreauc57f0e22009-05-10 13:12:33 +02007786 See section 7 about ACL usage.
Willy Tarreau62644772008-07-16 18:36:06 +02007787
Willy Tarreaue9656522010-08-17 15:40:09 +02007788 See also : "tcp-request connection", "tcp-request inspect-delay"
Willy Tarreau62644772008-07-16 18:36:06 +02007789
7790
7791tcp-request inspect-delay <timeout>
7792 Set the maximum allowed time to wait for data during content inspection
7793 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaufb356202010-08-03 14:02:05 +02007794 no | yes | yes | yes
Willy Tarreau62644772008-07-16 18:36:06 +02007795 Arguments :
7796 <timeout> is the timeout value specified in milliseconds by default, but
7797 can be in any other unit if the number is suffixed by the unit,
7798 as explained at the top of this document.
7799
7800 People using haproxy primarily as a TCP relay are often worried about the
7801 risk of passing any type of protocol to a server without any analysis. In
7802 order to be able to analyze the request contents, we must first withhold
7803 the data then analyze them. This statement simply enables withholding of
7804 data for at most the specified amount of time.
7805
Willy Tarreaufb356202010-08-03 14:02:05 +02007806 TCP content inspection applies very early when a connection reaches a
7807 frontend, then very early when the connection is forwarded to a backend. This
7808 means that a connection may experience a first delay in the frontend and a
7809 second delay in the backend if both have tcp-request rules.
7810
Willy Tarreau62644772008-07-16 18:36:06 +02007811 Note that when performing content inspection, haproxy will evaluate the whole
7812 rules for every new chunk which gets in, taking into account the fact that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01007813 those data are partial. If no rule matches before the aforementioned delay,
Willy Tarreau62644772008-07-16 18:36:06 +02007814 a last check is performed upon expiration, this time considering that the
Willy Tarreaud869b242009-03-15 14:43:58 +01007815 contents are definitive. If no delay is set, haproxy will not wait at all
7816 and will immediately apply a verdict based on the available information.
7817 Obviously this is unlikely to be very useful and might even be racy, so such
7818 setups are not recommended.
Willy Tarreau62644772008-07-16 18:36:06 +02007819
7820 As soon as a rule matches, the request is released and continues as usual. If
7821 the timeout is reached and no rule matches, the default policy will be to let
7822 it pass through unaffected.
7823
7824 For most protocols, it is enough to set it to a few seconds, as most clients
7825 send the full request immediately upon connection. Add 3 or more seconds to
7826 cover TCP retransmits but that's all. For some protocols, it may make sense
Willy Tarreaud72758d2010-01-12 10:42:19 +01007827 to use large values, for instance to ensure that the client never talks
Willy Tarreau62644772008-07-16 18:36:06 +02007828 before the server (eg: SMTP), or to wait for a client to talk before passing
7829 data to the server (eg: SSL). Note that the client timeout must cover at
Willy Tarreaub824b002010-09-29 16:36:16 +02007830 least the inspection delay, otherwise it will expire first. If the client
7831 closes the connection or if the buffer is full, the delay immediately expires
7832 since the contents will not be able to change anymore.
Willy Tarreau62644772008-07-16 18:36:06 +02007833
Willy Tarreau55165fe2009-05-10 12:02:55 +02007834 See also : "tcp-request content accept", "tcp-request content reject",
Willy Tarreau62644772008-07-16 18:36:06 +02007835 "timeout client".
7836
7837
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007838tcp-response content <action> [{if | unless} <condition>]
7839 Perform an action on a session response depending on a layer 4-7 condition
7840 May be used in sections : defaults | frontend | listen | backend
7841 no | no | yes | yes
7842 Arguments :
7843 <action> defines the action to perform if the condition applies. Valid
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007844 actions include : "accept", "close", "reject".
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007845
7846 <condition> is a standard layer 4-7 ACL-based condition (see section 7).
7847
7848 Response contents can be analysed at an early stage of response processing
7849 called "TCP content inspection". During this stage, ACL-based rules are
7850 evaluated every time the response contents are updated, until either an
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007851 "accept", "close" or a "reject" rule matches, or a TCP response inspection
7852 delay is set and expires with no matching rule.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007853
7854 Most often, these decisions will consider a protocol recognition or validity.
7855
7856 Content-based rules are evaluated in their exact declaration order. If no
7857 rule matches or if there is no rule, the default action is to accept the
7858 contents. There is no specific limit to the number of rules which may be
7859 inserted.
7860
7861 Two types of actions are supported :
7862 - accept :
7863 accepts the response if the condition is true (when used with "if")
7864 or false (when used with "unless"). The first such rule executed ends
7865 the rules evaluation.
7866
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007867 - close :
7868 immediately closes the connection with the server if the condition is
7869 true (when used with "if"), or false (when used with "unless"). The
7870 first such rule executed ends the rules evaluation. The main purpose of
7871 this action is to force a connection to be finished between a client
7872 and a server after an exchange when the application protocol expects
7873 some long time outs to elapse first. The goal is to eliminate idle
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03007874 connections which take significant resources on servers with certain
Willy Tarreaucc1e04b2013-09-11 23:20:29 +02007875 protocols.
7876
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007877 - reject :
7878 rejects the response if the condition is true (when used with "if")
7879 or false (when used with "unless"). The first such rule executed ends
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007880 the rules evaluation. Rejected session are immediately closed.
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007881
7882 Note that the "if/unless" condition is optional. If no condition is set on
7883 the action, it is simply performed unconditionally. That can be useful for
7884 for changing the default action to a reject.
7885
Jamie Gloudonaaa21002012-08-25 00:18:33 -04007886 It is perfectly possible to match layer 7 contents with "tcp-response
7887 content" rules, but then it is important to ensure that a full response has
7888 been buffered, otherwise no contents will match. In order to achieve this,
7889 the best solution involves detecting the HTTP protocol during the inspection
Emeric Brun0a3b67f2010-09-24 15:34:53 +02007890 period.
7891
7892 See section 7 about ACL usage.
7893
7894 See also : "tcp-request content", "tcp-response inspect-delay"
7895
7896
7897tcp-response inspect-delay <timeout>
7898 Set the maximum allowed time to wait for a response during content inspection
7899 May be used in sections : defaults | frontend | listen | backend
7900 no | no | yes | yes
7901 Arguments :
7902 <timeout> is the timeout value specified in milliseconds by default, but
7903 can be in any other unit if the number is suffixed by the unit,
7904 as explained at the top of this document.
7905
7906 See also : "tcp-response content", "tcp-request inspect-delay".
7907
7908
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007909timeout check <timeout>
7910 Set additional check timeout, but only after a connection has been already
7911 established.
7912
7913 May be used in sections: defaults | frontend | listen | backend
7914 yes | no | yes | yes
7915 Arguments:
7916 <timeout> is the timeout value specified in milliseconds by default, but
7917 can be in any other unit if the number is suffixed by the unit,
7918 as explained at the top of this document.
7919
7920 If set, haproxy uses min("timeout connect", "inter") as a connect timeout
7921 for check and "timeout check" as an additional read timeout. The "min" is
7922 used so that people running with *very* long "timeout connect" (eg. those
7923 who needed this due to the queue or tarpit) do not slow down their checks.
Willy Tarreaud7550a22010-02-10 05:10:19 +01007924 (Please also note that there is no valid reason to have such long connect
7925 timeouts, because "timeout queue" and "timeout tarpit" can always be used to
7926 avoid that).
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007927
7928 If "timeout check" is not set haproxy uses "inter" for complete check
7929 timeout (connect + read) exactly like all <1.3.15 version.
7930
7931 In most cases check request is much simpler and faster to handle than normal
7932 requests and people may want to kick out laggy servers so this timeout should
Willy Tarreau41a340d2008-01-22 12:25:31 +01007933 be smaller than "timeout server".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007934
7935 This parameter is specific to backends, but can be specified once for all in
7936 "defaults" sections. This is in fact one of the easiest solutions not to
7937 forget about it.
7938
Willy Tarreau41a340d2008-01-22 12:25:31 +01007939 See also: "timeout connect", "timeout queue", "timeout server",
7940 "timeout tarpit".
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01007941
7942
Willy Tarreau0ba27502007-12-24 16:55:16 +01007943timeout client <timeout>
7944timeout clitimeout <timeout> (deprecated)
7945 Set the maximum inactivity time on the client side.
7946 May be used in sections : defaults | frontend | listen | backend
7947 yes | yes | yes | no
7948 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01007949 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01007950 can be in any other unit if the number is suffixed by the unit,
7951 as explained at the top of this document.
7952
7953 The inactivity timeout applies when the client is expected to acknowledge or
7954 send data. In HTTP mode, this timeout is particularly important to consider
7955 during the first phase, when the client sends the request, and during the
7956 response while it is reading data sent by the server. The value is specified
7957 in milliseconds by default, but can be in any other unit if the number is
7958 suffixed by the unit, as specified at the top of this document. In TCP mode
7959 (and to a lesser extent, in HTTP mode), it is highly recommended that the
7960 client timeout remains equal to the server timeout in order to avoid complex
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01007961 situations to debug. It is a good practice to cover one or several TCP packet
Willy Tarreau0ba27502007-12-24 16:55:16 +01007962 losses by specifying timeouts that are slightly above multiples of 3 seconds
Willy Tarreauce887fd2012-05-12 12:50:00 +02007963 (eg: 4 or 5 seconds). If some long-lived sessions are mixed with short-lived
7964 sessions (eg: WebSocket and HTTP), it's worth considering "timeout tunnel",
Willy Tarreau05cdd962014-05-10 14:30:07 +02007965 which overrides "timeout client" and "timeout server" for tunnels, as well as
7966 "timeout client-fin" for half-closed connections.
Willy Tarreau0ba27502007-12-24 16:55:16 +01007967
7968 This parameter is specific to frontends, but can be specified once for all in
7969 "defaults" sections. This is in fact one of the easiest solutions not to
7970 forget about it. An unspecified timeout results in an infinite timeout, which
7971 is not recommended. Such a usage is accepted and works but reports a warning
7972 during startup because it may results in accumulation of expired sessions in
7973 the system if the system's timeouts are not configured either.
7974
7975 This parameter replaces the old, deprecated "clitimeout". It is recommended
7976 to use it to write new configurations. The form "timeout clitimeout" is
7977 provided only by backwards compatibility but its use is strongly discouraged.
7978
Willy Tarreauce887fd2012-05-12 12:50:00 +02007979 See also : "clitimeout", "timeout server", "timeout tunnel".
Willy Tarreau0ba27502007-12-24 16:55:16 +01007980
7981
Willy Tarreau05cdd962014-05-10 14:30:07 +02007982timeout client-fin <timeout>
7983 Set the inactivity timeout on the client side for half-closed connections.
7984 May be used in sections : defaults | frontend | listen | backend
7985 yes | yes | yes | no
7986 Arguments :
7987 <timeout> is the timeout value specified in milliseconds by default, but
7988 can be in any other unit if the number is suffixed by the unit,
7989 as explained at the top of this document.
7990
7991 The inactivity timeout applies when the client is expected to acknowledge or
7992 send data while one direction is already shut down. This timeout is different
7993 from "timeout client" in that it only applies to connections which are closed
7994 in one direction. This is particularly useful to avoid keeping connections in
7995 FIN_WAIT state for too long when clients do not disconnect cleanly. This
7996 problem is particularly common long connections such as RDP or WebSocket.
7997 Note that this timeout can override "timeout tunnel" when a connection shuts
7998 down in one direction.
7999
8000 This parameter is specific to frontends, but can be specified once for all in
8001 "defaults" sections. By default it is not set, so half-closed connections
8002 will use the other timeouts (timeout.client or timeout.tunnel).
8003
8004 See also : "timeout client", "timeout server-fin", and "timeout tunnel".
8005
8006
Willy Tarreau0ba27502007-12-24 16:55:16 +01008007timeout connect <timeout>
8008timeout contimeout <timeout> (deprecated)
8009 Set the maximum time to wait for a connection attempt to a server to succeed.
8010 May be used in sections : defaults | frontend | listen | backend
8011 yes | no | yes | yes
8012 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008013 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau0ba27502007-12-24 16:55:16 +01008014 can be in any other unit if the number is suffixed by the unit,
8015 as explained at the top of this document.
8016
8017 If the server is located on the same LAN as haproxy, the connection should be
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008018 immediate (less than a few milliseconds). Anyway, it is a good practice to
Willy Tarreaud72758d2010-01-12 10:42:19 +01008019 cover one or several TCP packet losses by specifying timeouts that are
Willy Tarreau0ba27502007-12-24 16:55:16 +01008020 slightly above multiples of 3 seconds (eg: 4 or 5 seconds). By default, the
Krzysztof Piotr Oledzki5259dfe2008-01-21 01:54:06 +01008021 connect timeout also presets both queue and tarpit timeouts to the same value
8022 if these have not been specified.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008023
8024 This parameter is specific to backends, but can be specified once for all in
8025 "defaults" sections. This is in fact one of the easiest solutions not to
8026 forget about it. An unspecified timeout results in an infinite timeout, which
8027 is not recommended. Such a usage is accepted and works but reports a warning
8028 during startup because it may results in accumulation of failed sessions in
8029 the system if the system's timeouts are not configured either.
8030
8031 This parameter replaces the old, deprecated "contimeout". It is recommended
8032 to use it to write new configurations. The form "timeout contimeout" is
8033 provided only by backwards compatibility but its use is strongly discouraged.
8034
Willy Tarreau41a340d2008-01-22 12:25:31 +01008035 See also: "timeout check", "timeout queue", "timeout server", "contimeout",
8036 "timeout tarpit".
Willy Tarreau0ba27502007-12-24 16:55:16 +01008037
8038
Willy Tarreaub16a5742010-01-10 14:46:16 +01008039timeout http-keep-alive <timeout>
8040 Set the maximum allowed time to wait for a new HTTP request to appear
8041 May be used in sections : defaults | frontend | listen | backend
8042 yes | yes | yes | yes
8043 Arguments :
8044 <timeout> is the timeout value specified in milliseconds by default, but
8045 can be in any other unit if the number is suffixed by the unit,
8046 as explained at the top of this document.
8047
8048 By default, the time to wait for a new request in case of keep-alive is set
8049 by "timeout http-request". However this is not always convenient because some
8050 people want very short keep-alive timeouts in order to release connections
8051 faster, and others prefer to have larger ones but still have short timeouts
8052 once the request has started to present itself.
8053
8054 The "http-keep-alive" timeout covers these needs. It will define how long to
8055 wait for a new HTTP request to start coming after a response was sent. Once
8056 the first byte of request has been seen, the "http-request" timeout is used
8057 to wait for the complete request to come. Note that empty lines prior to a
8058 new request do not refresh the timeout and are not counted as a new request.
8059
8060 There is also another difference between the two timeouts : when a connection
8061 expires during timeout http-keep-alive, no error is returned, the connection
8062 just closes. If the connection expires in "http-request" while waiting for a
8063 connection to complete, a HTTP 408 error is returned.
8064
8065 In general it is optimal to set this value to a few tens to hundreds of
8066 milliseconds, to allow users to fetch all objects of a page at once but
8067 without waiting for further clicks. Also, if set to a very small value (eg:
8068 1 millisecond) it will probably only accept pipelined requests but not the
8069 non-pipelined ones. It may be a nice trade-off for very large sites running
Patrick Mézard2382ad62010-05-09 10:43:32 +02008070 with tens to hundreds of thousands of clients.
Willy Tarreaub16a5742010-01-10 14:46:16 +01008071
8072 If this parameter is not set, the "http-request" timeout applies, and if both
8073 are not set, "timeout client" still applies at the lower level. It should be
8074 set in the frontend to take effect, unless the frontend is in TCP mode, in
8075 which case the HTTP backend's timeout will be used.
8076
8077 See also : "timeout http-request", "timeout client".
8078
8079
Willy Tarreau036fae02008-01-06 13:24:40 +01008080timeout http-request <timeout>
8081 Set the maximum allowed time to wait for a complete HTTP request
8082 May be used in sections : defaults | frontend | listen | backend
Willy Tarreaucd7afc02009-07-12 10:03:17 +02008083 yes | yes | yes | yes
Willy Tarreau036fae02008-01-06 13:24:40 +01008084 Arguments :
Willy Tarreau844e3c52008-01-11 16:28:18 +01008085 <timeout> is the timeout value specified in milliseconds by default, but
Willy Tarreau036fae02008-01-06 13:24:40 +01008086 can be in any other unit if the number is suffixed by the unit,
8087 as explained at the top of this document.
8088
8089 In order to offer DoS protection, it may be required to lower the maximum
8090 accepted time to receive a complete HTTP request without affecting the client
8091 timeout. This helps protecting against established connections on which
8092 nothing is sent. The client timeout cannot offer a good protection against
8093 this abuse because it is an inactivity timeout, which means that if the
8094 attacker sends one character every now and then, the timeout will not
8095 trigger. With the HTTP request timeout, no matter what speed the client
Willy Tarreau2705a612014-05-23 17:38:34 +02008096 types, the request will be aborted if it does not complete in time. When the
8097 timeout expires, an HTTP 408 response is sent to the client to inform it
8098 about the problem, and the connection is closed. The logs will report
8099 termination codes "cR". Some recent browsers are having problems with this
8100 standard, well-documented behaviour, so it might be needed to hide the 408
8101 code using "errorfile 408 /dev/null". See more details in the explanations of
8102 the "cR" termination code in section 8.5.
Willy Tarreau036fae02008-01-06 13:24:40 +01008103
8104 Note that this timeout only applies to the header part of the request, and
8105 not to any data. As soon as the empty line is received, this timeout is not
Willy Tarreaub16a5742010-01-10 14:46:16 +01008106 used anymore. It is used again on keep-alive connections to wait for a second
8107 request if "timeout http-keep-alive" is not set.
Willy Tarreau036fae02008-01-06 13:24:40 +01008108
8109 Generally it is enough to set it to a few seconds, as most clients send the
8110 full request immediately upon connection. Add 3 or more seconds to cover TCP
8111 retransmits but that's all. Setting it to very low values (eg: 50 ms) will
8112 generally work on local networks as long as there are no packet losses. This
8113 will prevent people from sending bare HTTP requests using telnet.
8114
8115 If this parameter is not set, the client timeout still applies between each
Willy Tarreaucd7afc02009-07-12 10:03:17 +02008116 chunk of the incoming request. It should be set in the frontend to take
8117 effect, unless the frontend is in TCP mode, in which case the HTTP backend's
8118 timeout will be used.
Willy Tarreau036fae02008-01-06 13:24:40 +01008119
Willy Tarreau2705a612014-05-23 17:38:34 +02008120 See also : "errorfile", "timeout http-keep-alive", "timeout client".
Willy Tarreau036fae02008-01-06 13:24:40 +01008121
Willy Tarreau844e3c52008-01-11 16:28:18 +01008122
8123timeout queue <timeout>
8124 Set the maximum time to wait in the queue for a connection slot to be free
8125 May be used in sections : defaults | frontend | listen | backend
8126 yes | no | yes | yes
8127 Arguments :
8128 <timeout> is the timeout value specified in milliseconds by default, but
8129 can be in any other unit if the number is suffixed by the unit,
8130 as explained at the top of this document.
8131
8132 When a server's maxconn is reached, connections are left pending in a queue
8133 which may be server-specific or global to the backend. In order not to wait
8134 indefinitely, a timeout is applied to requests pending in the queue. If the
8135 timeout is reached, it is considered that the request will almost never be
8136 served, so it is dropped and a 503 error is returned to the client.
8137
8138 The "timeout queue" statement allows to fix the maximum time for a request to
8139 be left pending in a queue. If unspecified, the same value as the backend's
8140 connection timeout ("timeout connect") is used, for backwards compatibility
8141 with older versions with no "timeout queue" parameter.
8142
8143 See also : "timeout connect", "contimeout".
8144
8145
8146timeout server <timeout>
8147timeout srvtimeout <timeout> (deprecated)
8148 Set the maximum inactivity time on the server side.
8149 May be used in sections : defaults | frontend | listen | backend
8150 yes | no | yes | yes
8151 Arguments :
8152 <timeout> is the timeout value specified in milliseconds by default, but
8153 can be in any other unit if the number is suffixed by the unit,
8154 as explained at the top of this document.
8155
8156 The inactivity timeout applies when the server is expected to acknowledge or
8157 send data. In HTTP mode, this timeout is particularly important to consider
8158 during the first phase of the server's response, when it has to send the
8159 headers, as it directly represents the server's processing time for the
8160 request. To find out what value to put there, it's often good to start with
8161 what would be considered as unacceptable response times, then check the logs
8162 to observe the response time distribution, and adjust the value accordingly.
8163
8164 The value is specified in milliseconds by default, but can be in any other
8165 unit if the number is suffixed by the unit, as specified at the top of this
8166 document. In TCP mode (and to a lesser extent, in HTTP mode), it is highly
8167 recommended that the client timeout remains equal to the server timeout in
8168 order to avoid complex situations to debug. Whatever the expected server
Willy Tarreaud2a4aa22008-01-31 15:28:22 +01008169 response times, it is a good practice to cover at least one or several TCP
Willy Tarreau844e3c52008-01-11 16:28:18 +01008170 packet losses by specifying timeouts that are slightly above multiples of 3
Willy Tarreauce887fd2012-05-12 12:50:00 +02008171 seconds (eg: 4 or 5 seconds minimum). If some long-lived sessions are mixed
8172 with short-lived sessions (eg: WebSocket and HTTP), it's worth considering
8173 "timeout tunnel", which overrides "timeout client" and "timeout server" for
8174 tunnels.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008175
8176 This parameter is specific to backends, but can be specified once for all in
8177 "defaults" sections. This is in fact one of the easiest solutions not to
8178 forget about it. An unspecified timeout results in an infinite timeout, which
8179 is not recommended. Such a usage is accepted and works but reports a warning
8180 during startup because it may results in accumulation of expired sessions in
8181 the system if the system's timeouts are not configured either.
8182
8183 This parameter replaces the old, deprecated "srvtimeout". It is recommended
8184 to use it to write new configurations. The form "timeout srvtimeout" is
8185 provided only by backwards compatibility but its use is strongly discouraged.
8186
Willy Tarreauce887fd2012-05-12 12:50:00 +02008187 See also : "srvtimeout", "timeout client" and "timeout tunnel".
Willy Tarreau844e3c52008-01-11 16:28:18 +01008188
Willy Tarreau05cdd962014-05-10 14:30:07 +02008189
8190timeout server-fin <timeout>
8191 Set the inactivity timeout on the server side for half-closed connections.
8192 May be used in sections : defaults | frontend | listen | backend
8193 yes | no | yes | yes
8194 Arguments :
8195 <timeout> is the timeout value specified in milliseconds by default, but
8196 can be in any other unit if the number is suffixed by the unit,
8197 as explained at the top of this document.
8198
8199 The inactivity timeout applies when the server is expected to acknowledge or
8200 send data while one direction is already shut down. This timeout is different
8201 from "timeout server" in that it only applies to connections which are closed
8202 in one direction. This is particularly useful to avoid keeping connections in
8203 FIN_WAIT state for too long when a remote server does not disconnect cleanly.
8204 This problem is particularly common long connections such as RDP or WebSocket.
8205 Note that this timeout can override "timeout tunnel" when a connection shuts
8206 down in one direction. This setting was provided for completeness, but in most
8207 situations, it should not be needed.
8208
8209 This parameter is specific to backends, but can be specified once for all in
8210 "defaults" sections. By default it is not set, so half-closed connections
8211 will use the other timeouts (timeout.server or timeout.tunnel).
8212
8213 See also : "timeout client-fin", "timeout server", and "timeout tunnel".
8214
Willy Tarreau844e3c52008-01-11 16:28:18 +01008215
8216timeout tarpit <timeout>
Cyril Bonté78caf842010-03-10 22:41:43 +01008217 Set the duration for which tarpitted connections will be maintained
Willy Tarreau844e3c52008-01-11 16:28:18 +01008218 May be used in sections : defaults | frontend | listen | backend
8219 yes | yes | yes | yes
8220 Arguments :
8221 <timeout> is the tarpit duration specified in milliseconds by default, but
8222 can be in any other unit if the number is suffixed by the unit,
8223 as explained at the top of this document.
8224
8225 When a connection is tarpitted using "reqtarpit", it is maintained open with
8226 no activity for a certain amount of time, then closed. "timeout tarpit"
8227 defines how long it will be maintained open.
8228
8229 The value is specified in milliseconds by default, but can be in any other
8230 unit if the number is suffixed by the unit, as specified at the top of this
8231 document. If unspecified, the same value as the backend's connection timeout
8232 ("timeout connect") is used, for backwards compatibility with older versions
Cyril Bonté78caf842010-03-10 22:41:43 +01008233 with no "timeout tarpit" parameter.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008234
8235 See also : "timeout connect", "contimeout".
8236
8237
Willy Tarreauce887fd2012-05-12 12:50:00 +02008238timeout tunnel <timeout>
8239 Set the maximum inactivity time on the client and server side for tunnels.
8240 May be used in sections : defaults | frontend | listen | backend
8241 yes | no | yes | yes
8242 Arguments :
8243 <timeout> is the timeout value specified in milliseconds by default, but
8244 can be in any other unit if the number is suffixed by the unit,
8245 as explained at the top of this document.
8246
Jamie Gloudonaaa21002012-08-25 00:18:33 -04008247 The tunnel timeout applies when a bidirectional connection is established
Willy Tarreauce887fd2012-05-12 12:50:00 +02008248 between a client and a server, and the connection remains inactive in both
8249 directions. This timeout supersedes both the client and server timeouts once
8250 the connection becomes a tunnel. In TCP, this timeout is used as soon as no
8251 analyser remains attached to either connection (eg: tcp content rules are
8252 accepted). In HTTP, this timeout is used when a connection is upgraded (eg:
8253 when switching to the WebSocket protocol, or forwarding a CONNECT request
8254 to a proxy), or after the first response when no keepalive/close option is
8255 specified.
8256
Willy Tarreau05cdd962014-05-10 14:30:07 +02008257 Since this timeout is usually used in conjunction with long-lived connections,
8258 it usually is a good idea to also set "timeout client-fin" to handle the
8259 situation where a client suddenly disappears from the net and does not
8260 acknowledge a close, or sends a shutdown and does not acknowledge pending
8261 data anymore. This can happen in lossy networks where firewalls are present,
8262 and is detected by the presence of large amounts of sessions in a FIN_WAIT
8263 state.
8264
Willy Tarreauce887fd2012-05-12 12:50:00 +02008265 The value is specified in milliseconds by default, but can be in any other
8266 unit if the number is suffixed by the unit, as specified at the top of this
8267 document. Whatever the expected normal idle time, it is a good practice to
8268 cover at least one or several TCP packet losses by specifying timeouts that
8269 are slightly above multiples of 3 seconds (eg: 4 or 5 seconds minimum).
8270
8271 This parameter is specific to backends, but can be specified once for all in
8272 "defaults" sections. This is in fact one of the easiest solutions not to
8273 forget about it.
8274
8275 Example :
8276 defaults http
8277 option http-server-close
8278 timeout connect 5s
8279 timeout client 30s
Willy Tarreau05cdd962014-05-10 14:30:07 +02008280 timeout client-fin 30s
Willy Tarreauce887fd2012-05-12 12:50:00 +02008281 timeout server 30s
8282 timeout tunnel 1h # timeout to use with WebSocket and CONNECT
8283
Willy Tarreau05cdd962014-05-10 14:30:07 +02008284 See also : "timeout client", "timeout client-fin", "timeout server".
Willy Tarreauce887fd2012-05-12 12:50:00 +02008285
8286
Willy Tarreau844e3c52008-01-11 16:28:18 +01008287transparent (deprecated)
8288 Enable client-side transparent proxying
8289 May be used in sections : defaults | frontend | listen | backend
Willy Tarreau4b1f8592008-12-23 23:13:55 +01008290 yes | no | yes | yes
Willy Tarreau844e3c52008-01-11 16:28:18 +01008291 Arguments : none
8292
8293 This keyword was introduced in order to provide layer 7 persistence to layer
8294 3 load balancers. The idea is to use the OS's ability to redirect an incoming
8295 connection for a remote address to a local process (here HAProxy), and let
8296 this process know what address was initially requested. When this option is
8297 used, sessions without cookies will be forwarded to the original destination
8298 IP address of the incoming request (which should match that of another
8299 equipment), while requests with cookies will still be forwarded to the
8300 appropriate server.
8301
8302 The "transparent" keyword is deprecated, use "option transparent" instead.
8303
8304 Note that contrary to a common belief, this option does NOT make HAProxy
8305 present the client's IP to the server when establishing the connection.
8306
Willy Tarreau844e3c52008-01-11 16:28:18 +01008307 See also: "option transparent"
8308
William Lallemanda73203e2012-03-12 12:48:57 +01008309unique-id-format <string>
8310 Generate a unique ID for each request.
8311 May be used in sections : defaults | frontend | listen | backend
8312 yes | yes | yes | no
8313 Arguments :
8314 <string> is a log-format string.
8315
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008316 This keyword creates a ID for each request using the custom log format. A
8317 unique ID is useful to trace a request passing through many components of
8318 a complex infrastructure. The newly created ID may also be logged using the
8319 %ID tag the log-format string.
William Lallemanda73203e2012-03-12 12:48:57 +01008320
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008321 The format should be composed from elements that are guaranteed to be
8322 unique when combined together. For instance, if multiple haproxy instances
8323 are involved, it might be important to include the node name. It is often
8324 needed to log the incoming connection's source and destination addresses
8325 and ports. Note that since multiple requests may be performed over the same
8326 connection, including a request counter may help differentiate them.
8327 Similarly, a timestamp may protect against a rollover of the counter.
8328 Logging the process ID will avoid collisions after a service restart.
William Lallemanda73203e2012-03-12 12:48:57 +01008329
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008330 It is recommended to use hexadecimal notation for many fields since it
8331 makes them more compact and saves space in logs.
William Lallemanda73203e2012-03-12 12:48:57 +01008332
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008333 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01008334
Julien Vehentf21be322014-03-07 08:27:34 -05008335 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01008336
8337 will generate:
8338
8339 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
8340
8341 See also: "unique-id-header"
8342
8343unique-id-header <name>
8344 Add a unique ID header in the HTTP request.
8345 May be used in sections : defaults | frontend | listen | backend
8346 yes | yes | yes | no
8347 Arguments :
8348 <name> is the name of the header.
8349
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008350 Add a unique-id header in the HTTP request sent to the server, using the
8351 unique-id-format. It can't work if the unique-id-format doesn't exist.
William Lallemanda73203e2012-03-12 12:48:57 +01008352
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008353 Example:
William Lallemanda73203e2012-03-12 12:48:57 +01008354
Julien Vehentf21be322014-03-07 08:27:34 -05008355 unique-id-format %{+X}o\ %ci:%cp_%fi:%fp_%Ts_%rt:%pid
William Lallemanda73203e2012-03-12 12:48:57 +01008356 unique-id-header X-Unique-ID
8357
8358 will generate:
8359
8360 X-Unique-ID: 7F000001:8296_7F00001E:1F90_4F7B0A69_0003:790A
8361
8362 See also: "unique-id-format"
Willy Tarreau844e3c52008-01-11 16:28:18 +01008363
Willy Tarreauf51658d2014-04-23 01:21:56 +02008364use_backend <backend> [{if | unless} <condition>]
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02008365 Switch to a specific backend if/unless an ACL-based condition is matched.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008366 May be used in sections : defaults | frontend | listen | backend
8367 no | yes | yes | no
8368 Arguments :
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008369 <backend> is the name of a valid backend or "listen" section, or a
8370 "log-format" string resolving to a backend name.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008371
Willy Tarreauf51658d2014-04-23 01:21:56 +02008372 <condition> is a condition composed of ACLs, as described in section 7. If
8373 it is omitted, the rule is unconditionally applied.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008374
8375 When doing content-switching, connections arrive on a frontend and are then
8376 dispatched to various backends depending on a number of conditions. The
8377 relation between the conditions and the backends is described with the
Willy Tarreau1d0dfb12009-07-07 15:10:31 +02008378 "use_backend" keyword. While it is normally used with HTTP processing, it can
8379 also be used in pure TCP, either without content using stateless ACLs (eg:
8380 source address validation) or combined with a "tcp-request" rule to wait for
8381 some payload.
Willy Tarreau844e3c52008-01-11 16:28:18 +01008382
8383 There may be as many "use_backend" rules as desired. All of these rules are
8384 evaluated in their declaration order, and the first one which matches will
8385 assign the backend.
8386
8387 In the first form, the backend will be used if the condition is met. In the
8388 second form, the backend will be used if the condition is not met. If no
8389 condition is valid, the backend defined with "default_backend" will be used.
8390 If no default backend is defined, either the servers in the same section are
8391 used (in case of a "listen" section) or, in case of a frontend, no server is
8392 used and a 503 service unavailable response is returned.
8393
Willy Tarreau51aecc72009-07-12 09:47:04 +02008394 Note that it is possible to switch from a TCP frontend to an HTTP backend. In
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01008395 this case, either the frontend has already checked that the protocol is HTTP,
Willy Tarreau51aecc72009-07-12 09:47:04 +02008396 and backend processing will immediately follow, or the backend will wait for
8397 a complete HTTP request to get in. This feature is useful when a frontend
8398 must decode several protocols on a unique port, one of them being HTTP.
8399
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008400 When <backend> is a simple name, it is resolved at configuration time, and an
8401 error is reported if the specified backend does not exist. If <backend> is
8402 a log-format string instead, no check may be done at configuration time, so
8403 the backend name is resolved dynamically at run time. If the resulting
8404 backend name does not correspond to any valid backend, no other rule is
8405 evaluated, and the default_backend directive is applied instead. Note that
8406 when using dynamic backend names, it is highly recommended to use a prefix
8407 that no other backend uses in order to ensure that an unauthorized backend
8408 cannot be forced from the request.
8409
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008410 It is worth mentioning that "use_backend" rules with an explicit name are
Bertrand Jacquin702d44f2013-11-19 11:43:06 +01008411 used to detect the association between frontends and backends to compute the
8412 backend's "fullconn" setting. This cannot be done for dynamic names.
8413
8414 See also: "default_backend", "tcp-request", "fullconn", "log-format", and
8415 section 7 about ACLs.
Willy Tarreaud72758d2010-01-12 10:42:19 +01008416
Willy Tarreau036fae02008-01-06 13:24:40 +01008417
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008418use-server <server> if <condition>
8419use-server <server> unless <condition>
8420 Only use a specific server if/unless an ACL-based condition is matched.
8421 May be used in sections : defaults | frontend | listen | backend
8422 no | no | yes | yes
8423 Arguments :
Cyril Bonté108cf6e2012-04-21 23:30:29 +02008424 <server> is the name of a valid server in the same backend section.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008425
8426 <condition> is a condition composed of ACLs, as described in section 7.
8427
8428 By default, connections which arrive to a backend are load-balanced across
8429 the available servers according to the configured algorithm, unless a
8430 persistence mechanism such as a cookie is used and found in the request.
8431
8432 Sometimes it is desirable to forward a particular request to a specific
8433 server without having to declare a dedicated backend for this server. This
8434 can be achieved using the "use-server" rules. These rules are evaluated after
8435 the "redirect" rules and before evaluating cookies, and they have precedence
8436 on them. There may be as many "use-server" rules as desired. All of these
8437 rules are evaluated in their declaration order, and the first one which
8438 matches will assign the server.
8439
8440 If a rule designates a server which is down, and "option persist" is not used
8441 and no force-persist rule was validated, it is ignored and evaluation goes on
8442 with the next rules until one matches.
8443
8444 In the first form, the server will be used if the condition is met. In the
8445 second form, the server will be used if the condition is not met. If no
8446 condition is valid, the processing continues and the server will be assigned
8447 according to other persistence mechanisms.
8448
8449 Note that even if a rule is matched, cookie processing is still performed but
8450 does not assign the server. This allows prefixed cookies to have their prefix
8451 stripped.
8452
8453 The "use-server" statement works both in HTTP and TCP mode. This makes it
8454 suitable for use with content-based inspection. For instance, a server could
8455 be selected in a farm according to the TLS SNI field. And if these servers
8456 have their weight set to zero, they will not be used for other traffic.
8457
8458 Example :
8459 # intercept incoming TLS requests based on the SNI field
8460 use-server www if { req_ssl_sni -i www.example.com }
8461 server www 192.168.0.1:443 weight 0
8462 use-server mail if { req_ssl_sni -i mail.example.com }
8463 server mail 192.168.0.1:587 weight 0
8464 use-server imap if { req_ssl_sni -i imap.example.com }
8465 server mail 192.168.0.1:993 weight 0
8466 # all the rest is forwarded to this server
8467 server default 192.168.0.2:443 check
8468
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008469 See also: "use_backend", section 5 about server and section 7 about ACLs.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008470
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008471
84725. Bind and Server options
8473--------------------------
8474
8475The "bind", "server" and "default-server" keywords support a number of settings
8476depending on some build options and on the system HAProxy was built on. These
8477settings generally each consist in one word sometimes followed by a value,
8478written on the same line as the "bind" or "server" line. All these options are
8479described in this section.
8480
8481
84825.1. Bind options
8483-----------------
8484
8485The "bind" keyword supports a certain number of settings which are all passed
8486as arguments on the same line. The order in which those arguments appear makes
8487no importance, provided that they appear after the bind address. All of these
8488parameters are optional. Some of them consist in a single words (booleans),
8489while other ones expect a value after them. In this case, the value must be
8490provided immediately after the setting name.
8491
8492The currently supported settings are the following ones.
8493
8494accept-proxy
8495 Enforces the use of the PROXY protocol over any connection accepted by any of
Willy Tarreau77992672014-06-14 11:06:17 +02008496 the sockets declared on the same line. Versions 1 and 2 of the PROXY protocol
8497 are supported and correctly detected. The PROXY protocol dictates the layer
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008498 3/4 addresses of the incoming connection to be used everywhere an address is
8499 used, with the only exception of "tcp-request connection" rules which will
8500 only see the real connection address. Logs will reflect the addresses
8501 indicated in the protocol, unless it is violated, in which case the real
8502 address will still be used. This keyword combined with support from external
8503 components can be used as an efficient and reliable alternative to the
8504 X-Forwarded-For mechanism which is not always reliable and not even always
Willy Tarreau4f0d9192013-06-11 20:40:55 +02008505 usable. See also "tcp-request connection expect-proxy" for a finer-grained
8506 setting of which client is allowed to use the protocol.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008507
Willy Tarreauab861d32013-04-02 02:30:41 +02008508alpn <protocols>
8509 This enables the TLS ALPN extension and advertises the specified protocol
8510 list as supported on top of ALPN. The protocol list consists in a comma-
8511 delimited list of protocol names, for instance: "http/1.1,http/1.0" (without
8512 quotes). This requires that the SSL library is build with support for TLS
8513 extensions enabled (check with haproxy -vv). The ALPN extension replaces the
8514 initial NPN extension.
8515
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008516backlog <backlog>
8517 Sets the socket's backlog to this value. If unspecified, the frontend's
8518 backlog is used instead, which generally defaults to the maxconn value.
8519
Emeric Brun7fb34422012-09-28 15:26:15 +02008520ecdhe <named curve>
8521 This setting is only available when support for OpenSSL was built in. It sets
Emeric Brun6924ef82013-03-06 14:08:53 +01008522 the named curve (RFC 4492) used to generate ECDH ephemeral keys. By default,
8523 used named curve is prime256v1.
Emeric Brun7fb34422012-09-28 15:26:15 +02008524
Emeric Brunfd33a262012-10-11 16:28:27 +02008525ca-file <cafile>
Emeric Brun1a073b42012-09-28 17:07:34 +02008526 This setting is only available when support for OpenSSL was built in. It
8527 designates a PEM file from which to load CA certificates used to verify
8528 client's certificate.
8529
Emeric Brunb6dc9342012-09-28 17:55:37 +02008530ca-ignore-err [all|<errorID>,...]
8531 This setting is only available when support for OpenSSL was built in.
8532 Sets a comma separated list of errorIDs to ignore during verify at depth > 0.
8533 If set to 'all', all errors are ignored. SSL handshake is not aborted if an
8534 error is ignored.
8535
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008536ciphers <ciphers>
8537 This setting is only available when support for OpenSSL was built in. It sets
8538 the string describing the list of cipher algorithms ("cipher suite") that are
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008539 negotiated during the SSL/TLS handshake. The format of the string is defined
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008540 in "man 1 ciphers" from OpenSSL man pages, and can be for instance a string
8541 such as "AES:ALL:!aNULL:!eNULL:+RC4:@STRENGTH" (without quotes).
8542
Emeric Brunfd33a262012-10-11 16:28:27 +02008543crl-file <crlfile>
Emeric Brun1a073b42012-09-28 17:07:34 +02008544 This setting is only available when support for OpenSSL was built in. It
8545 designates a PEM file from which to load certificate revocation list used
8546 to verify client's certificate.
8547
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008548crt <cert>
Alex Davies0fbf0162013-03-02 16:04:50 +00008549 This setting is only available when support for OpenSSL was built in. It
8550 designates a PEM file containing both the required certificates and any
8551 associated private keys. This file can be built by concatenating multiple
8552 PEM files into one (e.g. cat cert.pem key.pem > combined.pem). If your CA
8553 requires an intermediate certificate, this can also be concatenated into this
8554 file.
8555
8556 If the OpenSSL used supports Diffie-Hellman, parameters present in this file
8557 are loaded.
8558
8559 If a directory name is used instead of a PEM file, then all files found in
Cyril Bonté3180f7b2015-01-25 00:16:08 +01008560 that directory will be loaded in alphabetic order unless their name ends with
8561 '.issuer' or '.ocsp' (reserved extensions). This directive may be specified
8562 multiple times in order to load certificates from multiple files or
8563 directories. The certificates will be presented to clients who provide a valid
8564 TLS Server Name Indication field matching one of their CN or alt subjects.
8565 Wildcards are supported, where a wildcard character '*' is used instead of the
8566 first hostname component (eg: *.example.org matches www.example.org but not
Alex Davies0fbf0162013-03-02 16:04:50 +00008567 www.sub.example.org).
8568
8569 If no SNI is provided by the client or if the SSL library does not support
8570 TLS extensions, or if the client provides an SNI hostname which does not
8571 match any certificate, then the first loaded certificate will be presented.
8572 This means that when loading certificates from a directory, it is highly
Cyril Bonté3180f7b2015-01-25 00:16:08 +01008573 recommended to load the default one first as a file or to ensure that it will
8574 always be the first one in the directory.
Alex Davies0fbf0162013-03-02 16:04:50 +00008575
Emeric Brune032bfa2012-09-28 13:01:45 +02008576 Note that the same cert may be loaded multiple times without side effects.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008577
Alex Davies0fbf0162013-03-02 16:04:50 +00008578 Some CAs (such as Godaddy) offer a drop down list of server types that do not
8579 include HAProxy when obtaining a certificate. If this happens be sure to
Godbach8bf60a12014-04-21 21:42:41 +08008580 choose a webserver that the CA believes requires an intermediate CA (for
Alex Davies0fbf0162013-03-02 16:04:50 +00008581 Godaddy, selection Apache Tomcat will get the correct bundle, but many
8582 others, e.g. nginx, result in a wrong bundle that will not work for some
8583 clients).
8584
Emeric Brun4147b2e2014-06-16 18:36:30 +02008585 For each PEM file, haproxy checks for the presence of file at the same path
8586 suffixed by ".ocsp". If such file is found, support for the TLS Certificate
8587 Status Request extension (also known as "OCSP stapling") is automatically
8588 enabled. The content of this file is optional. If not empty, it must contain
8589 a valid OCSP Response in DER format. In order to be valid an OCSP Response
8590 must comply with the following rules: it has to indicate a good status,
8591 it has to be a single response for the certificate of the PEM file, and it
8592 has to be valid at the moment of addition. If these rules are not respected
8593 the OCSP Response is ignored and a warning is emitted. In order to identify
8594 which certificate an OCSP Response applies to, the issuer's certificate is
8595 necessary. If the issuer's certificate is not found in the PEM file, it will
8596 be loaded from a file at the same path as the PEM file suffixed by ".issuer"
8597 if it exists otherwise it will fail with an error.
8598
Emeric Brunb6dc9342012-09-28 17:55:37 +02008599crt-ignore-err <errors>
Alex Davies0fbf0162013-03-02 16:04:50 +00008600 This setting is only available when support for OpenSSL was built in. Sets a
8601 comma separated list of errorIDs to ignore during verify at depth == 0. If
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008602 set to 'all', all errors are ignored. SSL handshake is not aborted if an error
Alex Davies0fbf0162013-03-02 16:04:50 +00008603 is ignored.
Emeric Brunb6dc9342012-09-28 17:55:37 +02008604
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01008605crt-list <file>
8606 This setting is only available when support for OpenSSL was built in. It
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02008607 designates a list of PEM file with an optional list of SNI filter per
8608 certificate, with the following format for each line :
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01008609
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02008610 <crtfile> [[!]<snifilter> ...]
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01008611
Emmanuel Hocdet7c41a1b2013-05-07 20:20:06 +02008612 Wildcards are supported in the SNI filter. Negative filter are also supported,
8613 only useful in combination with a wildcard filter to exclude a particular SNI.
8614 The certificates will be presented to clients who provide a valid TLS Server
8615 Name Indication field matching one of the SNI filters. If no SNI filter is
8616 specified, the CN and alt subjects are used. This directive may be specified
8617 multiple times. See the "crt" option for more information. The default
8618 certificate is still needed to meet OpenSSL expectations. If it is not used,
8619 the 'strict-sni' option may be used.
Emmanuel Hocdetfe616562013-01-22 15:31:15 +01008620
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008621defer-accept
8622 Is an optional keyword which is supported only on certain Linux kernels. It
8623 states that a connection will only be accepted once some data arrive on it,
8624 or at worst after the first retransmit. This should be used only on protocols
8625 for which the client talks first (eg: HTTP). It can slightly improve
8626 performance by ensuring that most of the request is already available when
8627 the connection is accepted. On the other hand, it will not be able to detect
8628 connections which don't talk. It is important to note that this option is
8629 broken in all kernels up to 2.6.31, as the connection is never accepted until
8630 the client talks. This can cause issues with front firewalls which would see
8631 an established connection while the proxy will only see it in SYN_RECV. This
8632 option is only supported on TCPv4/TCPv6 sockets and ignored by other ones.
8633
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008634force-sslv3
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008635 This option enforces use of SSLv3 only on SSL connections instantiated from
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008636 this listener. SSLv3 is generally less expensive than the TLS counterparts
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008637 for high connection rates. This option is also available on global statement
8638 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008639
8640force-tlsv10
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008641 This option enforces use of TLSv1.0 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008642 this listener. This option is also available on global statement
8643 "ssl-default-bind-options". See also "no-tlsv*" and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008644
8645force-tlsv11
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008646 This option enforces use of TLSv1.1 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008647 this listener. This option is also available on global statement
8648 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008649
8650force-tlsv12
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008651 This option enforces use of TLSv1.2 only on SSL connections instantiated from
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008652 this listener. This option is also available on global statement
8653 "ssl-default-bind-options". See also "no-tlsv*", and "no-sslv3".
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008654
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008655gid <gid>
8656 Sets the group of the UNIX sockets to the designated system gid. It can also
8657 be set by default in the global section's "unix-bind" statement. Note that
8658 some platforms simply ignore this. This setting is equivalent to the "group"
8659 setting except that the group ID is used instead of its name. This setting is
8660 ignored by non UNIX sockets.
8661
8662group <group>
8663 Sets the group of the UNIX sockets to the designated system group. It can
8664 also be set by default in the global section's "unix-bind" statement. Note
8665 that some platforms simply ignore this. This setting is equivalent to the
8666 "gid" setting except that the group name is used instead of its gid. This
8667 setting is ignored by non UNIX sockets.
8668
8669id <id>
8670 Fixes the socket ID. By default, socket IDs are automatically assigned, but
8671 sometimes it is more convenient to fix them to ease monitoring. This value
8672 must be strictly positive and unique within the listener/frontend. This
8673 option can only be used when defining only a single socket.
8674
8675interface <interface>
Lukas Tribusfce2e962013-02-12 22:13:19 +01008676 Restricts the socket to a specific interface. When specified, only packets
8677 received from that particular interface are processed by the socket. This is
8678 currently only supported on Linux. The interface must be a primary system
8679 interface, not an aliased interface. It is also possible to bind multiple
8680 frontends to the same address if they are bound to different interfaces. Note
8681 that binding to a network interface requires root privileges. This parameter
8682 is only compatible with TCPv4/TCPv6 sockets.
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008683
Willy Tarreauabb175f2012-09-24 12:43:26 +02008684level <level>
8685 This setting is used with the stats sockets only to restrict the nature of
8686 the commands that can be issued on the socket. It is ignored by other
8687 sockets. <level> can be one of :
8688 - "user" is the least privileged level ; only non-sensitive stats can be
8689 read, and no change is allowed. It would make sense on systems where it
8690 is not easy to restrict access to the socket.
8691 - "operator" is the default level and fits most common uses. All data can
8692 be read, and only non-sensitive changes are permitted (eg: clear max
8693 counters).
8694 - "admin" should be used with care, as everything is permitted (eg: clear
8695 all counters).
8696
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008697maxconn <maxconn>
8698 Limits the sockets to this number of concurrent connections. Extraneous
8699 connections will remain in the system's backlog until a connection is
8700 released. If unspecified, the limit will be the same as the frontend's
8701 maxconn. Note that in case of port ranges or multiple addresses, the same
8702 value will be applied to each socket. This setting enables different
8703 limitations on expensive sockets, for instance SSL entries which may easily
8704 eat all memory.
8705
8706mode <mode>
8707 Sets the octal mode used to define access permissions on the UNIX socket. It
8708 can also be set by default in the global section's "unix-bind" statement.
8709 Note that some platforms simply ignore this. This setting is ignored by non
8710 UNIX sockets.
8711
8712mss <maxseg>
8713 Sets the TCP Maximum Segment Size (MSS) value to be advertised on incoming
8714 connections. This can be used to force a lower MSS for certain specific
8715 ports, for instance for connections passing through a VPN. Note that this
8716 relies on a kernel feature which is theoretically supported under Linux but
8717 was buggy in all versions prior to 2.6.28. It may or may not work on other
8718 operating systems. It may also not change the advertised value but change the
8719 effective size of outgoing segments. The commonly advertised value for TCPv4
8720 over Ethernet networks is 1460 = 1500(MTU) - 40(IP+TCP). If this value is
8721 positive, it will be used as the advertised MSS. If it is negative, it will
8722 indicate by how much to reduce the incoming connection's advertised MSS for
8723 outgoing segments. This parameter is only compatible with TCP v4/v6 sockets.
8724
8725name <name>
8726 Sets an optional name for these sockets, which will be reported on the stats
8727 page.
8728
8729nice <nice>
8730 Sets the 'niceness' of connections initiated from the socket. Value must be
8731 in the range -1024..1024 inclusive, and defaults to zero. Positive values
8732 means that such connections are more friendly to others and easily offer
8733 their place in the scheduler. On the opposite, negative values mean that
8734 connections want to run with a higher priority than others. The difference
8735 only happens under high loads when the system is close to saturation.
8736 Negative values are appropriate for low-latency or administration services,
8737 and high values are generally recommended for CPU intensive tasks such as SSL
8738 processing or bulk transfers which are less sensible to latency. For example,
8739 it may make sense to use a positive value for an SMTP socket and a negative
8740 one for an RDP socket.
8741
Emeric Brun9b3009b2012-10-05 11:55:06 +02008742no-sslv3
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008743 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008744 disables support for SSLv3 on any sockets instantiated from the listener when
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008745 SSL is supported. Note that SSLv2 is forced disabled in the code and cannot
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008746 be enabled using any configuration option. This option is also available on
8747 global statement "ssl-default-bind-options". See also "force-tls*",
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008748 and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008749
Emeric Brun90ad8722012-10-02 14:00:59 +02008750no-tls-tickets
8751 This setting is only available when support for OpenSSL was built in. It
8752 disables the stateless session resumption (RFC 5077 TLS Ticket
8753 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008754 session resumption is more expensive in CPU usage. This option is also
8755 available on global statement "ssl-default-bind-options".
Emeric Brun90ad8722012-10-02 14:00:59 +02008756
Emeric Brun9b3009b2012-10-05 11:55:06 +02008757no-tlsv10
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008758 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008759 disables support for TLSv1.0 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008760 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008761 cannot be enabled using any configuration option. This option is also
8762 available on global statement "ssl-default-bind-options". See also
8763 "force-tlsv*", and "force-sslv3".
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008764
Emeric Brun9b3009b2012-10-05 11:55:06 +02008765no-tlsv11
Emeric Brunf5da4932012-09-28 19:42:54 +02008766 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008767 disables support for TLSv1.1 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008768 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008769 cannot be enabled using any configuration option. This option is also
8770 available on global statement "ssl-default-bind-options". See also
8771 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02008772
Emeric Brun9b3009b2012-10-05 11:55:06 +02008773no-tlsv12
Emeric Brunf5da4932012-09-28 19:42:54 +02008774 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008775 disables support for TLSv1.2 on any sockets instantiated from the listener
Emeric Brun2cb7ae52012-10-05 14:14:21 +02008776 when SSL is supported. Note that SSLv2 is forced disabled in the code and
Emeric Brun2c86cbf2014-10-30 15:56:50 +01008777 cannot be enabled using any configuration option. This option is also
8778 available on global statement "ssl-default-bind-options". See also
8779 "force-tlsv*", and "force-sslv3".
Emeric Brunf5da4932012-09-28 19:42:54 +02008780
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02008781npn <protocols>
8782 This enables the NPN TLS extension and advertises the specified protocol list
8783 as supported on top of NPN. The protocol list consists in a comma-delimited
8784 list of protocol names, for instance: "http/1.1,http/1.0" (without quotes).
8785 This requires that the SSL library is build with support for TLS extensions
Willy Tarreauab861d32013-04-02 02:30:41 +02008786 enabled (check with haproxy -vv). Note that the NPN extension has been
8787 replaced with the ALPN extension (see the "alpn" keyword).
Willy Tarreau6c9a3d52012-10-18 18:57:14 +02008788
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02008789process [ all | odd | even | <number 1-64>[-<number 1-64>] ]
8790 This restricts the list of processes on which this listener is allowed to
8791 run. It does not enforce any process but eliminates those which do not match.
8792 If the frontend uses a "bind-process" setting, the intersection between the
8793 two is applied. If in the end the listener is not allowed to run on any
8794 remaining process, a warning is emitted, and the listener will either run on
8795 the first process of the listener if a single process was specified, or on
8796 all of its processes if multiple processes were specified. For the unlikely
Willy Tarreauae302532014-05-07 19:22:24 +02008797 case where several ranges are needed, this directive may be repeated. The
8798 main purpose of this directive is to be used with the stats sockets and have
8799 one different socket per process. The second purpose is to have multiple bind
8800 lines sharing the same IP:port but not the same process in a listener, so
8801 that the system can distribute the incoming connections into multiple queues
8802 and allow a smoother inter-process load balancing. Currently Linux 3.9 and
8803 above is known for supporting this. See also "bind-process" and "nbproc".
Willy Tarreau6ae1ba62014-05-07 19:01:58 +02008804
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008805ssl
8806 This setting is only available when support for OpenSSL was built in. It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008807 enables SSL deciphering on connections instantiated from this listener. A
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008808 certificate is necessary (see "crt" above). All contents in the buffers will
8809 appear in clear text, so that ACLs and HTTP processing will only have access
8810 to deciphered contents.
8811
Emmanuel Hocdet65623372013-01-24 17:17:15 +01008812strict-sni
8813 This setting is only available when support for OpenSSL was built in. The
8814 SSL/TLS negotiation is allow only if the client provided an SNI which match
8815 a certificate. The default certificate is not used.
8816 See the "crt" option for more information.
8817
Willy Tarreau1c862c52012-10-05 16:21:00 +02008818tfo
Lukas Tribus0defb902013-02-13 23:35:39 +01008819 Is an optional keyword which is supported only on Linux kernels >= 3.7. It
Willy Tarreau1c862c52012-10-05 16:21:00 +02008820 enables TCP Fast Open on the listening socket, which means that clients which
8821 support this feature will be able to send a request and receive a response
8822 during the 3-way handshake starting from second connection, thus saving one
8823 round-trip after the first connection. This only makes sense with protocols
8824 that use high connection rates and where each round trip matters. This can
8825 possibly cause issues with many firewalls which do not accept data on SYN
8826 packets, so this option should only be enabled once well tested. This option
Lukas Tribus0999f762013-04-02 16:43:24 +02008827 is only supported on TCPv4/TCPv6 sockets and ignored by other ones. You may
8828 need to build HAProxy with USE_TFO=1 if your libc doesn't define
8829 TCP_FASTOPEN.
Willy Tarreau1c862c52012-10-05 16:21:00 +02008830
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008831transparent
8832 Is an optional keyword which is supported only on certain Linux kernels. It
8833 indicates that the addresses will be bound even if they do not belong to the
8834 local machine, and that packets targeting any of these addresses will be
8835 intercepted just as if the addresses were locally configured. This normally
8836 requires that IP forwarding is enabled. Caution! do not use this with the
8837 default address '*', as it would redirect any traffic for the specified port.
8838 This keyword is available only when HAProxy is built with USE_LINUX_TPROXY=1.
8839 This parameter is only compatible with TCPv4 and TCPv6 sockets, depending on
8840 kernel version. Some distribution kernels include backports of the feature,
8841 so check for support with your vendor.
8842
Willy Tarreau77e3af92012-11-24 15:07:23 +01008843v4v6
8844 Is an optional keyword which is supported only on most recent systems
8845 including Linux kernels >= 2.4.21. It is used to bind a socket to both IPv4
8846 and IPv6 when it uses the default address. Doing so is sometimes necessary
8847 on systems which bind to IPv6 only by default. It has no effect on non-IPv6
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03008848 sockets, and is overridden by the "v6only" option.
Willy Tarreau77e3af92012-11-24 15:07:23 +01008849
Willy Tarreau9b6700f2012-11-24 11:55:28 +01008850v6only
8851 Is an optional keyword which is supported only on most recent systems
8852 including Linux kernels >= 2.4.21. It is used to bind a socket to IPv6 only
8853 when it uses the default address. Doing so is sometimes preferred to doing it
Willy Tarreau77e3af92012-11-24 15:07:23 +01008854 system-wide as it is per-listener. It has no effect on non-IPv6 sockets and
8855 has precedence over the "v4v6" option.
Willy Tarreau9b6700f2012-11-24 11:55:28 +01008856
Willy Tarreaub6205fd2012-09-24 12:27:33 +02008857uid <uid>
8858 Sets the owner of the UNIX sockets to the designated system uid. It can also
8859 be set by default in the global section's "unix-bind" statement. Note that
8860 some platforms simply ignore this. This setting is equivalent to the "user"
8861 setting except that the user numeric ID is used instead of its name. This
8862 setting is ignored by non UNIX sockets.
8863
8864user <user>
8865 Sets the owner of the UNIX sockets to the designated system user. It can also
8866 be set by default in the global section's "unix-bind" statement. Note that
8867 some platforms simply ignore this. This setting is equivalent to the "uid"
8868 setting except that the user name is used instead of its uid. This setting is
8869 ignored by non UNIX sockets.
8870
Emeric Brun1a073b42012-09-28 17:07:34 +02008871verify [none|optional|required]
8872 This setting is only available when support for OpenSSL was built in. If set
8873 to 'none', client certificate is not requested. This is the default. In other
8874 cases, a client certificate is requested. If the client does not provide a
8875 certificate after the request and if 'verify' is set to 'required', then the
8876 handshake is aborted, while it would have succeeded if set to 'optional'. The
Emeric Brunfd33a262012-10-11 16:28:27 +02008877 certificate provided by the client is always verified using CAs from
8878 'ca-file' and optional CRLs from 'crl-file'. On verify failure the handshake
8879 is aborted, regardless of the 'verify' option, unless the error code exactly
8880 matches one of those listed with 'ca-ignore-err' or 'crt-ignore-err'.
Willy Tarreau4a5cade2012-04-05 21:09:48 +02008881
Willy Tarreaub6205fd2012-09-24 12:27:33 +020088825.2. Server and default-server options
Cyril Bontéf0c60612010-02-06 14:44:47 +01008883------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02008884
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01008885The "server" and "default-server" keywords support a certain number of settings
8886which are all passed as arguments on the server line. The order in which those
8887arguments appear does not count, and they are all optional. Some of those
8888settings are single words (booleans) while others expect one or several values
8889after them. In this case, the values must immediately follow the setting name.
8890Except default-server, all those settings must be specified after the server's
8891address if they are used:
Willy Tarreau6a06a402007-07-15 20:15:28 +02008892
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008893 server <name> <address>[:port] [settings ...]
Krzysztof Piotr Oledzkic6df0662010-01-05 16:38:49 +01008894 default-server [settings ...]
Willy Tarreau6a06a402007-07-15 20:15:28 +02008895
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008896The currently supported settings are the following ones.
Willy Tarreau0ba27502007-12-24 16:55:16 +01008897
Willy Tarreauceb4ac92012-04-28 00:41:46 +02008898addr <ipv4|ipv6>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008899 Using the "addr" parameter, it becomes possible to use a different IP address
8900 to send health-checks. On some servers, it may be desirable to dedicate an IP
8901 address to specific component able to perform complex tests which are more
8902 suitable to health-checks than the application. This parameter is ignored if
8903 the "check" parameter is not set. See also the "port" parameter.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008904
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008905 Supported in default-server: No
8906
Simon Hormand60d6912013-11-25 10:46:36 +09008907agent-check
8908 Enable an auxiliary agent check which is run independently of a regular
Willy Tarreau81f5d942013-12-09 20:51:51 +01008909 health check. An agent health check is performed by making a TCP connection
8910 to the port set by the "agent-port" parameter and reading an ASCII string.
8911 The string is made of a series of words delimited by spaces, tabs or commas
8912 in any order, optionally terminated by '\r' and/or '\n', each consisting of :
Simon Hormand60d6912013-11-25 10:46:36 +09008913
Willy Tarreau81f5d942013-12-09 20:51:51 +01008914 - An ASCII representation of a positive integer percentage, e.g. "75%".
Simon Hormand60d6912013-11-25 10:46:36 +09008915 Values in this format will set the weight proportional to the initial
Willy Tarreauc5af3a62014-10-07 15:27:33 +02008916 weight of a server as configured when haproxy starts. Note that a zero
8917 weight is reported on the stats page as "DRAIN" since it has the same
8918 effect on the server (it's removed from the LB farm).
Simon Hormand60d6912013-11-25 10:46:36 +09008919
Willy Tarreau81f5d942013-12-09 20:51:51 +01008920 - The word "ready". This will turn the server's administrative state to the
8921 READY mode, thus cancelling any DRAIN or MAINT state
Simon Hormand60d6912013-11-25 10:46:36 +09008922
Willy Tarreau81f5d942013-12-09 20:51:51 +01008923 - The word "drain". This will turn the server's administrative state to the
8924 DRAIN mode, thus it will not accept any new connections other than those
8925 that are accepted via persistence.
Simon Hormand60d6912013-11-25 10:46:36 +09008926
Willy Tarreau81f5d942013-12-09 20:51:51 +01008927 - The word "maint". This will turn the server's administrative state to the
8928 MAINT mode, thus it will not accept any new connections at all, and health
8929 checks will be stopped.
Simon Hormand60d6912013-11-25 10:46:36 +09008930
Willy Tarreau81f5d942013-12-09 20:51:51 +01008931 - The words "down", "failed", or "stopped", optionally followed by a
8932 description string after a sharp ('#'). All of these mark the server's
8933 operating state as DOWN, but since the word itself is reported on the stats
8934 page, the difference allows an administrator to know if the situation was
8935 expected or not : the service may intentionally be stopped, may appear up
8936 but fail some validity tests, or may be seen as down (eg: missing process,
8937 or port not responding).
Simon Hormand60d6912013-11-25 10:46:36 +09008938
Willy Tarreau81f5d942013-12-09 20:51:51 +01008939 - The word "up" sets back the server's operating state as UP if health checks
8940 also report that the service is accessible.
Simon Hormand60d6912013-11-25 10:46:36 +09008941
Willy Tarreau81f5d942013-12-09 20:51:51 +01008942 Parameters which are not advertised by the agent are not changed. For
8943 example, an agent might be designed to monitor CPU usage and only report a
8944 relative weight and never interact with the operating status. Similarly, an
8945 agent could be designed as an end-user interface with 3 radio buttons
8946 allowing an administrator to change only the administrative state. However,
8947 it is important to consider that only the agent may revert its own actions,
8948 so if a server is set to DRAIN mode or to DOWN state using the agent, the
8949 agent must implement the other equivalent actions to bring the service into
8950 operations again.
Simon Hormand60d6912013-11-25 10:46:36 +09008951
Simon Horman2f1f9552013-11-25 10:46:37 +09008952 Failure to connect to the agent is not considered an error as connectivity
8953 is tested by the regular health check which is enabled by the "check"
Willy Tarreau81f5d942013-12-09 20:51:51 +01008954 parameter. Warning though, it is not a good idea to stop an agent after it
8955 reports "down", since only an agent reporting "up" will be able to turn the
8956 server up again. Note that the CLI on the Unix stats socket is also able to
8957 force an agent's result in order to workaround a bogus agent if needed.
Simon Horman2f1f9552013-11-25 10:46:37 +09008958
Willy Tarreau81f5d942013-12-09 20:51:51 +01008959 Requires the "agent-port" parameter to be set. See also the "agent-inter"
8960 parameter.
Simon Hormand60d6912013-11-25 10:46:36 +09008961
8962 Supported in default-server: No
8963
8964agent-inter <delay>
8965 The "agent-inter" parameter sets the interval between two agent checks
8966 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
8967
8968 Just as with every other time-based parameter, it may be entered in any
8969 other explicit unit among { us, ms, s, m, h, d }. The "agent-inter"
8970 parameter also serves as a timeout for agent checks "timeout check" is
8971 not set. In order to reduce "resonance" effects when multiple servers are
8972 hosted on the same hardware, the agent and health checks of all servers
8973 are started with a small time offset between them. It is also possible to
8974 add some random noise in the agent and health checks interval using the
8975 global "spread-checks" keyword. This makes sense for instance when a lot
8976 of backends use the same servers.
8977
8978 See also the "agent-check" and "agent-port" parameters.
8979
8980 Supported in default-server: Yes
8981
8982agent-port <port>
8983 The "agent-port" parameter sets the TCP port used for agent checks.
8984
8985 See also the "agent-check" and "agent-inter" parameters.
8986
8987 Supported in default-server: Yes
8988
Willy Tarreauc57f0e22009-05-10 13:12:33 +02008989backup
8990 When "backup" is present on a server line, the server is only used in load
8991 balancing when all other non-backup servers are unavailable. Requests coming
8992 with a persistence cookie referencing the server will always be served
8993 though. By default, only the first operational backup server is used, unless
8994 the "allbackups" option is set in the backend. See also the "allbackups"
8995 option.
Willy Tarreau6a06a402007-07-15 20:15:28 +02008996
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01008997 Supported in default-server: No
8998
Emeric Brunef42d922012-10-11 16:11:36 +02008999ca-file <cafile>
9000 This setting is only available when support for OpenSSL was built in. It
9001 designates a PEM file from which to load CA certificates used to verify
9002 server's certificate.
9003
9004 Supported in default-server: No
9005
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009006check
9007 This option enables health checks on the server. By default, a server is
Patrick Mézardb7aeec62012-01-22 16:01:22 +01009008 always considered available. If "check" is set, the server is available when
9009 accepting periodic TCP connections, to ensure that it is really able to serve
9010 requests. The default address and port to send the tests to are those of the
9011 server, and the default source is the same as the one defined in the
9012 backend. It is possible to change the address using the "addr" parameter, the
9013 port using the "port" parameter, the source address using the "source"
9014 address, and the interval and timers using the "inter", "rise" and "fall"
Simon Hormanafc47ee2013-11-25 10:46:35 +09009015 parameters. The request method is define in the backend using the "httpchk",
9016 "smtpchk", "mysql-check", "pgsql-check" and "ssl-hello-chk" options. Please
9017 refer to those options and parameters for more information.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009018
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009019 Supported in default-server: No
9020
Willy Tarreau6c16adc2012-10-05 00:04:16 +02009021check-send-proxy
9022 This option forces emission of a PROXY protocol line with outgoing health
9023 checks, regardless of whether the server uses send-proxy or not for the
9024 normal traffic. By default, the PROXY protocol is enabled for health checks
9025 if it is already enabled for normal traffic and if no "port" nor "addr"
9026 directive is present. However, if such a directive is present, the
9027 "check-send-proxy" option needs to be used to force the use of the
9028 protocol. See also the "send-proxy" option for more information.
9029
9030 Supported in default-server: No
9031
Willy Tarreau763a95b2012-10-04 23:15:39 +02009032check-ssl
9033 This option forces encryption of all health checks over SSL, regardless of
9034 whether the server uses SSL or not for the normal traffic. This is generally
9035 used when an explicit "port" or "addr" directive is specified and SSL health
9036 checks are not inherited. It is important to understand that this option
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009037 inserts an SSL transport layer below the checks, so that a simple TCP connect
Willy Tarreau763a95b2012-10-04 23:15:39 +02009038 check becomes an SSL connect, which replaces the old ssl-hello-chk. The most
9039 common use is to send HTTPS checks by combining "httpchk" with SSL checks.
9040 All SSL settings are common to health checks and traffic (eg: ciphers).
9041 See the "ssl" option for more information.
9042
9043 Supported in default-server: No
9044
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009045ciphers <ciphers>
9046 This option sets the string describing the list of cipher algorithms that is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009047 is negotiated during the SSL/TLS handshake with the server. The format of the
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009048 string is defined in "man 1 ciphers". When SSL is used to communicate with
9049 servers on the local network, it is common to see a weaker set of algorithms
9050 than what is used over the internet. Doing so reduces CPU usage on both the
9051 server and haproxy while still keeping it compatible with deployed software.
9052 Some algorithms such as RC4-SHA1 are reasonably cheap. If no security at all
9053 is needed and just connectivity, using DES can be appropriate.
9054
Willy Tarreau763a95b2012-10-04 23:15:39 +02009055 Supported in default-server: No
9056
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009057cookie <value>
9058 The "cookie" parameter sets the cookie value assigned to the server to
9059 <value>. This value will be checked in incoming requests, and the first
9060 operational server possessing the same value will be selected. In return, in
9061 cookie insertion or rewrite modes, this value will be assigned to the cookie
9062 sent to the client. There is nothing wrong in having several servers sharing
9063 the same cookie value, and it is in fact somewhat common between normal and
9064 backup servers. See also the "cookie" keyword in backend section.
9065
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009066 Supported in default-server: No
9067
Emeric Brunef42d922012-10-11 16:11:36 +02009068crl-file <crlfile>
9069 This setting is only available when support for OpenSSL was built in. It
9070 designates a PEM file from which to load certificate revocation list used
9071 to verify server's certificate.
9072
9073 Supported in default-server: No
9074
Emeric Bruna7aa3092012-10-26 12:58:00 +02009075crt <cert>
9076 This setting is only available when support for OpenSSL was built in.
9077 It designates a PEM file from which to load both a certificate and the
9078 associated private key. This file can be built by concatenating both PEM
9079 files into one. This certificate will be sent if the server send a client
9080 certificate request.
9081
9082 Supported in default-server: No
9083
Willy Tarreau96839092010-03-29 10:02:24 +02009084disabled
9085 The "disabled" keyword starts the server in the "disabled" state. That means
9086 that it is marked down in maintenance mode, and no connection other than the
9087 ones allowed by persist mode will reach it. It is very well suited to setup
9088 new servers, because normal traffic will never reach them, while it is still
9089 possible to test the service by making use of the force-persist mechanism.
9090
9091 Supported in default-server: No
9092
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009093error-limit <count>
Willy Tarreau983e01e2010-01-11 18:42:06 +01009094 If health observing is enabled, the "error-limit" parameter specifies the
9095 number of consecutive errors that triggers event selected by the "on-error"
9096 option. By default it is set to 10 consecutive errors.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009097
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009098 Supported in default-server: Yes
9099
9100 See also the "check", "error-limit" and "on-error".
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009101
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009102fall <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009103 The "fall" parameter states that a server will be considered as dead after
9104 <count> consecutive unsuccessful health checks. This value defaults to 3 if
9105 unspecified. See also the "check", "inter" and "rise" parameters.
9106
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009107 Supported in default-server: Yes
9108
Emeric Brun8694b9a2012-10-05 14:39:07 +02009109force-sslv3
9110 This option enforces use of SSLv3 only when SSL is used to communicate with
9111 the server. SSLv3 is generally less expensive than the TLS counterparts for
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009112 high connection rates. This option is also available on global statement
9113 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009114
9115 Supported in default-server: No
9116
9117force-tlsv10
9118 This option enforces use of TLSv1.0 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009119 the server. This option is also available on global statement
9120 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009121
9122 Supported in default-server: No
9123
9124force-tlsv11
9125 This option enforces use of TLSv1.1 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009126 the server. This option is also available on global statement
9127 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009128
9129 Supported in default-server: No
9130
9131force-tlsv12
9132 This option enforces use of TLSv1.2 only when SSL is used to communicate with
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009133 the server. This option is also available on global statement
9134 "ssl-default-server-options". See also "no-tlsv*", "no-sslv3".
Emeric Brun8694b9a2012-10-05 14:39:07 +02009135
9136 Supported in default-server: No
9137
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009138id <value>
Willy Tarreau53fb4ae2009-10-04 23:04:08 +02009139 Set a persistent ID for the server. This ID must be positive and unique for
9140 the proxy. An unused ID will automatically be assigned if unset. The first
9141 assigned value will be 1. This ID is currently only returned in statistics.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009142
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009143 Supported in default-server: No
9144
9145inter <delay>
9146fastinter <delay>
9147downinter <delay>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009148 The "inter" parameter sets the interval between two consecutive health checks
9149 to <delay> milliseconds. If left unspecified, the delay defaults to 2000 ms.
9150 It is also possible to use "fastinter" and "downinter" to optimize delays
9151 between checks depending on the server state :
9152
9153 Server state | Interval used
9154 ---------------------------------+-----------------------------------------
9155 UP 100% (non-transitional) | "inter"
9156 ---------------------------------+-----------------------------------------
9157 Transitionally UP (going down), |
9158 Transitionally DOWN (going up), | "fastinter" if set, "inter" otherwise.
9159 or yet unchecked. |
9160 ---------------------------------+-----------------------------------------
9161 DOWN 100% (non-transitional) | "downinter" if set, "inter" otherwise.
9162 ---------------------------------+-----------------------------------------
Willy Tarreaud72758d2010-01-12 10:42:19 +01009163
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009164 Just as with every other time-based parameter, they can be entered in any
9165 other explicit unit among { us, ms, s, m, h, d }. The "inter" parameter also
9166 serves as a timeout for health checks sent to servers if "timeout check" is
9167 not set. In order to reduce "resonance" effects when multiple servers are
Simon Hormand60d6912013-11-25 10:46:36 +09009168 hosted on the same hardware, the agent and health checks of all servers
9169 are started with a small time offset between them. It is also possible to
9170 add some random noise in the agent and health checks interval using the
9171 global "spread-checks" keyword. This makes sense for instance when a lot
9172 of backends use the same servers.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009173
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009174 Supported in default-server: Yes
9175
9176maxconn <maxconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009177 The "maxconn" parameter specifies the maximal number of concurrent
9178 connections that will be sent to this server. If the number of incoming
9179 concurrent requests goes higher than this value, they will be queued, waiting
9180 for a connection to be released. This parameter is very important as it can
9181 save fragile servers from going down under extreme loads. If a "minconn"
9182 parameter is specified, the limit becomes dynamic. The default value is "0"
9183 which means unlimited. See also the "minconn" and "maxqueue" parameters, and
9184 the backend's "fullconn" keyword.
9185
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009186 Supported in default-server: Yes
9187
9188maxqueue <maxqueue>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009189 The "maxqueue" parameter specifies the maximal number of connections which
9190 will wait in the queue for this server. If this limit is reached, next
9191 requests will be redispatched to other servers instead of indefinitely
9192 waiting to be served. This will break persistence but may allow people to
9193 quickly re-log in when the server they try to connect to is dying. The
9194 default value is "0" which means the queue is unlimited. See also the
9195 "maxconn" and "minconn" parameters.
9196
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009197 Supported in default-server: Yes
9198
9199minconn <minconn>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009200 When the "minconn" parameter is set, the maxconn limit becomes a dynamic
9201 limit following the backend's load. The server will always accept at least
9202 <minconn> connections, never more than <maxconn>, and the limit will be on
9203 the ramp between both values when the backend has less than <fullconn>
9204 concurrent connections. This makes it possible to limit the load on the
9205 server during normal loads, but push it further for important loads without
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009206 overloading the server during exceptional loads. See also the "maxconn"
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009207 and "maxqueue" parameters, as well as the "fullconn" backend keyword.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009208
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009209 Supported in default-server: Yes
9210
Emeric Brun9b3009b2012-10-05 11:55:06 +02009211no-sslv3
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009212 This option disables support for SSLv3 when SSL is used to communicate with
9213 the server. Note that SSLv2 is disabled in the code and cannot be enabled
Emeric Brun8694b9a2012-10-05 14:39:07 +02009214 using any configuration option. See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009215
Willy Tarreau763a95b2012-10-04 23:15:39 +02009216 Supported in default-server: No
9217
Emeric Brunf9c5c472012-10-11 15:28:34 +02009218no-tls-tickets
9219 This setting is only available when support for OpenSSL was built in. It
9220 disables the stateless session resumption (RFC 5077 TLS Ticket
9221 extension) and force to use stateful session resumption. Stateless
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009222 session resumption is more expensive in CPU usage for servers. This option
9223 is also available on global statement "ssl-default-server-options".
Emeric Brunf9c5c472012-10-11 15:28:34 +02009224
9225 Supported in default-server: No
9226
Emeric Brun9b3009b2012-10-05 11:55:06 +02009227no-tlsv10
Emeric Brun8694b9a2012-10-05 14:39:07 +02009228 This option disables support for TLSv1.0 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02009229 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9230 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009231 often makes sense to disable it when communicating with local servers. This
9232 option is also available on global statement "ssl-default-server-options".
9233 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02009234
Willy Tarreau763a95b2012-10-04 23:15:39 +02009235 Supported in default-server: No
9236
Emeric Brun9b3009b2012-10-05 11:55:06 +02009237no-tlsv11
Emeric Brun8694b9a2012-10-05 14:39:07 +02009238 This option disables support for TLSv1.1 when SSL is used to communicate with
Emeric Brunf5da4932012-09-28 19:42:54 +02009239 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9240 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009241 often makes sense to disable it when communicating with local servers. This
9242 option is also available on global statement "ssl-default-server-options".
9243 See also "force-sslv3", "force-tlsv*".
Emeric Brunf5da4932012-09-28 19:42:54 +02009244
Willy Tarreau763a95b2012-10-04 23:15:39 +02009245 Supported in default-server: No
9246
Emeric Brun9b3009b2012-10-05 11:55:06 +02009247no-tlsv12
Emeric Brun8694b9a2012-10-05 14:39:07 +02009248 This option disables support for TLSv1.2 when SSL is used to communicate with
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009249 the server. Note that SSLv2 is disabled in the code and cannot be enabled
9250 using any configuration option. TLSv1 is more expensive than SSLv3 so it
Emeric Brun2c86cbf2014-10-30 15:56:50 +01009251 often makes sense to disable it when communicating with local servers. This
9252 option is also available on global statement "ssl-default-server-options".
9253 See also "force-sslv3", "force-tlsv*".
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009254
Willy Tarreau763a95b2012-10-04 23:15:39 +02009255 Supported in default-server: No
9256
Simon Hormanfa461682011-06-25 09:39:49 +09009257non-stick
9258 Never add connections allocated to this sever to a stick-table.
9259 This may be used in conjunction with backup to ensure that
9260 stick-table persistence is disabled for backup servers.
9261
Willy Tarreau763a95b2012-10-04 23:15:39 +02009262 Supported in default-server: No
9263
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009264observe <mode>
9265 This option enables health adjusting based on observing communication with
9266 the server. By default this functionality is disabled and enabling it also
9267 requires to enable health checks. There are two supported modes: "layer4" and
9268 "layer7". In layer4 mode, only successful/unsuccessful tcp connections are
9269 significant. In layer7, which is only allowed for http proxies, responses
9270 received from server are verified, like valid/wrong http code, unparsable
Willy Tarreau150d1462012-03-10 08:19:02 +01009271 headers, a timeout, etc. Valid status codes include 100 to 499, 501 and 505.
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009272
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009273 Supported in default-server: No
9274
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009275 See also the "check", "on-error" and "error-limit".
9276
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009277on-error <mode>
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009278 Select what should happen when enough consecutive errors are detected.
9279 Currently, four modes are available:
9280 - fastinter: force fastinter
9281 - fail-check: simulate a failed check, also forces fastinter (default)
9282 - sudden-death: simulate a pre-fatal failed health check, one more failed
9283 check will mark a server down, forces fastinter
9284 - mark-down: mark the server immediately down and force fastinter
9285
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009286 Supported in default-server: Yes
9287
Krzysztof Piotr Oledzki97f07b82009-12-15 22:31:24 +01009288 See also the "check", "observe" and "error-limit".
9289
Simon Hormane0d1bfb2011-06-21 14:34:58 +09009290on-marked-down <action>
9291 Modify what occurs when a server is marked down.
9292 Currently one action is available:
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07009293 - shutdown-sessions: Shutdown peer sessions. When this setting is enabled,
9294 all connections to the server are immediately terminated when the server
9295 goes down. It might be used if the health check detects more complex cases
9296 than a simple connection status, and long timeouts would cause the service
9297 to remain unresponsive for too long a time. For instance, a health check
9298 might detect that a database is stuck and that there's no chance to reuse
9299 existing connections anymore. Connections killed this way are logged with
9300 a 'D' termination code (for "Down").
Simon Hormane0d1bfb2011-06-21 14:34:58 +09009301
9302 Actions are disabled by default
9303
9304 Supported in default-server: Yes
9305
Justin Karnegeseb2c24a2012-05-24 15:28:52 -07009306on-marked-up <action>
9307 Modify what occurs when a server is marked up.
9308 Currently one action is available:
9309 - shutdown-backup-sessions: Shutdown sessions on all backup servers. This is
9310 done only if the server is not in backup state and if it is not disabled
9311 (it must have an effective weight > 0). This can be used sometimes to force
9312 an active server to take all the traffic back after recovery when dealing
9313 with long sessions (eg: LDAP, SQL, ...). Doing this can cause more trouble
9314 than it tries to solve (eg: incomplete transactions), so use this feature
9315 with extreme care. Sessions killed because a server comes up are logged
9316 with an 'U' termination code (for "Up").
9317
9318 Actions are disabled by default
9319
9320 Supported in default-server: Yes
9321
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009322port <port>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009323 Using the "port" parameter, it becomes possible to use a different port to
9324 send health-checks. On some servers, it may be desirable to dedicate a port
9325 to a specific component able to perform complex tests which are more suitable
9326 to health-checks than the application. It is common to run a simple script in
9327 inetd for instance. This parameter is ignored if the "check" parameter is not
9328 set. See also the "addr" parameter.
9329
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009330 Supported in default-server: Yes
9331
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009332redir <prefix>
9333 The "redir" parameter enables the redirection mode for all GET and HEAD
9334 requests addressing this server. This means that instead of having HAProxy
9335 forward the request to the server, it will send an "HTTP 302" response with
9336 the "Location" header composed of this prefix immediately followed by the
9337 requested URI beginning at the leading '/' of the path component. That means
9338 that no trailing slash should be used after <prefix>. All invalid requests
9339 will be rejected, and all non-GET or HEAD requests will be normally served by
9340 the server. Note that since the response is completely forged, no header
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009341 mangling nor cookie insertion is possible in the response. However, cookies in
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009342 requests are still analysed, making this solution completely usable to direct
9343 users to a remote location in case of local disaster. Main use consists in
9344 increasing bandwidth for static servers by having the clients directly
9345 connect to them. Note: never use a relative location here, it would cause a
9346 loop between the client and HAProxy!
9347
9348 Example : server srv1 192.168.1.1:80 redir http://image1.mydomain.com check
9349
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009350 Supported in default-server: No
9351
9352rise <count>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009353 The "rise" parameter states that a server will be considered as operational
9354 after <count> consecutive successful health checks. This value defaults to 2
9355 if unspecified. See also the "check", "inter" and "fall" parameters.
9356
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009357 Supported in default-server: Yes
9358
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01009359send-proxy
9360 The "send-proxy" parameter enforces use of the PROXY protocol over any
9361 connection established to this server. The PROXY protocol informs the other
9362 end about the layer 3/4 addresses of the incoming connection, so that it can
9363 know the client's address or the public address it accessed to, whatever the
9364 upper layer protocol. For connections accepted by an "accept-proxy" listener,
9365 the advertised address will be used. Only TCPv4 and TCPv6 address families
9366 are supported. Other families such as Unix sockets, will report an UNKNOWN
9367 family. Servers using this option can fully be chained to another instance of
9368 haproxy listening with an "accept-proxy" setting. This setting must not be
Willy Tarreau6c16adc2012-10-05 00:04:16 +02009369 used if the server isn't aware of the protocol. When health checks are sent
9370 to the server, the PROXY protocol is automatically used when this option is
9371 set, unless there is an explicit "port" or "addr" directive, in which case an
9372 explicit "check-send-proxy" directive would also be needed to use the PROXY
9373 protocol. See also the "accept-proxy" option of the "bind" keyword.
Willy Tarreau5ab04ec2011-03-20 10:32:26 +01009374
9375 Supported in default-server: No
9376
David Safb76832014-05-08 23:42:08 -04009377send-proxy-v2
9378 The "send-proxy-v2" parameter enforces use of the PROXY protocol version 2
9379 over any connection established to this server. The PROXY protocol informs
9380 the other end about the layer 3/4 addresses of the incoming connection, so
9381 that it can know the client's address or the public address it accessed to,
9382 whatever the upper layer protocol. This setting must not be used if the
9383 server isn't aware of this version of the protocol. See also the "send-proxy"
9384 option of the "bind" keyword.
9385
9386 Supported in default-server: No
9387
9388send-proxy-v2-ssl
9389 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
9390 2 over any connection established to this server. The PROXY protocol informs
9391 the other end about the layer 3/4 addresses of the incoming connection, so
9392 that it can know the client's address or the public address it accessed to,
9393 whatever the upper layer protocol. In addition, the SSL information extension
9394 of the PROXY protocol is added to the PROXY protocol header. This setting
9395 must not be used if the server isn't aware of this version of the protocol.
9396 See also the "send-proxy-v2" option of the "bind" keyword.
9397
9398 Supported in default-server: No
9399
9400send-proxy-v2-ssl-cn
9401 The "send-proxy-v2-ssl" parameter enforces use of the PROXY protocol version
9402 2 over any connection established to this server. The PROXY protocol informs
9403 the other end about the layer 3/4 addresses of the incoming connection, so
9404 that it can know the client's address or the public address it accessed to,
9405 whatever the upper layer protocol. In addition, the SSL information extension
9406 of the PROXY protocol, along along with the Common Name from the subject of
9407 the client certificate (if any), is added to the PROXY protocol header. This
9408 setting must not be used if the server isn't aware of this version of the
9409 protocol. See also the "send-proxy-v2" option of the "bind" keyword.
9410
9411 Supported in default-server: No
9412
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009413slowstart <start_time_in_ms>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009414 The "slowstart" parameter for a server accepts a value in milliseconds which
9415 indicates after how long a server which has just come back up will run at
9416 full speed. Just as with every other time-based parameter, it can be entered
9417 in any other explicit unit among { us, ms, s, m, h, d }. The speed grows
9418 linearly from 0 to 100% during this time. The limitation applies to two
9419 parameters :
9420
9421 - maxconn: the number of connections accepted by the server will grow from 1
9422 to 100% of the usual dynamic limit defined by (minconn,maxconn,fullconn).
9423
9424 - weight: when the backend uses a dynamic weighted algorithm, the weight
9425 grows linearly from 1 to 100%. In this case, the weight is updated at every
9426 health-check. For this reason, it is important that the "inter" parameter
9427 is smaller than the "slowstart", in order to maximize the number of steps.
9428
9429 The slowstart never applies when haproxy starts, otherwise it would cause
9430 trouble to running servers. It only applies when a server has been previously
9431 seen as failed.
9432
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009433 Supported in default-server: Yes
9434
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009435source <addr>[:<pl>[-<ph>]] [usesrc { <addr2>[:<port2>] | client | clientip } ]
Willy Tarreaubce70882009-09-07 11:51:47 +02009436source <addr>[:<port>] [usesrc { <addr2>[:<port2>] | hdr_ip(<hdr>[,<occ>]) } ]
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009437source <addr>[:<pl>[-<ph>]] [interface <name>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009438 The "source" parameter sets the source address which will be used when
9439 connecting to the server. It follows the exact same parameters and principle
9440 as the backend "source" keyword, except that it only applies to the server
9441 referencing it. Please consult the "source" keyword for details.
9442
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02009443 Additionally, the "source" statement on a server line allows one to specify a
9444 source port range by indicating the lower and higher bounds delimited by a
9445 dash ('-'). Some operating systems might require a valid IP address when a
9446 source port range is specified. It is permitted to have the same IP/range for
9447 several servers. Doing so makes it possible to bypass the maximum of 64k
9448 total concurrent connections. The limit will then reach 64k connections per
9449 server.
9450
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009451 Supported in default-server: No
9452
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009453ssl
Willy Tarreau44f65392013-06-25 07:56:20 +02009454 This option enables SSL ciphering on outgoing connections to the server. It
9455 is critical to verify server certificates using "verify" when using SSL to
9456 connect to servers, otherwise the communication is prone to trivial man in
9457 the-middle attacks rendering SSL useless. When this option is used, health
9458 checks are automatically sent in SSL too unless there is a "port" or an
9459 "addr" directive indicating the check should be sent to a different location.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009460 See the "check-ssl" option to force SSL health checks.
Willy Tarreau763a95b2012-10-04 23:15:39 +02009461
9462 Supported in default-server: No
Willy Tarreaua0ee1d02012-09-10 09:01:23 +02009463
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009464track [<proxy>/]<server>
Willy Tarreau32091232014-05-16 13:52:00 +02009465 This option enables ability to set the current state of the server by tracking
9466 another one. It is possible to track a server which itself tracks another
9467 server, provided that at the end of the chain, a server has health checks
9468 enabled. If <proxy> is omitted the current one is used. If disable-on-404 is
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009469 used, it has to be enabled on both proxies.
9470
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009471 Supported in default-server: No
9472
Emeric Brunef42d922012-10-11 16:11:36 +02009473verify [none|required]
9474 This setting is only available when support for OpenSSL was built in. If set
Emeric Brun850efd52014-01-29 12:24:34 +01009475 to 'none', server certificate is not verified. In the other case, The
9476 certificate provided by the server is verified using CAs from 'ca-file'
9477 and optional CRLs from 'crl-file'. If 'ssl_server_verify' is not specified
9478 in global section, this is the default. On verify failure the handshake
Willy Tarreau44f65392013-06-25 07:56:20 +02009479 is aborted. It is critically important to verify server certificates when
9480 using SSL to connect to servers, otherwise the communication is prone to
9481 trivial man-in-the-middle attacks rendering SSL totally useless.
Emeric Brunef42d922012-10-11 16:11:36 +02009482
9483 Supported in default-server: No
9484
Evan Broderbe554312013-06-27 00:05:25 -07009485verifyhost <hostname>
9486 This setting is only available when support for OpenSSL was built in, and
9487 only takes effect if 'verify required' is also specified. When set, the
9488 hostnames in the subject and subjectAlternateNames of the certificate
9489 provided by the server are checked. If none of the hostnames in the
9490 certificate match the specified hostname, the handshake is aborted. The
9491 hostnames in the server-provided certificate may include wildcards.
9492
9493 Supported in default-server: No
9494
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009495weight <weight>
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009496 The "weight" parameter is used to adjust the server's weight relative to
9497 other servers. All servers will receive a load proportional to their weight
9498 relative to the sum of all weights, so the higher the weight, the higher the
Willy Tarreau6704d672009-06-15 10:56:05 +02009499 load. The default weight is 1, and the maximal value is 256. A value of 0
9500 means the server will not participate in load-balancing but will still accept
9501 persistent connections. If this parameter is used to distribute the load
9502 according to server's capacity, it is recommended to start with values which
9503 can both grow and shrink, for instance between 10 and 100 to leave enough
9504 room above and below for later adjustments.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009505
Krzysztof Piotr Oledzkic53601c2010-01-06 10:50:42 +01009506 Supported in default-server: Yes
9507
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009508
95096. HTTP header manipulation
9510---------------------------
9511
9512In HTTP mode, it is possible to rewrite, add or delete some of the request and
9513response headers based on regular expressions. It is also possible to block a
9514request or a response if a particular header matches a regular expression,
9515which is enough to stop most elementary protocol attacks, and to protect
Willy Tarreau70dffda2014-01-30 03:07:23 +01009516against information leak from the internal network.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009517
Willy Tarreau70dffda2014-01-30 03:07:23 +01009518If HAProxy encounters an "Informational Response" (status code 1xx), it is able
9519to process all rsp* rules which can allow, deny, rewrite or delete a header,
9520but it will refuse to add a header to any such messages as this is not
9521HTTP-compliant. The reason for still processing headers in such responses is to
9522stop and/or fix any possible information leak which may happen, for instance
9523because another downstream equipment would unconditionally add a header, or if
9524a server name appears there. When such messages are seen, normal processing
9525still occurs on the next non-informational messages.
Willy Tarreau816b9792009-09-15 21:25:21 +02009526
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009527This section covers common usage of the following keywords, described in detail
9528in section 4.2 :
9529
9530 - reqadd <string>
9531 - reqallow <search>
9532 - reqiallow <search>
9533 - reqdel <search>
9534 - reqidel <search>
9535 - reqdeny <search>
9536 - reqideny <search>
9537 - reqpass <search>
9538 - reqipass <search>
9539 - reqrep <search> <replace>
9540 - reqirep <search> <replace>
9541 - reqtarpit <search>
9542 - reqitarpit <search>
9543 - rspadd <string>
9544 - rspdel <search>
9545 - rspidel <search>
9546 - rspdeny <search>
9547 - rspideny <search>
9548 - rsprep <search> <replace>
9549 - rspirep <search> <replace>
9550
9551With all these keywords, the same conventions are used. The <search> parameter
9552is a POSIX extended regular expression (regex) which supports grouping through
9553parenthesis (without the backslash). Spaces and other delimiters must be
9554prefixed with a backslash ('\') to avoid confusion with a field delimiter.
9555Other characters may be prefixed with a backslash to change their meaning :
9556
9557 \t for a tab
9558 \r for a carriage return (CR)
9559 \n for a new line (LF)
9560 \ to mark a space and differentiate it from a delimiter
9561 \# to mark a sharp and differentiate it from a comment
9562 \\ to use a backslash in a regex
9563 \\\\ to use a backslash in the text (*2 for regex, *2 for haproxy)
9564 \xXX to write the ASCII hex code XX as in the C language
9565
9566The <replace> parameter contains the string to be used to replace the largest
9567portion of text matching the regex. It can make use of the special characters
9568above, and can reference a substring which is delimited by parenthesis in the
9569regex, by writing a backslash ('\') immediately followed by one digit from 0 to
95709 indicating the group position (0 designating the entire line). This practice
9571is very common to users of the "sed" program.
9572
9573The <string> parameter represents the string which will systematically be added
9574after the last header line. It can also use special character sequences above.
9575
9576Notes related to these keywords :
9577---------------------------------
9578 - these keywords are not always convenient to allow/deny based on header
9579 contents. It is strongly recommended to use ACLs with the "block" keyword
9580 instead, resulting in far more flexible and manageable rules.
9581
9582 - lines are always considered as a whole. It is not possible to reference
9583 a header name only or a value only. This is important because of the way
9584 headers are written (notably the number of spaces after the colon).
9585
9586 - the first line is always considered as a header, which makes it possible to
9587 rewrite or filter HTTP requests URIs or response codes, but in turn makes
9588 it harder to distinguish between headers and request line. The regex prefix
9589 ^[^\ \t]*[\ \t] matches any HTTP method followed by a space, and the prefix
9590 ^[^ \t:]*: matches any header name followed by a colon.
9591
9592 - for performances reasons, the number of characters added to a request or to
9593 a response is limited at build time to values between 1 and 4 kB. This
9594 should normally be far more than enough for most usages. If it is too short
9595 on occasional usages, it is possible to gain some space by removing some
9596 useless headers before adding new ones.
9597
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +01009598 - keywords beginning with "reqi" and "rspi" are the same as their counterpart
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009599 without the 'i' letter except that they ignore case when matching patterns.
9600
9601 - when a request passes through a frontend then a backend, all req* rules
9602 from the frontend will be evaluated, then all req* rules from the backend
9603 will be evaluated. The reverse path is applied to responses.
9604
9605 - req* statements are applied after "block" statements, so that "block" is
9606 always the first one, but before "use_backend" in order to permit rewriting
Willy Tarreaud72758d2010-01-12 10:42:19 +01009607 before switching.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009608
9609
Willy Tarreau74ca5042013-06-11 23:12:07 +020096107. Using ACLs and fetching samples
9611----------------------------------
9612
9613Haproxy is capable of extracting data from request or response streams, from
9614client or server information, from tables, environmental information etc...
9615The action of extracting such data is called fetching a sample. Once retrieved,
9616these samples may be used for various purposes such as a key to a stick-table,
9617but most common usages consist in matching them against predefined constant
9618data called patterns.
9619
9620
96217.1. ACL basics
9622---------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009623
9624The use of Access Control Lists (ACL) provides a flexible solution to perform
9625content switching and generally to take decisions based on content extracted
9626from the request, the response or any environmental status. The principle is
9627simple :
9628
Willy Tarreau74ca5042013-06-11 23:12:07 +02009629 - extract a data sample from a stream, table or the environment
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009630 - optionally apply some format conversion to the extracted sample
Willy Tarreau74ca5042013-06-11 23:12:07 +02009631 - apply one or multiple pattern matching methods on this sample
9632 - perform actions only when a pattern matches the sample
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009633
Willy Tarreau74ca5042013-06-11 23:12:07 +02009634The actions generally consist in blocking a request, selecting a backend, or
9635adding a header.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009636
9637In order to define a test, the "acl" keyword is used. The syntax is :
9638
Willy Tarreau74ca5042013-06-11 23:12:07 +02009639 acl <aclname> <criterion> [flags] [operator] [<value>] ...
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009640
9641This creates a new ACL <aclname> or completes an existing one with new tests.
9642Those tests apply to the portion of request/response specified in <criterion>
9643and may be adjusted with optional flags [flags]. Some criteria also support
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009644an operator which may be specified before the set of values. Optionally some
9645conversion operators may be applied to the sample, and they will be specified
9646as a comma-delimited list of keywords just after the first keyword. The values
9647are of the type supported by the criterion, and are separated by spaces.
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009648
9649ACL names must be formed from upper and lower case letters, digits, '-' (dash),
9650'_' (underscore) , '.' (dot) and ':' (colon). ACL names are case-sensitive,
9651which means that "my_acl" and "My_Acl" are two different ACLs.
9652
9653There is no enforced limit to the number of ACLs. The unused ones do not affect
9654performance, they just consume a small amount of memory.
9655
Willy Tarreau74ca5042013-06-11 23:12:07 +02009656The criterion generally is the name of a sample fetch method, or one of its ACL
9657specific declinations. The default test method is implied by the output type of
9658this sample fetch method. The ACL declinations can describe alternate matching
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009659methods of a same sample fetch method. The sample fetch methods are the only
9660ones supporting a conversion.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009661
9662Sample fetch methods return data which can be of the following types :
9663 - boolean
9664 - integer (signed or unsigned)
9665 - IPv4 or IPv6 address
9666 - string
9667 - data block
9668
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009669Converters transform any of these data into any of these. For example, some
9670converters might convert a string to a lower-case string while other ones
9671would turn a string to an IPv4 address, or apply a netmask to an IP address.
9672The resulting sample is of the type of the last converter applied to the list,
9673which defaults to the type of the sample fetch method.
9674
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009675Each sample or converter returns data of a specific type, specified with its
9676keyword in this documentation. When an ACL is declared using a standard sample
9677fetch method, certain types automatically involved a default matching method
9678which are summarized in the table below :
9679
9680 +---------------------+-----------------+
9681 | Sample or converter | Default |
9682 | output type | matching method |
9683 +---------------------+-----------------+
9684 | boolean | bool |
9685 +---------------------+-----------------+
9686 | integer | int |
9687 +---------------------+-----------------+
9688 | ip | ip |
9689 +---------------------+-----------------+
9690 | string | str |
9691 +---------------------+-----------------+
9692 | binary | none, use "-m" |
9693 +---------------------+-----------------+
9694
9695Note that in order to match a binary samples, it is mandatory to specify a
9696matching method, see below.
9697
Willy Tarreau74ca5042013-06-11 23:12:07 +02009698The ACL engine can match these types against patterns of the following types :
9699 - boolean
9700 - integer or integer range
9701 - IP address / network
9702 - string (exact, substring, suffix, prefix, subdir, domain)
9703 - regular expression
9704 - hex block
9705
Willy Tarreauc57f0e22009-05-10 13:12:33 +02009706The following ACL flags are currently supported :
9707
Willy Tarreau2b5285d2010-05-09 23:45:24 +02009708 -i : ignore case during matching of all subsequent patterns.
9709 -f : load patterns from a file.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009710 -m : use a specific pattern matching method
Thierry FOURNIERb7729c92014-02-11 16:24:41 +01009711 -n : forbid the DNS resolutions
Thierry FOURNIER9860c412014-01-29 14:23:29 +01009712 -M : load the file pointed by -f like a map file.
Thierry FOURNIER3534d882014-01-20 17:01:44 +01009713 -u : force the unique id of the ACL
Willy Tarreau6a06a402007-07-15 20:15:28 +02009714 -- : force end of flags. Useful when a string looks like one of the flags.
9715
Willy Tarreau74ca5042013-06-11 23:12:07 +02009716The "-f" flag is followed by the name of a file from which all lines will be
9717read as individual values. It is even possible to pass multiple "-f" arguments
9718if the patterns are to be loaded from multiple files. Empty lines as well as
9719lines beginning with a sharp ('#') will be ignored. All leading spaces and tabs
9720will be stripped. If it is absolutely necessary to insert a valid pattern
9721beginning with a sharp, just prefix it with a space so that it is not taken for
9722a comment. Depending on the data type and match method, haproxy may load the
9723lines into a binary tree, allowing very fast lookups. This is true for IPv4 and
9724exact string matching. In this case, duplicates will automatically be removed.
9725
Thierry FOURNIER9860c412014-01-29 14:23:29 +01009726The "-M" flag allows an ACL to use a map file. If this flag is set, the file is
9727parsed as two column file. The first column contains the patterns used by the
9728ACL, and the second column contain the samples. The sample can be used later by
9729a map. This can be useful in some rare cases where an ACL would just be used to
9730check for the existence of a pattern in a map before a mapping is applied.
9731
Thierry FOURNIER3534d882014-01-20 17:01:44 +01009732The "-u" flag forces the unique id of the ACL. This unique id is used with the
9733socket interface to identify ACL and dynamically change its values. Note that a
9734file is always identified by its name even if an id is set.
9735
Willy Tarreau74ca5042013-06-11 23:12:07 +02009736Also, note that the "-i" flag applies to subsequent entries and not to entries
9737loaded from files preceding it. For instance :
9738
9739 acl valid-ua hdr(user-agent) -f exact-ua.lst -i -f generic-ua.lst test
9740
9741In this example, each line of "exact-ua.lst" will be exactly matched against
9742the "user-agent" header of the request. Then each line of "generic-ua" will be
9743case-insensitively matched. Then the word "test" will be insensitively matched
9744as well.
9745
9746The "-m" flag is used to select a specific pattern matching method on the input
9747sample. All ACL-specific criteria imply a pattern matching method and generally
9748do not need this flag. However, this flag is useful with generic sample fetch
9749methods to describe how they're going to be matched against the patterns. This
9750is required for sample fetches which return data type for which there is no
9751obvious matching method (eg: string or binary). When "-m" is specified and
9752followed by a pattern matching method name, this method is used instead of the
9753default one for the criterion. This makes it possible to match contents in ways
9754that were not initially planned, or with sample fetch methods which return a
9755string. The matching method also affects the way the patterns are parsed.
9756
Thierry FOURNIERb7729c92014-02-11 16:24:41 +01009757The "-n" flag forbids the dns resolutions. It is used with the load of ip files.
9758By default, if the parser cannot parse ip address it considers that the parsed
9759string is maybe a domain name and try dns resolution. The flag "-n" disable this
9760resolution. It is useful for detecting malformed ip lists. Note that if the DNS
9761server is not reachable, the haproxy configuration parsing may last many minutes
9762waiting fir the timeout. During this time no error messages are displayed. The
9763flag "-n" disable this behavior. Note also that during the runtime, this
9764function is disabled for the dynamic acl modifications.
9765
Willy Tarreau74ca5042013-06-11 23:12:07 +02009766There are some restrictions however. Not all methods can be used with all
9767sample fetch methods. Also, if "-m" is used in conjunction with "-f", it must
9768be placed first. The pattern matching method must be one of the following :
Willy Tarreau5adeda12013-03-31 22:13:34 +02009769
9770 - "found" : only check if the requested sample could be found in the stream,
9771 but do not compare it against any pattern. It is recommended not
Willy Tarreau74ca5042013-06-11 23:12:07 +02009772 to pass any pattern to avoid confusion. This matching method is
9773 particularly useful to detect presence of certain contents such
9774 as headers, cookies, etc... even if they are empty and without
9775 comparing them to anything nor counting them.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009776
9777 - "bool" : check the value as a boolean. It can only be applied to fetches
9778 which return a boolean or integer value, and takes no pattern.
Willy Tarreau74ca5042013-06-11 23:12:07 +02009779 Value zero or false does not match, all other values do match.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009780
9781 - "int" : match the value as an integer. It can be used with integer and
Willy Tarreau74ca5042013-06-11 23:12:07 +02009782 boolean samples. Boolean false is integer 0, true is integer 1.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009783
9784 - "ip" : match the value as an IPv4 or IPv6 address. It is compatible
Willy Tarreau74ca5042013-06-11 23:12:07 +02009785 with IP address samples only, so it is implied and never needed.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009786
9787 - "bin" : match the contents against an hexadecimal string representing a
9788 binary sequence. This may be used with binary or string samples.
9789
9790 - "len" : match the sample's length as an integer. This may be used with
9791 binary or string samples.
9792
Willy Tarreau74ca5042013-06-11 23:12:07 +02009793 - "str" : exact match : match the contents against a string. This may be
9794 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009795
Willy Tarreau74ca5042013-06-11 23:12:07 +02009796 - "sub" : substring match : check that the contents contain at least one of
9797 the provided string patterns. This may be used with binary or
9798 string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009799
Willy Tarreau74ca5042013-06-11 23:12:07 +02009800 - "reg" : regex match : match the contents against a list of regular
9801 expressions. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009802
Willy Tarreau74ca5042013-06-11 23:12:07 +02009803 - "beg" : prefix match : check that the contents begin like the provided
9804 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009805
Willy Tarreau74ca5042013-06-11 23:12:07 +02009806 - "end" : suffix match : check that the contents end like the provided
9807 string patterns. This may be used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009808
Willy Tarreau74ca5042013-06-11 23:12:07 +02009809 - "dir" : subdir match : check that a slash-delimited portion of the
9810 contents exactly matches one of the provided string patterns.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009811 This may be used with binary or string samples.
9812
Willy Tarreau74ca5042013-06-11 23:12:07 +02009813 - "dom" : domain match : check that a dot-delimited portion of the contents
9814 exactly match one of the provided string patterns. This may be
9815 used with binary or string samples.
Willy Tarreau5adeda12013-03-31 22:13:34 +02009816
9817For example, to quickly detect the presence of cookie "JSESSIONID" in an HTTP
9818request, it is possible to do :
9819
9820 acl jsess_present cook(JSESSIONID) -m found
9821
9822In order to apply a regular expression on the 500 first bytes of data in the
9823buffer, one would use the following acl :
9824
9825 acl script_tag payload(0,500) -m reg -i <script>
9826
Willy Tarreaue6b11e42013-11-26 19:02:32 +01009827On systems where the regex library is much slower when using "-i", it is
9828possible to convert the sample to lowercase before matching, like this :
9829
9830 acl script_tag payload(0,500),lower -m reg <script>
9831
Willy Tarreau74ca5042013-06-11 23:12:07 +02009832All ACL-specific criteria imply a default matching method. Most often, these
9833criteria are composed by concatenating the name of the original sample fetch
9834method and the matching method. For example, "hdr_beg" applies the "beg" match
9835to samples retrieved using the "hdr" fetch method. Since all ACL-specific
9836criteria rely on a sample fetch method, it is always possible instead to use
9837the original sample fetch method and the explicit matching method using "-m".
Willy Tarreau2b5285d2010-05-09 23:45:24 +02009838
Willy Tarreau74ca5042013-06-11 23:12:07 +02009839If an alternate match is specified using "-m" on an ACL-specific criterion,
Jarno Huuskonen0e82b922014-04-12 18:22:19 +03009840the matching method is simply applied to the underlying sample fetch method.
9841For example, all ACLs below are exact equivalent :
Willy Tarreau2b5285d2010-05-09 23:45:24 +02009842
Willy Tarreau74ca5042013-06-11 23:12:07 +02009843 acl short_form hdr_beg(host) www.
9844 acl alternate1 hdr_beg(host) -m beg www.
9845 acl alternate2 hdr_dom(host) -m beg www.
9846 acl alternate3 hdr(host) -m beg www.
Willy Tarreau2b5285d2010-05-09 23:45:24 +02009847
Willy Tarreau2b5285d2010-05-09 23:45:24 +02009848
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009849The table below summarizes the compatibility matrix between sample or converter
9850types and the pattern types to fetch against. It indicates for each compatible
9851combination the name of the matching method to be used, surrounded with angle
9852brackets ">" and "<" when the method is the default one and will work by
9853default without "-m".
Willy Tarreau0ba27502007-12-24 16:55:16 +01009854
Willy Tarreau74ca5042013-06-11 23:12:07 +02009855 +-------------------------------------------------+
9856 | Input sample type |
9857 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009858 | pattern type | boolean | integer | ip | string | binary |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009859 +----------------------+---------+---------+---------+---------+---------+
9860 | none (presence only) | found | found | found | found | found |
9861 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009862 | none (boolean value) |> bool <| bool | | bool | |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009863 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009864 | integer (value) | int |> int <| int | int | |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009865 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009866 | integer (length) | len | len | len | len | len |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009867 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009868 | IP address | | |> ip <| ip | ip |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009869 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIER2a06e392014-05-11 15:49:55 +02009870 | exact string | str | str | str |> str <| str |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009871 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009872 | prefix | beg | beg | beg | beg | beg |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009873 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009874 | suffix | end | end | end | end | end |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009875 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009876 | substring | sub | sub | sub | sub | sub |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009877 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009878 | subdir | dir | dir | dir | dir | dir |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009879 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009880 | domain | dom | dom | dom | dom | dom |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009881 +----------------------+---------+---------+---------+---------+---------+
Thierry FOURNIERe3ded592013-12-06 15:36:54 +01009882 | regex | reg | reg | reg | reg | reg |
Willy Tarreau74ca5042013-06-11 23:12:07 +02009883 +----------------------+---------+---------+---------+---------+---------+
9884 | hex block | | | | bin | bin |
9885 +----------------------+---------+---------+---------+---------+---------+
Willy Tarreau6a06a402007-07-15 20:15:28 +02009886
9887
Willy Tarreau74ca5042013-06-11 23:12:07 +020098887.1.1. Matching booleans
9889------------------------
9890
9891In order to match a boolean, no value is needed and all values are ignored.
9892Boolean matching is used by default for all fetch methods of type "boolean".
9893When boolean matching is used, the fetched value is returned as-is, which means
9894that a boolean "true" will always match and a boolean "false" will never match.
9895
9896Boolean matching may also be enforced using "-m bool" on fetch methods which
9897return an integer value. Then, integer value 0 is converted to the boolean
9898"false" and all other values are converted to "true".
9899
Willy Tarreau6a06a402007-07-15 20:15:28 +02009900
Willy Tarreau74ca5042013-06-11 23:12:07 +020099017.1.2. Matching integers
9902------------------------
9903
9904Integer matching applies by default to integer fetch methods. It can also be
9905enforced on boolean fetches using "-m int". In this case, "false" is converted
9906to the integer 0, and "true" is converted to the integer 1.
9907
9908Integer matching also supports integer ranges and operators. Note that integer
9909matching only applies to positive values. A range is a value expressed with a
9910lower and an upper bound separated with a colon, both of which may be omitted.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009911
9912For instance, "1024:65535" is a valid range to represent a range of
9913unprivileged ports, and "1024:" would also work. "0:1023" is a valid
9914representation of privileged ports, and ":1023" would also work.
9915
Willy Tarreau62644772008-07-16 18:36:06 +02009916As a special case, some ACL functions support decimal numbers which are in fact
9917two integers separated by a dot. This is used with some version checks for
9918instance. All integer properties apply to those decimal numbers, including
9919ranges and operators.
9920
Willy Tarreau6a06a402007-07-15 20:15:28 +02009921For an easier usage, comparison operators are also supported. Note that using
Willy Tarreau0ba27502007-12-24 16:55:16 +01009922operators with ranges does not make much sense and is strongly discouraged.
9923Similarly, it does not make much sense to perform order comparisons with a set
9924of values.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009925
Willy Tarreau0ba27502007-12-24 16:55:16 +01009926Available operators for integer matching are :
Willy Tarreau6a06a402007-07-15 20:15:28 +02009927
9928 eq : true if the tested value equals at least one value
9929 ge : true if the tested value is greater than or equal to at least one value
9930 gt : true if the tested value is greater than at least one value
9931 le : true if the tested value is less than or equal to at least one value
9932 lt : true if the tested value is less than at least one value
9933
Willy Tarreau0ba27502007-12-24 16:55:16 +01009934For instance, the following ACL matches any negative Content-Length header :
Willy Tarreau6a06a402007-07-15 20:15:28 +02009935
9936 acl negative-length hdr_val(content-length) lt 0
9937
Willy Tarreau62644772008-07-16 18:36:06 +02009938This one matches SSL versions between 3.0 and 3.1 (inclusive) :
9939
9940 acl sslv3 req_ssl_ver 3:3.1
9941
Willy Tarreau6a06a402007-07-15 20:15:28 +02009942
Willy Tarreau74ca5042013-06-11 23:12:07 +020099437.1.3. Matching strings
9944-----------------------
9945
9946String matching applies to string or binary fetch methods, and exists in 6
9947different forms :
9948
9949 - exact match (-m str) : the extracted string must exactly match the
9950 patterns ;
9951
9952 - substring match (-m sub) : the patterns are looked up inside the
9953 extracted string, and the ACL matches if any of them is found inside ;
9954
9955 - prefix match (-m beg) : the patterns are compared with the beginning of
9956 the extracted string, and the ACL matches if any of them matches.
9957
9958 - suffix match (-m end) : the patterns are compared with the end of the
9959 extracted string, and the ACL matches if any of them matches.
9960
9961 - subdir match (-m sub) : the patterns are looked up inside the extracted
9962 string, delimited with slashes ("/"), and the ACL matches if any of them
9963 matches.
9964
9965 - domain match (-m dom) : the patterns are looked up inside the extracted
9966 string, delimited with dots ("."), and the ACL matches if any of them
9967 matches.
Willy Tarreau6a06a402007-07-15 20:15:28 +02009968
9969String matching applies to verbatim strings as they are passed, with the
9970exception of the backslash ("\") which makes it possible to escape some
9971characters such as the space. If the "-i" flag is passed before the first
9972string, then the matching will be performed ignoring the case. In order
9973to match the string "-i", either set it second, or pass the "--" flag
Willy Tarreau0ba27502007-12-24 16:55:16 +01009974before the first string. Same applies of course to match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02009975
9976
Willy Tarreau74ca5042013-06-11 23:12:07 +020099777.1.4. Matching regular expressions (regexes)
9978---------------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +02009979
9980Just like with string matching, regex matching applies to verbatim strings as
9981they are passed, with the exception of the backslash ("\") which makes it
9982possible to escape some characters such as the space. If the "-i" flag is
9983passed before the first regex, then the matching will be performed ignoring
9984the case. In order to match the string "-i", either set it second, or pass
Willy Tarreau0ba27502007-12-24 16:55:16 +01009985the "--" flag before the first string. Same principle applies of course to
9986match the string "--".
Willy Tarreau6a06a402007-07-15 20:15:28 +02009987
9988
Willy Tarreau74ca5042013-06-11 23:12:07 +020099897.1.5. Matching arbitrary data blocks
9990-------------------------------------
9991
9992It is possible to match some extracted samples against a binary block which may
9993not safely be represented as a string. For this, the patterns must be passed as
9994a series of hexadecimal digits in an even number, when the match method is set
9995to binary. Each sequence of two digits will represent a byte. The hexadecimal
9996digits may be used upper or lower case.
9997
9998Example :
9999 # match "Hello\n" in the input stream (\x48 \x65 \x6c \x6c \x6f \x0a)
10000 acl hello payload(0,6) -m bin 48656c6c6f0a
10001
10002
100037.1.6. Matching IPv4 and IPv6 addresses
10004---------------------------------------
Willy Tarreau6a06a402007-07-15 20:15:28 +020010005
10006IPv4 addresses values can be specified either as plain addresses or with a
10007netmask appended, in which case the IPv4 address matches whenever it is
10008within the network. Plain addresses may also be replaced with a resolvable
Willy Tarreaud2a4aa22008-01-31 15:28:22 +010010009host name, but this practice is generally discouraged as it makes it more
Willy Tarreau0ba27502007-12-24 16:55:16 +010010010difficult to read and debug configurations. If hostnames are used, you should
10011at least ensure that they are present in /etc/hosts so that the configuration
10012does not depend on any random DNS match at the moment the configuration is
10013parsed.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010014
Willy Tarreauceb4ac92012-04-28 00:41:46 +020010015IPv6 may be entered in their usual form, with or without a netmask appended.
10016Only bit counts are accepted for IPv6 netmasks. In order to avoid any risk of
10017trouble with randomly resolved IP addresses, host names are never allowed in
10018IPv6 patterns.
10019
10020HAProxy is also able to match IPv4 addresses with IPv6 addresses in the
10021following situations :
10022 - tested address is IPv4, pattern address is IPv4, the match applies
10023 in IPv4 using the supplied mask if any.
10024 - tested address is IPv6, pattern address is IPv6, the match applies
10025 in IPv6 using the supplied mask if any.
10026 - tested address is IPv6, pattern address is IPv4, the match applies in IPv4
10027 using the pattern's mask if the IPv6 address matches with 2002:IPV4::,
10028 ::IPV4 or ::ffff:IPV4, otherwise it fails.
10029 - tested address is IPv4, pattern address is IPv6, the IPv4 address is first
10030 converted to IPv6 by prefixing ::ffff: in front of it, then the match is
10031 applied in IPv6 using the supplied IPv6 mask.
10032
Willy Tarreau74ca5042013-06-11 23:12:07 +020010033
100347.2. Using ACLs to form conditions
10035----------------------------------
10036
10037Some actions are only performed upon a valid condition. A condition is a
10038combination of ACLs with operators. 3 operators are supported :
10039
10040 - AND (implicit)
10041 - OR (explicit with the "or" keyword or the "||" operator)
10042 - Negation with the exclamation mark ("!")
Willy Tarreau6a06a402007-07-15 20:15:28 +020010043
Willy Tarreau74ca5042013-06-11 23:12:07 +020010044A condition is formed as a disjunctive form:
Willy Tarreau6a06a402007-07-15 20:15:28 +020010045
Willy Tarreau74ca5042013-06-11 23:12:07 +020010046 [!]acl1 [!]acl2 ... [!]acln { or [!]acl1 [!]acl2 ... [!]acln } ...
Willy Tarreaubef91e72013-03-31 23:14:46 +020010047
Willy Tarreau74ca5042013-06-11 23:12:07 +020010048Such conditions are generally used after an "if" or "unless" statement,
10049indicating when the condition will trigger the action.
Willy Tarreaubef91e72013-03-31 23:14:46 +020010050
Willy Tarreau74ca5042013-06-11 23:12:07 +020010051For instance, to block HTTP requests to the "*" URL with methods other than
10052"OPTIONS", as well as POST requests without content-length, and GET or HEAD
10053requests with a content-length greater than 0, and finally every request which
10054is not either GET/HEAD/POST/OPTIONS !
10055
10056 acl missing_cl hdr_cnt(Content-length) eq 0
10057 block if HTTP_URL_STAR !METH_OPTIONS || METH_POST missing_cl
10058 block if METH_GET HTTP_CONTENT
10059 block unless METH_GET or METH_POST or METH_OPTIONS
10060
10061To select a different backend for requests to static contents on the "www" site
10062and to every request on the "img", "video", "download" and "ftp" hosts :
10063
10064 acl url_static path_beg /static /images /img /css
10065 acl url_static path_end .gif .png .jpg .css .js
10066 acl host_www hdr_beg(host) -i www
10067 acl host_static hdr_beg(host) -i img. video. download. ftp.
10068
10069 # now use backend "static" for all static-only hosts, and for static urls
10070 # of host "www". Use backend "www" for the rest.
10071 use_backend static if host_static or host_www url_static
10072 use_backend www if host_www
10073
10074It is also possible to form rules using "anonymous ACLs". Those are unnamed ACL
10075expressions that are built on the fly without needing to be declared. They must
10076be enclosed between braces, with a space before and after each brace (because
10077the braces must be seen as independent words). Example :
10078
10079 The following rule :
10080
10081 acl missing_cl hdr_cnt(Content-length) eq 0
10082 block if METH_POST missing_cl
10083
10084 Can also be written that way :
10085
10086 block if METH_POST { hdr_cnt(Content-length) eq 0 }
10087
10088It is generally not recommended to use this construct because it's a lot easier
10089to leave errors in the configuration when written that way. However, for very
10090simple rules matching only one source IP address for instance, it can make more
10091sense to use them than to declare ACLs with random names. Another example of
10092good use is the following :
10093
10094 With named ACLs :
10095
10096 acl site_dead nbsrv(dynamic) lt 2
10097 acl site_dead nbsrv(static) lt 2
10098 monitor fail if site_dead
10099
10100 With anonymous ACLs :
10101
10102 monitor fail if { nbsrv(dynamic) lt 2 } || { nbsrv(static) lt 2 }
10103
10104See section 4.2 for detailed help on the "block" and "use_backend" keywords.
10105
10106
101077.3. Fetching samples
10108---------------------
10109
10110Historically, sample fetch methods were only used to retrieve data to match
10111against patterns using ACLs. With the arrival of stick-tables, a new class of
10112sample fetch methods was created, most often sharing the same syntax as their
10113ACL counterpart. These sample fetch methods are also known as "fetches". As
10114of now, ACLs and fetches have converged. All ACL fetch methods have been made
10115available as fetch methods, and ACLs may use any sample fetch method as well.
10116
10117This section details all available sample fetch methods and their output type.
10118Some sample fetch methods have deprecated aliases that are used to maintain
10119compatibility with existing configurations. They are then explicitly marked as
10120deprecated and should not be used in new setups.
10121
10122The ACL derivatives are also indicated when available, with their respective
10123matching methods. These ones all have a well defined default pattern matching
10124method, so it is never necessary (though allowed) to pass the "-m" option to
10125indicate how the sample will be matched using ACLs.
10126
10127As indicated in the sample type versus matching compatibility matrix above,
10128when using a generic sample fetch method in an ACL, the "-m" option is
10129mandatory unless the sample type is one of boolean, integer, IPv4 or IPv6. When
10130the same keyword exists as an ACL keyword and as a standard fetch method, the
10131ACL engine will automatically pick the ACL-only one by default.
10132
10133Some of these keywords support one or multiple mandatory arguments, and one or
10134multiple optional arguments. These arguments are strongly typed and are checked
10135when the configuration is parsed so that there is no risk of running with an
10136incorrect argument (eg: an unresolved backend name). Fetch function arguments
10137are passed between parenthesis and are delimited by commas. When an argument
10138is optional, it will be indicated below between square brackets ('[ ]'). When
10139all arguments are optional, the parenthesis may be omitted.
10140
10141Thus, the syntax of a standard sample fetch method is one of the following :
10142 - name
10143 - name(arg1)
10144 - name(arg1,arg2)
10145
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010146
101477.3.1. Converters
10148-----------------
10149
Willy Tarreaue6b11e42013-11-26 19:02:32 +010010150Sample fetch methods may be combined with transformations to be applied on top
10151of the fetched sample (also called "converters"). These combinations form what
10152is called "sample expressions" and the result is a "sample". Initially this
10153was only supported by "stick on" and "stick store-request" directives but this
10154has now be extended to all places where samples may be used (acls, log-format,
10155unique-id-format, add-header, ...).
10156
10157These transformations are enumerated as a series of specific keywords after the
10158sample fetch method. These keywords may equally be appended immediately after
10159the fetch keyword's argument, delimited by a comma. These keywords can also
10160support some arguments (eg: a netmask) which must be passed in parenthesis.
Willy Tarreau0ba27502007-12-24 16:55:16 +010010161
Willy Tarreau97707872015-01-27 15:12:13 +010010162A certain category of converters are bitwise and arithmetic operators which
10163support performing basic operations on integers. Some bitwise operations are
10164supported (and, or, xor, cpl) and some arithmetic operations are supported
10165(add, sub, mul, div, mod, neg). Some comparators are provided (odd, even, not,
10166bool) which make it possible to report a match without having to write an ACL.
10167
Willy Tarreau74ca5042013-06-11 23:12:07 +020010168The currently available list of transformation keywords include :
Willy Tarreau0ba27502007-12-24 16:55:16 +010010169
Willy Tarreau97707872015-01-27 15:12:13 +010010170add(<value>)
10171 Adds <value> to the input value of type unsigned integer, and returns the
10172 result as an unsigned integer.
10173
10174and(<value>)
10175 Performs a bitwise "AND" between <value> and the input value of type unsigned
10176 integer, and returns the result as an unsigned integer.
10177
Emeric Brun53d1a982014-04-30 18:21:37 +020010178base64
10179 Converts a binary input sample to a base64 string. It is used to log or
10180 transfer binary content in a way that can be reliably transferred (eg:
10181 an SSL ID can be copied in a header).
10182
Willy Tarreau97707872015-01-27 15:12:13 +010010183bool
10184 Returns a boolean TRUE if the input value of type unsigned integer is
10185 non-null, otherwise returns FALSE. Used in conjunction with and(), it can be
10186 used to report true/false for bit testing on input values (eg: verify the
10187 presence of a flag).
10188
Emeric Brun54c4ac82014-11-03 15:32:43 +010010189bytes(<offset>[,<length>])
10190 Extracts some bytes from an input binary sample. The result is a binary
10191 sample starting at an offset (in bytes) of the original sample and
10192 optionnaly truncated at the given length.
10193
Willy Tarreau97707872015-01-27 15:12:13 +010010194cpl
10195 Takes the input value of type unsigned integer, applies a twos-complement
10196 (flips all bits) and returns the result as an unsigned integer.
10197
Willy Tarreau80599772015-01-20 19:35:24 +010010198crc32([<avalanche>])
10199 Hashes a binary input sample into an unsigned 32-bit quantity using the CRC32
10200 hash function. Optionally, it is possible to apply a full avalanche hash
10201 function to the output if the optional <avalanche> argument equals 1. This
10202 converter uses the same functions as used by the various hash-based load
10203 balancing algorithms, so it will provide exactly the same results. It is
10204 provided for compatibility with other software which want a CRC32 to be
10205 computed on some input keys, so it follows the most common implementation as
10206 found in Ethernet, Gzip, PNG, etc... It is slower than the other algorithms
10207 but may provide a better or at least less predictable distribution. It must
10208 not be used for security purposes as a 32-bit hash is trivial to break. See
10209 also "djb2", "sdbm", "wt6" and the "hash-type" directive.
10210
Willy Tarreau97707872015-01-27 15:12:13 +010010211div(<value>)
10212 Divides the input value of type unsigned integer by <value>, and returns the
10213 result as an unsigned integer. If <value> is null, the largest unsigned
10214 integer is returned (typically 2^32-1).
10215
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010216djb2([<avalanche>])
10217 Hashes a binary input sample into an unsigned 32-bit quantity using the DJB2
10218 hash function. Optionally, it is possible to apply a full avalanche hash
10219 function to the output if the optional <avalanche> argument equals 1. This
10220 converter uses the same functions as used by the various hash-based load
10221 balancing algorithms, so it will provide exactly the same results. It is
10222 mostly intended for debugging, but can be used as a stick-table entry to
10223 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010010224 32-bit hash is trivial to break. See also "crc32", "sdbm", "wt6" and the
10225 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010226
Willy Tarreau97707872015-01-27 15:12:13 +010010227even
10228 Returns a boolean TRUE if the input value of type unsigned integer is even
10229 otherwise returns FALSE. It is functionally equivalent to "not,and(1),bool".
10230
Emeric Brunf399b0d2014-11-03 17:07:03 +010010231field(<index>,<delimiters>)
10232 Extracts the substring at the given index considering given delimiters from
10233 an input string. Indexes start at 1 and delimiters are a string formatted
10234 list of chars.
10235
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010236hex
10237 Converts a binary input sample to an hex string containing two hex digits per
10238 input byte. It is used to log or transfer hex dumps of some binary input data
10239 in a way that can be reliably transferred (eg: an SSL ID can be copied in a
10240 header).
Thierry FOURNIER2f49d6d2014-03-12 15:01:52 +010010241
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010242http_date([<offset>])
10243 Converts an integer supposed to contain a date since epoch to a string
10244 representing this date in a format suitable for use in HTTP header fields. If
10245 an offset value is specified, then it is a number of seconds that is added to
10246 the date before the conversion is operated. This is particularly useful to
10247 emit Date header fields, Expires values in responses when combined with a
10248 positive offset, or Last-Modified values when the offset is negative.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010249
Willy Tarreaud9f316a2014-07-10 14:03:38 +020010250in_table(<table>)
10251 Uses the string representation of the input sample to perform a look up in
10252 the specified table. If the key is not found in the table, a boolean false
10253 is returned. Otherwise a boolean true is returned. This can be used to verify
10254 the presence of a certain key in a table tracking some elements (eg: whether
10255 or not a source IP address or an Authorization header was already seen).
10256
Willy Tarreauffcb2e42014-07-10 16:29:08 +020010257ipmask(<mask>)
10258 Apply a mask to an IPv4 address, and use the result for lookups and storage.
10259 This can be used to make all hosts within a certain mask to share the same
10260 table entries and as such use the same server. The mask can be passed in
10261 dotted form (eg: 255.255.255.0) or in CIDR form (eg: 24).
10262
Thierry FOURNIER317e1c42014-08-12 10:20:47 +020010263json([<input-code>])
10264 Escapes the input string and produces an ASCII ouput string ready to use as a
10265 JSON string. The converter tries to decode the input string according to the
10266 <input-code> parameter. It can be "ascii", "utf8", "utf8s", "utf8"" or
10267 "utf8ps". The "ascii" decoder never fails. The "utf8" decoder detects 3 types
10268 of errors:
10269 - bad UTF-8 sequence (lone continuation byte, bad number of continuation
10270 bytes, ...)
10271 - invalid range (the decoded value is within a UTF-8 prohibited range),
10272 - code overlong (the value is encoded with more bytes than necessary).
10273
10274 The UTF-8 JSON encoding can produce a "too long value" error when the UTF-8
10275 character is greater than 0xffff because the JSON string escape specification
10276 only authorizes 4 hex digits for the value encoding. The UTF-8 decoder exists
10277 in 4 variants designated by a combination of two suffix letters : "p" for
10278 "permissive" and "s" for "silently ignore". The behaviors of the decoders
10279 are :
10280 - "ascii" : never fails ;
10281 - "utf8" : fails on any detected errors ;
10282 - "utf8s" : never fails, but removes characters corresponding to errors ;
10283 - "utf8p" : accepts and fixes the overlong errors, but fails on any other
10284 error ;
10285 - "utf8ps" : never fails, accepts and fixes the overlong errors, but removes
10286 characters corresponding to the other errors.
10287
10288 This converter is particularly useful for building properly escaped JSON for
10289 logging to servers which consume JSON-formated traffic logs.
10290
10291 Example:
10292 capture request header user-agent len 150
10293 capture request header Host len 15
10294 log-format {"ip":"%[src]","user-agent":"%[capture.req.hdr(1),json]"}
10295
10296 Input request from client 127.0.0.1:
10297 GET / HTTP/1.0
10298 User-Agent: Very "Ugly" UA 1/2
10299
10300 Output log:
10301 {"ip":"127.0.0.1","user-agent":"Very \"Ugly\" UA 1\/2"}
10302
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010303language(<value>[,<default>])
10304 Returns the value with the highest q-factor from a list as extracted from the
10305 "accept-language" header using "req.fhdr". Values with no q-factor have a
10306 q-factor of 1. Values with a q-factor of 0 are dropped. Only values which
10307 belong to the list of semi-colon delimited <values> will be considered. The
10308 argument <value> syntax is "lang[;lang[;lang[;...]]]". If no value matches the
10309 given list and a default value is provided, it is returned. Note that language
10310 names may have a variant after a dash ('-'). If this variant is present in the
10311 list, it will be matched, but if it is not, only the base language is checked.
10312 The match is case-sensitive, and the output string is always one of those
10313 provided in arguments. The ordering of arguments is meaningless, only the
10314 ordering of the values in the request counts, as the first value among
10315 multiple sharing the same q-factor is used.
Thierry FOURNIERad903512014-04-11 17:51:01 +020010316
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010317 Example :
Thierry FOURNIERad903512014-04-11 17:51:01 +020010318
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010319 # this configuration switches to the backend matching a
10320 # given language based on the request :
Thierry FOURNIERad903512014-04-11 17:51:01 +020010321
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010322 acl es req.fhdr(accept-language),language(es;fr;en) -m str es
10323 acl fr req.fhdr(accept-language),language(es;fr;en) -m str fr
10324 acl en req.fhdr(accept-language),language(es;fr;en) -m str en
10325 use_backend spanish if es
10326 use_backend french if fr
10327 use_backend english if en
10328 default_backend choose_your_language
Thierry FOURNIERad903512014-04-11 17:51:01 +020010329
Willy Tarreauffcb2e42014-07-10 16:29:08 +020010330lower
10331 Convert a string sample to lower case. This can only be placed after a string
10332 sample fetch function or after a transformation keyword returning a string
10333 type. The result is of type string.
10334
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020010335ltime(<format>[,<offset>])
10336 Converts an integer supposed to contain a date since epoch to a string
10337 representing this date in local time using a format defined by the <format>
10338 string using strftime(3). The purpose is to allow any date format to be used
10339 in logs. An optional <offset> in seconds may be applied to the input date
10340 (positive or negative). See the strftime() man page for the format supported
10341 by your operating system. See also the utime converter.
10342
10343 Example :
10344
10345 # Emit two colons, one with the local time and another with ip:port
10346 # Eg: 20140710162350 127.0.0.1:57325
10347 log-format %[date,ltime(%Y%m%d%H%M%S)]\ %ci:%cp
10348
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010349map(<map_file>[,<default_value>])
10350map_<match_type>(<map_file>[,<default_value>])
10351map_<match_type>_<output_type>(<map_file>[,<default_value>])
10352 Search the input value from <map_file> using the <match_type> matching method,
10353 and return the associated value converted to the type <output_type>. If the
10354 input value cannot be found in the <map_file>, the converter returns the
10355 <default_value>. If the <default_value> is not set, the converter fails and
10356 acts as if no input value could be fetched. If the <match_type> is not set, it
10357 defaults to "str". Likewise, if the <output_type> is not set, it defaults to
10358 "str". For convenience, the "map" keyword is an alias for "map_str" and maps a
10359 string to another string.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010360
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010361 It is important to avoid overlapping between the keys : IP addresses and
10362 strings are stored in trees, so the first of the finest match will be used.
10363 Other keys are stored in lists, so the first matching occurrence will be used.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010364
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010365 The following array contains the list of all map functions avalaible sorted by
10366 input type, match type and output type.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010367
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010368 input type | match method | output type str | output type int | output type ip
10369 -----------+--------------+-----------------+-----------------+---------------
10370 str | str | map_str | map_str_int | map_str_ip
10371 -----------+--------------+-----------------+-----------------+---------------
Willy Tarreau787a4c02014-05-10 07:55:30 +020010372 str | beg | map_beg | map_beg_int | map_end_ip
10373 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010374 str | sub | map_sub | map_sub_int | map_sub_ip
10375 -----------+--------------+-----------------+-----------------+---------------
10376 str | dir | map_dir | map_dir_int | map_dir_ip
10377 -----------+--------------+-----------------+-----------------+---------------
10378 str | dom | map_dom | map_dom_int | map_dom_ip
10379 -----------+--------------+-----------------+-----------------+---------------
10380 str | end | map_end | map_end_int | map_end_ip
10381 -----------+--------------+-----------------+-----------------+---------------
10382 str | reg | map_reg | map_reg_int | map_reg_ip
10383 -----------+--------------+-----------------+-----------------+---------------
10384 int | int | map_int | map_int_int | map_int_ip
10385 -----------+--------------+-----------------+-----------------+---------------
10386 ip | ip | map_ip | map_ip_int | map_ip_ip
10387 -----------+--------------+-----------------+-----------------+---------------
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010388
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010389 The file contains one key + value per line. Lines which start with '#' are
10390 ignored, just like empty lines. Leading tabs and spaces are stripped. The key
10391 is then the first "word" (series of non-space/tabs characters), and the value
10392 is what follows this series of space/tab till the end of the line excluding
10393 trailing spaces/tabs.
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010394
Thierry FOURNIER060762e2014-04-23 13:29:15 +020010395 Example :
10396
10397 # this is a comment and is ignored
10398 2.22.246.0/23 United Kingdom \n
10399 <-><-----------><--><------------><---->
10400 | | | | `- trailing spaces ignored
10401 | | | `---------- value
10402 | | `-------------------- middle spaces ignored
10403 | `---------------------------- key
10404 `------------------------------------ leading spaces ignored
10405
Willy Tarreau97707872015-01-27 15:12:13 +010010406mod(<value>)
10407 Divides the input value of type unsigned integer by <value>, and returns the
10408 remainder as an unsigned integer. If <value> is null, then zero is returned.
10409
10410mul(<value>)
10411 Multiplies the input value of type unsigned integer by <value>, and returns
10412 the product as an unsigned integer. In case of overflow, the higher bits are
10413 lost, leading to seemingly strange values.
10414
10415neg
10416 Takes the input value of type unsigned integer, computes the opposite value,
10417 and returns the remainder as an unsigned integer. 0 is identity. This
10418 operator is provided for reversed subtracts : in order to subtract the input
10419 from a constant, simply perform a "neg,add(value)".
10420
10421not
10422 Returns a boolean FALSE if the input value of type unsigned integer is
10423 non-null, otherwise returns TRUE. Used in conjunction with and(), it can be
10424 used to report true/false for bit testing on input values (eg: verify the
10425 absence of a flag).
10426
10427odd
10428 Returns a boolean TRUE if the input value of type unsigned integer is odd
10429 otherwise returns FALSE. It is functionally equivalent to "and(1),bool".
10430
10431or(<value>)
10432 Performs a bitwise "OR" between <value> and the input value of type unsigned
10433 integer, and returns the result as an unsigned integer.
10434
Willy Tarreauc4dc3502015-01-23 20:39:28 +010010435regsub(<regex>,<subst>[,<flags>])
Willy Tarreau7eda8492015-01-20 19:47:06 +010010436 Applies a regex-based substitution to the input string. It does the same
10437 operation as the well-known "sed" utility with "s/<regex>/<subst>/". By
10438 default it will replace in the input string the first occurrence of the
10439 largest part matching the regular expression <regex> with the substitution
10440 string <subst>. It is possible to replace all occurrences instead by adding
10441 the flag "g" in the third argument <flags>. It is also possible to make the
10442 regex case insensitive by adding the flag "i" in <flags>. Since <flags> is a
10443 string, it is made up from the concatenation of all desired flags. Thus if
10444 both "i" and "g" are desired, using "gi" or "ig" will have the same effect.
10445 It is important to note that due to the current limitations of the
10446 configuration parser, some characters such as closing parenthesis or comma
10447 are not possible to use in the arguments. The first use of this converter is
10448 to replace certain characters or sequence of characters with other ones.
10449
10450 Example :
10451
10452 # de-duplicate "/" in header "x-path".
10453 # input: x-path: /////a///b/c/xzxyz/
10454 # output: x-path: /a/b/c/xzxyz/
10455 http-request set-header x-path %[hdr(x-path),regsub(/+,/,g)]
10456
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010457sdbm([<avalanche>])
10458 Hashes a binary input sample into an unsigned 32-bit quantity using the SDBM
10459 hash function. Optionally, it is possible to apply a full avalanche hash
10460 function to the output if the optional <avalanche> argument equals 1. This
10461 converter uses the same functions as used by the various hash-based load
10462 balancing algorithms, so it will provide exactly the same results. It is
10463 mostly intended for debugging, but can be used as a stick-table entry to
10464 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010010465 32-bit hash is trivial to break. See also "crc32", "djb2", "wt6" and the
10466 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010467
Willy Tarreau97707872015-01-27 15:12:13 +010010468sub(<value>)
10469 Subtracts <value> from the input value of type unsigned integer, and returns
10470 the result as an unsigned integer. Note: in order to subtract the input from
10471 a constant, simply perform a "neg,add(value)".
10472
Willy Tarreaud9f316a2014-07-10 14:03:38 +020010473table_bytes_in_rate(<table>)
10474 Uses the string representation of the input sample to perform a look up in
10475 the specified table. If the key is not found in the table, integer value zero
10476 is returned. Otherwise the converter returns the average client-to-server
10477 bytes rate associated with the input sample in the designated table, measured
10478 in amount of bytes over the period configured in the table. See also the
10479 sc_bytes_in_rate sample fetch keyword.
10480
10481
10482table_bytes_out_rate(<table>)
10483 Uses the string representation of the input sample to perform a look up in
10484 the specified table. If the key is not found in the table, integer value zero
10485 is returned. Otherwise the converter returns the average server-to-client
10486 bytes rate associated with the input sample in the designated table, measured
10487 in amount of bytes over the period configured in the table. See also the
10488 sc_bytes_out_rate sample fetch keyword.
10489
10490table_conn_cnt(<table>)
10491 Uses the string representation of the input sample to perform a look up in
10492 the specified table. If the key is not found in the table, integer value zero
10493 is returned. Otherwise the converter returns the cumulated amount of incoming
10494 connections associated with the input sample in the designated table. See
10495 also the sc_conn_cnt sample fetch keyword.
10496
10497table_conn_cur(<table>)
10498 Uses the string representation of the input sample to perform a look up in
10499 the specified table. If the key is not found in the table, integer value zero
10500 is returned. Otherwise the converter returns the current amount of concurrent
10501 tracked connections associated with the input sample in the designated table.
10502 See also the sc_conn_cur sample fetch keyword.
10503
10504table_conn_rate(<table>)
10505 Uses the string representation of the input sample to perform a look up in
10506 the specified table. If the key is not found in the table, integer value zero
10507 is returned. Otherwise the converter returns the average incoming connection
10508 rate associated with the input sample in the designated table. See also the
10509 sc_conn_rate sample fetch keyword.
10510
10511table_gpc0(<table>)
10512 Uses the string representation of the input sample to perform a look up in
10513 the specified table. If the key is not found in the table, integer value zero
10514 is returned. Otherwise the converter returns the current value of the first
10515 general purpose counter associated with the input sample in the designated
10516 table. See also the sc_get_gpc0 sample fetch keyword.
10517
10518table_gpc0_rate(<table>)
10519 Uses the string representation of the input sample to perform a look up in
10520 the specified table. If the key is not found in the table, integer value zero
10521 is returned. Otherwise the converter returns the frequency which the gpc0
10522 counter was incremented over the configured period in the table, associated
10523 with the input sample in the designated table. See also the sc_get_gpc0_rate
10524 sample fetch keyword.
10525
10526table_http_err_cnt(<table>)
10527 Uses the string representation of the input sample to perform a look up in
10528 the specified table. If the key is not found in the table, integer value zero
10529 is returned. Otherwise the converter returns the cumulated amount of HTTP
10530 errors associated with the input sample in the designated table. See also the
10531 sc_http_err_cnt sample fetch keyword.
10532
10533table_http_err_rate(<table>)
10534 Uses the string representation of the input sample to perform a look up in
10535 the specified table. If the key is not found in the table, integer value zero
10536 is returned. Otherwise the average rate of HTTP errors associated with the
10537 input sample in the designated table, measured in amount of errors over the
10538 period configured in the table. See also the sc_http_err_rate sample fetch
10539 keyword.
10540
10541table_http_req_cnt(<table>)
10542 Uses the string representation of the input sample to perform a look up in
10543 the specified table. If the key is not found in the table, integer value zero
10544 is returned. Otherwise the converter returns the cumulated amount of HTTP
10545 requests associated with the input sample in the designated table. See also
10546 the sc_http_req_cnt sample fetch keyword.
10547
10548table_http_req_rate(<table>)
10549 Uses the string representation of the input sample to perform a look up in
10550 the specified table. If the key is not found in the table, integer value zero
10551 is returned. Otherwise the average rate of HTTP requests associated with the
10552 input sample in the designated table, measured in amount of requests over the
10553 period configured in the table. See also the sc_http_req_rate sample fetch
10554 keyword.
10555
10556table_kbytes_in(<table>)
10557 Uses the string representation of the input sample to perform a look up in
10558 the specified table. If the key is not found in the table, integer value zero
10559 is returned. Otherwise the converter returns the cumulated amount of client-
10560 to-server data associated with the input sample in the designated table,
10561 measured in kilobytes. The test is currently performed on 32-bit integers,
10562 which limits values to 4 terabytes. See also the sc_kbytes_in sample fetch
10563 keyword.
10564
10565table_kbytes_out(<table>)
10566 Uses the string representation of the input sample to perform a look up in
10567 the specified table. If the key is not found in the table, integer value zero
10568 is returned. Otherwise the converter returns the cumulated amount of server-
10569 to-client data associated with the input sample in the designated table,
10570 measured in kilobytes. The test is currently performed on 32-bit integers,
10571 which limits values to 4 terabytes. See also the sc_kbytes_out sample fetch
10572 keyword.
10573
10574table_server_id(<table>)
10575 Uses the string representation of the input sample to perform a look up in
10576 the specified table. If the key is not found in the table, integer value zero
10577 is returned. Otherwise the converter returns the server ID associated with
10578 the input sample in the designated table. A server ID is associated to a
10579 sample by a "stick" rule when a connection to a server succeeds. A server ID
10580 zero means that no server is associated with this key.
10581
10582table_sess_cnt(<table>)
10583 Uses the string representation of the input sample to perform a look up in
10584 the specified table. If the key is not found in the table, integer value zero
10585 is returned. Otherwise the converter returns the cumulated amount of incoming
10586 sessions associated with the input sample in the designated table. Note that
10587 a session here refers to an incoming connection being accepted by the
10588 "tcp-request connection" rulesets. See also the sc_sess_cnt sample fetch
10589 keyword.
10590
10591table_sess_rate(<table>)
10592 Uses the string representation of the input sample to perform a look up in
10593 the specified table. If the key is not found in the table, integer value zero
10594 is returned. Otherwise the converter returns the average incoming session
10595 rate associated with the input sample in the designated table. Note that a
10596 session here refers to an incoming connection being accepted by the
10597 "tcp-request connection" rulesets. See also the sc_sess_rate sample fetch
10598 keyword.
10599
10600table_trackers(<table>)
10601 Uses the string representation of the input sample to perform a look up in
10602 the specified table. If the key is not found in the table, integer value zero
10603 is returned. Otherwise the converter returns the current amount of concurrent
10604 connections tracking the same key as the input sample in the designated
10605 table. It differs from table_conn_cur in that it does not rely on any stored
10606 information but on the table's reference count (the "use" value which is
10607 returned by "show table" on the CLI). This may sometimes be more suited for
10608 layer7 tracking. It can be used to tell a server how many concurrent
10609 connections there are from a given address for example. See also the
10610 sc_trackers sample fetch keyword.
10611
Willy Tarreauffcb2e42014-07-10 16:29:08 +020010612upper
10613 Convert a string sample to upper case. This can only be placed after a string
10614 sample fetch function or after a transformation keyword returning a string
10615 type. The result is of type string.
10616
Willy Tarreau0dbfdba2014-07-10 16:37:47 +020010617utime(<format>[,<offset>])
10618 Converts an integer supposed to contain a date since epoch to a string
10619 representing this date in UTC time using a format defined by the <format>
10620 string using strftime(3). The purpose is to allow any date format to be used
10621 in logs. An optional <offset> in seconds may be applied to the input date
10622 (positive or negative). See the strftime() man page for the format supported
10623 by your operating system. See also the ltime converter.
10624
10625 Example :
10626
10627 # Emit two colons, one with the UTC time and another with ip:port
10628 # Eg: 20140710162350 127.0.0.1:57325
10629 log-format %[date,utime(%Y%m%d%H%M%S)]\ %ci:%cp
10630
Emeric Brunc9a0f6d2014-11-25 14:09:01 +010010631word(<index>,<delimiters>)
10632 Extracts the nth word considering given delimiters from an input string.
10633 Indexes start at 1 and delimiters are a string formatted list of chars.
10634
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010635wt6([<avalanche>])
10636 Hashes a binary input sample into an unsigned 32-bit quantity using the WT6
10637 hash function. Optionally, it is possible to apply a full avalanche hash
10638 function to the output if the optional <avalanche> argument equals 1. This
10639 converter uses the same functions as used by the various hash-based load
10640 balancing algorithms, so it will provide exactly the same results. It is
10641 mostly intended for debugging, but can be used as a stick-table entry to
10642 collect rough statistics. It must not be used for security purposes as a
Willy Tarreau80599772015-01-20 19:35:24 +010010643 32-bit hash is trivial to break. See also "crc32", "djb2", "sdbm", and the
10644 "hash-type" directive.
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020010645
Willy Tarreau97707872015-01-27 15:12:13 +010010646xor(<value>)
10647 Performs a bitwise "XOR" (exclusive OR) between <value> and the input value
10648 of type unsigned integer, and returns the result as an unsigned integer.
10649
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010010650
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200106517.3.2. Fetching samples from internal states
Willy Tarreau74ca5042013-06-11 23:12:07 +020010652--------------------------------------------
10653
10654A first set of sample fetch methods applies to internal information which does
10655not even relate to any client information. These ones are sometimes used with
10656"monitor-fail" directives to report an internal status to external watchers.
10657The sample fetch methods described in this section are usable anywhere.
10658
10659always_false : boolean
10660 Always returns the boolean "false" value. It may be used with ACLs as a
10661 temporary replacement for another one when adjusting configurations.
10662
10663always_true : boolean
10664 Always returns the boolean "true" value. It may be used with ACLs as a
10665 temporary replacement for another one when adjusting configurations.
10666
10667avg_queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010010668 Returns the total number of queued connections of the designated backend
Willy Tarreau74ca5042013-06-11 23:12:07 +020010669 divided by the number of active servers. The current backend is used if no
10670 backend is specified. This is very similar to "queue" except that the size of
10671 the farm is considered, in order to give a more accurate measurement of the
10672 time it may take for a new connection to be processed. The main usage is with
10673 ACL to return a sorry page to new users when it becomes certain they will get
10674 a degraded service, or to pass to the backend servers in a header so that
10675 they decide to work in degraded mode or to disable some functions to speed up
10676 the processing a bit. Note that in the event there would not be any active
10677 server anymore, twice the number of queued connections would be considered as
10678 the measured value. This is a fair estimate, as we expect one server to get
10679 back soon anyway, but we still prefer to send new traffic to another backend
10680 if in better shape. See also the "queue", "be_conn", and "be_sess_rate"
10681 sample fetches.
Krzysztof Piotr Oledzki346f76d2010-01-12 21:59:30 +010010682
Willy Tarreau74ca5042013-06-11 23:12:07 +020010683be_conn([<backend>]) : integer
Willy Tarreaua36af912009-10-10 12:02:45 +020010684 Applies to the number of currently established connections on the backend,
10685 possibly including the connection being evaluated. If no backend name is
10686 specified, the current one is used. But it is also possible to check another
10687 backend. It can be used to use a specific farm when the nominal one is full.
10688 See also the "fe_conn", "queue" and "be_sess_rate" criteria.
Willy Tarreau6a06a402007-07-15 20:15:28 +020010689
Willy Tarreau74ca5042013-06-11 23:12:07 +020010690be_sess_rate([<backend>]) : integer
10691 Returns an integer value corresponding to the sessions creation rate on the
10692 backend, in number of new sessions per second. This is used with ACLs to
10693 switch to an alternate backend when an expensive or fragile one reaches too
10694 high a session rate, or to limit abuse of service (eg. prevent sucking of an
10695 online dictionary). It can also be useful to add this element to logs using a
10696 log-format directive.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010697
10698 Example :
10699 # Redirect to an error page if the dictionary is requested too often
10700 backend dynamic
10701 mode http
10702 acl being_scanned be_sess_rate gt 100
10703 redirect location /denied.html if being_scanned
Willy Tarreau0ba27502007-12-24 16:55:16 +010010704
Willy Tarreau74ca5042013-06-11 23:12:07 +020010705connslots([<backend>]) : integer
10706 Returns an integer value corresponding to the number of connection slots
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010707 still available in the backend, by totaling the maximum amount of
Willy Tarreau74ca5042013-06-11 23:12:07 +020010708 connections on all servers and the maximum queue size. This is probably only
10709 used with ACLs.
Tait Clarridge7896d522012-12-05 21:39:31 -050010710
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080010711 The basic idea here is to be able to measure the number of connection "slots"
Willy Tarreau55165fe2009-05-10 12:02:55 +020010712 still available (connection + queue), so that anything beyond that (intended
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080010713 usage; see "use_backend" keyword) can be redirected to a different backend.
10714
Willy Tarreau55165fe2009-05-10 12:02:55 +020010715 'connslots' = number of available server connection slots, + number of
10716 available server queue slots.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080010717
Willy Tarreaua36af912009-10-10 12:02:45 +020010718 Note that while "fe_conn" may be used, "connslots" comes in especially
Willy Tarreau55165fe2009-05-10 12:02:55 +020010719 useful when you have a case of traffic going to one single ip, splitting into
Willy Tarreau74ca5042013-06-11 23:12:07 +020010720 multiple backends (perhaps using ACLs to do name-based load balancing) and
Willy Tarreau55165fe2009-05-10 12:02:55 +020010721 you want to be able to differentiate between different backends, and their
10722 available "connslots". Also, whereas "nbsrv" only measures servers that are
Willy Tarreau74ca5042013-06-11 23:12:07 +020010723 actually *down*, this fetch is more fine-grained and looks into the number of
Willy Tarreaua36af912009-10-10 12:02:45 +020010724 available connection slots as well. See also "queue" and "avg_queue".
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080010725
Willy Tarreau55165fe2009-05-10 12:02:55 +020010726 OTHER CAVEATS AND NOTES: at this point in time, the code does not take care
10727 of dynamic connections. Also, if any of the server maxconn, or maxqueue is 0,
Willy Tarreau74ca5042013-06-11 23:12:07 +020010728 then this fetch clearly does not make sense, in which case the value returned
Willy Tarreau55165fe2009-05-10 12:02:55 +020010729 will be -1.
Jeffrey 'jf' Lim5051d7b2008-09-04 01:03:03 +080010730
Willy Tarreau6236d3a2013-07-25 14:28:25 +020010731date([<offset>]) : integer
10732 Returns the current date as the epoch (number of seconds since 01/01/1970).
10733 If an offset value is specified, then it is a number of seconds that is added
10734 to the current date before returning the value. This is particularly useful
10735 to compute relative dates, as both positive and negative offsets are allowed.
Willy Tarreau276fae92013-07-25 14:36:01 +020010736 It is useful combined with the http_date converter.
10737
10738 Example :
10739
10740 # set an expires header to now+1 hour in every response
10741 http-response set-header Expires %[date(3600),http_date]
Willy Tarreau6236d3a2013-07-25 14:28:25 +020010742
Willy Tarreau595ec542013-06-12 21:34:28 +020010743env(<name>) : string
10744 Returns a string containing the value of environment variable <name>. As a
10745 reminder, environment variables are per-process and are sampled when the
10746 process starts. This can be useful to pass some information to a next hop
10747 server, or with ACLs to take specific action when the process is started a
10748 certain way.
10749
10750 Examples :
10751 # Pass the Via header to next hop with the local hostname in it
10752 http-request add-header Via 1.1\ %[env(HOSTNAME)]
10753
10754 # reject cookie-less requests when the STOP environment variable is set
10755 http-request deny if !{ cook(SESSIONID) -m found } { env(STOP) -m found }
10756
Willy Tarreau74ca5042013-06-11 23:12:07 +020010757fe_conn([<frontend>]) : integer
10758 Returns the number of currently established connections on the frontend,
Willy Tarreaud63335a2010-02-26 12:56:52 +010010759 possibly including the connection being evaluated. If no frontend name is
10760 specified, the current one is used. But it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020010761 frontend. It can be used to return a sorry page before hard-blocking, or to
10762 use a specific backend to drain new requests when the farm is considered
10763 full. This is mostly used with ACLs but can also be used to pass some
10764 statistics to servers in HTTP headers. See also the "dst_conn", "be_conn",
10765 "fe_sess_rate" fetches.
Willy Tarreaua36af912009-10-10 12:02:45 +020010766
Willy Tarreau74ca5042013-06-11 23:12:07 +020010767fe_sess_rate([<frontend>]) : integer
10768 Returns an integer value corresponding to the sessions creation rate on the
10769 frontend, in number of new sessions per second. This is used with ACLs to
10770 limit the incoming session rate to an acceptable range in order to prevent
10771 abuse of service at the earliest moment, for example when combined with other
10772 layer 4 ACLs in order to force the clients to wait a bit for the rate to go
10773 down below the limit. It can also be useful to add this element to logs using
10774 a log-format directive. See also the "rate-limit sessions" directive for use
10775 in frontends.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010010776
10777 Example :
10778 # This frontend limits incoming mails to 10/s with a max of 100
10779 # concurrent connections. We accept any connection below 10/s, and
10780 # force excess clients to wait for 100 ms. Since clients are limited to
10781 # 100 max, there cannot be more than 10 incoming mails per second.
10782 frontend mail
10783 bind :25
10784 mode tcp
10785 maxconn 100
10786 acl too_fast fe_sess_rate ge 10
10787 tcp-request inspect-delay 100ms
10788 tcp-request content accept if ! too_fast
10789 tcp-request content accept if WAIT_END
Willy Tarreaud72758d2010-01-12 10:42:19 +010010790
Willy Tarreau0f30d262014-11-24 16:02:05 +010010791nbproc : integer
10792 Returns an integer value corresponding to the number of processes that were
10793 started (it equals the global "nbproc" setting). This is useful for logging
10794 and debugging purposes.
10795
Willy Tarreau74ca5042013-06-11 23:12:07 +020010796nbsrv([<backend>]) : integer
10797 Returns an integer value corresponding to the number of usable servers of
10798 either the current backend or the named backend. This is mostly used with
10799 ACLs but can also be useful when added to logs. This is normally used to
Willy Tarreaud63335a2010-02-26 12:56:52 +010010800 switch to an alternate backend when the number of servers is too low to
10801 to handle some load. It is useful to report a failure when combined with
10802 "monitor fail".
Willy Tarreau079ff0a2009-03-05 21:34:28 +010010803
Willy Tarreau0f30d262014-11-24 16:02:05 +010010804proc : integer
10805 Returns an integer value corresponding to the position of the process calling
10806 the function, between 1 and global.nbproc. This is useful for logging and
10807 debugging purposes.
10808
Willy Tarreau74ca5042013-06-11 23:12:07 +020010809queue([<backend>]) : integer
Willy Tarreaud63335a2010-02-26 12:56:52 +010010810 Returns the total number of queued connections of the designated backend,
10811 including all the connections in server queues. If no backend name is
10812 specified, the current one is used, but it is also possible to check another
Willy Tarreau74ca5042013-06-11 23:12:07 +020010813 one. This is useful with ACLs or to pass statistics to backend servers. This
10814 can be used to take actions when queuing goes above a known level, generally
10815 indicating a surge of traffic or a massive slowdown on the servers. One
10816 possible action could be to reject new users but still accept old ones. See
10817 also the "avg_queue", "be_conn", and "be_sess_rate" fetches.
10818
Willy Tarreau84310e22014-02-14 11:59:04 +010010819rand([<range>]) : integer
10820 Returns a random integer value within a range of <range> possible values,
10821 starting at zero. If the range is not specified, it defaults to 2^32, which
10822 gives numbers between 0 and 4294967295. It can be useful to pass some values
10823 needed to take some routing decisions for example, or just for debugging
10824 purposes. This random must not be used for security purposes.
10825
Willy Tarreau74ca5042013-06-11 23:12:07 +020010826srv_conn([<backend>/]<server>) : integer
10827 Returns an integer value corresponding to the number of currently established
10828 connections on the designated server, possibly including the connection being
10829 evaluated. If <backend> is omitted, then the server is looked up in the
10830 current backend. It can be used to use a specific farm when one server is
10831 full, or to inform the server about our view of the number of active
10832 connections with it. See also the "fe_conn", "be_conn" and "queue" fetch
10833 methods.
10834
10835srv_is_up([<backend>/]<server>) : boolean
10836 Returns true when the designated server is UP, and false when it is either
10837 DOWN or in maintenance mode. If <backend> is omitted, then the server is
10838 looked up in the current backend. It is mainly used to take action based on
10839 an external status reported via a health check (eg: a geographical site's
10840 availability). Another possible use which is more of a hack consists in
10841 using dummy servers as boolean variables that can be enabled or disabled from
10842 the CLI, so that rules depending on those ACLs can be tweaked in realtime.
10843
10844srv_sess_rate([<backend>/]<server>) : integer
10845 Returns an integer corresponding to the sessions creation rate on the
10846 designated server, in number of new sessions per second. If <backend> is
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030010847 omitted, then the server is looked up in the current backend. This is mostly
Willy Tarreau74ca5042013-06-11 23:12:07 +020010848 used with ACLs but can make sense with logs too. This is used to switch to an
10849 alternate backend when an expensive or fragile one reaches too high a session
10850 rate, or to limit abuse of service (eg. prevent latent requests from
10851 overloading servers).
10852
10853 Example :
10854 # Redirect to a separate back
10855 acl srv1_full srv_sess_rate(be1/srv1) gt 50
10856 acl srv2_full srv_sess_rate(be1/srv2) gt 50
10857 use_backend be2 if srv1_full or srv2_full
10858
Willy Tarreau0f30d262014-11-24 16:02:05 +010010859stopping : boolean
10860 Returns TRUE if the process calling the function is currently stopping. This
10861 can be useful for logging, or for relaxing certain checks or helping close
10862 certain connections upon graceful shutdown.
10863
Willy Tarreau74ca5042013-06-11 23:12:07 +020010864table_avl([<table>]) : integer
10865 Returns the total number of available entries in the current proxy's
10866 stick-table or in the designated stick-table. See also table_cnt.
10867
10868table_cnt([<table>]) : integer
10869 Returns the total number of entries currently in use in the current proxy's
10870 stick-table or in the designated stick-table. See also src_conn_cnt and
10871 table_avl for other entry counting methods.
10872
10873
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200108747.3.3. Fetching samples at Layer 4
Willy Tarreau74ca5042013-06-11 23:12:07 +020010875----------------------------------
10876
10877The layer 4 usually describes just the transport layer which in haproxy is
10878closest to the connection, where no content is yet made available. The fetch
10879methods described here are usable as low as the "tcp-request connection" rule
10880sets unless they require some future information. Those generally include
10881TCP/IP addresses and ports, as well as elements from stick-tables related to
Willy Tarreau4d4149c2013-07-23 19:33:46 +020010882the incoming connection. For retrieving a value from a sticky counters, the
10883counter number can be explicitly set as 0, 1, or 2 using the pre-defined
10884"sc0_", "sc1_", or "sc2_" prefix, or it can be specified as the first integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010885argument when using the "sc_" prefix. An optional table may be specified with
10886the "sc*" form, in which case the currently tracked key will be looked up into
10887this alternate table instead of the table currently being tracked.
Willy Tarreau74ca5042013-06-11 23:12:07 +020010888
10889be_id : integer
10890 Returns an integer containing the current backend's id. It can be used in
10891 frontends with responses to check which backend processed the request.
10892
10893dst : ip
10894 This is the destination IPv4 address of the connection on the client side,
10895 which is the address the client connected to. It can be useful when running
10896 in transparent mode. It is of type IP and works on both IPv4 and IPv6 tables.
10897 On IPv6 tables, IPv4 address is mapped to its IPv6 equivalent, according to
10898 RFC 4291.
10899
10900dst_conn : integer
10901 Returns an integer value corresponding to the number of currently established
10902 connections on the same socket including the one being evaluated. It is
10903 normally used with ACLs but can as well be used to pass the information to
10904 servers in an HTTP header or in logs. It can be used to either return a sorry
10905 page before hard-blocking, or to use a specific backend to drain new requests
10906 when the socket is considered saturated. This offers the ability to assign
10907 different limits to different listening ports or addresses. See also the
10908 "fe_conn" and "be_conn" fetches.
Willy Tarreaud63335a2010-02-26 12:56:52 +010010909
Willy Tarreau74ca5042013-06-11 23:12:07 +020010910dst_port : integer
10911 Returns an integer value corresponding to the destination TCP port of the
10912 connection on the client side, which is the port the client connected to.
10913 This might be used when running in transparent mode, when assigning dynamic
10914 ports to some clients for a whole application session, to stick all users to
10915 a same server, or to pass the destination port information to a server using
10916 an HTTP header.
10917
10918fe_id : integer
10919 Returns an integer containing the current frontend's id. It can be used in
10920 backends to check from which backend it was called, or to stick all users
10921 coming via a same frontend to the same server.
10922
Cyril Bonté62ba8702014-04-22 23:52:25 +020010923sc_bytes_in_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010924sc0_bytes_in_rate([<table>]) : integer
10925sc1_bytes_in_rate([<table>]) : integer
10926sc2_bytes_in_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010927 Returns the average client-to-server bytes rate from the currently tracked
10928 counters, measured in amount of bytes over the period configured in the
10929 table. See also src_bytes_in_rate.
10930
Cyril Bonté62ba8702014-04-22 23:52:25 +020010931sc_bytes_out_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010932sc0_bytes_out_rate([<table>]) : integer
10933sc1_bytes_out_rate([<table>]) : integer
10934sc2_bytes_out_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010935 Returns the average server-to-client bytes rate from the currently tracked
10936 counters, measured in amount of bytes over the period configured in the
10937 table. See also src_bytes_out_rate.
10938
Cyril Bonté62ba8702014-04-22 23:52:25 +020010939sc_clr_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010940sc0_clr_gpc0([<table>]) : integer
10941sc1_clr_gpc0([<table>]) : integer
10942sc2_clr_gpc0([<table>]) : integer
Willy Tarreauf73cd112011-08-13 01:45:16 +020010943 Clears the first General Purpose Counter associated to the currently tracked
10944 counters, and returns its previous value. Before the first invocation, the
Willy Tarreau869948b2013-01-04 14:14:57 +010010945 stored value is zero, so first invocation will always return zero. This is
10946 typically used as a second ACL in an expression in order to mark a connection
10947 when a first ACL was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020010948
10949 # block if 5 consecutive requests continue to come faster than 10 sess
10950 # per second, and reset the counter as soon as the traffic slows down.
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020010951 acl abuse sc0_http_req_rate gt 10
10952 acl kill sc0_inc_gpc0 gt 5
10953 acl save sc0_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020010954 tcp-request connection accept if !abuse save
10955 tcp-request connection reject if abuse kill
10956
Cyril Bonté62ba8702014-04-22 23:52:25 +020010957sc_conn_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010958sc0_conn_cnt([<table>]) : integer
10959sc1_conn_cnt([<table>]) : integer
10960sc2_conn_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010961 Returns the cumulated number of incoming connections from currently tracked
10962 counters. See also src_conn_cnt.
10963
Cyril Bonté62ba8702014-04-22 23:52:25 +020010964sc_conn_cur(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010965sc0_conn_cur([<table>]) : integer
10966sc1_conn_cur([<table>]) : integer
10967sc2_conn_cur([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010968 Returns the current amount of concurrent connections tracking the same
10969 tracked counters. This number is automatically incremented when tracking
10970 begins and decremented when tracking stops. See also src_conn_cur.
10971
Cyril Bonté62ba8702014-04-22 23:52:25 +020010972sc_conn_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010973sc0_conn_rate([<table>]) : integer
10974sc1_conn_rate([<table>]) : integer
10975sc2_conn_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010976 Returns the average connection rate from the currently tracked counters,
10977 measured in amount of connections over the period configured in the table.
10978 See also src_conn_rate.
10979
Cyril Bonté62ba8702014-04-22 23:52:25 +020010980sc_get_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010981sc0_get_gpc0([<table>]) : integer
10982sc1_get_gpc0([<table>]) : integer
10983sc2_get_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020010984 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020010985 currently tracked counters. See also src_get_gpc0 and sc/sc0/sc1/sc2_inc_gpc0.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020010986
Cyril Bonté62ba8702014-04-22 23:52:25 +020010987sc_gpc0_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010988sc0_gpc0_rate([<table>]) : integer
10989sc1_gpc0_rate([<table>]) : integer
10990sc2_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020010991 Returns the average increment rate of the first General Purpose Counter
10992 associated to the currently tracked counters. It reports the frequency
10993 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020010994 src_gpc0_rate, sc/sc0/sc1/sc2_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
10995 that the "gpc0_rate" counter must be stored in the stick-table for a value to
10996 be returned, as "gpc0" only holds the event count.
Willy Tarreaue9656522010-08-17 15:40:09 +020010997
Cyril Bonté62ba8702014-04-22 23:52:25 +020010998sc_http_err_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020010999sc0_http_err_cnt([<table>]) : integer
11000sc1_http_err_cnt([<table>]) : integer
11001sc2_http_err_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011002 Returns the cumulated number of HTTP errors from the currently tracked
11003 counters. This includes the both request errors and 4xx error responses.
11004 See also src_http_err_cnt.
11005
Cyril Bonté62ba8702014-04-22 23:52:25 +020011006sc_http_err_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011007sc0_http_err_rate([<table>]) : integer
11008sc1_http_err_rate([<table>]) : integer
11009sc2_http_err_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011010 Returns the average rate of HTTP errors from the currently tracked counters,
11011 measured in amount of errors over the period configured in the table. This
11012 includes the both request errors and 4xx error responses. See also
11013 src_http_err_rate.
11014
Cyril Bonté62ba8702014-04-22 23:52:25 +020011015sc_http_req_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011016sc0_http_req_cnt([<table>]) : integer
11017sc1_http_req_cnt([<table>]) : integer
11018sc2_http_req_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011019 Returns the cumulated number of HTTP requests from the currently tracked
11020 counters. This includes every started request, valid or not. See also
11021 src_http_req_cnt.
11022
Cyril Bonté62ba8702014-04-22 23:52:25 +020011023sc_http_req_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011024sc0_http_req_rate([<table>]) : integer
11025sc1_http_req_rate([<table>]) : integer
11026sc2_http_req_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011027 Returns the average rate of HTTP requests from the currently tracked
11028 counters, measured in amount of requests over the period configured in
11029 the table. This includes every started request, valid or not. See also
11030 src_http_req_rate.
11031
Cyril Bonté62ba8702014-04-22 23:52:25 +020011032sc_inc_gpc0(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011033sc0_inc_gpc0([<table>]) : integer
11034sc1_inc_gpc0([<table>]) : integer
11035sc2_inc_gpc0([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011036 Increments the first General Purpose Counter associated to the currently
Willy Tarreau869948b2013-01-04 14:14:57 +010011037 tracked counters, and returns its new value. Before the first invocation,
11038 the stored value is zero, so first invocation will increase it to 1 and will
11039 return 1. This is typically used as a second ACL in an expression in order
11040 to mark a connection when a first ACL was verified :
Willy Tarreaue9656522010-08-17 15:40:09 +020011041
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011042 acl abuse sc0_http_req_rate gt 10
11043 acl kill sc0_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020011044 tcp-request connection reject if abuse kill
11045
Cyril Bonté62ba8702014-04-22 23:52:25 +020011046sc_kbytes_in(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011047sc0_kbytes_in([<table>]) : integer
11048sc1_kbytes_in([<table>]) : integer
11049sc2_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011050 Returns the total amount of client-to-server data from the currently tracked
11051 counters, measured in kilobytes. The test is currently performed on 32-bit
11052 integers, which limits values to 4 terabytes. See also src_kbytes_in.
Willy Tarreaue9656522010-08-17 15:40:09 +020011053
Cyril Bonté62ba8702014-04-22 23:52:25 +020011054sc_kbytes_out(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011055sc0_kbytes_out([<table>]) : integer
11056sc1_kbytes_out([<table>]) : integer
11057sc2_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011058 Returns the total amount of server-to-client data from the currently tracked
11059 counters, measured in kilobytes. The test is currently performed on 32-bit
11060 integers, which limits values to 4 terabytes. See also src_kbytes_out.
Willy Tarreaue9656522010-08-17 15:40:09 +020011061
Cyril Bonté62ba8702014-04-22 23:52:25 +020011062sc_sess_cnt(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011063sc0_sess_cnt([<table>]) : integer
11064sc1_sess_cnt([<table>]) : integer
11065sc2_sess_cnt([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011066 Returns the cumulated number of incoming connections that were transformed
11067 into sessions, which means that they were accepted by a "tcp-request
11068 connection" rule, from the currently tracked counters. A backend may count
11069 more sessions than connections because each connection could result in many
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011070 backend sessions if some HTTP keep-alive is performed over the connection
Willy Tarreaue9656522010-08-17 15:40:09 +020011071 with the client. See also src_sess_cnt.
11072
Cyril Bonté62ba8702014-04-22 23:52:25 +020011073sc_sess_rate(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011074sc0_sess_rate([<table>]) : integer
11075sc1_sess_rate([<table>]) : integer
11076sc2_sess_rate([<table>]) : integer
Willy Tarreaue9656522010-08-17 15:40:09 +020011077 Returns the average session rate from the currently tracked counters,
11078 measured in amount of sessions over the period configured in the table. A
11079 session is a connection that got past the early "tcp-request connection"
11080 rules. A backend may count more sessions than connections because each
11081 connection could result in many backend sessions if some HTTP keep-alive is
Jamie Gloudonaaa21002012-08-25 00:18:33 -040011082 performed over the connection with the client. See also src_sess_rate.
Willy Tarreaue9656522010-08-17 15:40:09 +020011083
Cyril Bonté62ba8702014-04-22 23:52:25 +020011084sc_tracked(<ctr>[,<table>]) : boolean
Willy Tarreau0f791d42013-07-23 19:56:43 +020011085sc0_tracked([<table>]) : boolean
11086sc1_tracked([<table>]) : boolean
11087sc2_tracked([<table>]) : boolean
Willy Tarreau6f1615f2013-06-03 15:15:22 +020011088 Returns true if the designated session counter is currently being tracked by
11089 the current session. This can be useful when deciding whether or not we want
11090 to set some values in a header passed to the server.
11091
Cyril Bonté62ba8702014-04-22 23:52:25 +020011092sc_trackers(<ctr>[,<table>]) : integer
Willy Tarreau0f791d42013-07-23 19:56:43 +020011093sc0_trackers([<table>]) : integer
11094sc1_trackers([<table>]) : integer
11095sc2_trackers([<table>]) : integer
Willy Tarreau2406db42012-12-09 12:16:43 +010011096 Returns the current amount of concurrent connections tracking the same
11097 tracked counters. This number is automatically incremented when tracking
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011098 begins and decremented when tracking stops. It differs from sc0_conn_cur in
Willy Tarreau2406db42012-12-09 12:16:43 +010011099 that it does not rely on any stored information but on the table's reference
11100 count (the "use" value which is returned by "show table" on the CLI). This
Willy Tarreau74ca5042013-06-11 23:12:07 +020011101 may sometimes be more suited for layer7 tracking. It can be used to tell a
11102 server how many concurrent connections there are from a given address for
11103 example.
Willy Tarreau2406db42012-12-09 12:16:43 +010011104
Willy Tarreau74ca5042013-06-11 23:12:07 +020011105so_id : integer
11106 Returns an integer containing the current listening socket's id. It is useful
11107 in frontends involving many "bind" lines, or to stick all users coming via a
11108 same socket to the same server.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011109
Willy Tarreau74ca5042013-06-11 23:12:07 +020011110src : ip
11111 This is the source IPv4 address of the client of the session. It is of type
11112 IP and works on both IPv4 and IPv6 tables. On IPv6 tables, IPv4 addresses are
11113 mapped to their IPv6 equivalent, according to RFC 4291. Note that it is the
11114 TCP-level source address which is used, and not the address of a client
11115 behind a proxy. However if the "accept-proxy" bind directive is used, it can
11116 be the address of a client behind another PROXY-protocol compatible component
11117 for all rule sets except "tcp-request connection" which sees the real address.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011118
Thierry FOURNIERd5f624d2013-11-26 11:52:33 +010011119 Example:
11120 # add an HTTP header in requests with the originating address' country
11121 http-request set-header X-Country %[src,map_ip(geoip.lst)]
11122
Willy Tarreau74ca5042013-06-11 23:12:07 +020011123src_bytes_in_rate([<table>]) : integer
11124 Returns the average bytes rate from the incoming connection's source address
11125 in the current proxy's stick-table or in the designated stick-table, measured
11126 in amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011127 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_in_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011128
Willy Tarreau74ca5042013-06-11 23:12:07 +020011129src_bytes_out_rate([<table>]) : integer
11130 Returns the average bytes rate to the incoming connection's source address in
11131 the current proxy's stick-table or in the designated stick-table, measured in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011132 amount of bytes over the period configured in the table. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011133 not found, zero is returned. See also sc/sc0/sc1/sc2_bytes_out_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011134
Willy Tarreau74ca5042013-06-11 23:12:07 +020011135src_clr_gpc0([<table>]) : integer
11136 Clears the first General Purpose Counter associated to the incoming
11137 connection's source address in the current proxy's stick-table or in the
11138 designated stick-table, and returns its previous value. If the address is not
11139 found, an entry is created and 0 is returned. This is typically used as a
11140 second ACL in an expression in order to mark a connection when a first ACL
11141 was verified :
Willy Tarreauf73cd112011-08-13 01:45:16 +020011142
11143 # block if 5 consecutive requests continue to come faster than 10 sess
11144 # per second, and reset the counter as soon as the traffic slows down.
11145 acl abuse src_http_req_rate gt 10
11146 acl kill src_inc_gpc0 gt 5
Willy Tarreau869948b2013-01-04 14:14:57 +010011147 acl save src_clr_gpc0 ge 0
Willy Tarreauf73cd112011-08-13 01:45:16 +020011148 tcp-request connection accept if !abuse save
11149 tcp-request connection reject if abuse kill
11150
Willy Tarreau74ca5042013-06-11 23:12:07 +020011151src_conn_cnt([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011152 Returns the cumulated number of connections initiated from the current
Willy Tarreau74ca5042013-06-11 23:12:07 +020011153 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011154 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011155 See also sc/sc0/sc1/sc2_conn_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011156
Willy Tarreau74ca5042013-06-11 23:12:07 +020011157src_conn_cur([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011158 Returns the current amount of concurrent connections initiated from the
Willy Tarreau74ca5042013-06-11 23:12:07 +020011159 current incoming connection's source address in the current proxy's
11160 stick-table or in the designated stick-table. If the address is not found,
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011161 zero is returned. See also sc/sc0/sc1/sc2_conn_cur.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011162
Willy Tarreau74ca5042013-06-11 23:12:07 +020011163src_conn_rate([<table>]) : integer
11164 Returns the average connection rate from the incoming connection's source
11165 address in the current proxy's stick-table or in the designated stick-table,
11166 measured in amount of connections over the period configured in the table. If
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011167 the address is not found, zero is returned. See also sc/sc0/sc1/sc2_conn_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011168
Willy Tarreau74ca5042013-06-11 23:12:07 +020011169src_get_gpc0([<table>]) : integer
Willy Tarreauc9705a12010-07-27 20:05:50 +020011170 Returns the value of the first General Purpose Counter associated to the
Willy Tarreau74ca5042013-06-11 23:12:07 +020011171 incoming connection's source address in the current proxy's stick-table or in
Willy Tarreauc9705a12010-07-27 20:05:50 +020011172 the designated stick-table. If the address is not found, zero is returned.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011173 See also sc/sc0/sc1/sc2_get_gpc0 and src_inc_gpc0.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011174
Willy Tarreau74ca5042013-06-11 23:12:07 +020011175src_gpc0_rate([<table>]) : integer
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011176 Returns the average increment rate of the first General Purpose Counter
Willy Tarreau74ca5042013-06-11 23:12:07 +020011177 associated to the incoming connection's source address in the current proxy's
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011178 stick-table or in the designated stick-table. It reports the frequency
11179 which the gpc0 counter was incremented over the configured period. See also
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011180 sc/sc0/sc1/sc2_gpc0_rate, src_get_gpc0, and sc/sc0/sc1/sc2_inc_gpc0. Note
11181 that the "gpc0_rate" counter must be stored in the stick-table for a value to
11182 be returned, as "gpc0" only holds the event count.
Willy Tarreauba2ffd12013-05-29 15:54:14 +020011183
Willy Tarreau74ca5042013-06-11 23:12:07 +020011184src_http_err_cnt([<table>]) : integer
11185 Returns the cumulated number of HTTP errors from the incoming connection's
11186 source address in the current proxy's stick-table or in the designated
Willy Tarreauc9705a12010-07-27 20:05:50 +020011187 stick-table. This includes the both request errors and 4xx error responses.
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011188 See also sc/sc0/sc1/sc2_http_err_cnt. If the address is not found, zero is
Willy Tarreau74ca5042013-06-11 23:12:07 +020011189 returned.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011190
Willy Tarreau74ca5042013-06-11 23:12:07 +020011191src_http_err_rate([<table>]) : integer
11192 Returns the average rate of HTTP errors from the incoming connection's source
11193 address in the current proxy's stick-table or in the designated stick-table,
11194 measured in amount of errors over the period configured in the table. This
11195 includes the both request errors and 4xx error responses. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011196 not found, zero is returned. See also sc/sc0/sc1/sc2_http_err_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011197
Willy Tarreau74ca5042013-06-11 23:12:07 +020011198src_http_req_cnt([<table>]) : integer
11199 Returns the cumulated number of HTTP requests from the incoming connection's
11200 source address in the current proxy's stick-table or in the designated stick-
11201 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011202 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011203
Willy Tarreau74ca5042013-06-11 23:12:07 +020011204src_http_req_rate([<table>]) : integer
11205 Returns the average rate of HTTP requests from the incoming connection's
11206 source address in the current proxy's stick-table or in the designated stick-
11207 table, measured in amount of requests over the period configured in the
Willy Tarreauc9705a12010-07-27 20:05:50 +020011208 table. This includes every started request, valid or not. If the address is
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011209 not found, zero is returned. See also sc/sc0/sc1/sc2_http_req_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011210
Willy Tarreau74ca5042013-06-11 23:12:07 +020011211src_inc_gpc0([<table>]) : integer
11212 Increments the first General Purpose Counter associated to the incoming
11213 connection's source address in the current proxy's stick-table or in the
11214 designated stick-table, and returns its new value. If the address is not
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011215 found, an entry is created and 1 is returned. See also sc0/sc2/sc2_inc_gpc0.
Willy Tarreau74ca5042013-06-11 23:12:07 +020011216 This is typically used as a second ACL in an expression in order to mark a
11217 connection when a first ACL was verified :
Willy Tarreauc9705a12010-07-27 20:05:50 +020011218
11219 acl abuse src_http_req_rate gt 10
Willy Tarreau869948b2013-01-04 14:14:57 +010011220 acl kill src_inc_gpc0 gt 0
Willy Tarreaue9656522010-08-17 15:40:09 +020011221 tcp-request connection reject if abuse kill
Willy Tarreauc9705a12010-07-27 20:05:50 +020011222
Willy Tarreau74ca5042013-06-11 23:12:07 +020011223src_kbytes_in([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011224 Returns the total amount of data received from the incoming connection's
11225 source address in the current proxy's stick-table or in the designated
11226 stick-table, measured in kilobytes. If the address is not found, zero is
11227 returned. The test is currently performed on 32-bit integers, which limits
11228 values to 4 terabytes. See also sc/sc0/sc1/sc2_kbytes_in.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011229
Willy Tarreau74ca5042013-06-11 23:12:07 +020011230src_kbytes_out([<table>]) : integer
Willy Tarreaua01b9742014-07-10 15:29:24 +020011231 Returns the total amount of data sent to the incoming connection's source
11232 address in the current proxy's stick-table or in the designated stick-table,
11233 measured in kilobytes. If the address is not found, zero is returned. The
11234 test is currently performed on 32-bit integers, which limits values to 4
11235 terabytes. See also sc/sc0/sc1/sc2_kbytes_out.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011236
Willy Tarreau74ca5042013-06-11 23:12:07 +020011237src_port : integer
11238 Returns an integer value corresponding to the TCP source port of the
11239 connection on the client side, which is the port the client connected from.
11240 Usage of this function is very limited as modern protocols do not care much
11241 about source ports nowadays.
Willy Tarreau079ff0a2009-03-05 21:34:28 +010011242
Willy Tarreau74ca5042013-06-11 23:12:07 +020011243src_sess_cnt([<table>]) : integer
11244 Returns the cumulated number of connections initiated from the incoming
Willy Tarreauc9705a12010-07-27 20:05:50 +020011245 connection's source IPv4 address in the current proxy's stick-table or in the
11246 designated stick-table, that were transformed into sessions, which means that
11247 they were accepted by "tcp-request" rules. If the address is not found, zero
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011248 is returned. See also sc/sc0/sc1/sc2_sess_cnt.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011249
Willy Tarreau74ca5042013-06-11 23:12:07 +020011250src_sess_rate([<table>]) : integer
11251 Returns the average session rate from the incoming connection's source
11252 address in the current proxy's stick-table or in the designated stick-table,
11253 measured in amount of sessions over the period configured in the table. A
11254 session is a connection that went past the early "tcp-request" rules. If the
Willy Tarreau4d4149c2013-07-23 19:33:46 +020011255 address is not found, zero is returned. See also sc/sc0/sc1/sc2_sess_rate.
Willy Tarreauc9705a12010-07-27 20:05:50 +020011256
Willy Tarreau74ca5042013-06-11 23:12:07 +020011257src_updt_conn_cnt([<table>]) : integer
11258 Creates or updates the entry associated to the incoming connection's source
11259 address in the current proxy's stick-table or in the designated stick-table.
11260 This table must be configured to store the "conn_cnt" data type, otherwise
11261 the match will be ignored. The current count is incremented by one, and the
11262 expiration timer refreshed. The updated count is returned, so this match
11263 can't return zero. This was used to reject service abusers based on their
11264 source address. Note: it is recommended to use the more complete "track-sc*"
11265 actions in "tcp-request" rules instead.
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011266
11267 Example :
11268 # This frontend limits incoming SSH connections to 3 per 10 second for
11269 # each source address, and rejects excess connections until a 10 second
11270 # silence is observed. At most 20 addresses are tracked.
11271 listen ssh
11272 bind :22
11273 mode tcp
11274 maxconn 100
Willy Tarreauc9705a12010-07-27 20:05:50 +020011275 stick-table type ip size 20 expire 10s store conn_cnt
Willy Tarreau74ca5042013-06-11 23:12:07 +020011276 tcp-request content reject if { src_updt_conn_cnt gt 3 }
Willy Tarreaua975b8f2010-06-05 19:13:27 +020011277 server local 127.0.0.1:22
11278
Willy Tarreau74ca5042013-06-11 23:12:07 +020011279srv_id : integer
11280 Returns an integer containing the server's id when processing the response.
11281 While it's almost only used with ACLs, it may be used for logging or
11282 debugging.
Hervé COMMOWICKdaa824e2011-08-05 12:09:44 +020011283
Hervé COMMOWICK35ed8012010-12-15 14:04:51 +010011284
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200112857.3.4. Fetching samples at Layer 5
Willy Tarreau74ca5042013-06-11 23:12:07 +020011286----------------------------------
Willy Tarreau0b1cd942010-05-16 22:18:27 +020011287
Willy Tarreau74ca5042013-06-11 23:12:07 +020011288The layer 5 usually describes just the session layer which in haproxy is
11289closest to the session once all the connection handshakes are finished, but
11290when no content is yet made available. The fetch methods described here are
11291usable as low as the "tcp-request content" rule sets unless they require some
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011292future information. Those generally include the results of SSL negotiations.
Willy Tarreauc735a072011-03-29 00:57:02 +020011293
Emeric Brun645ae792014-04-30 14:21:06 +020011294ssl_bc : boolean
11295 Returns true when the back connection was made via an SSL/TLS transport
11296 layer and is locally deciphered. This means the outgoing connection was made
11297 other a server with the "ssl" option.
11298
11299ssl_bc_alg_keysize : integer
11300 Returns the symmetric cipher key size supported in bits when the outgoing
11301 connection was made over an SSL/TLS transport layer.
11302
11303ssl_bc_cipher : string
11304 Returns the name of the used cipher when the outgoing connection was made
11305 over an SSL/TLS transport layer.
11306
11307ssl_bc_protocol : string
11308 Returns the name of the used protocol when the outgoing connection was made
11309 over an SSL/TLS transport layer.
11310
Emeric Brunb73a9b02014-04-30 18:49:19 +020011311ssl_bc_unique_id : binary
Emeric Brun645ae792014-04-30 14:21:06 +020011312 When the outgoing connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020011313 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
11314 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
Emeric Brun645ae792014-04-30 14:21:06 +020011315
11316ssl_bc_session_id : binary
11317 Returns the SSL ID of the back connection when the outgoing connection was
11318 made over an SSL/TLS transport layer. It is useful to log if we want to know
11319 if session was reused or not.
11320
11321ssl_bc_use_keysize : integer
11322 Returns the symmetric cipher key size used in bits when the outgoing
11323 connection was made over an SSL/TLS transport layer.
11324
Willy Tarreau74ca5042013-06-11 23:12:07 +020011325ssl_c_ca_err : integer
11326 When the incoming connection was made over an SSL/TLS transport layer,
11327 returns the ID of the first error detected during verification of the client
11328 certificate at depth > 0, or 0 if no error was encountered during this
11329 verification process. Please refer to your SSL library's documentation to
11330 find the exhaustive list of error codes.
Willy Tarreauc735a072011-03-29 00:57:02 +020011331
Willy Tarreau74ca5042013-06-11 23:12:07 +020011332ssl_c_ca_err_depth : integer
11333 When the incoming connection was made over an SSL/TLS transport layer,
11334 returns the depth in the CA chain of the first error detected during the
11335 verification of the client certificate. If no error is encountered, 0 is
11336 returned.
Willy Tarreau0ba27502007-12-24 16:55:16 +010011337
Emeric Brun43e79582014-10-29 19:03:26 +010011338ssl_c_der : binary
11339 Returns the DER formatted certificate presented by the client when the
11340 incoming connection was made over an SSL/TLS transport layer. When used for
11341 an ACL, the value(s) to match against can be passed in hexadecimal form.
11342
Willy Tarreau74ca5042013-06-11 23:12:07 +020011343ssl_c_err : integer
11344 When the incoming connection was made over an SSL/TLS transport layer,
11345 returns the ID of the first error detected during verification at depth 0, or
11346 0 if no error was encountered during this verification process. Please refer
11347 to your SSL library's documentation to find the exhaustive list of error
11348 codes.
Willy Tarreau62644772008-07-16 18:36:06 +020011349
Willy Tarreau74ca5042013-06-11 23:12:07 +020011350ssl_c_i_dn([<entry>[,<occ>]]) : string
11351 When the incoming connection was made over an SSL/TLS transport layer,
11352 returns the full distinguished name of the issuer of the certificate
11353 presented by the client when no <entry> is specified, or the value of the
11354 first given entry found from the beginning of the DN. If a positive/negative
11355 occurrence number is specified as the optional second argument, it returns
11356 the value of the nth given entry value from the beginning/end of the DN.
11357 For instance, "ssl_c_i_dn(OU,2)" the second organization unit, and
11358 "ssl_c_i_dn(CN)" retrieves the common name.
Willy Tarreau62644772008-07-16 18:36:06 +020011359
Willy Tarreau74ca5042013-06-11 23:12:07 +020011360ssl_c_key_alg : string
11361 Returns the name of the algorithm used to generate the key of the certificate
11362 presented by the client when the incoming connection was made over an SSL/TLS
11363 transport layer.
Willy Tarreau62644772008-07-16 18:36:06 +020011364
Willy Tarreau74ca5042013-06-11 23:12:07 +020011365ssl_c_notafter : string
11366 Returns the end date presented by the client as a formatted string
11367 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11368 transport layer.
Emeric Brunbede3d02009-06-30 17:54:00 +020011369
Willy Tarreau74ca5042013-06-11 23:12:07 +020011370ssl_c_notbefore : string
11371 Returns the start date presented by the client as a formatted string
11372 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11373 transport layer.
Willy Tarreaub6672b52011-12-12 17:23:41 +010011374
Willy Tarreau74ca5042013-06-11 23:12:07 +020011375ssl_c_s_dn([<entry>[,<occ>]]) : string
11376 When the incoming connection was made over an SSL/TLS transport layer,
11377 returns the full distinguished name of the subject of the certificate
11378 presented by the client when no <entry> is specified, or the value of the
11379 first given entry found from the beginning of the DN. If a positive/negative
11380 occurrence number is specified as the optional second argument, it returns
11381 the value of the nth given entry value from the beginning/end of the DN.
11382 For instance, "ssl_c_s_dn(OU,2)" the second organization unit, and
11383 "ssl_c_s_dn(CN)" retrieves the common name.
Willy Tarreaub6672b52011-12-12 17:23:41 +010011384
Willy Tarreau74ca5042013-06-11 23:12:07 +020011385ssl_c_serial : binary
11386 Returns the serial of the certificate presented by the client when the
11387 incoming connection was made over an SSL/TLS transport layer. When used for
11388 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011389
Willy Tarreau74ca5042013-06-11 23:12:07 +020011390ssl_c_sha1 : binary
11391 Returns the SHA-1 fingerprint of the certificate presented by the client when
11392 the incoming connection was made over an SSL/TLS transport layer. This can be
11393 used to stick a client to a server, or to pass this information to a server.
Willy Tarreau2d0caa32014-07-02 19:01:22 +020011394 Note that the output is binary, so if you want to pass that signature to the
11395 server, you need to encode it in hex or base64, such as in the example below:
11396
11397 http-request set-header X-SSL-Client-SHA1 %[ssl_c_sha1,hex]
Emeric Brun2525b6b2012-10-18 15:59:43 +020011398
Willy Tarreau74ca5042013-06-11 23:12:07 +020011399ssl_c_sig_alg : string
11400 Returns the name of the algorithm used to sign the certificate presented by
11401 the client when the incoming connection was made over an SSL/TLS transport
11402 layer.
Emeric Brun87855892012-10-17 17:39:35 +020011403
Willy Tarreau74ca5042013-06-11 23:12:07 +020011404ssl_c_used : boolean
11405 Returns true if current SSL session uses a client certificate even if current
11406 connection uses SSL session resumption. See also "ssl_fc_has_crt".
Emeric Brun7f56e742012-10-19 18:15:40 +020011407
Willy Tarreau74ca5042013-06-11 23:12:07 +020011408ssl_c_verify : integer
11409 Returns the verify result error ID when the incoming connection was made over
11410 an SSL/TLS transport layer, otherwise zero if no error is encountered. Please
11411 refer to your SSL library's documentation for an exhaustive list of error
11412 codes.
Emeric Brunce5ad802012-10-22 14:11:22 +020011413
Willy Tarreau74ca5042013-06-11 23:12:07 +020011414ssl_c_version : integer
11415 Returns the version of the certificate presented by the client when the
11416 incoming connection was made over an SSL/TLS transport layer.
Emeric Brunce5ad802012-10-22 14:11:22 +020011417
Emeric Brun43e79582014-10-29 19:03:26 +010011418ssl_f_der : binary
11419 Returns the DER formatted certificate presented by the frontend when the
11420 incoming connection was made over an SSL/TLS transport layer. When used for
11421 an ACL, the value(s) to match against can be passed in hexadecimal form.
11422
Willy Tarreau74ca5042013-06-11 23:12:07 +020011423ssl_f_i_dn([<entry>[,<occ>]]) : string
11424 When the incoming connection was made over an SSL/TLS transport layer,
11425 returns the full distinguished name of the issuer of the certificate
11426 presented by the frontend when no <entry> is specified, or the value of the
11427 first given entry found from the beginning of the DN. If a positive/negative
Emeric Brun87855892012-10-17 17:39:35 +020011428 occurrence number is specified as the optional second argument, it returns
Willy Tarreau74ca5042013-06-11 23:12:07 +020011429 the value of the nth given entry value from the beginning/end of the DN.
11430 For instance, "ssl_f_i_dn(OU,2)" the second organization unit, and
11431 "ssl_f_i_dn(CN)" retrieves the common name.
Emeric Brun87855892012-10-17 17:39:35 +020011432
Willy Tarreau74ca5042013-06-11 23:12:07 +020011433ssl_f_key_alg : string
11434 Returns the name of the algorithm used to generate the key of the certificate
11435 presented by the frontend when the incoming connection was made over an
11436 SSL/TLS transport layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020011437
Willy Tarreau74ca5042013-06-11 23:12:07 +020011438ssl_f_notafter : string
11439 Returns the end date presented by the frontend as a formatted string
11440 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11441 transport layer.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011442
Willy Tarreau74ca5042013-06-11 23:12:07 +020011443ssl_f_notbefore : string
11444 Returns the start date presented by the frontend as a formatted string
11445 YYMMDDhhmmss[Z] when the incoming connection was made over an SSL/TLS
11446 transport layer.
Emeric Brun87855892012-10-17 17:39:35 +020011447
Willy Tarreau74ca5042013-06-11 23:12:07 +020011448ssl_f_s_dn([<entry>[,<occ>]]) : string
11449 When the incoming connection was made over an SSL/TLS transport layer,
11450 returns the full distinguished name of the subject of the certificate
11451 presented by the frontend when no <entry> is specified, or the value of the
11452 first given entry found from the beginning of the DN. If a positive/negative
11453 occurrence number is specified as the optional second argument, it returns
11454 the value of the nth given entry value from the beginning/end of the DN.
11455 For instance, "ssl_f_s_dn(OU,2)" the second organization unit, and
11456 "ssl_f_s_dn(CN)" retrieves the common name.
Emeric Brunce5ad802012-10-22 14:11:22 +020011457
Willy Tarreau74ca5042013-06-11 23:12:07 +020011458ssl_f_serial : binary
11459 Returns the serial of the certificate presented by the frontend when the
11460 incoming connection was made over an SSL/TLS transport layer. When used for
11461 an ACL, the value(s) to match against can be passed in hexadecimal form.
Emeric Brun87855892012-10-17 17:39:35 +020011462
Emeric Brun55f4fa82014-04-30 17:11:25 +020011463ssl_f_sha1 : binary
11464 Returns the SHA-1 fingerprint of the certificate presented by the frontend
11465 when the incoming connection was made over an SSL/TLS transport layer. This
11466 can be used to know which certificate was chosen using SNI.
11467
Willy Tarreau74ca5042013-06-11 23:12:07 +020011468ssl_f_sig_alg : string
11469 Returns the name of the algorithm used to sign the certificate presented by
11470 the frontend when the incoming connection was made over an SSL/TLS transport
11471 layer.
Emeric Brun7f56e742012-10-19 18:15:40 +020011472
Willy Tarreau74ca5042013-06-11 23:12:07 +020011473ssl_f_version : integer
11474 Returns the version of the certificate presented by the frontend when the
11475 incoming connection was made over an SSL/TLS transport layer.
11476
11477ssl_fc : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020011478 Returns true when the front connection was made via an SSL/TLS transport
11479 layer and is locally deciphered. This means it has matched a socket declared
11480 with a "bind" line having the "ssl" option.
11481
Willy Tarreau74ca5042013-06-11 23:12:07 +020011482 Example :
11483 # This passes "X-Proto: https" to servers when client connects over SSL
11484 listen http-https
11485 bind :80
11486 bind :443 ssl crt /etc/haproxy.pem
11487 http-request add-header X-Proto https if { ssl_fc }
11488
11489ssl_fc_alg_keysize : integer
11490 Returns the symmetric cipher key size supported in bits when the incoming
11491 connection was made over an SSL/TLS transport layer.
11492
11493ssl_fc_alpn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011494 This extracts the Application Layer Protocol Negotiation field from an
Willy Tarreau74ca5042013-06-11 23:12:07 +020011495 incoming connection made via a TLS transport layer and locally deciphered by
11496 haproxy. The result is a string containing the protocol name advertised by
11497 the client. The SSL library must have been built with support for TLS
11498 extensions enabled (check haproxy -vv). Note that the TLS ALPN extension is
11499 not advertised unless the "alpn" keyword on the "bind" line specifies a
11500 protocol list. Also, nothing forces the client to pick a protocol from this
11501 list, any other one may be requested. The TLS ALPN extension is meant to
11502 replace the TLS NPN extension. See also "ssl_fc_npn".
11503
Willy Tarreau74ca5042013-06-11 23:12:07 +020011504ssl_fc_cipher : string
11505 Returns the name of the used cipher when the incoming connection was made
11506 over an SSL/TLS transport layer.
Willy Tarreauab861d32013-04-02 02:30:41 +020011507
Willy Tarreau74ca5042013-06-11 23:12:07 +020011508ssl_fc_has_crt : boolean
Emeric Brun2525b6b2012-10-18 15:59:43 +020011509 Returns true if a client certificate is present in an incoming connection over
11510 SSL/TLS transport layer. Useful if 'verify' statement is set to 'optional'.
Emeric Brun9143d372012-12-20 15:44:16 +010011511 Note: on SSL session resumption with Session ID or TLS ticket, client
11512 certificate is not present in the current connection but may be retrieved
11513 from the cache or the ticket. So prefer "ssl_c_used" if you want to check if
11514 current SSL session uses a client certificate.
Emeric Brun2525b6b2012-10-18 15:59:43 +020011515
Willy Tarreau74ca5042013-06-11 23:12:07 +020011516ssl_fc_has_sni : boolean
11517 This checks for the presence of a Server Name Indication TLS extension (SNI)
Willy Tarreauf7bc57c2012-10-03 00:19:48 +020011518 in an incoming connection was made over an SSL/TLS transport layer. Returns
11519 true when the incoming connection presents a TLS SNI field. This requires
11520 that the SSL library is build with support for TLS extensions enabled (check
11521 haproxy -vv).
Willy Tarreau7875d092012-09-10 08:20:03 +020011522
Willy Tarreau74ca5042013-06-11 23:12:07 +020011523ssl_fc_npn : string
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011524 This extracts the Next Protocol Negotiation field from an incoming connection
Willy Tarreau74ca5042013-06-11 23:12:07 +020011525 made via a TLS transport layer and locally deciphered by haproxy. The result
11526 is a string containing the protocol name advertised by the client. The SSL
11527 library must have been built with support for TLS extensions enabled (check
11528 haproxy -vv). Note that the TLS NPN extension is not advertised unless the
11529 "npn" keyword on the "bind" line specifies a protocol list. Also, nothing
11530 forces the client to pick a protocol from this list, any other one may be
11531 requested. Please note that the TLS NPN extension was replaced with ALPN.
Willy Tarreaua33c6542012-10-15 13:19:06 +020011532
Willy Tarreau74ca5042013-06-11 23:12:07 +020011533ssl_fc_protocol : string
11534 Returns the name of the used protocol when the incoming connection was made
11535 over an SSL/TLS transport layer.
Willy Tarreau7875d092012-09-10 08:20:03 +020011536
Emeric Brunb73a9b02014-04-30 18:49:19 +020011537ssl_fc_unique_id : binary
David Sc1ad52e2014-04-08 18:48:47 -040011538 When the incoming connection was made over an SSL/TLS transport layer,
Emeric Brunb73a9b02014-04-30 18:49:19 +020011539 returns the TLS unique ID as defined in RFC5929 section 3. The unique id
11540 can be encoded to base64 using the converter: "ssl_bc_unique_id,base64".
David Sc1ad52e2014-04-08 18:48:47 -040011541
Willy Tarreau74ca5042013-06-11 23:12:07 +020011542ssl_fc_session_id : binary
11543 Returns the SSL ID of the front connection when the incoming connection was
11544 made over an SSL/TLS transport layer. It is useful to stick a given client to
11545 a server. It is important to note that some browsers refresh their session ID
11546 every few minutes.
Willy Tarreau7875d092012-09-10 08:20:03 +020011547
Willy Tarreau74ca5042013-06-11 23:12:07 +020011548ssl_fc_sni : string
11549 This extracts the Server Name Indication TLS extension (SNI) field from an
11550 incoming connection made via an SSL/TLS transport layer and locally
11551 deciphered by haproxy. The result (when present) typically is a string
11552 matching the HTTPS host name (253 chars or less). The SSL library must have
11553 been built with support for TLS extensions enabled (check haproxy -vv).
11554
11555 This fetch is different from "req_ssl_sni" above in that it applies to the
11556 connection being deciphered by haproxy and not to SSL contents being blindly
11557 forwarded. See also "ssl_fc_sni_end" and "ssl_fc_sni_reg" below. This
Cyril Bonté9c1eb1e2012-10-09 22:45:34 +020011558 requires that the SSL library is build with support for TLS extensions
11559 enabled (check haproxy -vv).
Willy Tarreau62644772008-07-16 18:36:06 +020011560
Willy Tarreau74ca5042013-06-11 23:12:07 +020011561 ACL derivatives :
Willy Tarreau74ca5042013-06-11 23:12:07 +020011562 ssl_fc_sni_end : suffix match
11563 ssl_fc_sni_reg : regex match
Emeric Brun589fcad2012-10-16 14:13:26 +020011564
Willy Tarreau74ca5042013-06-11 23:12:07 +020011565ssl_fc_use_keysize : integer
11566 Returns the symmetric cipher key size used in bits when the incoming
11567 connection was made over an SSL/TLS transport layer.
Willy Tarreaub6fb4202008-07-20 11:18:28 +020011568
Willy Tarreaub6fb4202008-07-20 11:18:28 +020011569
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200115707.3.5. Fetching samples from buffer contents (Layer 6)
Willy Tarreau74ca5042013-06-11 23:12:07 +020011571------------------------------------------------------
Willy Tarreaub6fb4202008-07-20 11:18:28 +020011572
Willy Tarreau74ca5042013-06-11 23:12:07 +020011573Fetching samples from buffer contents is a bit different from the previous
11574sample fetches above because the sampled data are ephemeral. These data can
11575only be used when they're available and will be lost when they're forwarded.
11576For this reason, samples fetched from buffer contents during a request cannot
11577be used in a response for example. Even while the data are being fetched, they
11578can change. Sometimes it is necessary to set some delays or combine multiple
11579sample fetch methods to ensure that the expected data are complete and usable,
11580for example through TCP request content inspection. Please see the "tcp-request
11581content" keyword for more detailed information on the subject.
Willy Tarreau62644772008-07-16 18:36:06 +020011582
Willy Tarreau74ca5042013-06-11 23:12:07 +020011583payload(<offset>,<length>) : binary (deprecated)
11584 This is an alias for "req.payload" when used in the context of a request (eg:
11585 "stick on", "stick match"), and for "res.payload" when used in the context of
11586 a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011587
Willy Tarreau74ca5042013-06-11 23:12:07 +020011588payload_lv(<offset1>,<length>[,<offset2>]) : binary (deprecated)
11589 This is an alias for "req.payload_lv" when used in the context of a request
11590 (eg: "stick on", "stick match"), and for "res.payload_lv" when used in the
11591 context of a response such as in "stick store response".
Willy Tarreau0ba27502007-12-24 16:55:16 +010011592
Willy Tarreau74ca5042013-06-11 23:12:07 +020011593req.len : integer
11594req_len : integer (deprecated)
11595 Returns an integer value corresponding to the number of bytes present in the
11596 request buffer. This is mostly used in ACL. It is important to understand
11597 that this test does not return false as long as the buffer is changing. This
11598 means that a check with equality to zero will almost always immediately match
11599 at the beginning of the session, while a test for more data will wait for
11600 that data to come in and return false only when haproxy is certain that no
11601 more data will come in. This test was designed to be used with TCP request
11602 content inspection.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011603
Willy Tarreau74ca5042013-06-11 23:12:07 +020011604req.payload(<offset>,<length>) : binary
11605 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020011606 in the request buffer. As a special case, if the <length> argument is zero,
11607 the the whole buffer from <offset> to the end is extracted. This can be used
11608 with ACLs in order to check for the presence of some content in a buffer at
11609 any location.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011610
Willy Tarreau74ca5042013-06-11 23:12:07 +020011611 ACL alternatives :
11612 payload(<offset>,<length>) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011613
Willy Tarreau74ca5042013-06-11 23:12:07 +020011614req.payload_lv(<offset1>,<length>[,<offset2>]) : binary
11615 This extracts a binary block whose size is specified at <offset1> for <length>
11616 bytes, and which starts at <offset2> if specified or just after the length in
11617 the request buffer. The <offset2> parameter also supports relative offsets if
11618 prepended with a '+' or '-' sign.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011619
Willy Tarreau74ca5042013-06-11 23:12:07 +020011620 ACL alternatives :
11621 payload_lv(<offset1>,<length>[,<offset2>]) : hex binary match
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011622
Willy Tarreau74ca5042013-06-11 23:12:07 +020011623 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011624
Willy Tarreau74ca5042013-06-11 23:12:07 +020011625req.proto_http : boolean
11626req_proto_http : boolean (deprecated)
11627 Returns true when data in the request buffer look like HTTP and correctly
11628 parses as such. It is the same parser as the common HTTP request parser which
11629 is used so there should be no surprises. The test does not match until the
11630 request is complete, failed or timed out. This test may be used to report the
11631 protocol in TCP logs, but the biggest use is to block TCP request analysis
11632 until a complete HTTP request is present in the buffer, for example to track
11633 a header.
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011634
Willy Tarreau74ca5042013-06-11 23:12:07 +020011635 Example:
11636 # track request counts per "base" (concatenation of Host+URL)
11637 tcp-request inspect-delay 10s
11638 tcp-request content reject if !HTTP
Willy Tarreaube4a3ef2013-06-17 15:04:07 +020011639 tcp-request content track-sc0 base table req-rate
Willy Tarreaua7ad50c2012-04-29 15:39:40 +020011640
Willy Tarreau74ca5042013-06-11 23:12:07 +020011641req.rdp_cookie([<name>]) : string
11642rdp_cookie([<name>]) : string (deprecated)
11643 When the request buffer looks like the RDP protocol, extracts the RDP cookie
11644 <name>, or any cookie if unspecified. The parser only checks for the first
11645 cookie, as illustrated in the RDP protocol specification. The cookie name is
11646 case insensitive. Generally the "MSTS" cookie name will be used, as it can
11647 contain the user name of the client connecting to the server if properly
11648 configured on the client. The "MSTSHASH" cookie is often used as well for
11649 session stickiness to servers.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011650
Willy Tarreau74ca5042013-06-11 23:12:07 +020011651 This differs from "balance rdp-cookie" in that any balancing algorithm may be
11652 used and thus the distribution of clients to backend servers is not linked to
11653 a hash of the RDP cookie. It is envisaged that using a balancing algorithm
11654 such as "balance roundrobin" or "balance leastconn" will lead to a more even
11655 distribution of clients to backend servers than the hash used by "balance
11656 rdp-cookie".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011657
Willy Tarreau74ca5042013-06-11 23:12:07 +020011658 ACL derivatives :
11659 req_rdp_cookie([<name>]) : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011660
Willy Tarreau74ca5042013-06-11 23:12:07 +020011661 Example :
11662 listen tse-farm
11663 bind 0.0.0.0:3389
11664 # wait up to 5s for an RDP cookie in the request
11665 tcp-request inspect-delay 5s
11666 tcp-request content accept if RDP_COOKIE
11667 # apply RDP cookie persistence
11668 persist rdp-cookie
11669 # Persist based on the mstshash cookie
11670 # This is only useful makes sense if
11671 # balance rdp-cookie is not used
11672 stick-table type string size 204800
11673 stick on req.rdp_cookie(mstshash)
11674 server srv1 1.1.1.1:3389
11675 server srv1 1.1.1.2:3389
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011676
Willy Tarreau74ca5042013-06-11 23:12:07 +020011677 See also : "balance rdp-cookie", "persist rdp-cookie", "tcp-request" and the
11678 "req_rdp_cookie" ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011679
Willy Tarreau74ca5042013-06-11 23:12:07 +020011680req.rdp_cookie_cnt([name]) : integer
11681rdp_cookie_cnt([name]) : integer (deprecated)
11682 Tries to parse the request buffer as RDP protocol, then returns an integer
11683 corresponding to the number of RDP cookies found. If an optional cookie name
11684 is passed, only cookies matching this name are considered. This is mostly
11685 used in ACL.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011686
Willy Tarreau74ca5042013-06-11 23:12:07 +020011687 ACL derivatives :
11688 req_rdp_cookie_cnt([<name>]) : integer match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011689
Willy Tarreau74ca5042013-06-11 23:12:07 +020011690req.ssl_hello_type : integer
11691req_ssl_hello_type : integer (deprecated)
11692 Returns an integer value containing the type of the SSL hello message found
11693 in the request buffer if the buffer contains data that parse as a complete
11694 SSL (v3 or superior) client hello message. Note that this only applies to raw
11695 contents found in the request buffer and not to contents deciphered via an
11696 SSL data layer, so this will not work with "bind" lines having the "ssl"
11697 option. This is mostly used in ACL to detect presence of an SSL hello message
11698 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011699
Willy Tarreau74ca5042013-06-11 23:12:07 +020011700req.ssl_sni : string
11701req_ssl_sni : string (deprecated)
11702 Returns a string containing the value of the Server Name TLS extension sent
11703 by a client in a TLS stream passing through the request buffer if the buffer
11704 contains data that parse as a complete SSL (v3 or superior) client hello
11705 message. Note that this only applies to raw contents found in the request
11706 buffer and not to contents deciphered via an SSL data layer, so this will not
11707 work with "bind" lines having the "ssl" option. SNI normally contains the
11708 name of the host the client tries to connect to (for recent browsers). SNI is
11709 useful for allowing or denying access to certain hosts when SSL/TLS is used
11710 by the client. This test was designed to be used with TCP request content
11711 inspection. If content switching is needed, it is recommended to first wait
11712 for a complete client hello (type 1), like in the example below. See also
11713 "ssl_fc_sni".
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011714
Willy Tarreau74ca5042013-06-11 23:12:07 +020011715 ACL derivatives :
11716 req_ssl_sni : exact string match
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011717
Willy Tarreau74ca5042013-06-11 23:12:07 +020011718 Examples :
11719 # Wait for a client hello for at most 5 seconds
11720 tcp-request inspect-delay 5s
11721 tcp-request content accept if { req_ssl_hello_type 1 }
11722 use_backend bk_allow if { req_ssl_sni -f allowed_sites }
11723 default_backend bk_sorry_page
Willy Tarreau04aa6a92012-04-06 18:57:55 +020011724
Willy Tarreau74ca5042013-06-11 23:12:07 +020011725res.ssl_hello_type : integer
11726rep_ssl_hello_type : integer (deprecated)
11727 Returns an integer value containing the type of the SSL hello message found
11728 in the response buffer if the buffer contains data that parses as a complete
11729 SSL (v3 or superior) hello message. Note that this only applies to raw
11730 contents found in the response buffer and not to contents deciphered via an
11731 SSL data layer, so this will not work with "server" lines having the "ssl"
11732 option. This is mostly used in ACL to detect presence of an SSL hello message
11733 that is supposed to contain an SSL session ID usable for stickiness.
Willy Tarreau51539362012-05-08 12:46:28 +020011734
Willy Tarreau74ca5042013-06-11 23:12:07 +020011735req.ssl_ver : integer
11736req_ssl_ver : integer (deprecated)
11737 Returns an integer value containing the version of the SSL/TLS protocol of a
11738 stream present in the request buffer. Both SSLv2 hello messages and SSLv3
11739 messages are supported. TLSv1 is announced as SSL version 3.1. The value is
11740 composed of the major version multiplied by 65536, added to the minor
11741 version. Note that this only applies to raw contents found in the request
11742 buffer and not to contents deciphered via an SSL data layer, so this will not
11743 work with "bind" lines having the "ssl" option. The ACL version of the test
11744 matches against a decimal notation in the form MAJOR.MINOR (eg: 3.1). This
11745 fetch is mostly used in ACL.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011746
Willy Tarreau74ca5042013-06-11 23:12:07 +020011747 ACL derivatives :
11748 req_ssl_ver : decimal match
Willy Tarreaud63335a2010-02-26 12:56:52 +010011749
Willy Tarreau47e8eba2013-09-11 23:28:46 +020011750res.len : integer
11751 Returns an integer value corresponding to the number of bytes present in the
11752 response buffer. This is mostly used in ACL. It is important to understand
11753 that this test does not return false as long as the buffer is changing. This
11754 means that a check with equality to zero will almost always immediately match
11755 at the beginning of the session, while a test for more data will wait for
11756 that data to come in and return false only when haproxy is certain that no
11757 more data will come in. This test was designed to be used with TCP response
11758 content inspection.
11759
Willy Tarreau74ca5042013-06-11 23:12:07 +020011760res.payload(<offset>,<length>) : binary
11761 This extracts a binary block of <length> bytes and starting at byte <offset>
Willy Tarreau00f00842013-08-02 11:07:32 +020011762 in the response buffer. As a special case, if the <length> argument is zero,
11763 the the whole buffer from <offset> to the end is extracted. This can be used
11764 with ACLs in order to check for the presence of some content in a buffer at
11765 any location.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011766
Willy Tarreau74ca5042013-06-11 23:12:07 +020011767res.payload_lv(<offset1>,<length>[,<offset2>]) : binary
11768 This extracts a binary block whose size is specified at <offset1> for <length>
11769 bytes, and which starts at <offset2> if specified or just after the length in
11770 the response buffer. The <offset2> parameter also supports relative offsets
11771 if prepended with a '+' or '-' sign.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011772
Willy Tarreau74ca5042013-06-11 23:12:07 +020011773 Example : please consult the example from the "stick store-response" keyword.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011774
Willy Tarreau74ca5042013-06-11 23:12:07 +020011775wait_end : boolean
11776 This fetch either returns true when the inspection period is over, or does
11777 not fetch. It is only used in ACLs, in conjunction with content analysis to
11778 avoid returning a wrong verdict early. It may also be used to delay some
11779 actions, such as a delayed reject for some special addresses. Since it either
11780 stops the rules evaluation or immediately returns true, it is recommended to
11781 use this acl as the last one in a rule. Please note that the default ACL
11782 "WAIT_END" is always usable without prior declaration. This test was designed
11783 to be used with TCP request content inspection.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011784
Willy Tarreau74ca5042013-06-11 23:12:07 +020011785 Examples :
11786 # delay every incoming request by 2 seconds
11787 tcp-request inspect-delay 2s
11788 tcp-request content accept if WAIT_END
Willy Tarreaud63335a2010-02-26 12:56:52 +010011789
Willy Tarreau74ca5042013-06-11 23:12:07 +020011790 # don't immediately tell bad guys they are rejected
11791 tcp-request inspect-delay 10s
11792 acl goodguys src 10.0.0.0/24
11793 acl badguys src 10.0.1.0/24
11794 tcp-request content accept if goodguys
11795 tcp-request content reject if badguys WAIT_END
11796 tcp-request content reject
11797
11798
Thierry FOURNIER060762e2014-04-23 13:29:15 +0200117997.3.6. Fetching HTTP samples (Layer 7)
Willy Tarreau74ca5042013-06-11 23:12:07 +020011800--------------------------------------
11801
11802It is possible to fetch samples from HTTP contents, requests and responses.
11803This application layer is also called layer 7. It is only possible to fetch the
11804data in this section when a full HTTP request or response has been parsed from
11805its respective request or response buffer. This is always the case with all
11806HTTP specific rules and for sections running with "mode http". When using TCP
11807content inspection, it may be necessary to support an inspection delay in order
11808to let the request or response come in first. These fetches may require a bit
11809more CPU resources than the layer 4 ones, but not much since the request and
11810response are indexed.
11811
11812base : string
11813 This returns the concatenation of the first Host header and the path part of
11814 the request, which starts at the first slash and ends before the question
11815 mark. It can be useful in virtual hosted environments to detect URL abuses as
11816 well as to improve shared caches efficiency. Using this with a limited size
11817 stick table also allows one to collect statistics about most commonly
11818 requested objects by host/path. With ACLs it can allow simple content
11819 switching rules involving the host and the path at the same time, such as
11820 "www.example.com/favicon.ico". See also "path" and "uri".
11821
11822 ACL derivatives :
11823 base : exact string match
11824 base_beg : prefix match
11825 base_dir : subdir match
11826 base_dom : domain match
11827 base_end : suffix match
11828 base_len : length match
11829 base_reg : regex match
11830 base_sub : substring match
11831
11832base32 : integer
11833 This returns a 32-bit hash of the value returned by the "base" fetch method
11834 above. This is useful to track per-URL activity on high traffic sites without
11835 having to store all URLs. Instead a shorter hash is stored, saving a lot of
Willy Tarreau23ec4ca2014-07-15 20:15:37 +020011836 memory. The output type is an unsigned integer. The hash function used is
11837 SDBM with full avalanche on the output. Technically, base32 is exactly equal
11838 to "base,sdbm(1)".
Willy Tarreau74ca5042013-06-11 23:12:07 +020011839
11840base32+src : binary
11841 This returns the concatenation of the base32 fetch above and the src fetch
11842 below. The resulting type is of type binary, with a size of 8 or 20 bytes
11843 depending on the source address family. This can be used to track per-IP,
11844 per-URL counters.
11845
William Lallemand65ad6e12014-01-31 15:08:02 +010011846capture.req.hdr(<idx>) : string
11847 This extracts the content of the header captured by the "capture request
11848 header", idx is the position of the capture keyword in the configuration.
11849 The first entry is an index of 0. See also: "capture request header".
11850
11851capture.req.method : string
11852 This extracts the METHOD of an HTTP request. It can be used in both request
11853 and response. Unlike "method", it can be used in both request and response
11854 because it's allocated.
11855
11856capture.req.uri : string
11857 This extracts the request's URI, which starts at the first slash and ends
11858 before the first space in the request (without the host part). Unlike "path"
11859 and "url", it can be used in both request and response because it's
11860 allocated.
11861
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020011862capture.req.ver : string
11863 This extracts the request's HTTP version and returns either "HTTP/1.0" or
11864 "HTTP/1.1". Unlike "req.ver", it can be used in both request, response, and
11865 logs because it relies on a persistent flag.
11866
William Lallemand65ad6e12014-01-31 15:08:02 +010011867capture.res.hdr(<idx>) : string
11868 This extracts the content of the header captured by the "capture response
11869 header", idx is the position of the capture keyword in the configuration.
11870 The first entry is an index of 0.
11871 See also: "capture response header"
11872
Willy Tarreau3c1b5ec2014-04-24 23:41:57 +020011873capture.res.ver : string
11874 This extracts the response's HTTP version and returns either "HTTP/1.0" or
11875 "HTTP/1.1". Unlike "res.ver", it can be used in logs because it relies on a
11876 persistent flag.
11877
Willy Tarreau74ca5042013-06-11 23:12:07 +020011878req.cook([<name>]) : string
11879cook([<name>]) : string (deprecated)
11880 This extracts the last occurrence of the cookie name <name> on a "Cookie"
11881 header line from the request, and returns its value as string. If no name is
11882 specified, the first cookie value is returned. When used with ACLs, all
11883 matching cookies are evaluated. Spaces around the name and the value are
11884 ignored as requested by the Cookie header specification (RFC6265). The cookie
11885 name is case-sensitive. Empty cookies are valid, so an empty cookie may very
11886 well return an empty value if it is present. Use the "found" match to detect
11887 presence. Use the res.cook() variant for response cookies sent by the server.
11888
11889 ACL derivatives :
11890 cook([<name>]) : exact string match
11891 cook_beg([<name>]) : prefix match
11892 cook_dir([<name>]) : subdir match
11893 cook_dom([<name>]) : domain match
11894 cook_end([<name>]) : suffix match
11895 cook_len([<name>]) : length match
11896 cook_reg([<name>]) : regex match
11897 cook_sub([<name>]) : substring match
Willy Tarreaud63335a2010-02-26 12:56:52 +010011898
Willy Tarreau74ca5042013-06-11 23:12:07 +020011899req.cook_cnt([<name>]) : integer
11900cook_cnt([<name>]) : integer (deprecated)
11901 Returns an integer value representing the number of occurrences of the cookie
11902 <name> in the request, or all cookies if <name> is not specified.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011903
Willy Tarreau74ca5042013-06-11 23:12:07 +020011904req.cook_val([<name>]) : integer
11905cook_val([<name>]) : integer (deprecated)
11906 This extracts the last occurrence of the cookie name <name> on a "Cookie"
11907 header line from the request, and converts its value to an integer which is
11908 returned. If no name is specified, the first cookie value is returned. When
11909 used in ACLs, all matching names are iterated over until a value matches.
Willy Tarreau0e698542011-09-16 08:32:32 +020011910
Willy Tarreau74ca5042013-06-11 23:12:07 +020011911cookie([<name>]) : string (deprecated)
11912 This extracts the last occurrence of the cookie name <name> on a "Cookie"
11913 header line from the request, or a "Set-Cookie" header from the response, and
11914 returns its value as a string. A typical use is to get multiple clients
11915 sharing a same profile use the same server. This can be similar to what
11916 "appsession" does with the "request-learn" statement, but with support for
11917 multi-peer synchronization and state keeping across restarts. If no name is
11918 specified, the first cookie value is returned. This fetch should not be used
11919 anymore and should be replaced by req.cook() or res.cook() instead as it
11920 ambiguously uses the direction based on the context where it is used.
11921 See also : "appsession".
Willy Tarreaud63335a2010-02-26 12:56:52 +010011922
Willy Tarreau74ca5042013-06-11 23:12:07 +020011923hdr([<name>[,<occ>]]) : string
11924 This is equivalent to req.hdr() when used on requests, and to res.hdr() when
11925 used on responses. Please refer to these respective fetches for more details.
11926 In case of doubt about the fetch direction, please use the explicit ones.
11927 Note that contrary to the hdr() sample fetch method, the hdr_* ACL keywords
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030011928 unambiguously apply to the request headers.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011929
Willy Tarreau74ca5042013-06-11 23:12:07 +020011930req.fhdr(<name>[,<occ>]) : string
11931 This extracts the last occurrence of header <name> in an HTTP request. When
11932 used from an ACL, all occurrences are iterated over until a match is found.
11933 Optionally, a specific occurrence might be specified as a position number.
11934 Positive values indicate a position from the first occurrence, with 1 being
11935 the first one. Negative values indicate positions relative to the last one,
11936 with -1 being the last one. It differs from req.hdr() in that any commas
11937 present in the value are returned and are not used as delimiters. This is
11938 sometimes useful with headers such as User-Agent.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011939
Willy Tarreau74ca5042013-06-11 23:12:07 +020011940req.fhdr_cnt([<name>]) : integer
11941 Returns an integer value representing the number of occurrences of request
11942 header field name <name>, or the total number of header fields if <name> is
11943 not specified. Contrary to its req.hdr_cnt() cousin, this function returns
11944 the number of full line headers and does not stop on commas.
Willy Tarreaud63335a2010-02-26 12:56:52 +010011945
Willy Tarreau74ca5042013-06-11 23:12:07 +020011946req.hdr([<name>[,<occ>]]) : string
11947 This extracts the last occurrence of header <name> in an HTTP request. When
11948 used from an ACL, all occurrences are iterated over until a match is found.
11949 Optionally, a specific occurrence might be specified as a position number.
11950 Positive values indicate a position from the first occurrence, with 1 being
11951 the first one. Negative values indicate positions relative to the last one,
11952 with -1 being the last one. A typical use is with the X-Forwarded-For header
11953 once converted to IP, associated with an IP stick-table. The function
11954 considers any comma as a delimiter for distinct values. If full-line headers
11955 are desired instead, use req.fhdr(). Please carefully check RFC2616 to know
11956 how certain headers are supposed to be parsed. Also, some of them are case
11957 insensitive (eg: Connection).
Willy Tarreaud63335a2010-02-26 12:56:52 +010011958
Willy Tarreau74ca5042013-06-11 23:12:07 +020011959 ACL derivatives :
11960 hdr([<name>[,<occ>]]) : exact string match
11961 hdr_beg([<name>[,<occ>]]) : prefix match
11962 hdr_dir([<name>[,<occ>]]) : subdir match
11963 hdr_dom([<name>[,<occ>]]) : domain match
11964 hdr_end([<name>[,<occ>]]) : suffix match
11965 hdr_len([<name>[,<occ>]]) : length match
11966 hdr_reg([<name>[,<occ>]]) : regex match
11967 hdr_sub([<name>[,<occ>]]) : substring match
11968
11969req.hdr_cnt([<name>]) : integer
11970hdr_cnt([<header>]) : integer (deprecated)
11971 Returns an integer value representing the number of occurrences of request
11972 header field name <name>, or the total number of header field values if
11973 <name> is not specified. It is important to remember that one header line may
11974 count as several headers if it has several values. The function considers any
11975 comma as a delimiter for distinct values. If full-line headers are desired
11976 instead, req.fhdr_cnt() should be used instead. With ACLs, it can be used to
11977 detect presence, absence or abuse of a specific header, as well as to block
11978 request smuggling attacks by rejecting requests which contain more than one
11979 of certain headers. See "req.hdr" for more information on header matching.
11980
11981req.hdr_ip([<name>[,<occ>]]) : ip
11982hdr_ip([<name>[,<occ>]]) : ip (deprecated)
11983 This extracts the last occurrence of header <name> in an HTTP request,
11984 converts it to an IPv4 or IPv6 address and returns this address. When used
11985 with ACLs, all occurrences are checked, and if <name> is omitted, every value
11986 of every header is checked. Optionally, a specific occurrence might be
11987 specified as a position number. Positive values indicate a position from the
11988 first occurrence, with 1 being the first one. Negative values indicate
11989 positions relative to the last one, with -1 being the last one. A typical use
11990 is with the X-Forwarded-For and X-Client-IP headers.
11991
11992req.hdr_val([<name>[,<occ>]]) : integer
11993hdr_val([<name>[,<occ>]]) : integer (deprecated)
11994 This extracts the last occurrence of header <name> in an HTTP request, and
11995 converts it to an integer value. When used with ACLs, all occurrences are
11996 checked, and if <name> is omitted, every value of every header is checked.
11997 Optionally, a specific occurrence might be specified as a position number.
11998 Positive values indicate a position from the first occurrence, with 1 being
11999 the first one. Negative values indicate positions relative to the last one,
12000 with -1 being the last one. A typical use is with the X-Forwarded-For header.
12001
12002http_auth(<userlist>) : boolean
12003 Returns a boolean indicating whether the authentication data received from
12004 the client match a username & password stored in the specified userlist. This
12005 fetch function is not really useful outside of ACLs. Currently only http
12006 basic auth is supported.
12007
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010012008http_auth_group(<userlist>) : string
12009 Returns a string corresponding to the user name found in the authentication
12010 data received from the client if both the user name and password are valid
12011 according to the specified userlist. The main purpose is to use it in ACLs
12012 where it is then checked whether the user belongs to any group within a list.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012013 This fetch function is not really useful outside of ACLs. Currently only http
12014 basic auth is supported.
12015
12016 ACL derivatives :
Thierry FOURNIER9eec0a62014-01-22 18:38:02 +010012017 http_auth_group(<userlist>) : group ...
12018 Returns true when the user extracted from the request and whose password is
12019 valid according to the specified userlist belongs to at least one of the
12020 groups.
Willy Tarreau74ca5042013-06-11 23:12:07 +020012021
12022http_first_req : boolean
Willy Tarreau7f18e522010-10-22 20:04:13 +020012023 Returns true when the request being processed is the first one of the
12024 connection. This can be used to add or remove headers that may be missing
Willy Tarreau74ca5042013-06-11 23:12:07 +020012025 from some requests when a request is not the first one, or to help grouping
12026 requests in the logs.
Willy Tarreau7f18e522010-10-22 20:04:13 +020012027
Willy Tarreau74ca5042013-06-11 23:12:07 +020012028method : integer + string
12029 Returns an integer value corresponding to the method in the HTTP request. For
12030 example, "GET" equals 1 (check sources to establish the matching). Value 9
12031 means "other method" and may be converted to a string extracted from the
12032 stream. This should not be used directly as a sample, this is only meant to
12033 be used from ACLs, which transparently convert methods from patterns to these
12034 integer + string values. Some predefined ACL already check for most common
12035 methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012036
Willy Tarreau74ca5042013-06-11 23:12:07 +020012037 ACL derivatives :
12038 method : case insensitive method match
Willy Tarreau6a06a402007-07-15 20:15:28 +020012039
Willy Tarreau74ca5042013-06-11 23:12:07 +020012040 Example :
12041 # only accept GET and HEAD requests
12042 acl valid_method method GET HEAD
12043 http-request deny if ! valid_method
Willy Tarreau6a06a402007-07-15 20:15:28 +020012044
Willy Tarreau74ca5042013-06-11 23:12:07 +020012045path : string
12046 This extracts the request's URL path, which starts at the first slash and
12047 ends before the question mark (without the host part). A typical use is with
12048 prefetch-capable caches, and with portals which need to aggregate multiple
12049 information from databases and keep them in caches. Note that with outgoing
12050 caches, it would be wiser to use "url" instead. With ACLs, it's typically
12051 used to match exact file names (eg: "/login.php"), or directory parts using
12052 the derivative forms. See also the "url" and "base" fetch methods.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012053
Willy Tarreau74ca5042013-06-11 23:12:07 +020012054 ACL derivatives :
12055 path : exact string match
12056 path_beg : prefix match
12057 path_dir : subdir match
12058 path_dom : domain match
12059 path_end : suffix match
12060 path_len : length match
12061 path_reg : regex match
12062 path_sub : substring match
Willy Tarreau6a06a402007-07-15 20:15:28 +020012063
Willy Tarreau49ad95c2015-01-19 15:06:26 +010012064query : string
12065 This extracts the request's query string, which starts after the first
12066 question mark. If no question mark is present, this fetch returns nothing. If
12067 a question mark is present but nothing follows, it returns an empty string.
12068 This means it's possible to easily know whether a query string is present
12069 using the "found" matching method. This fetch is the completemnt of "path"
12070 which stops before the question mark.
12071
Willy Tarreau74ca5042013-06-11 23:12:07 +020012072req.ver : string
12073req_ver : string (deprecated)
12074 Returns the version string from the HTTP request, for example "1.1". This can
12075 be useful for logs, but is mostly there for ACL. Some predefined ACL already
12076 check for versions 1.0 and 1.1.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012077
Willy Tarreau74ca5042013-06-11 23:12:07 +020012078 ACL derivatives :
12079 req_ver : exact string match
Willy Tarreau0e698542011-09-16 08:32:32 +020012080
Willy Tarreau74ca5042013-06-11 23:12:07 +020012081res.comp : boolean
12082 Returns the boolean "true" value if the response has been compressed by
12083 HAProxy, otherwise returns boolean "false". This may be used to add
12084 information in the logs.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012085
Willy Tarreau74ca5042013-06-11 23:12:07 +020012086res.comp_algo : string
12087 Returns a string containing the name of the algorithm used if the response
12088 was compressed by HAProxy, for example : "deflate". This may be used to add
12089 some information in the logs.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012090
Willy Tarreau74ca5042013-06-11 23:12:07 +020012091res.cook([<name>]) : string
12092scook([<name>]) : string (deprecated)
12093 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12094 header line from the response, and returns its value as string. If no name is
12095 specified, the first cookie value is returned.
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020012096
Willy Tarreau74ca5042013-06-11 23:12:07 +020012097 ACL derivatives :
12098 scook([<name>] : exact string match
Willy Tarreau0ce3aa02012-04-25 18:46:33 +020012099
Willy Tarreau74ca5042013-06-11 23:12:07 +020012100res.cook_cnt([<name>]) : integer
12101scook_cnt([<name>]) : integer (deprecated)
12102 Returns an integer value representing the number of occurrences of the cookie
12103 <name> in the response, or all cookies if <name> is not specified. This is
12104 mostly useful when combined with ACLs to detect suspicious responses.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012105
Willy Tarreau74ca5042013-06-11 23:12:07 +020012106res.cook_val([<name>]) : integer
12107scook_val([<name>]) : integer (deprecated)
12108 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12109 header line from the response, and converts its value to an integer which is
12110 returned. If no name is specified, the first cookie value is returned.
Willy Tarreaud63335a2010-02-26 12:56:52 +010012111
Willy Tarreau74ca5042013-06-11 23:12:07 +020012112res.fhdr([<name>[,<occ>]]) : string
12113 This extracts the last occurrence of header <name> in an HTTP response, or of
12114 the last header if no <name> is specified. Optionally, a specific occurrence
12115 might be specified as a position number. Positive values indicate a position
12116 from the first occurrence, with 1 being the first one. Negative values
12117 indicate positions relative to the last one, with -1 being the last one. It
12118 differs from res.hdr() in that any commas present in the value are returned
12119 and are not used as delimiters. If this is not desired, the res.hdr() fetch
12120 should be used instead. This is sometimes useful with headers such as Date or
12121 Expires.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012122
Willy Tarreau74ca5042013-06-11 23:12:07 +020012123res.fhdr_cnt([<name>]) : integer
12124 Returns an integer value representing the number of occurrences of response
12125 header field name <name>, or the total number of header fields if <name> is
12126 not specified. Contrary to its res.hdr_cnt() cousin, this function returns
12127 the number of full line headers and does not stop on commas. If this is not
12128 desired, the res.hdr_cnt() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012129
Willy Tarreau74ca5042013-06-11 23:12:07 +020012130res.hdr([<name>[,<occ>]]) : string
12131shdr([<name>[,<occ>]]) : string (deprecated)
12132 This extracts the last occurrence of header <name> in an HTTP response, or of
12133 the last header if no <name> is specified. Optionally, a specific occurrence
12134 might be specified as a position number. Positive values indicate a position
12135 from the first occurrence, with 1 being the first one. Negative values
12136 indicate positions relative to the last one, with -1 being the last one. This
12137 can be useful to learn some data into a stick-table. The function considers
12138 any comma as a delimiter for distinct values. If this is not desired, the
12139 res.fhdr() fetch should be used instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012140
Willy Tarreau74ca5042013-06-11 23:12:07 +020012141 ACL derivatives :
12142 shdr([<name>[,<occ>]]) : exact string match
12143 shdr_beg([<name>[,<occ>]]) : prefix match
12144 shdr_dir([<name>[,<occ>]]) : subdir match
12145 shdr_dom([<name>[,<occ>]]) : domain match
12146 shdr_end([<name>[,<occ>]]) : suffix match
12147 shdr_len([<name>[,<occ>]]) : length match
12148 shdr_reg([<name>[,<occ>]]) : regex match
12149 shdr_sub([<name>[,<occ>]]) : substring match
12150
12151res.hdr_cnt([<name>]) : integer
12152shdr_cnt([<name>]) : integer (deprecated)
12153 Returns an integer value representing the number of occurrences of response
12154 header field name <name>, or the total number of header fields if <name> is
12155 not specified. The function considers any comma as a delimiter for distinct
12156 values. If this is not desired, the res.fhdr_cnt() fetch should be used
12157 instead.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012158
Willy Tarreau74ca5042013-06-11 23:12:07 +020012159res.hdr_ip([<name>[,<occ>]]) : ip
12160shdr_ip([<name>[,<occ>]]) : ip (deprecated)
12161 This extracts the last occurrence of header <name> in an HTTP response,
12162 convert it to an IPv4 or IPv6 address and returns this address. Optionally, a
12163 specific occurrence might be specified as a position number. Positive values
12164 indicate a position from the first occurrence, with 1 being the first one.
12165 Negative values indicate positions relative to the last one, with -1 being
12166 the last one. This can be useful to learn some data into a stick table.
Willy Tarreau6a06a402007-07-15 20:15:28 +020012167
Willy Tarreau74ca5042013-06-11 23:12:07 +020012168res.hdr_val([<name>[,<occ>]]) : integer
12169shdr_val([<name>[,<occ>]]) : integer (deprecated)
12170 This extracts the last occurrence of header <name> in an HTTP response, and
12171 converts it to an integer value. Optionally, a specific occurrence might be
12172 specified as a position number. Positive values indicate a position from the
12173 first occurrence, with 1 being the first one. Negative values indicate
12174 positions relative to the last one, with -1 being the last one. This can be
12175 useful to learn some data into a stick table.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010012176
Willy Tarreau74ca5042013-06-11 23:12:07 +020012177res.ver : string
12178resp_ver : string (deprecated)
12179 Returns the version string from the HTTP response, for example "1.1". This
12180 can be useful for logs, but is mostly there for ACL.
Willy Tarreau0e698542011-09-16 08:32:32 +020012181
Willy Tarreau74ca5042013-06-11 23:12:07 +020012182 ACL derivatives :
12183 resp_ver : exact string match
Alexandre Cassen5eb1a902007-11-29 15:43:32 +010012184
Willy Tarreau74ca5042013-06-11 23:12:07 +020012185set-cookie([<name>]) : string (deprecated)
12186 This extracts the last occurrence of the cookie name <name> on a "Set-Cookie"
12187 header line from the response and uses the corresponding value to match. This
12188 can be comparable to what "appsession" does with default options, but with
12189 support for multi-peer synchronization and state keeping across restarts.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010012190
Willy Tarreau74ca5042013-06-11 23:12:07 +020012191 This fetch function is deprecated and has been superseded by the "res.cook"
12192 fetch. This keyword will disappear soon.
Krzysztof Piotr Oledzki6b35ce12010-02-01 23:35:44 +010012193
Willy Tarreau74ca5042013-06-11 23:12:07 +020012194 See also : "appsession"
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012195
Willy Tarreau74ca5042013-06-11 23:12:07 +020012196status : integer
12197 Returns an integer containing the HTTP status code in the HTTP response, for
12198 example, 302. It is mostly used within ACLs and integer ranges, for example,
12199 to remove any Location header if the response is not a 3xx.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012200
Willy Tarreau74ca5042013-06-11 23:12:07 +020012201url : string
12202 This extracts the request's URL as presented in the request. A typical use is
12203 with prefetch-capable caches, and with portals which need to aggregate
12204 multiple information from databases and keep them in caches. With ACLs, using
12205 "path" is preferred over using "url", because clients may send a full URL as
12206 is normally done with proxies. The only real use is to match "*" which does
12207 not match in "path", and for which there is already a predefined ACL. See
12208 also "path" and "base".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012209
Willy Tarreau74ca5042013-06-11 23:12:07 +020012210 ACL derivatives :
12211 url : exact string match
12212 url_beg : prefix match
12213 url_dir : subdir match
12214 url_dom : domain match
12215 url_end : suffix match
12216 url_len : length match
12217 url_reg : regex match
12218 url_sub : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012219
Willy Tarreau74ca5042013-06-11 23:12:07 +020012220url_ip : ip
12221 This extracts the IP address from the request's URL when the host part is
12222 presented as an IP address. Its use is very limited. For instance, a
12223 monitoring system might use this field as an alternative for the source IP in
12224 order to test what path a given source address would follow, or to force an
12225 entry in a table for a given source address. With ACLs it can be used to
12226 restrict access to certain systems through a proxy, for example when combined
12227 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012228
Willy Tarreau74ca5042013-06-11 23:12:07 +020012229url_port : integer
12230 This extracts the port part from the request's URL. Note that if the port is
12231 not specified in the request, port 80 is assumed. With ACLs it can be used to
12232 restrict access to certain systems through a proxy, for example when combined
12233 with option "http_proxy".
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012234
Willy Tarreau74ca5042013-06-11 23:12:07 +020012235urlp(<name>[,<delim>]) : string
12236url_param(<name>[,<delim>]) : string
12237 This extracts the first occurrence of the parameter <name> in the query
12238 string, which begins after either '?' or <delim>, and which ends before '&',
12239 ';' or <delim>. The parameter name is case-sensitive. The result is a string
12240 corresponding to the value of the parameter <name> as presented in the
12241 request (no URL decoding is performed). This can be used for session
12242 stickiness based on a client ID, to extract an application cookie passed as a
12243 URL parameter, or in ACLs to apply some checks. Note that the ACL version of
12244 this fetch do not iterate over multiple parameters and stop at the first one
12245 as well.
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012246
Willy Tarreau74ca5042013-06-11 23:12:07 +020012247 ACL derivatives :
12248 urlp(<name>[,<delim>]) : exact string match
12249 urlp_beg(<name>[,<delim>]) : prefix match
12250 urlp_dir(<name>[,<delim>]) : subdir match
12251 urlp_dom(<name>[,<delim>]) : domain match
12252 urlp_end(<name>[,<delim>]) : suffix match
12253 urlp_len(<name>[,<delim>]) : length match
12254 urlp_reg(<name>[,<delim>]) : regex match
12255 urlp_sub(<name>[,<delim>]) : substring match
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012256
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012257
Willy Tarreau74ca5042013-06-11 23:12:07 +020012258 Example :
12259 # match http://example.com/foo?PHPSESSIONID=some_id
12260 stick on urlp(PHPSESSIONID)
12261 # match http://example.com/foo;JSESSIONID=some_id
12262 stick on urlp(JSESSIONID,;)
Willy Tarreau25c1ebc2012-04-25 16:21:44 +020012263
Willy Tarreau74ca5042013-06-11 23:12:07 +020012264urlp_val(<name>[,<delim>]) : integer
12265 See "urlp" above. This one extracts the URL parameter <name> in the request
12266 and converts it to an integer value. This can be used for session stickiness
12267 based on a user ID for example, or with ACLs to match a page number or price.
Willy Tarreaua9fddca2012-07-31 07:51:48 +020012268
Willy Tarreau198a7442008-01-17 12:05:32 +010012269
Willy Tarreau74ca5042013-06-11 23:12:07 +0200122707.4. Pre-defined ACLs
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012271---------------------
Willy Tarreauced27012008-01-17 20:35:34 +010012272
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012273Some predefined ACLs are hard-coded so that they do not have to be declared in
12274every frontend which needs them. They all have their names in upper case in
Patrick Mézard2382ad62010-05-09 10:43:32 +020012275order to avoid confusion. Their equivalence is provided below.
Willy Tarreauced27012008-01-17 20:35:34 +010012276
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012277ACL name Equivalent to Usage
12278---------------+-----------------------------+---------------------------------
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012279FALSE always_false never match
Willy Tarreau2492d5b2009-07-11 00:06:00 +020012280HTTP req_proto_http match if protocol is valid HTTP
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012281HTTP_1.0 req_ver 1.0 match HTTP version 1.0
12282HTTP_1.1 req_ver 1.1 match HTTP version 1.1
Willy Tarreaud63335a2010-02-26 12:56:52 +010012283HTTP_CONTENT hdr_val(content-length) gt 0 match an existing content-length
12284HTTP_URL_ABS url_reg ^[^/:]*:// match absolute URL with scheme
12285HTTP_URL_SLASH url_beg / match URL beginning with "/"
12286HTTP_URL_STAR url * match URL equal to "*"
12287LOCALHOST src 127.0.0.1/8 match connection from local host
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012288METH_CONNECT method CONNECT match HTTP CONNECT method
12289METH_GET method GET HEAD match HTTP GET or HEAD method
12290METH_HEAD method HEAD match HTTP HEAD method
12291METH_OPTIONS method OPTIONS match HTTP OPTIONS method
12292METH_POST method POST match HTTP POST method
12293METH_TRACE method TRACE match HTTP TRACE method
Emeric Brunbede3d02009-06-30 17:54:00 +020012294RDP_COOKIE req_rdp_cookie_cnt gt 0 match presence of an RDP cookie
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012295REQ_CONTENT req_len gt 0 match data in the request buffer
Willy Tarreaud63335a2010-02-26 12:56:52 +010012296TRUE always_true always match
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012297WAIT_END wait_end wait for end of content analysis
12298---------------+-----------------------------+---------------------------------
Willy Tarreauced27012008-01-17 20:35:34 +010012299
Willy Tarreaub937b7e2010-01-12 15:27:54 +010012300
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200123018. Logging
12302----------
Willy Tarreau844e3c52008-01-11 16:28:18 +010012303
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012304One of HAProxy's strong points certainly lies is its precise logs. It probably
12305provides the finest level of information available for such a product, which is
12306very important for troubleshooting complex environments. Standard information
12307provided in logs include client ports, TCP/HTTP state timers, precise session
12308state at termination and precise termination cause, information about decisions
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010012309to direct traffic to a server, and of course the ability to capture arbitrary
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012310headers.
12311
12312In order to improve administrators reactivity, it offers a great transparency
12313about encountered problems, both internal and external, and it is possible to
12314send logs to different sources at the same time with different level filters :
12315
12316 - global process-level logs (system errors, start/stop, etc..)
12317 - per-instance system and internal errors (lack of resource, bugs, ...)
12318 - per-instance external troubles (servers up/down, max connections)
12319 - per-instance activity (client connections), either at the establishment or
12320 at the termination.
12321
12322The ability to distribute different levels of logs to different log servers
12323allow several production teams to interact and to fix their problems as soon
12324as possible. For example, the system team might monitor system-wide errors,
12325while the application team might be monitoring the up/down for their servers in
12326real time, and the security team might analyze the activity logs with one hour
12327delay.
12328
12329
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200123308.1. Log levels
12331---------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012332
Simon Hormandf791f52011-05-29 15:01:10 +090012333TCP and HTTP connections can be logged with information such as the date, time,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012334source IP address, destination address, connection duration, response times,
Simon Hormandf791f52011-05-29 15:01:10 +090012335HTTP request, HTTP return code, number of bytes transmitted, conditions
12336in which the session ended, and even exchanged cookies values. For example
12337track a particular user's problems. All messages may be sent to up to two
12338syslog servers. Check the "log" keyword in section 4.2 for more information
12339about log facilities.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012340
12341
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200123428.2. Log formats
12343----------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012344
William Lallemand48940402012-01-30 16:47:22 +010012345HAProxy supports 5 log formats. Several fields are common between these formats
Simon Hormandf791f52011-05-29 15:01:10 +090012346and will be detailed in the following sections. A few of them may vary
12347slightly with the configuration, due to indicators specific to certain
12348options. The supported formats are as follows :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012349
12350 - the default format, which is very basic and very rarely used. It only
12351 provides very basic information about the incoming connection at the moment
12352 it is accepted : source IP:port, destination IP:port, and frontend-name.
12353 This mode will eventually disappear so it will not be described to great
12354 extents.
12355
12356 - the TCP format, which is more advanced. This format is enabled when "option
12357 tcplog" is set on the frontend. HAProxy will then usually wait for the
12358 connection to terminate before logging. This format provides much richer
12359 information, such as timers, connection counts, queue size, etc... This
12360 format is recommended for pure TCP proxies.
12361
12362 - the HTTP format, which is the most advanced for HTTP proxying. This format
12363 is enabled when "option httplog" is set on the frontend. It provides the
12364 same information as the TCP format with some HTTP-specific fields such as
12365 the request, the status code, and captures of headers and cookies. This
12366 format is recommended for HTTP proxies.
12367
Emeric Brun3a058f32009-06-30 18:26:00 +020012368 - the CLF HTTP format, which is equivalent to the HTTP format, but with the
12369 fields arranged in the same order as the CLF format. In this mode, all
12370 timers, captures, flags, etc... appear one per field after the end of the
12371 common fields, in the same order they appear in the standard HTTP format.
12372
William Lallemand48940402012-01-30 16:47:22 +010012373 - the custom log format, allows you to make your own log line.
12374
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012375Next sections will go deeper into details for each of these formats. Format
12376specification will be performed on a "field" basis. Unless stated otherwise, a
12377field is a portion of text delimited by any number of spaces. Since syslog
12378servers are susceptible of inserting fields at the beginning of a line, it is
12379always assumed that the first field is the one containing the process name and
12380identifier.
12381
12382Note : Since log lines may be quite long, the log examples in sections below
12383 might be broken into multiple lines. The example log lines will be
12384 prefixed with 3 closing angle brackets ('>>>') and each time a log is
12385 broken into multiple lines, each non-final line will end with a
12386 backslash ('\') and the next line will start indented by two characters.
12387
12388
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200123898.2.1. Default log format
12390-------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012391
12392This format is used when no specific option is set. The log is emitted as soon
12393as the connection is accepted. One should note that this currently is the only
12394format which logs the request's destination IP and ports.
12395
12396 Example :
12397 listen www
12398 mode http
12399 log global
12400 server srv1 127.0.0.1:8000
12401
12402 >>> Feb 6 12:12:09 localhost \
12403 haproxy[14385]: Connect from 10.0.1.2:33312 to 10.0.3.31:8012 \
12404 (www/HTTP)
12405
12406 Field Format Extract from the example above
12407 1 process_name '[' pid ']:' haproxy[14385]:
12408 2 'Connect from' Connect from
12409 3 source_ip ':' source_port 10.0.1.2:33312
12410 4 'to' to
12411 5 destination_ip ':' destination_port 10.0.3.31:8012
12412 6 '(' frontend_name '/' mode ')' (www/HTTP)
12413
12414Detailed fields description :
12415 - "source_ip" is the IP address of the client which initiated the connection.
12416 - "source_port" is the TCP port of the client which initiated the connection.
12417 - "destination_ip" is the IP address the client connected to.
12418 - "destination_port" is the TCP port the client connected to.
12419 - "frontend_name" is the name of the frontend (or listener) which received
12420 and processed the connection.
12421 - "mode is the mode the frontend is operating (TCP or HTTP).
12422
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012423In case of a UNIX socket, the source and destination addresses are marked as
12424"unix:" and the ports reflect the internal ID of the socket which accepted the
12425connection (the same ID as reported in the stats).
12426
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012427It is advised not to use this deprecated format for newer installations as it
12428will eventually disappear.
12429
12430
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200124318.2.2. TCP log format
12432---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012433
12434The TCP format is used when "option tcplog" is specified in the frontend, and
12435is the recommended format for pure TCP proxies. It provides a lot of precious
12436information for troubleshooting. Since this format includes timers and byte
12437counts, the log is normally emitted at the end of the session. It can be
12438emitted earlier if "option logasap" is specified, which makes sense in most
12439environments with long sessions such as remote terminals. Sessions which match
12440the "monitor" rules are never logged. It is also possible not to emit logs for
12441sessions for which no data were exchanged between the client and the server, by
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020012442specifying "option dontlognull" in the frontend. Successful connections will
12443not be logged if "option dontlog-normal" is specified in the frontend. A few
12444fields may slightly vary depending on some configuration options, those are
12445marked with a star ('*') after the field name below.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012446
12447 Example :
12448 frontend fnt
12449 mode tcp
12450 option tcplog
12451 log global
12452 default_backend bck
12453
12454 backend bck
12455 server srv1 127.0.0.1:8000
12456
12457 >>> Feb 6 12:12:56 localhost \
12458 haproxy[14387]: 10.0.1.2:33313 [06/Feb/2009:12:12:51.443] fnt \
12459 bck/srv1 0/0/5007 212 -- 0/0/0/0/3 0/0
12460
12461 Field Format Extract from the example above
12462 1 process_name '[' pid ']:' haproxy[14387]:
12463 2 client_ip ':' client_port 10.0.1.2:33313
12464 3 '[' accept_date ']' [06/Feb/2009:12:12:51.443]
12465 4 frontend_name fnt
12466 5 backend_name '/' server_name bck/srv1
12467 6 Tw '/' Tc '/' Tt* 0/0/5007
12468 7 bytes_read* 212
12469 8 termination_state --
12470 9 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 0/0/0/0/3
12471 10 srv_queue '/' backend_queue 0/0
12472
12473Detailed fields description :
12474 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012475 connection to haproxy. If the connection was accepted on a UNIX socket
12476 instead, the IP address would be replaced with the word "unix". Note that
12477 when the connection is accepted on a socket configured with "accept-proxy"
12478 and the PROXY protocol is correctly used, then the logs will reflect the
12479 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012480
12481 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012482 If the connection was accepted on a UNIX socket instead, the port would be
12483 replaced with the ID of the accepting socket, which is also reported in the
12484 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012485
12486 - "accept_date" is the exact date when the connection was received by haproxy
12487 (which might be very slightly different from the date observed on the
12488 network if there was some queuing in the system's backlog). This is usually
12489 the same date which may appear in any upstream firewall's log.
12490
12491 - "frontend_name" is the name of the frontend (or listener) which received
12492 and processed the connection.
12493
12494 - "backend_name" is the name of the backend (or listener) which was selected
12495 to manage the connection to the server. This will be the same as the
12496 frontend if no switching rule has been applied, which is common for TCP
12497 applications.
12498
12499 - "server_name" is the name of the last server to which the connection was
12500 sent, which might differ from the first one if there were connection errors
12501 and a redispatch occurred. Note that this server belongs to the backend
12502 which processed the request. If the connection was aborted before reaching
12503 a server, "<NOSRV>" is indicated instead of a server name.
12504
12505 - "Tw" is the total time in milliseconds spent waiting in the various queues.
12506 It can be "-1" if the connection was aborted before reaching the queue.
12507 See "Timers" below for more details.
12508
12509 - "Tc" is the total time in milliseconds spent waiting for the connection to
12510 establish to the final server, including retries. It can be "-1" if the
12511 connection was aborted before a connection could be established. See
12512 "Timers" below for more details.
12513
12514 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012515 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012516 "option logasap" was specified, then the time counting stops at the moment
12517 the log is emitted. In this case, a '+' sign is prepended before the value,
12518 indicating that the final one will be larger. See "Timers" below for more
12519 details.
12520
12521 - "bytes_read" is the total number of bytes transmitted from the server to
12522 the client when the log is emitted. If "option logasap" is specified, the
12523 this value will be prefixed with a '+' sign indicating that the final one
12524 may be larger. Please note that this value is a 64-bit counter, so log
12525 analysis tools must be able to handle it without overflowing.
12526
12527 - "termination_state" is the condition the session was in when the session
12528 ended. This indicates the session state, which side caused the end of
12529 session to happen, and for what reason (timeout, error, ...). The normal
12530 flags should be "--", indicating the session was closed by either end with
12531 no data remaining in buffers. See below "Session state at disconnection"
12532 for more details.
12533
12534 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012535 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012536 limits have been reached. For instance, if actconn is close to 512 when
12537 multiple connection errors occur, chances are high that the system limits
12538 the process to use a maximum of 1024 file descriptors and that all of them
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012539 are used. See section 3 "Global parameters" to find how to tune the system.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012540
12541 - "feconn" is the total number of concurrent connections on the frontend when
12542 the session was logged. It is useful to estimate the amount of resource
12543 required to sustain high loads, and to detect when the frontend's "maxconn"
12544 has been reached. Most often when this value increases by huge jumps, it is
12545 because there is congestion on the backend servers, but sometimes it can be
12546 caused by a denial of service attack.
12547
12548 - "beconn" is the total number of concurrent connections handled by the
12549 backend when the session was logged. It includes the total number of
12550 concurrent connections active on servers as well as the number of
12551 connections pending in queues. It is useful to estimate the amount of
12552 additional servers needed to support high loads for a given application.
12553 Most often when this value increases by huge jumps, it is because there is
12554 congestion on the backend servers, but sometimes it can be caused by a
12555 denial of service attack.
12556
12557 - "srv_conn" is the total number of concurrent connections still active on
12558 the server when the session was logged. It can never exceed the server's
12559 configured "maxconn" parameter. If this value is very often close or equal
12560 to the server's "maxconn", it means that traffic regulation is involved a
12561 lot, meaning that either the server's maxconn value is too low, or that
12562 there aren't enough servers to process the load with an optimal response
12563 time. When only one of the server's "srv_conn" is high, it usually means
12564 that this server has some trouble causing the connections to take longer to
12565 be processed than on other servers.
12566
12567 - "retries" is the number of connection retries experienced by this session
12568 when trying to connect to the server. It must normally be zero, unless a
12569 server is being stopped at the same moment the connection was attempted.
12570 Frequent retries generally indicate either a network problem between
12571 haproxy and the server, or a misconfigured system backlog on the server
12572 preventing new connections from being queued. This field may optionally be
12573 prefixed with a '+' sign, indicating that the session has experienced a
12574 redispatch after the maximal retry count has been reached on the initial
12575 server. In this case, the server name appearing in the log is the one the
12576 connection was redispatched to, and not the first one, though both may
12577 sometimes be the same in case of hashing for instance. So as a general rule
12578 of thumb, when a '+' is present in front of the retry count, this count
12579 should not be attributed to the logged server.
12580
12581 - "srv_queue" is the total number of requests which were processed before
12582 this one in the server queue. It is zero when the request has not gone
12583 through the server queue. It makes it possible to estimate the approximate
12584 server's response time by dividing the time spent in queue by the number of
12585 requests in the queue. It is worth noting that if a session experiences a
12586 redispatch and passes through two server queues, their positions will be
12587 cumulated. A request should not pass through both the server queue and the
12588 backend queue unless a redispatch occurs.
12589
12590 - "backend_queue" is the total number of requests which were processed before
12591 this one in the backend's global queue. It is zero when the request has not
12592 gone through the global queue. It makes it possible to estimate the average
12593 queue length, which easily translates into a number of missing servers when
12594 divided by a server's "maxconn" parameter. It is worth noting that if a
12595 session experiences a redispatch, it may pass twice in the backend's queue,
12596 and then both positions will be cumulated. A request should not pass
12597 through both the server queue and the backend queue unless a redispatch
12598 occurs.
12599
12600
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200126018.2.3. HTTP log format
12602----------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012603
12604The HTTP format is the most complete and the best suited for HTTP proxies. It
12605is enabled by when "option httplog" is specified in the frontend. It provides
12606the same level of information as the TCP format with additional features which
12607are specific to the HTTP protocol. Just like the TCP format, the log is usually
12608emitted at the end of the session, unless "option logasap" is specified, which
12609generally only makes sense for download sites. A session which matches the
12610"monitor" rules will never logged. It is also possible not to log sessions for
12611which no data were sent by the client by specifying "option dontlognull" in the
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020012612frontend. Successful connections will not be logged if "option dontlog-normal"
12613is specified in the frontend.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012614
12615Most fields are shared with the TCP log, some being different. A few fields may
12616slightly vary depending on some configuration options. Those ones are marked
12617with a star ('*') after the field name below.
12618
12619 Example :
12620 frontend http-in
12621 mode http
12622 option httplog
12623 log global
12624 default_backend bck
12625
12626 backend static
12627 server srv1 127.0.0.1:8000
12628
12629 >>> Feb 6 12:14:14 localhost \
12630 haproxy[14389]: 10.0.1.2:33317 [06/Feb/2009:12:14:14.655] http-in \
12631 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 +010012632 {} "GET /index.html HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012633
12634 Field Format Extract from the example above
12635 1 process_name '[' pid ']:' haproxy[14389]:
12636 2 client_ip ':' client_port 10.0.1.2:33317
12637 3 '[' accept_date ']' [06/Feb/2009:12:14:14.655]
12638 4 frontend_name http-in
12639 5 backend_name '/' server_name static/srv1
12640 6 Tq '/' Tw '/' Tc '/' Tr '/' Tt* 10/0/30/69/109
12641 7 status_code 200
12642 8 bytes_read* 2750
12643 9 captured_request_cookie -
12644 10 captured_response_cookie -
12645 11 termination_state ----
12646 12 actconn '/' feconn '/' beconn '/' srv_conn '/' retries* 1/1/1/1/0
12647 13 srv_queue '/' backend_queue 0/0
12648 14 '{' captured_request_headers* '}' {haproxy.1wt.eu}
12649 15 '{' captured_response_headers* '}' {}
12650 16 '"' http_request '"' "GET /index.html HTTP/1.1"
Willy Tarreaud72758d2010-01-12 10:42:19 +010012651
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012652
12653Detailed fields description :
12654 - "client_ip" is the IP address of the client which initiated the TCP
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012655 connection to haproxy. If the connection was accepted on a UNIX socket
12656 instead, the IP address would be replaced with the word "unix". Note that
12657 when the connection is accepted on a socket configured with "accept-proxy"
12658 and the PROXY protocol is correctly used, then the logs will reflect the
12659 forwarded connection's information.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012660
12661 - "client_port" is the TCP port of the client which initiated the connection.
Willy Tarreauceb24bc2010-11-09 12:46:41 +010012662 If the connection was accepted on a UNIX socket instead, the port would be
12663 replaced with the ID of the accepting socket, which is also reported in the
12664 stats interface.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012665
12666 - "accept_date" is the exact date when the TCP connection was received by
12667 haproxy (which might be very slightly different from the date observed on
12668 the network if there was some queuing in the system's backlog). This is
12669 usually the same date which may appear in any upstream firewall's log. This
12670 does not depend on the fact that the client has sent the request or not.
12671
12672 - "frontend_name" is the name of the frontend (or listener) which received
12673 and processed the connection.
12674
12675 - "backend_name" is the name of the backend (or listener) which was selected
12676 to manage the connection to the server. This will be the same as the
12677 frontend if no switching rule has been applied.
12678
12679 - "server_name" is the name of the last server to which the connection was
12680 sent, which might differ from the first one if there were connection errors
12681 and a redispatch occurred. Note that this server belongs to the backend
12682 which processed the request. If the request was aborted before reaching a
12683 server, "<NOSRV>" is indicated instead of a server name. If the request was
12684 intercepted by the stats subsystem, "<STATS>" is indicated instead.
12685
12686 - "Tq" is the total time in milliseconds spent waiting for the client to send
12687 a full HTTP request, not counting data. It can be "-1" if the connection
12688 was aborted before a complete request could be received. It should always
12689 be very small because a request generally fits in one single packet. Large
12690 times here generally indicate network trouble between the client and
12691 haproxy. See "Timers" below for more details.
12692
12693 - "Tw" is the total time in milliseconds spent waiting in the various queues.
12694 It can be "-1" if the connection was aborted before reaching the queue.
12695 See "Timers" below for more details.
12696
12697 - "Tc" is the total time in milliseconds spent waiting for the connection to
12698 establish to the final server, including retries. It can be "-1" if the
12699 request was aborted before a connection could be established. See "Timers"
12700 below for more details.
12701
12702 - "Tr" is the total time in milliseconds spent waiting for the server to send
12703 a full HTTP response, not counting data. It can be "-1" if the request was
12704 aborted before a complete response could be received. It generally matches
12705 the server's processing time for the request, though it may be altered by
12706 the amount of data sent by the client to the server. Large times here on
12707 "GET" requests generally indicate an overloaded server. See "Timers" below
12708 for more details.
12709
12710 - "Tt" is the total time in milliseconds elapsed between the accept and the
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012711 last close. It covers all possible processing. There is one exception, if
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012712 "option logasap" was specified, then the time counting stops at the moment
12713 the log is emitted. In this case, a '+' sign is prepended before the value,
12714 indicating that the final one will be larger. See "Timers" below for more
12715 details.
12716
12717 - "status_code" is the HTTP status code returned to the client. This status
12718 is generally set by the server, but it might also be set by haproxy when
12719 the server cannot be reached or when its response is blocked by haproxy.
12720
12721 - "bytes_read" is the total number of bytes transmitted to the client when
12722 the log is emitted. This does include HTTP headers. If "option logasap" is
12723 specified, the this value will be prefixed with a '+' sign indicating that
12724 the final one may be larger. Please note that this value is a 64-bit
12725 counter, so log analysis tools must be able to handle it without
12726 overflowing.
12727
12728 - "captured_request_cookie" is an optional "name=value" entry indicating that
12729 the client had this cookie in the request. The cookie name and its maximum
12730 length are defined by the "capture cookie" statement in the frontend
12731 configuration. The field is a single dash ('-') when the option is not
12732 set. Only one cookie may be captured, it is generally used to track session
12733 ID exchanges between a client and a server to detect session crossing
12734 between clients due to application bugs. For more details, please consult
12735 the section "Capturing HTTP headers and cookies" below.
12736
12737 - "captured_response_cookie" is an optional "name=value" entry indicating
12738 that the server has returned a cookie with its response. The cookie name
12739 and its maximum length are defined by the "capture cookie" statement in the
12740 frontend configuration. The field is a single dash ('-') when the option is
12741 not set. Only one cookie may be captured, it is generally used to track
12742 session ID exchanges between a client and a server to detect session
12743 crossing between clients due to application bugs. For more details, please
12744 consult the section "Capturing HTTP headers and cookies" below.
12745
12746 - "termination_state" is the condition the session was in when the session
12747 ended. This indicates the session state, which side caused the end of
12748 session to happen, for what reason (timeout, error, ...), just like in TCP
12749 logs, and information about persistence operations on cookies in the last
12750 two characters. The normal flags should begin with "--", indicating the
12751 session was closed by either end with no data remaining in buffers. See
12752 below "Session state at disconnection" for more details.
12753
12754 - "actconn" is the total number of concurrent connections on the process when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012755 the session was logged. It is useful to detect when some per-process system
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012756 limits have been reached. For instance, if actconn is close to 512 or 1024
12757 when multiple connection errors occur, chances are high that the system
12758 limits the process to use a maximum of 1024 file descriptors and that all
Willy Tarreauc57f0e22009-05-10 13:12:33 +020012759 of them are used. See section 3 "Global parameters" to find how to tune the
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012760 system.
12761
12762 - "feconn" is the total number of concurrent connections on the frontend when
12763 the session was logged. It is useful to estimate the amount of resource
12764 required to sustain high loads, and to detect when the frontend's "maxconn"
12765 has been reached. Most often when this value increases by huge jumps, it is
12766 because there is congestion on the backend servers, but sometimes it can be
12767 caused by a denial of service attack.
12768
12769 - "beconn" is the total number of concurrent connections handled by the
12770 backend when the session was logged. It includes the total number of
12771 concurrent connections active on servers as well as the number of
12772 connections pending in queues. It is useful to estimate the amount of
12773 additional servers needed to support high loads for a given application.
12774 Most often when this value increases by huge jumps, it is because there is
12775 congestion on the backend servers, but sometimes it can be caused by a
12776 denial of service attack.
12777
12778 - "srv_conn" is the total number of concurrent connections still active on
12779 the server when the session was logged. It can never exceed the server's
12780 configured "maxconn" parameter. If this value is very often close or equal
12781 to the server's "maxconn", it means that traffic regulation is involved a
12782 lot, meaning that either the server's maxconn value is too low, or that
12783 there aren't enough servers to process the load with an optimal response
12784 time. When only one of the server's "srv_conn" is high, it usually means
12785 that this server has some trouble causing the requests to take longer to be
12786 processed than on other servers.
12787
12788 - "retries" is the number of connection retries experienced by this session
12789 when trying to connect to the server. It must normally be zero, unless a
12790 server is being stopped at the same moment the connection was attempted.
12791 Frequent retries generally indicate either a network problem between
12792 haproxy and the server, or a misconfigured system backlog on the server
12793 preventing new connections from being queued. This field may optionally be
12794 prefixed with a '+' sign, indicating that the session has experienced a
12795 redispatch after the maximal retry count has been reached on the initial
12796 server. In this case, the server name appearing in the log is the one the
12797 connection was redispatched to, and not the first one, though both may
12798 sometimes be the same in case of hashing for instance. So as a general rule
12799 of thumb, when a '+' is present in front of the retry count, this count
12800 should not be attributed to the logged server.
12801
12802 - "srv_queue" is the total number of requests which were processed before
12803 this one in the server queue. It is zero when the request has not gone
12804 through the server queue. It makes it possible to estimate the approximate
12805 server's response time by dividing the time spent in queue by the number of
12806 requests in the queue. It is worth noting that if a session experiences a
12807 redispatch and passes through two server queues, their positions will be
12808 cumulated. A request should not pass through both the server queue and the
12809 backend queue unless a redispatch occurs.
12810
12811 - "backend_queue" is the total number of requests which were processed before
12812 this one in the backend's global queue. It is zero when the request has not
12813 gone through the global queue. It makes it possible to estimate the average
12814 queue length, which easily translates into a number of missing servers when
12815 divided by a server's "maxconn" parameter. It is worth noting that if a
12816 session experiences a redispatch, it may pass twice in the backend's queue,
12817 and then both positions will be cumulated. A request should not pass
12818 through both the server queue and the backend queue unless a redispatch
12819 occurs.
12820
12821 - "captured_request_headers" is a list of headers captured in the request due
12822 to the presence of the "capture request header" statement in the frontend.
12823 Multiple headers can be captured, they will be delimited by a vertical bar
12824 ('|'). When no capture is enabled, the braces do not appear, causing a
12825 shift of remaining fields. It is important to note that this field may
12826 contain spaces, and that using it requires a smarter log parser than when
12827 it's not used. Please consult the section "Capturing HTTP headers and
12828 cookies" below for more details.
12829
12830 - "captured_response_headers" is a list of headers captured in the response
12831 due to the presence of the "capture response header" statement in the
12832 frontend. Multiple headers can be captured, they will be delimited by a
12833 vertical bar ('|'). When no capture is enabled, the braces do not appear,
12834 causing a shift of remaining fields. It is important to note that this
12835 field may contain spaces, and that using it requires a smarter log parser
12836 than when it's not used. Please consult the section "Capturing HTTP headers
12837 and cookies" below for more details.
12838
12839 - "http_request" is the complete HTTP request line, including the method,
12840 request and HTTP version string. Non-printable characters are encoded (see
12841 below the section "Non-printable characters"). This is always the last
12842 field, and it is always delimited by quotes and is the only one which can
12843 contain quotes. If new fields are added to the log format, they will be
12844 added before this field. This field might be truncated if the request is
12845 huge and does not fit in the standard syslog buffer (1024 characters). This
12846 is the reason why this field must always remain the last one.
12847
12848
Cyril Bontédc4d9032012-04-08 21:57:39 +0200128498.2.4. Custom log format
12850------------------------
William Lallemand48940402012-01-30 16:47:22 +010012851
Willy Tarreau2beef582012-12-20 17:22:52 +010012852The directive log-format allows you to customize the logs in http mode and tcp
William Lallemandbddd4fd2012-02-27 11:23:10 +010012853mode. It takes a string as argument.
William Lallemand48940402012-01-30 16:47:22 +010012854
12855HAproxy understands some log format variables. % precedes log format variables.
12856Variables can take arguments using braces ('{}'), and multiple arguments are
12857separated by commas within the braces. Flags may be added or removed by
12858prefixing them with a '+' or '-' sign.
12859
12860Special variable "%o" may be used to propagate its flags to all other
12861variables on the same format string. This is particularly handy with quoted
12862string formats ("Q").
12863
Willy Tarreauc8368452012-12-21 00:09:23 +010012864If a variable is named between square brackets ('[' .. ']') then it is used
Willy Tarreaube722a22014-06-13 16:31:59 +020012865as a sample expression rule (see section 7.3). This it useful to add some
Willy Tarreauc8368452012-12-21 00:09:23 +010012866less common information such as the client's SSL certificate's DN, or to log
12867the key that would be used to store an entry into a stick table.
12868
William Lallemand48940402012-01-30 16:47:22 +010012869Note: spaces must be escaped. A space character is considered as a separator.
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030012870In order to emit a verbatim '%', it must be preceded by another '%' resulting
Willy Tarreau06d97f92013-12-02 17:45:48 +010012871in '%%'. HAProxy will automatically merge consecutive separators.
William Lallemand48940402012-01-30 16:47:22 +010012872
12873Flags are :
12874 * Q: quote a string
Jamie Gloudonaaa21002012-08-25 00:18:33 -040012875 * X: hexadecimal representation (IPs, Ports, %Ts, %rt, %pid)
William Lallemand48940402012-01-30 16:47:22 +010012876
12877 Example:
12878
12879 log-format %T\ %t\ Some\ Text
12880 log-format %{+Q}o\ %t\ %s\ %{-Q}r
12881
12882At the moment, the default HTTP format is defined this way :
12883
Willy Tarreau2beef582012-12-20 17:22:52 +010012884 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tq/%Tw/%Tc/%Tr/%Tt\ %ST\ %B\ %CC\ \
12885 %CS\ %tsc\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq\ %hr\ %hs\ %{+Q}r
William Lallemand48940402012-01-30 16:47:22 +010012886
William Lallemandbddd4fd2012-02-27 11:23:10 +010012887the default CLF format is defined this way :
William Lallemand48940402012-01-30 16:47:22 +010012888
Willy Tarreau2beef582012-12-20 17:22:52 +010012889 log-format %{+Q}o\ %{-Q}ci\ -\ -\ [%T]\ %r\ %ST\ %B\ \"\"\ \"\"\ %cp\ \
Willy Tarreau773d65f2012-10-12 14:56:11 +020012890 %ms\ %ft\ %b\ %s\ \%Tq\ %Tw\ %Tc\ %Tr\ %Tt\ %tsc\ %ac\ %fc\ \
Willy Tarreau2beef582012-12-20 17:22:52 +010012891 %bc\ %sc\ %rc\ %sq\ %bq\ %CC\ %CS\ \%hrl\ %hsl
William Lallemand48940402012-01-30 16:47:22 +010012892
William Lallemandbddd4fd2012-02-27 11:23:10 +010012893and the default TCP format is defined this way :
12894
Willy Tarreau2beef582012-12-20 17:22:52 +010012895 log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ \
William Lallemandbddd4fd2012-02-27 11:23:10 +010012896 %ac/%fc/%bc/%sc/%rc\ %sq/%bq
12897
William Lallemand48940402012-01-30 16:47:22 +010012898Please refer to the table below for currently defined variables :
12899
William Lallemandbddd4fd2012-02-27 11:23:10 +010012900 +---+------+-----------------------------------------------+-------------+
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012901 | R | var | field name (8.2.2 and 8.2.3 for description) | type |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012902 +---+------+-----------------------------------------------+-------------+
12903 | | %o | special variable, apply flags on all next var | |
12904 +---+------+-----------------------------------------------+-------------+
Willy Tarreau2beef582012-12-20 17:22:52 +010012905 | | %B | bytes_read (from server to client) | numeric |
12906 | H | %CC | captured_request_cookie | string |
12907 | H | %CS | captured_response_cookie | string |
William Lallemand5f232402012-04-05 18:02:55 +020012908 | | %H | hostname | string |
William Lallemanda73203e2012-03-12 12:48:57 +010012909 | | %ID | unique-id | string |
Willy Tarreau4bf99632014-06-13 12:21:40 +020012910 | | %ST | status_code | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020012911 | | %T | gmt_date_time | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012912 | | %Tc | Tc | numeric |
Yuxans Yao4e25b012012-10-19 10:36:09 +080012913 | | %Tl | local_date_time | date |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012914 | H | %Tq | Tq | numeric |
12915 | H | %Tr | Tr | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020012916 | | %Ts | timestamp | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012917 | | %Tt | Tt | numeric |
12918 | | %Tw | Tw | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010012919 | | %U | bytes_uploaded (from client to server) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012920 | | %ac | actconn | numeric |
12921 | | %b | backend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010012922 | | %bc | beconn (backend concurrent connections) | numeric |
12923 | | %bi | backend_source_ip (connecting address) | IP |
12924 | | %bp | backend_source_port (connecting address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012925 | | %bq | backend_queue | numeric |
Willy Tarreau2beef582012-12-20 17:22:52 +010012926 | | %ci | client_ip (accepted address) | IP |
12927 | | %cp | client_port (accepted address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012928 | | %f | frontend_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010012929 | | %fc | feconn (frontend concurrent connections) | numeric |
12930 | | %fi | frontend_ip (accepting address) | IP |
12931 | | %fp | frontend_port (accepting address) | numeric |
Willy Tarreau773d65f2012-10-12 14:56:11 +020012932 | | %ft | frontend_name_transport ('~' suffix for SSL) | string |
Willy Tarreau7346acb2014-08-28 15:03:15 +020012933 | | %lc | frontend_log_counter | numeric |
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020012934 | | %hr | captured_request_headers default style | string |
12935 | | %hrl | captured_request_headers CLF style | string list |
12936 | | %hs | captured_response_headers default style | string |
12937 | | %hsl | captured_response_headers CLF style | string list |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012938 | | %ms | accept date milliseconds | numeric |
William Lallemand5f232402012-04-05 18:02:55 +020012939 | | %pid | PID | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012940 | H | %r | http_request | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012941 | | %rc | retries | numeric |
Willy Tarreau1f0da242014-01-25 11:01:50 +010012942 | | %rt | request_counter (HTTP req or TCP session) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012943 | | %s | server_name | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010012944 | | %sc | srv_conn (server concurrent connections) | numeric |
12945 | | %si | server_IP (target address) | IP |
12946 | | %sp | server_port (target address) | numeric |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012947 | | %sq | srv_queue | numeric |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012948 | S | %sslc| ssl_ciphers (ex: AES-SHA) | string |
12949 | S | %sslv| ssl_version (ex: TLSv1) | string |
Willy Tarreau2beef582012-12-20 17:22:52 +010012950 | | %t | date_time (with millisecond resolution) | date |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012951 | | %ts | termination_state | string |
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012952 | H | %tsc | termination_state with cookie status | string |
William Lallemandbddd4fd2012-02-27 11:23:10 +010012953 +---+------+-----------------------------------------------+-------------+
William Lallemand48940402012-01-30 16:47:22 +010012954
Willy Tarreauffc3fcd2012-10-12 20:17:54 +020012955 R = Restrictions : H = mode http only ; S = SSL only
William Lallemand48940402012-01-30 16:47:22 +010012956
Willy Tarreau5f51e1a2012-12-03 18:40:10 +010012957
129588.2.5. Error log format
12959-----------------------
12960
12961When an incoming connection fails due to an SSL handshake or an invalid PROXY
12962protocol header, haproxy will log the event using a shorter, fixed line format.
12963By default, logs are emitted at the LOG_INFO level, unless the option
12964"log-separate-errors" is set in the backend, in which case the LOG_ERR level
12965will be used. Connections on which no data are exchanged (eg: probes) are not
12966logged if the "dontlognull" option is set.
12967
12968The format looks like this :
12969
12970 >>> Dec 3 18:27:14 localhost \
12971 haproxy[6103]: 127.0.0.1:56059 [03/Dec/2012:17:35:10.380] frt/f1: \
12972 Connection error during SSL handshake
12973
12974 Field Format Extract from the example above
12975 1 process_name '[' pid ']:' haproxy[6103]:
12976 2 client_ip ':' client_port 127.0.0.1:56059
12977 3 '[' accept_date ']' [03/Dec/2012:17:35:10.380]
12978 4 frontend_name "/" bind_name ":" frt/f1:
12979 5 message Connection error during SSL handshake
12980
12981These fields just provide minimal information to help debugging connection
12982failures.
12983
12984
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200129858.3. Advanced logging options
12986-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012987
12988Some advanced logging options are often looked for but are not easy to find out
12989just by looking at the various options. Here is an entry point for the few
12990options which can enable better logging. Please refer to the keywords reference
12991for more information about their usage.
12992
12993
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200129948.3.1. Disabling logging of external tests
12995------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010012996
12997It is quite common to have some monitoring tools perform health checks on
12998haproxy. Sometimes it will be a layer 3 load-balancer such as LVS or any
12999commercial load-balancer, and sometimes it will simply be a more complete
13000monitoring system such as Nagios. When the tests are very frequent, users often
13001ask how to disable logging for those checks. There are three possibilities :
13002
13003 - if connections come from everywhere and are just TCP probes, it is often
13004 desired to simply disable logging of connections without data exchange, by
13005 setting "option dontlognull" in the frontend. It also disables logging of
13006 port scans, which may or may not be desired.
13007
13008 - if the connection come from a known source network, use "monitor-net" to
13009 declare this network as monitoring only. Any host in this network will then
13010 only be able to perform health checks, and their requests will not be
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013011 logged. This is generally appropriate to designate a list of equipment
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013012 such as other load-balancers.
13013
13014 - if the tests are performed on a known URI, use "monitor-uri" to declare
13015 this URI as dedicated to monitoring. Any host sending this request will
13016 only get the result of a health-check, and the request will not be logged.
13017
13018
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200130198.3.2. Logging before waiting for the session to terminate
13020----------------------------------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013021
13022The problem with logging at end of connection is that you have no clue about
13023what is happening during very long sessions, such as remote terminal sessions
13024or large file downloads. This problem can be worked around by specifying
13025"option logasap" in the frontend. Haproxy will then log as soon as possible,
13026just before data transfer begins. This means that in case of TCP, it will still
13027log the connection status to the server, and in case of HTTP, it will log just
13028after processing the server headers. In this case, the number of bytes reported
13029is the number of header bytes sent to the client. In order to avoid confusion
13030with normal logs, the total time field and the number of bytes are prefixed
13031with a '+' sign which means that real numbers are certainly larger.
13032
13033
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200130348.3.3. Raising log level upon errors
13035------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013036
13037Sometimes it is more convenient to separate normal traffic from errors logs,
13038for instance in order to ease error monitoring from log files. When the option
13039"log-separate-errors" is used, connections which experience errors, timeouts,
13040retries, redispatches or HTTP status codes 5xx will see their syslog level
13041raised from "info" to "err". This will help a syslog daemon store the log in
13042a separate file. It is very important to keep the errors in the normal traffic
13043file too, so that log ordering is not altered. You should also be careful if
13044you already have configured your syslog daemon to store all logs higher than
13045"notice" in an "admin" file, because the "err" level is higher than "notice".
13046
13047
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200130488.3.4. Disabling logging of successful connections
13049--------------------------------------------------
Willy Tarreauc9bd0cc2009-05-10 11:57:02 +020013050
13051Although this may sound strange at first, some large sites have to deal with
13052multiple thousands of logs per second and are experiencing difficulties keeping
13053them intact for a long time or detecting errors within them. If the option
13054"dontlog-normal" is set on the frontend, all normal connections will not be
13055logged. In this regard, a normal connection is defined as one without any
13056error, timeout, retry nor redispatch. In HTTP, the status code is checked too,
13057and a response with a status 5xx is not considered normal and will be logged
13058too. Of course, doing is is really discouraged as it will remove most of the
13059useful information from the logs. Do this only if you have no other
13060alternative.
13061
13062
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200130638.4. Timing events
13064------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013065
13066Timers provide a great help in troubleshooting network problems. All values are
13067reported in milliseconds (ms). These timers should be used in conjunction with
13068the session termination flags. In TCP mode with "option tcplog" set on the
13069frontend, 3 control points are reported under the form "Tw/Tc/Tt", and in HTTP
13070mode, 5 control points are reported under the form "Tq/Tw/Tc/Tr/Tt" :
13071
13072 - Tq: total time to get the client request (HTTP mode only). It's the time
13073 elapsed between the moment the client connection was accepted and the
13074 moment the proxy received the last HTTP header. The value "-1" indicates
13075 that the end of headers (empty line) has never been seen. This happens when
13076 the client closes prematurely or times out.
13077
13078 - Tw: total time spent in the queues waiting for a connection slot. It
13079 accounts for backend queue as well as the server queues, and depends on the
13080 queue size, and the time needed for the server to complete previous
13081 requests. The value "-1" means that the request was killed before reaching
13082 the queue, which is generally what happens with invalid or denied requests.
13083
13084 - Tc: total time to establish the TCP connection to the server. It's the time
13085 elapsed between the moment the proxy sent the connection request, and the
13086 moment it was acknowledged by the server, or between the TCP SYN packet and
13087 the matching SYN/ACK packet in return. The value "-1" means that the
13088 connection never established.
13089
13090 - Tr: server response time (HTTP mode only). It's the time elapsed between
13091 the moment the TCP connection was established to the server and the moment
13092 the server sent its complete response headers. It purely shows its request
13093 processing time, without the network overhead due to the data transmission.
13094 It is worth noting that when the client has data to send to the server, for
13095 instance during a POST request, the time already runs, and this can distort
13096 apparent response time. For this reason, it's generally wise not to trust
13097 too much this field for POST requests initiated from clients behind an
13098 untrusted network. A value of "-1" here means that the last the response
13099 header (empty line) was never seen, most likely because the server timeout
13100 stroke before the server managed to process the request.
13101
13102 - Tt: total session duration time, between the moment the proxy accepted it
13103 and the moment both ends were closed. The exception is when the "logasap"
13104 option is specified. In this case, it only equals (Tq+Tw+Tc+Tr), and is
13105 prefixed with a '+' sign. From this field, we can deduce "Td", the data
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013106 transmission time, by subtracting other timers when valid :
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013107
13108 Td = Tt - (Tq + Tw + Tc + Tr)
13109
13110 Timers with "-1" values have to be excluded from this equation. In TCP
13111 mode, "Tq" and "Tr" have to be excluded too. Note that "Tt" can never be
13112 negative.
13113
13114These timers provide precious indications on trouble causes. Since the TCP
13115protocol defines retransmit delays of 3, 6, 12... seconds, we know for sure
13116that timers close to multiples of 3s are nearly always related to lost packets
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013117due to network problems (wires, negotiation, congestion). Moreover, if "Tt" is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013118close to a timeout value specified in the configuration, it often means that a
13119session has been aborted on timeout.
13120
13121Most common cases :
13122
13123 - If "Tq" is close to 3000, a packet has probably been lost between the
13124 client and the proxy. This is very rare on local networks but might happen
13125 when clients are on far remote networks and send large requests. It may
13126 happen that values larger than usual appear here without any network cause.
13127 Sometimes, during an attack or just after a resource starvation has ended,
13128 haproxy may accept thousands of connections in a few milliseconds. The time
13129 spent accepting these connections will inevitably slightly delay processing
13130 of other connections, and it can happen that request times in the order of
13131 a few tens of milliseconds are measured after a few thousands of new
Patrick Mezard105faca2010-06-12 17:02:46 +020013132 connections have been accepted at once. Setting "option http-server-close"
13133 may display larger request times since "Tq" also measures the time spent
13134 waiting for additional requests.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013135
13136 - If "Tc" is close to 3000, a packet has probably been lost between the
13137 server and the proxy during the server connection phase. This value should
13138 always be very low, such as 1 ms on local networks and less than a few tens
13139 of ms on remote networks.
13140
Willy Tarreau55165fe2009-05-10 12:02:55 +020013141 - If "Tr" is nearly always lower than 3000 except some rare values which seem
13142 to be the average majored by 3000, there are probably some packets lost
13143 between the proxy and the server.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013144
13145 - If "Tt" is large even for small byte counts, it generally is because
13146 neither the client nor the server decides to close the connection, for
13147 instance because both have agreed on a keep-alive connection mode. In order
13148 to solve this issue, it will be needed to specify "option httpclose" on
13149 either the frontend or the backend. If the problem persists, it means that
13150 the server ignores the "close" connection mode and expects the client to
13151 close. Then it will be required to use "option forceclose". Having the
13152 smallest possible 'Tt' is important when connection regulation is used with
13153 the "maxconn" option on the servers, since no new connection will be sent
13154 to the server until another one is released.
13155
13156Other noticeable HTTP log cases ('xx' means any value to be ignored) :
13157
13158 Tq/Tw/Tc/Tr/+Tt The "option logasap" is present on the frontend and the log
13159 was emitted before the data phase. All the timers are valid
13160 except "Tt" which is shorter than reality.
13161
13162 -1/xx/xx/xx/Tt The client was not able to send a complete request in time
13163 or it aborted too early. Check the session termination flags
13164 then "timeout http-request" and "timeout client" settings.
13165
13166 Tq/-1/xx/xx/Tt It was not possible to process the request, maybe because
13167 servers were out of order, because the request was invalid
13168 or forbidden by ACL rules. Check the session termination
13169 flags.
13170
13171 Tq/Tw/-1/xx/Tt The connection could not establish on the server. Either it
13172 actively refused it or it timed out after Tt-(Tq+Tw) ms.
13173 Check the session termination flags, then check the
13174 "timeout connect" setting. Note that the tarpit action might
13175 return similar-looking patterns, with "Tw" equal to the time
13176 the client connection was maintained open.
13177
13178 Tq/Tw/Tc/-1/Tt The server has accepted the connection but did not return
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013179 a complete response in time, or it closed its connection
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013180 unexpectedly after Tt-(Tq+Tw+Tc) ms. Check the session
13181 termination flags, then check the "timeout server" setting.
13182
13183
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200131848.5. Session state at disconnection
13185-----------------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013186
13187TCP and HTTP logs provide a session termination indicator in the
13188"termination_state" field, just before the number of active connections. It is
131892-characters long in TCP mode, and is extended to 4 characters in HTTP mode,
13190each of which has a special meaning :
13191
13192 - On the first character, a code reporting the first event which caused the
13193 session to terminate :
13194
13195 C : the TCP session was unexpectedly aborted by the client.
13196
13197 S : the TCP session was unexpectedly aborted by the server, or the
13198 server explicitly refused it.
13199
13200 P : the session was prematurely aborted by the proxy, because of a
13201 connection limit enforcement, because a DENY filter was matched,
13202 because of a security check which detected and blocked a dangerous
13203 error in server response which might have caused information leak
Willy Tarreau570f2212013-06-10 16:42:09 +020013204 (eg: cacheable cookie).
13205
13206 L : the session was locally processed by haproxy and was not passed to
13207 a server. This is what happens for stats and redirects.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013208
13209 R : a resource on the proxy has been exhausted (memory, sockets, source
13210 ports, ...). Usually, this appears during the connection phase, and
13211 system logs should contain a copy of the precise error. If this
13212 happens, it must be considered as a very serious anomaly which
13213 should be fixed as soon as possible by any means.
13214
13215 I : an internal error was identified by the proxy during a self-check.
13216 This should NEVER happen, and you are encouraged to report any log
13217 containing this, because this would almost certainly be a bug. It
13218 would be wise to preventively restart the process after such an
13219 event too, in case it would be caused by memory corruption.
13220
Simon Horman752dc4a2011-06-21 14:34:59 +090013221 D : the session was killed by haproxy because the server was detected
13222 as down and was configured to kill all connections when going down.
13223
Justin Karnegeseb2c24a2012-05-24 15:28:52 -070013224 U : the session was killed by haproxy on this backup server because an
13225 active server was detected as up and was configured to kill all
13226 backup connections when going up.
13227
Willy Tarreaua2a64e92011-09-07 23:01:56 +020013228 K : the session was actively killed by an admin operating on haproxy.
13229
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013230 c : the client-side timeout expired while waiting for the client to
13231 send or receive data.
13232
13233 s : the server-side timeout expired while waiting for the server to
13234 send or receive data.
13235
13236 - : normal session completion, both the client and the server closed
13237 with nothing left in the buffers.
13238
13239 - on the second character, the TCP or HTTP session state when it was closed :
13240
Willy Tarreauf7b30a92010-12-06 22:59:17 +010013241 R : the proxy was waiting for a complete, valid REQUEST from the client
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013242 (HTTP mode only). Nothing was sent to any server.
13243
13244 Q : the proxy was waiting in the QUEUE for a connection slot. This can
13245 only happen when servers have a 'maxconn' parameter set. It can
13246 also happen in the global queue after a redispatch consecutive to
13247 a failed attempt to connect to a dying server. If no redispatch is
13248 reported, then no connection attempt was made to any server.
13249
13250 C : the proxy was waiting for the CONNECTION to establish on the
13251 server. The server might at most have noticed a connection attempt.
13252
13253 H : the proxy was waiting for complete, valid response HEADERS from the
13254 server (HTTP only).
13255
13256 D : the session was in the DATA phase.
13257
13258 L : the proxy was still transmitting LAST data to the client while the
13259 server had already finished. This one is very rare as it can only
13260 happen when the client dies while receiving the last packets.
13261
13262 T : the request was tarpitted. It has been held open with the client
13263 during the whole "timeout tarpit" duration or until the client
13264 closed, both of which will be reported in the "Tw" timer.
13265
13266 - : normal session completion after end of data transfer.
13267
13268 - the third character tells whether the persistence cookie was provided by
13269 the client (only in HTTP mode) :
13270
13271 N : the client provided NO cookie. This is usually the case for new
13272 visitors, so counting the number of occurrences of this flag in the
13273 logs generally indicate a valid trend for the site frequentation.
13274
13275 I : the client provided an INVALID cookie matching no known server.
13276 This might be caused by a recent configuration change, mixed
Cyril Bontéa8e7bbc2010-04-25 22:29:29 +020013277 cookies between HTTP/HTTPS sites, persistence conditionally
13278 ignored, or an attack.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013279
13280 D : the client provided a cookie designating a server which was DOWN,
13281 so either "option persist" was used and the client was sent to
13282 this server, or it was not set and the client was redispatched to
13283 another server.
13284
Willy Tarreau996a92c2010-10-13 19:30:47 +020013285 V : the client provided a VALID cookie, and was sent to the associated
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013286 server.
13287
Willy Tarreau996a92c2010-10-13 19:30:47 +020013288 E : the client provided a valid cookie, but with a last date which was
13289 older than what is allowed by the "maxidle" cookie parameter, so
13290 the cookie is consider EXPIRED and is ignored. The request will be
13291 redispatched just as if there was no cookie.
13292
13293 O : the client provided a valid cookie, but with a first date which was
13294 older than what is allowed by the "maxlife" cookie parameter, so
13295 the cookie is consider too OLD and is ignored. The request will be
13296 redispatched just as if there was no cookie.
13297
Willy Tarreauc89ccb62012-04-05 21:18:22 +020013298 U : a cookie was present but was not used to select the server because
13299 some other server selection mechanism was used instead (typically a
13300 "use-server" rule).
13301
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013302 - : does not apply (no cookie set in configuration).
13303
13304 - the last character reports what operations were performed on the persistence
13305 cookie returned by the server (only in HTTP mode) :
13306
13307 N : NO cookie was provided by the server, and none was inserted either.
13308
13309 I : no cookie was provided by the server, and the proxy INSERTED one.
13310 Note that in "cookie insert" mode, if the server provides a cookie,
13311 it will still be overwritten and reported as "I" here.
13312
Willy Tarreau996a92c2010-10-13 19:30:47 +020013313 U : the proxy UPDATED the last date in the cookie that was presented by
13314 the client. This can only happen in insert mode with "maxidle". It
Jarno Huuskonen0e82b922014-04-12 18:22:19 +030013315 happens every time there is activity at a different date than the
Willy Tarreau996a92c2010-10-13 19:30:47 +020013316 date indicated in the cookie. If any other change happens, such as
13317 a redispatch, then the cookie will be marked as inserted instead.
13318
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013319 P : a cookie was PROVIDED by the server and transmitted as-is.
13320
13321 R : the cookie provided by the server was REWRITTEN by the proxy, which
13322 happens in "cookie rewrite" or "cookie prefix" modes.
13323
13324 D : the cookie provided by the server was DELETED by the proxy.
13325
13326 - : does not apply (no cookie set in configuration).
13327
Willy Tarreau996a92c2010-10-13 19:30:47 +020013328The combination of the two first flags gives a lot of information about what
13329was happening when the session terminated, and why it did terminate. It can be
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013330helpful to detect server saturation, network troubles, local system resource
13331starvation, attacks, etc...
13332
13333The most common termination flags combinations are indicated below. They are
13334alphabetically sorted, with the lowercase set just after the upper case for
13335easier finding and understanding.
13336
13337 Flags Reason
13338
13339 -- Normal termination.
13340
13341 CC The client aborted before the connection could be established to the
13342 server. This can happen when haproxy tries to connect to a recently
13343 dead (or unchecked) server, and the client aborts while haproxy is
13344 waiting for the server to respond or for "timeout connect" to expire.
13345
13346 CD The client unexpectedly aborted during data transfer. This can be
13347 caused by a browser crash, by an intermediate equipment between the
13348 client and haproxy which decided to actively break the connection,
13349 by network routing issues between the client and haproxy, or by a
13350 keep-alive session between the server and the client terminated first
13351 by the client.
Willy Tarreaud72758d2010-01-12 10:42:19 +010013352
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013353 cD The client did not send nor acknowledge any data for as long as the
13354 "timeout client" delay. This is often caused by network failures on
Cyril Bontédc4d9032012-04-08 21:57:39 +020013355 the client side, or the client simply leaving the net uncleanly.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013356
13357 CH The client aborted while waiting for the server to start responding.
13358 It might be the server taking too long to respond or the client
13359 clicking the 'Stop' button too fast.
13360
13361 cH The "timeout client" stroke while waiting for client data during a
13362 POST request. This is sometimes caused by too large TCP MSS values
13363 for PPPoE networks which cannot transport full-sized packets. It can
13364 also happen when client timeout is smaller than server timeout and
13365 the server takes too long to respond.
13366
13367 CQ The client aborted while its session was queued, waiting for a server
13368 with enough empty slots to accept it. It might be that either all the
13369 servers were saturated or that the assigned server was taking too
13370 long a time to respond.
13371
13372 CR The client aborted before sending a full HTTP request. Most likely
13373 the request was typed by hand using a telnet client, and aborted
13374 too early. The HTTP status code is likely a 400 here. Sometimes this
13375 might also be caused by an IDS killing the connection between haproxy
13376 and the client.
13377
13378 cR The "timeout http-request" stroke before the client sent a full HTTP
13379 request. This is sometimes caused by too large TCP MSS values on the
13380 client side for PPPoE networks which cannot transport full-sized
13381 packets, or by clients sending requests by hand and not typing fast
13382 enough, or forgetting to enter the empty line at the end of the
Willy Tarreau2705a612014-05-23 17:38:34 +020013383 request. The HTTP status code is likely a 408 here. Note: recently,
13384 some browsers such as Google Chrome started to break the deployed Web
13385 infrastructure by aggressively implementing a new "pre-connect"
13386 feature, consisting in sending connections to sites recently visited
13387 without sending any request on them until the user starts to browse
13388 the site. This mechanism causes massive disruption among resource-
13389 limited servers, and causes a lot of 408 errors in HAProxy logs.
13390 Worse, some people report that sometimes the browser displays the 408
13391 error when the user expects to see the actual content (Mozilla fixed
13392 this bug in 2004, while Chrome users continue to report it in 2014),
13393 so in this case, using "errorfile 408 /dev/null" can be used as a
13394 workaround. More information on the subject is available here :
13395 https://bugzilla.mozilla.org/show_bug.cgi?id=248827
13396 https://code.google.com/p/chromium/issues/detail?id=85229
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013397
13398 CT The client aborted while its session was tarpitted. It is important to
13399 check if this happens on valid requests, in order to be sure that no
Willy Tarreau55165fe2009-05-10 12:02:55 +020013400 wrong tarpit rules have been written. If a lot of them happen, it
13401 might make sense to lower the "timeout tarpit" value to something
13402 closer to the average reported "Tw" timer, in order not to consume
13403 resources for just a few attackers.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013404
Willy Tarreau570f2212013-06-10 16:42:09 +020013405 LR The request was intercepted and locally handled by haproxy. Generally
13406 it means that this was a redirect or a stats request.
13407
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013408 SC The server or an equipment between it and haproxy explicitly refused
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013409 the TCP connection (the proxy received a TCP RST or an ICMP message
13410 in return). Under some circumstances, it can also be the network
13411 stack telling the proxy that the server is unreachable (eg: no route,
13412 or no ARP response on local network). When this happens in HTTP mode,
13413 the status code is likely a 502 or 503 here.
13414
13415 sC The "timeout connect" stroke before a connection to the server could
13416 complete. When this happens in HTTP mode, the status code is likely a
13417 503 or 504 here.
13418
13419 SD The connection to the server died with an error during the data
13420 transfer. This usually means that haproxy has received an RST from
13421 the server or an ICMP message from an intermediate equipment while
13422 exchanging data with the server. This can be caused by a server crash
13423 or by a network issue on an intermediate equipment.
13424
13425 sD The server did not send nor acknowledge any data for as long as the
13426 "timeout server" setting during the data phase. This is often caused
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013427 by too short timeouts on L4 equipments before the server (firewalls,
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013428 load-balancers, ...), as well as keep-alive sessions maintained
13429 between the client and the server expiring first on haproxy.
13430
13431 SH The server aborted before sending its full HTTP response headers, or
13432 it crashed while processing the request. Since a server aborting at
13433 this moment is very rare, it would be wise to inspect its logs to
13434 control whether it crashed and why. The logged request may indicate a
13435 small set of faulty requests, demonstrating bugs in the application.
13436 Sometimes this might also be caused by an IDS killing the connection
13437 between haproxy and the server.
13438
13439 sH The "timeout server" stroke before the server could return its
13440 response headers. This is the most common anomaly, indicating too
13441 long transactions, probably caused by server or database saturation.
13442 The immediate workaround consists in increasing the "timeout server"
13443 setting, but it is important to keep in mind that the user experience
13444 will suffer from these long response times. The only long term
13445 solution is to fix the application.
13446
13447 sQ The session spent too much time in queue and has been expired. See
13448 the "timeout queue" and "timeout connect" settings to find out how to
13449 fix this if it happens too often. If it often happens massively in
13450 short periods, it may indicate general problems on the affected
13451 servers due to I/O or database congestion, or saturation caused by
13452 external attacks.
13453
13454 PC The proxy refused to establish a connection to the server because the
13455 process' socket limit has been reached while attempting to connect.
Cyril Bontédc4d9032012-04-08 21:57:39 +020013456 The global "maxconn" parameter may be increased in the configuration
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013457 so that it does not happen anymore. This status is very rare and
13458 might happen when the global "ulimit-n" parameter is forced by hand.
13459
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013460 PD The proxy blocked an incorrectly formatted chunked encoded message in
13461 a request or a response, after the server has emitted its headers. In
13462 most cases, this will indicate an invalid message from the server to
Willy Tarreauf3a3e132013-08-31 08:16:26 +020013463 the client. Haproxy supports chunk sizes of up to 2GB - 1 (2147483647
13464 bytes). Any larger size will be considered as an error.
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013465
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013466 PH The proxy blocked the server's response, because it was invalid,
13467 incomplete, dangerous (cache control), or matched a security filter.
13468 In any case, an HTTP 502 error is sent to the client. One possible
13469 cause for this error is an invalid syntax in an HTTP header name
Willy Tarreaued2fd2d2010-12-29 11:23:27 +010013470 containing unauthorized characters. It is also possible but quite
13471 rare, that the proxy blocked a chunked-encoding request from the
13472 client due to an invalid syntax, before the server responded. In this
13473 case, an HTTP 400 error is sent to the client and reported in the
13474 logs.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013475
13476 PR The proxy blocked the client's HTTP request, either because of an
13477 invalid HTTP syntax, in which case it returned an HTTP 400 error to
13478 the client, or because a deny filter matched, in which case it
13479 returned an HTTP 403 error.
13480
13481 PT The proxy blocked the client's request and has tarpitted its
13482 connection before returning it a 500 server error. Nothing was sent
13483 to the server. The connection was maintained open for as long as
13484 reported by the "Tw" timer field.
13485
13486 RC A local resource has been exhausted (memory, sockets, source ports)
13487 preventing the connection to the server from establishing. The error
13488 logs will tell precisely what was missing. This is very rare and can
13489 only be solved by proper system tuning.
13490
Willy Tarreau996a92c2010-10-13 19:30:47 +020013491The combination of the two last flags gives a lot of information about how
13492persistence was handled by the client, the server and by haproxy. This is very
13493important to troubleshoot disconnections, when users complain they have to
13494re-authenticate. The commonly encountered flags are :
13495
13496 -- Persistence cookie is not enabled.
13497
13498 NN No cookie was provided by the client, none was inserted in the
13499 response. For instance, this can be in insert mode with "postonly"
13500 set on a GET request.
13501
13502 II A cookie designating an invalid server was provided by the client,
13503 a valid one was inserted in the response. This typically happens when
Jamie Gloudonaaa21002012-08-25 00:18:33 -040013504 a "server" entry is removed from the configuration, since its cookie
Willy Tarreau996a92c2010-10-13 19:30:47 +020013505 value can be presented by a client when no other server knows it.
13506
13507 NI No cookie was provided by the client, one was inserted in the
13508 response. This typically happens for first requests from every user
13509 in "insert" mode, which makes it an easy way to count real users.
13510
13511 VN A cookie was provided by the client, none was inserted in the
13512 response. This happens for most responses for which the client has
13513 already got a cookie.
13514
13515 VU A cookie was provided by the client, with a last visit date which is
13516 not completely up-to-date, so an updated cookie was provided in
13517 response. This can also happen if there was no date at all, or if
13518 there was a date but the "maxidle" parameter was not set, so that the
13519 cookie can be switched to unlimited time.
13520
13521 EI A cookie was provided by the client, with a last visit date which is
13522 too old for the "maxidle" parameter, so the cookie was ignored and a
13523 new cookie was inserted in the response.
13524
13525 OI A cookie was provided by the client, with a first visit date which is
13526 too old for the "maxlife" parameter, so the cookie was ignored and a
13527 new cookie was inserted in the response.
13528
13529 DI The server designated by the cookie was down, a new server was
13530 selected and a new cookie was emitted in the response.
13531
13532 VI The server designated by the cookie was not marked dead but could not
13533 be reached. A redispatch happened and selected another one, which was
13534 then advertised in the response.
13535
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013536
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135378.6. Non-printable characters
13538-----------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013539
13540In order not to cause trouble to log analysis tools or terminals during log
13541consulting, non-printable characters are not sent as-is into log files, but are
13542converted to the two-digits hexadecimal representation of their ASCII code,
13543prefixed by the character '#'. The only characters that can be logged without
13544being escaped are comprised between 32 and 126 (inclusive). Obviously, the
13545escape character '#' itself is also encoded to avoid any ambiguity ("#23"). It
13546is the same for the character '"' which becomes "#22", as well as '{', '|' and
13547'}' when logging headers.
13548
13549Note that the space character (' ') is not encoded in headers, which can cause
13550issues for tools relying on space count to locate fields. A typical header
13551containing spaces is "User-Agent".
13552
13553Last, it has been observed that some syslog daemons such as syslog-ng escape
13554the quote ('"') with a backslash ('\'). The reverse operation can safely be
13555performed since no quote may appear anywhere else in the logs.
13556
13557
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135588.7. Capturing HTTP cookies
13559---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013560
13561Cookie capture simplifies the tracking a complete user session. This can be
13562achieved using the "capture cookie" statement in the frontend. Please refer to
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013563section 4.2 for more details. Only one cookie can be captured, and the same
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013564cookie will simultaneously be checked in the request ("Cookie:" header) and in
13565the response ("Set-Cookie:" header). The respective values will be reported in
13566the HTTP logs at the "captured_request_cookie" and "captured_response_cookie"
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013567locations (see section 8.2.3 about HTTP log format). When either cookie is
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013568not seen, a dash ('-') replaces the value. This way, it's easy to detect when a
13569user switches to a new session for example, because the server will reassign it
13570a new cookie. It is also possible to detect if a server unexpectedly sets a
13571wrong cookie to a client, leading to session crossing.
13572
13573 Examples :
13574 # capture the first cookie whose name starts with "ASPSESSION"
13575 capture cookie ASPSESSION len 32
13576
13577 # capture the first cookie whose name is exactly "vgnvisitor"
13578 capture cookie vgnvisitor= len 32
13579
13580
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200135818.8. Capturing HTTP headers
13582---------------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013583
13584Header captures are useful to track unique request identifiers set by an upper
13585proxy, virtual host names, user-agents, POST content-length, referrers, etc. In
13586the response, one can search for information about the response length, how the
13587server asked the cache to behave, or an object location during a redirection.
13588
13589Header captures are performed using the "capture request header" and "capture
13590response header" statements in the frontend. Please consult their definition in
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013591section 4.2 for more details.
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013592
13593It is possible to include both request headers and response headers at the same
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013594time. Non-existent headers are logged as empty strings, and if one header
13595appears more than once, only its last occurrence will be logged. Request headers
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013596are grouped within braces '{' and '}' in the same order as they were declared,
13597and delimited with a vertical bar '|' without any space. Response headers
13598follow the same representation, but are displayed after a space following the
13599request headers block. These blocks are displayed just before the HTTP request
13600in the logs.
13601
Willy Tarreaud9ed3d22014-06-13 12:23:06 +020013602As a special case, it is possible to specify an HTTP header capture in a TCP
13603frontend. The purpose is to enable logging of headers which will be parsed in
13604an HTTP backend if the request is then switched to this HTTP backend.
13605
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013606 Example :
13607 # This instance chains to the outgoing proxy
13608 listen proxy-out
13609 mode http
13610 option httplog
13611 option logasap
13612 log global
13613 server cache1 192.168.1.1:3128
13614
13615 # log the name of the virtual server
13616 capture request header Host len 20
13617
13618 # log the amount of data uploaded during a POST
13619 capture request header Content-Length len 10
13620
13621 # log the beginning of the referrer
13622 capture request header Referer len 20
13623
13624 # server name (useful for outgoing proxies only)
13625 capture response header Server len 20
13626
13627 # logging the content-length is useful with "option logasap"
13628 capture response header Content-Length len 10
13629
13630 # log the expected cache behaviour on the response
13631 capture response header Cache-Control len 8
13632
13633 # the Via header will report the next proxy's name
13634 capture response header Via len 20
13635
13636 # log the URL location during a redirection
13637 capture response header Location len 20
13638
13639 >>> Aug 9 20:26:09 localhost \
13640 haproxy[2022]: 127.0.0.1:34014 [09/Aug/2004:20:26:09] proxy-out \
13641 proxy-out/cache1 0/0/0/162/+162 200 +350 - - ---- 0/0/0/0/0 0/0 \
13642 {fr.adserver.yahoo.co||http://fr.f416.mail.} {|864|private||} \
13643 "GET http://fr.adserver.yahoo.com/"
13644
13645 >>> Aug 9 20:30:46 localhost \
13646 haproxy[2022]: 127.0.0.1:34020 [09/Aug/2004:20:30:46] proxy-out \
13647 proxy-out/cache1 0/0/0/182/+182 200 +279 - - ---- 0/0/0/0/0 0/0 \
13648 {w.ods.org||} {Formilux/0.1.8|3495|||} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010013649 "GET http://trafic.1wt.eu/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013650
13651 >>> Aug 9 20:30:46 localhost \
13652 haproxy[2022]: 127.0.0.1:34028 [09/Aug/2004:20:30:46] proxy-out \
13653 proxy-out/cache1 0/0/2/126/+128 301 +223 - - ---- 0/0/0/0/0 0/0 \
13654 {www.sytadin.equipement.gouv.fr||http://trafic.1wt.eu/} \
13655 {Apache|230|||http://www.sytadin.} \
Willy Tarreaud72758d2010-01-12 10:42:19 +010013656 "GET http://www.sytadin.equipement.gouv.fr/ HTTP/1.1"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013657
13658
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200136598.9. Examples of logs
13660---------------------
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013661
13662These are real-world examples of logs accompanied with an explanation. Some of
13663them have been made up by hand. The syslog part has been removed for better
13664reading. Their sole purpose is to explain how to decipher them.
13665
13666 >>> haproxy[674]: 127.0.0.1:33318 [15/Oct/2003:08:31:57.130] px-http \
13667 px-http/srv1 6559/0/7/147/6723 200 243 - - ---- 5/3/3/1/0 0/0 \
13668 "HEAD / HTTP/1.0"
13669
13670 => long request (6.5s) entered by hand through 'telnet'. The server replied
13671 in 147 ms, and the session ended normally ('----')
13672
13673 >>> haproxy[674]: 127.0.0.1:33319 [15/Oct/2003:08:31:57.149] px-http \
13674 px-http/srv1 6559/1230/7/147/6870 200 243 - - ---- 324/239/239/99/0 \
13675 0/9 "HEAD / HTTP/1.0"
13676
13677 => Idem, but the request was queued in the global queue behind 9 other
13678 requests, and waited there for 1230 ms.
13679
13680 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.654] px-http \
13681 px-http/srv1 9/0/7/14/+30 200 +243 - - ---- 3/3/3/1/0 0/0 \
13682 "GET /image.iso HTTP/1.0"
13683
13684 => request for a long data transfer. The "logasap" option was specified, so
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010013685 the log was produced just before transferring data. The server replied in
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013686 14 ms, 243 bytes of headers were sent to the client, and total time from
13687 accept to first data byte is 30 ms.
13688
13689 >>> haproxy[674]: 127.0.0.1:33320 [15/Oct/2003:08:32:17.925] px-http \
13690 px-http/srv1 9/0/7/14/30 502 243 - - PH-- 3/2/2/0/0 0/0 \
13691 "GET /cgi-bin/bug.cgi? HTTP/1.0"
13692
13693 => the proxy blocked a server response either because of an "rspdeny" or
13694 "rspideny" filter, or because the response was improperly formatted and
Willy Tarreau3c92c5f2011-08-28 09:45:47 +020013695 not HTTP-compliant, or because it blocked sensitive information which
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013696 risked being cached. In this case, the response is replaced with a "502
13697 bad gateway". The flags ("PH--") tell us that it was haproxy who decided
13698 to return the 502 and not the server.
13699
13700 >>> haproxy[18113]: 127.0.0.1:34548 [15/Oct/2003:15:18:55.798] px-http \
Willy Tarreaud72758d2010-01-12 10:42:19 +010013701 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 +010013702
13703 => the client never completed its request and aborted itself ("C---") after
13704 8.5s, while the proxy was waiting for the request headers ("-R--").
13705 Nothing was sent to any server.
13706
13707 >>> haproxy[18113]: 127.0.0.1:34549 [15/Oct/2003:15:19:06.103] px-http \
13708 px-http/<NOSRV> -1/-1/-1/-1/50001 408 0 - - cR-- 2/2/2/0/0 0/0 ""
13709
13710 => The client never completed its request, which was aborted by the
13711 time-out ("c---") after 50s, while the proxy was waiting for the request
13712 headers ("-R--"). Nothing was sent to any server, but the proxy could
13713 send a 408 return code to the client.
13714
13715 >>> haproxy[18989]: 127.0.0.1:34550 [15/Oct/2003:15:24:28.312] px-tcp \
13716 px-tcp/srv1 0/0/5007 0 cD 0/0/0/0/0 0/0
13717
13718 => This log was produced with "option tcplog". The client timed out after
13719 5 seconds ("c----").
13720
13721 >>> haproxy[18989]: 10.0.0.1:34552 [15/Oct/2003:15:26:31.462] px-http \
13722 px-http/srv1 3183/-1/-1/-1/11215 503 0 - - SC-- 205/202/202/115/3 \
Willy Tarreaud72758d2010-01-12 10:42:19 +010013723 0/0 "HEAD / HTTP/1.0"
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013724
13725 => The request took 3s to complete (probably a network problem), and the
Willy Tarreauc57f0e22009-05-10 13:12:33 +020013726 connection to the server failed ('SC--') after 4 attempts of 2 seconds
Willy Tarreaucc6c8912009-02-22 10:53:55 +010013727 (config says 'retries 3'), and no redispatch (otherwise we would have
13728 seen "/+3"). Status code 503 was returned to the client. There were 115
13729 connections on this server, 202 connections on this proxy, and 205 on
13730 the global process. It is possible that the server refused the
13731 connection because of too many already established.
Willy Tarreau844e3c52008-01-11 16:28:18 +010013732
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013733
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200137349. Statistics and monitoring
13735----------------------------
13736
13737It is possible to query HAProxy about its status. The most commonly used
13738mechanism is the HTTP statistics page. This page also exposes an alternative
13739CSV output format for monitoring tools. The same format is provided on the
13740Unix socket.
13741
13742
137439.1. CSV format
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013744---------------
Krzysztof Piotr Oledzkif58a9622008-02-23 01:19:10 +010013745
Willy Tarreau7f062c42009-03-05 18:43:00 +010013746The statistics may be consulted either from the unix socket or from the HTTP
Willy Tarreaua3310dc2014-06-16 15:43:21 +020013747page. Both means provide a CSV format whose fields follow. The first line
13748begins with a sharp ('#') and has one word per comma-delimited field which
13749represents the title of the column. All other lines starting at the second one
13750use a classical CSV format using a comma as the delimiter, and the double quote
13751('"') as an optional text delimiter, but only if the enclosed text is ambiguous
13752(if it contains a quote or a comma). The double-quote character ('"') in the
13753text is doubled ('""'), which is the format that most tools recognize. Please
13754do not insert any column before these ones in order not to break tools which
13755use hard-coded column positions.
Willy Tarreau7f062c42009-03-05 18:43:00 +010013756
James Westbyebe62d62014-07-08 10:14:57 -040013757In brackets after each field name are the types which may have a value for
13758that field. The types are L (Listeners), F (Frontends), B (Backends), and
13759S (Servers).
13760
13761 0. pxname [LFBS]: proxy name
13762 1. svname [LFBS]: service name (FRONTEND for frontend, BACKEND for backend,
13763 any name for server/listener)
13764 2. qcur [..BS]: current queued requests. For the backend this reports the
13765 number queued without a server assigned.
13766 3. qmax [..BS]: max value of qcur
13767 4. scur [LFBS]: current sessions
13768 5. smax [LFBS]: max sessions
13769 6. slim [LFBS]: configured session limit
13770 7. stot [LFBS]: cumulative number of connections
13771 8. bin [LFBS]: bytes in
13772 9. bout [LFBS]: bytes out
13773 10. dreq [LFB.]: requests denied because of security concerns.
13774 - For tcp this is because of a matched tcp-request content rule.
13775 - For http this is because of a matched http-request or tarpit rule.
13776 11. dresp [LFBS]: responses denied because of security concerns.
13777 - For http this is because of a matched http-request rule, or
13778 "option checkcache".
13779 12. ereq [LF..]: request errors. Some of the possible causes are:
13780 - early termination from the client, before the request has been sent.
13781 - read error from the client
13782 - client timeout
13783 - client closed connection
13784 - various bad requests from the client.
13785 - request was tarpitted.
13786 13. econ [..BS]: number of requests that encountered an error trying to
13787 connect to a backend server. The backend stat is the sum of the stat
13788 for all servers of that backend, plus any connection errors not
13789 associated with a particular server (such as the backend having no
13790 active servers).
13791 14. eresp [..BS]: response errors. srv_abrt will be counted here also.
13792 Some other errors are:
13793 - write error on the client socket (won't be counted for the server stat)
13794 - failure applying filters to the response.
13795 15. wretr [..BS]: number of times a connection to a server was retried.
13796 16. wredis [..BS]: number of times a request was redispatched to another
13797 server. The server value counts the number of times that server was
13798 switched away from.
13799 17. status [LFBS]: status (UP/DOWN/NOLB/MAINT/MAINT(via)...)
13800 18. weight [..BS]: server weight (server), total weight (backend)
13801 19. act [..BS]: server is active (server), number of active servers (backend)
13802 20. bck [..BS]: server is backup (server), number of backup servers (backend)
13803 21. chkfail [...S]: number of failed checks. (Only counts checks failed when
13804 the server is up.)
13805 22. chkdown [..BS]: number of UP->DOWN transitions. The backend counter counts
13806 transitions to the whole backend being down, rather than the sum of the
13807 counters for each server.
13808 23. lastchg [..BS]: number of seconds since the last UP<->DOWN transition
13809 24. downtime [..BS]: total downtime (in seconds). The value for the backend
13810 is the downtime for the whole backend, not the sum of the server downtime.
13811 25. qlimit [...S]: configured maxqueue for the server, or nothing in the
13812 value is 0 (default, meaning no limit)
13813 26. pid [LFBS]: process id (0 for first instance, 1 for second, ...)
13814 27. iid [LFBS]: unique proxy id
13815 28. sid [L..S]: server id (unique inside a proxy)
13816 29. throttle [...S]: current throttle percentage for the server, when
13817 slowstart is active, or no value if not in slowstart.
13818 30. lbtot [..BS]: total number of times a server was selected, either for new
13819 sessions, or when re-dispatching. The server counter is the number
13820 of times that server was selected.
13821 31. tracked [...S]: id of proxy/server if tracking is enabled.
13822 32. type [LFBS]: (0=frontend, 1=backend, 2=server, 3=socket/listener)
13823 33. rate [.FBS]: number of sessions per second over last elapsed second
13824 34. rate_lim [.F..]: configured limit on new sessions per second
13825 35. rate_max [.FBS]: max number of new sessions per second
13826 36. check_status [...S]: status of last health check, one of:
Cyril Bontéf0c60612010-02-06 14:44:47 +010013827 UNK -> unknown
13828 INI -> initializing
13829 SOCKERR -> socket error
13830 L4OK -> check passed on layer 4, no upper layers testing enabled
13831 L4TMOUT -> layer 1-4 timeout
13832 L4CON -> layer 1-4 connection problem, for example
13833 "Connection refused" (tcp rst) or "No route to host" (icmp)
13834 L6OK -> check passed on layer 6
13835 L6TOUT -> layer 6 (SSL) timeout
13836 L6RSP -> layer 6 invalid response - protocol error
13837 L7OK -> check passed on layer 7
13838 L7OKC -> check conditionally passed on layer 7, for example 404 with
13839 disable-on-404
13840 L7TOUT -> layer 7 (HTTP/SMTP) timeout
13841 L7RSP -> layer 7 invalid response - protocol error
13842 L7STS -> layer 7 response error, for example HTTP 5xx
James Westbyebe62d62014-07-08 10:14:57 -040013843 37. check_code [...S]: layer5-7 code, if available
13844 38. check_duration [...S]: time in ms took to finish last health check
13845 39. hrsp_1xx [.FBS]: http responses with 1xx code
13846 40. hrsp_2xx [.FBS]: http responses with 2xx code
13847 41. hrsp_3xx [.FBS]: http responses with 3xx code
13848 42. hrsp_4xx [.FBS]: http responses with 4xx code
13849 43. hrsp_5xx [.FBS]: http responses with 5xx code
13850 44. hrsp_other [.FBS]: http responses with other codes (protocol error)
13851 45. hanafail [...S]: failed health checks details
13852 46. req_rate [.F..]: HTTP requests per second over last elapsed second
13853 47. req_rate_max [.F..]: max number of HTTP requests per second observed
13854 48. req_tot [.F..]: total number of HTTP requests received
13855 49. cli_abrt [..BS]: number of data transfers aborted by the client
13856 50. srv_abrt [..BS]: number of data transfers aborted by the server
13857 (inc. in eresp)
13858 51. comp_in [.FB.]: number of HTTP response bytes fed to the compressor
13859 52. comp_out [.FB.]: number of HTTP response bytes emitted by the compressor
13860 53. comp_byp [.FB.]: number of bytes that bypassed the HTTP compressor
13861 (CPU/BW limit)
13862 54. comp_rsp [.FB.]: number of HTTP responses that were compressed
13863 55. lastsess [..BS]: number of seconds since last session assigned to
13864 server/backend
13865 56. last_chk [...S]: last health check contents or textual error
13866 57. last_agt [...S]: last agent check contents or textual error
13867 58. qtime [..BS]: the average queue time in ms over the 1024 last requests
13868 59. ctime [..BS]: the average connect time in ms over the 1024 last requests
13869 60. rtime [..BS]: the average response time in ms over the 1024 last requests
13870 (0 for TCP)
13871 61. ttime [..BS]: the average total session time in ms over the 1024 last
13872 requests
Willy Tarreau844e3c52008-01-11 16:28:18 +010013873
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013874
Willy Tarreauc57f0e22009-05-10 13:12:33 +0200138759.2. Unix Socket commands
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013876-------------------------
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010013877
Willy Tarreau468f4932013-08-01 16:50:16 +020013878The stats socket is not enabled by default. In order to enable it, it is
13879necessary to add one line in the global section of the haproxy configuration.
13880A second line is recommended to set a larger timeout, always appreciated when
13881issuing commands by hand :
13882
13883 global
13884 stats socket /var/run/haproxy.sock mode 600 level admin
13885 stats timeout 2m
13886
13887It is also possible to add multiple instances of the stats socket by repeating
13888the line, and make them listen to a TCP port instead of a UNIX socket. This is
13889never done by default because this is dangerous, but can be handy in some
13890situations :
13891
13892 global
13893 stats socket /var/run/haproxy.sock mode 600 level admin
13894 stats socket ipv4@192.168.0.1:9999 level admin
13895 stats timeout 2m
13896
13897To access the socket, an external utility such as "socat" is required. Socat is a
13898swiss-army knife to connect anything to anything. We use it to connect terminals
13899to the socket, or a couple of stdin/stdout pipes to it for scripts. The two main
13900syntaxes we'll use are the following :
13901
13902 # socat /var/run/haproxy.sock stdio
13903 # socat /var/run/haproxy.sock readline
13904
13905The first one is used with scripts. It is possible to send the output of a
13906script to haproxy, and pass haproxy's output to another script. That's useful
13907for retrieving counters or attack traces for example.
13908
13909The second one is only useful for issuing commands by hand. It has the benefit
13910that the terminal is handled by the readline library which supports line
13911editing and history, which is very convenient when issuing repeated commands
13912(eg: watch a counter).
13913
13914The socket supports two operation modes :
13915 - interactive
13916 - non-interactive
13917
13918The non-interactive mode is the default when socat connects to the socket. In
13919this mode, a single line may be sent. It is processed as a whole, responses are
13920sent back, and the connection closes after the end of the response. This is the
13921mode that scripts and monitoring tools use. It is possible to send multiple
13922commands in this mode, they need to be delimited by a semi-colon (';'). For
13923example :
13924
13925 # echo "show info;show stat;show table" | socat /var/run/haproxy stdio
13926
13927The interactive mode displays a prompt ('>') and waits for commands to be
13928entered on the line, then processes them, and displays the prompt again to wait
13929for a new command. This mode is entered via the "prompt" command which must be
13930sent on the first line in non-interactive mode. The mode is a flip switch, if
13931"prompt" is sent in interactive mode, it is disabled and the connection closes
13932after processing the last command of the same line.
13933
13934For this reason, when debugging by hand, it's quite common to start with the
13935"prompt" command :
13936
13937 # socat /var/run/haproxy readline
13938 prompt
13939 > show info
13940 ...
13941 >
13942
13943Since multiple commands may be issued at once, haproxy uses the empty line as a
13944delimiter to mark an end of output for each command, and takes care of ensuring
13945that no command can emit an empty line on output. A script can thus easily
13946parse the output even when multiple commands were pipelined on a single line.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013947
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020013948It is important to understand that when multiple haproxy processes are started
13949on the same sockets, any process may pick up the request and will output its
13950own stats.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010013951
Willy Tarreau468f4932013-08-01 16:50:16 +020013952The list of commands currently supported on the stats socket is provided below.
13953If an unknown command is sent, haproxy displays the usage message which reminds
13954all supported commands. Some commands support a more complex syntax, generally
13955it will explain what part of the command is invalid when this happens.
13956
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010013957add acl <acl> <pattern>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010013958 Add an entry into the acl <acl>. <acl> is the #<id> or the <file> returned by
13959 "show acl". This command does not verify if the entry already exists. This
13960 command cannot be used if the reference <acl> is a file also used with a map.
13961 In this case, you must use the command "add map" in place of "add acl".
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010013962
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010013963add map <map> <key> <value>
13964 Add an entry into the map <map> to associate the value <value> to the key
13965 <key>. This command does not verify if the entry already exists. It is
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010013966 mainly used to fill a map after a clear operation. Note that if the reference
13967 <map> is a file and is shared with a map, this map will contain also a new
13968 pattern entry.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010013969
Willy Tarreaud63335a2010-02-26 12:56:52 +010013970clear counters
13971 Clear the max values of the statistics counters in each proxy (frontend &
13972 backend) and in each server. The cumulated counters are not affected. This
13973 can be used to get clean counters after an incident, without having to
13974 restart nor to clear traffic counters. This command is restricted and can
13975 only be issued on sockets configured for levels "operator" or "admin".
13976
13977clear counters all
13978 Clear all statistics counters in each proxy (frontend & backend) and in each
13979 server. This has the same effect as restarting. This command is restricted
13980 and can only be issued on sockets configured for level "admin".
13981
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010013982clear acl <acl>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010013983 Remove all entries from the acl <acl>. <acl> is the #<id> or the <file>
13984 returned by "show acl". Note that if the reference <acl> is a file and is
13985 shared with a map, this map will be also cleared.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010013986
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010013987clear map <map>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010013988 Remove all entries from the map <map>. <map> is the #<id> or the <file>
13989 returned by "show map". Note that if the reference <map> is a file and is
13990 shared with a acl, this acl will be also cleared.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010013991
Simon Hormanc88b8872011-06-15 15:18:49 +090013992clear table <table> [ data.<type> <operator> <value> ] | [ key <key> ]
13993 Remove entries from the stick-table <table>.
13994
13995 This is typically used to unblock some users complaining they have been
13996 abusively denied access to a service, but this can also be used to clear some
13997 stickiness entries matching a server that is going to be replaced (see "show
13998 table" below for details). Note that sometimes, removal of an entry will be
13999 refused because it is currently tracked by a session. Retrying a few seconds
14000 later after the session ends is usual enough.
14001
14002 In the case where no options arguments are given all entries will be removed.
14003
14004 When the "data." form is used entries matching a filter applied using the
14005 stored data (see "stick-table" in section 4.2) are removed. A stored data
14006 type must be specified in <type>, and this data type must be stored in the
14007 table otherwise an error is reported. The data is compared according to
14008 <operator> with the 64-bit integer <value>. Operators are the same as with
14009 the ACLs :
14010
14011 - eq : match entries whose data is equal to this value
14012 - ne : match entries whose data is not equal to this value
14013 - le : match entries whose data is less than or equal to this value
14014 - ge : match entries whose data is greater than or equal to this value
14015 - lt : match entries whose data is less than this value
14016 - gt : match entries whose data is greater than this value
14017
14018 When the key form is used the entry <key> is removed. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090014019 same type as the table, which currently is limited to IPv4, IPv6, integer and
14020 string.
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014021
14022 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014023 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020014024 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020014025 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
14026 bytes_out_rate(60000)=187
14027 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14028 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014029
14030 $ echo "clear table http_proxy key 127.0.0.1" | socat stdio /tmp/sock1
14031
14032 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Emeric Brun7c6b82e2010-09-24 16:34:28 +020014033 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau62a36c42010-08-17 15:53:10 +020014034 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14035 bytes_out_rate(60000)=191
Simon Hormanc88b8872011-06-15 15:18:49 +090014036 $ echo "clear table http_proxy data.gpc0 eq 1" | socat stdio /tmp/sock1
14037 $ echo "show table http_proxy" | socat stdio /tmp/sock1
14038 >>> # table: http_proxy, type: ip, size:204800, used:1
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014039
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014040del acl <acl> [<key>|#<ref>]
14041 Delete all the acl entries from the acl <acl> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014042 <acl> is the #<id> or the <file> returned by "show acl". If the <ref> is used,
14043 this command delete only the listed reference. The reference can be found with
14044 listing the content of the acl. Note that if the reference <acl> is a file and
14045 is shared with a map, the entry will be also deleted in the map.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014046
14047del map <map> [<key>|#<ref>]
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010014048 Delete all the map entries from the map <map> corresponding to the key <key>.
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014049 <map> is the #<id> or the <file> returned by "show map". If the <ref> is used,
14050 this command delete only the listed reference. The reference can be found with
14051 listing the content of the map. Note that if the reference <map> is a file and
14052 is shared with a acl, the entry will be also deleted in the map.
Thierry FOURNIER0b90f312014-01-29 20:40:18 +010014053
14054disable agent <backend>/<server>
Simon Horman671b6f02013-11-25 10:46:39 +090014055 Mark the auxiliary agent check as temporarily stopped.
14056
14057 In the case where an agent check is being run as a auxiliary check, due
14058 to the agent-check parameter of a server directive, new checks are only
14059 initialised when the agent is in the enabled. Thus, disable agent will
14060 prevent any new agent checks from begin initiated until the agent
14061 re-enabled using enable agent.
14062
14063 When an agent is disabled the processing of an auxiliary agent check that
14064 was initiated while the agent was set as enabled is as follows: All
14065 results that would alter the weight, specifically "drain" or a weight
14066 returned by the agent, are ignored. The processing of agent check is
14067 otherwise unchanged.
14068
14069 The motivation for this feature is to allow the weight changing effects
14070 of the agent checks to be paused to allow the weight of a server to be
14071 configured using set weight without being overridden by the agent.
14072
14073 This command is restricted and can only be issued on sockets configured for
14074 level "admin".
14075
Willy Tarreau532a4502011-09-07 22:37:44 +020014076disable frontend <frontend>
14077 Mark the frontend as temporarily stopped. This corresponds to the mode which
14078 is used during a soft restart : the frontend releases the port but can be
14079 enabled again if needed. This should be used with care as some non-Linux OSes
14080 are unable to enable it back. This is intended to be used in environments
14081 where stopping a proxy is not even imaginable but a misconfigured proxy must
14082 be fixed. That way it's possible to release the port and bind it into another
14083 process to restore operations. The frontend will appear with status "STOP"
14084 on the stats page.
14085
14086 The frontend may be specified either by its name or by its numeric ID,
14087 prefixed with a sharp ('#').
14088
14089 This command is restricted and can only be issued on sockets configured for
14090 level "admin".
14091
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020014092disable health <backend>/<server>
14093 Mark the primary health check as temporarily stopped. This will disable
14094 sending of health checks, and the last health check result will be ignored.
14095 The server will be in unchecked state and considered UP unless an auxiliary
14096 agent check forces it down.
14097
14098 This command is restricted and can only be issued on sockets configured for
14099 level "admin".
14100
Willy Tarreaud63335a2010-02-26 12:56:52 +010014101disable server <backend>/<server>
14102 Mark the server DOWN for maintenance. In this mode, no more checks will be
14103 performed on the server until it leaves maintenance.
14104 If the server is tracked by other servers, those servers will be set to DOWN
14105 during the maintenance.
14106
14107 In the statistics page, a server DOWN for maintenance will appear with a
14108 "MAINT" status, its tracking servers with the "MAINT(via)" one.
14109
14110 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020014111 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014112
14113 This command is restricted and can only be issued on sockets configured for
14114 level "admin".
14115
Simon Horman671b6f02013-11-25 10:46:39 +090014116enable agent <backend>/<server>
14117 Resume auxiliary agent check that was temporarily stopped.
14118
14119 See "disable agent" for details of the effect of temporarily starting
14120 and stopping an auxiliary agent.
14121
14122 This command is restricted and can only be issued on sockets configured for
14123 level "admin".
14124
Willy Tarreau532a4502011-09-07 22:37:44 +020014125enable frontend <frontend>
14126 Resume a frontend which was temporarily stopped. It is possible that some of
14127 the listening ports won't be able to bind anymore (eg: if another process
14128 took them since the 'disable frontend' operation). If this happens, an error
14129 is displayed. Some operating systems might not be able to resume a frontend
14130 which was disabled.
14131
14132 The frontend may be specified either by its name or by its numeric ID,
14133 prefixed with a sharp ('#').
14134
14135 This command is restricted and can only be issued on sockets configured for
14136 level "admin".
14137
Willy Tarreau9b5aecd2014-05-23 11:53:10 +020014138enable health <backend>/<server>
14139 Resume a primary health check that was temporarily stopped. This will enable
14140 sending of health checks again. Please see "disable health" for details.
14141
14142 This command is restricted and can only be issued on sockets configured for
14143 level "admin".
14144
Willy Tarreaud63335a2010-02-26 12:56:52 +010014145enable server <backend>/<server>
14146 If the server was previously marked as DOWN for maintenance, this marks the
14147 server UP and checks are re-enabled.
14148
14149 Both the backend and the server may be specified either by their name or by
Willy Tarreauf5f31922011-08-02 11:32:07 +020014150 their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014151
14152 This command is restricted and can only be issued on sockets configured for
14153 level "admin".
14154
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010014155get map <map> <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014156get acl <acl> <value>
14157 Lookup the value <value> in the map <map> or in the ACL <acl>. <map> or <acl>
14158 are the #<id> or the <file> returned by "show map" or "show acl". This command
14159 returns all the matching patterns associated with this map. This is useful for
14160 debugging maps and ACLs. The output format is composed by one line par
14161 matching type. Each line is composed by space-delimited series of words.
Thierry FOURNIER5b16df72014-02-26 18:07:38 +010014162
14163 The first two words are:
14164
14165 <match method>: The match method applied. It can be "found", "bool",
14166 "int", "ip", "bin", "len", "str", "beg", "sub", "dir",
14167 "dom", "end" or "reg".
14168
14169 <match result>: The result. Can be "match" or "no-match".
14170
14171 The following words are returned only if the pattern matches an entry.
14172
14173 <index type>: "tree" or "list". The internal lookup algorithm.
14174
14175 <case>: "case-insensitive" or "case-sensitive". The
14176 interpretation of the case.
14177
14178 <entry matched>: match="<entry>". Return the matched pattern. It is
14179 useful with regular expressions.
14180
14181 The two last word are used to show the returned value and its type. With the
14182 "acl" case, the pattern doesn't exist.
14183
14184 return=nothing: No return because there are no "map".
14185 return="<value>": The value returned in the string format.
14186 return=cannot-display: The value cannot be converted as string.
14187
14188 type="<type>": The type of the returned sample.
14189
Willy Tarreaud63335a2010-02-26 12:56:52 +010014190get weight <backend>/<server>
14191 Report the current weight and the initial weight of server <server> in
14192 backend <backend> or an error if either doesn't exist. The initial weight is
14193 the one that appears in the configuration file. Both are normally equal
14194 unless the current weight has been changed. Both the backend and the server
14195 may be specified either by their name or by their numeric ID, prefixed with a
Willy Tarreauf5f31922011-08-02 11:32:07 +020014196 sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014197
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014198help
14199 Print the list of known keywords and their basic usage. The same help screen
14200 is also displayed for unknown commands.
Willy Tarreau3dfe6cd2008-12-07 22:29:48 +010014201
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014202prompt
14203 Toggle the prompt at the beginning of the line and enter or leave interactive
14204 mode. In interactive mode, the connection is not closed after a command
14205 completes. Instead, the prompt will appear again, indicating the user that
14206 the interpreter is waiting for a new command. The prompt consists in a right
14207 angle bracket followed by a space "> ". This mode is particularly convenient
14208 when one wants to periodically check information such as stats or errors.
14209 It is also a good idea to enter interactive mode before issuing a "help"
14210 command.
14211
14212quit
14213 Close the connection when in interactive mode.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010014214
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014215set map <map> [<key>|#<ref>] <value>
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014216 Modify the value corresponding to each key <key> in a map <map>. <map> is the
14217 #<id> or <file> returned by "show map". If the <ref> is used in place of
14218 <key>, only the entry pointed by <ref> is changed. The new value is <value>.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014219
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020014220set maxconn frontend <frontend> <value>
Willy Tarreau3c7a79d2012-09-26 21:07:15 +020014221 Dynamically change the specified frontend's maxconn setting. Any positive
14222 value is allowed including zero, but setting values larger than the global
14223 maxconn does not make much sense. If the limit is increased and connections
14224 were pending, they will immediately be accepted. If it is lowered to a value
14225 below the current number of connections, new connections acceptation will be
Willy Tarreau2a0f4d22011-08-02 11:49:05 +020014226 delayed until the threshold is reached. The frontend might be specified by
14227 either its name or its numeric ID prefixed with a sharp ('#').
14228
Willy Tarreau91886b62011-09-07 14:38:31 +020014229set maxconn global <maxconn>
14230 Dynamically change the global maxconn setting within the range defined by the
14231 initial global maxconn setting. If it is increased and connections were
14232 pending, they will immediately be accepted. If it is lowered to a value below
14233 the current number of connections, new connections acceptation will be
14234 delayed until the threshold is reached. A value of zero restores the initial
14235 setting.
14236
Willy Tarreauf5b22872011-09-07 16:13:44 +020014237set rate-limit connections global <value>
14238 Change the process-wide connection rate limit, which is set by the global
14239 'maxconnrate' setting. A value of zero disables the limitation. This limit
14240 applies to all frontends and the change has an immediate effect. The value
14241 is passed in number of connections per second.
14242
William Lallemandd85f9172012-11-09 17:05:39 +010014243set rate-limit http-compression global <value>
14244 Change the maximum input compression rate, which is set by the global
14245 'maxcomprate' setting. A value of zero disables the limitation. The value is
William Lallemand096f5542012-11-19 17:26:05 +010014246 passed in number of kilobytes per second. The value is available in the "show
14247 info" on the line "CompressBpsRateLim" in bytes.
William Lallemandd85f9172012-11-09 17:05:39 +010014248
Willy Tarreau93e7c002013-10-07 18:51:07 +020014249set rate-limit sessions global <value>
14250 Change the process-wide session rate limit, which is set by the global
14251 'maxsessrate' setting. A value of zero disables the limitation. This limit
14252 applies to all frontends and the change has an immediate effect. The value
14253 is passed in number of sessions per second.
14254
Willy Tarreaue43d5322013-10-07 20:01:52 +020014255set rate-limit ssl-sessions global <value>
14256 Change the process-wide SSL session rate limit, which is set by the global
14257 'maxsslrate' setting. A value of zero disables the limitation. This limit
14258 applies to all frontends and the change has an immediate effect. The value
14259 is passed in number of sessions per second sent to the SSL stack. It applies
14260 before the handshake in order to protect the stack against handshake abuses.
14261
Willy Tarreau2a4b70f2014-05-22 18:42:35 +020014262set server <backend>/<server> agent [ up | down ]
14263 Force a server's agent to a new state. This can be useful to immediately
14264 switch a server's state regardless of some slow agent checks for example.
14265 Note that the change is propagated to tracking servers if any.
14266
14267set server <backend>/<server> health [ up | stopping | down ]
14268 Force a server's health to a new state. This can be useful to immediately
14269 switch a server's state regardless of some slow health checks for example.
14270 Note that the change is propagated to tracking servers if any.
14271
14272set server <backend>/<server> state [ ready | drain | maint ]
14273 Force a server's administrative state to a new state. This can be useful to
14274 disable load balancing and/or any traffic to a server. Setting the state to
14275 "ready" puts the server in normal mode, and the command is the equivalent of
14276 the "enable server" command. Setting the state to "maint" disables any traffic
14277 to the server as well as any health checks. This is the equivalent of the
14278 "disable server" command. Setting the mode to "drain" only removes the server
14279 from load balancing but still allows it to be checked and to accept new
14280 persistent connections. Changes are propagated to tracking servers if any.
14281
14282set server <backend>/<server> weight <weight>[%]
14283 Change a server's weight to the value passed in argument. This is the exact
14284 equivalent of the "set weight" command below.
14285
Emeric Brun4147b2e2014-06-16 18:36:30 +020014286set ssl ocsp-response <response>
14287 This command is used to update an OCSP Response for a certificate (see "crt"
14288 on "bind" lines). Same controls are performed as during the initial loading of
14289 the response. The <response> must be passed as a base64 encoded string of the
14290 DER encoded response from the OCSP server.
14291
14292 Example:
14293 openssl ocsp -issuer issuer.pem -cert server.pem \
14294 -host ocsp.issuer.com:80 -respout resp.der
14295 echo "set ssl ocsp-response $(base64 -w 10000 resp.der)" | \
14296 socat stdio /var/run/haproxy.stat
14297
Willy Tarreau47060b62013-08-01 21:11:42 +020014298set table <table> key <key> [data.<data_type> <value>]*
Willy Tarreau654694e2012-06-07 01:03:16 +020014299 Create or update a stick-table entry in the table. If the key is not present,
14300 an entry is inserted. See stick-table in section 4.2 to find all possible
14301 values for <data_type>. The most likely use consists in dynamically entering
14302 entries for source IP addresses, with a flag in gpc0 to dynamically block an
Willy Tarreau47060b62013-08-01 21:11:42 +020014303 IP address or affect its quality of service. It is possible to pass multiple
14304 data_types in a single call.
Willy Tarreau654694e2012-06-07 01:03:16 +020014305
Willy Tarreaud63335a2010-02-26 12:56:52 +010014306set timeout cli <delay>
14307 Change the CLI interface timeout for current connection. This can be useful
14308 during long debugging sessions where the user needs to constantly inspect
14309 some indicators without being disconnected. The delay is passed in seconds.
14310
14311set weight <backend>/<server> <weight>[%]
14312 Change a server's weight to the value passed in argument. If the value ends
14313 with the '%' sign, then the new weight will be relative to the initially
Simon Horman58b5d292013-02-12 10:45:52 +090014314 configured weight. Absolute weights are permitted between 0 and 256.
14315 Relative weights must be positive with the resulting absolute weight is
14316 capped at 256. Servers which are part of a farm running a static
14317 load-balancing algorithm have stricter limitations because the weight
14318 cannot change once set. Thus for these servers, the only accepted values
14319 are 0 and 100% (or 0 and the initial weight). Changes take effect
14320 immediately, though certain LB algorithms require a certain amount of
14321 requests to consider changes. A typical usage of this command is to
14322 disable a server during an update by setting its weight to zero, then to
14323 enable it again after the update by setting it back to 100%. This command
14324 is restricted and can only be issued on sockets configured for level
14325 "admin". Both the backend and the server may be specified either by their
14326 name or by their numeric ID, prefixed with a sharp ('#').
Willy Tarreaud63335a2010-02-26 12:56:52 +010014327
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014328show errors [<iid>]
14329 Dump last known request and response errors collected by frontends and
14330 backends. If <iid> is specified, the limit the dump to errors concerning
Willy Tarreau6162db22009-10-10 17:13:00 +020014331 either frontend or backend whose ID is <iid>. This command is restricted
14332 and can only be issued on sockets configured for levels "operator" or
14333 "admin".
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014334
14335 The errors which may be collected are the last request and response errors
14336 caused by protocol violations, often due to invalid characters in header
14337 names. The report precisely indicates what exact character violated the
14338 protocol. Other important information such as the exact date the error was
14339 detected, frontend and backend names, the server name (when known), the
14340 internal session ID and the source address which has initiated the session
14341 are reported too.
14342
14343 All characters are returned, and non-printable characters are encoded. The
14344 most common ones (\t = 9, \n = 10, \r = 13 and \e = 27) are encoded as one
14345 letter following a backslash. The backslash itself is encoded as '\\' to
14346 avoid confusion. Other non-printable characters are encoded '\xNN' where
14347 NN is the two-digits hexadecimal representation of the character's ASCII
14348 code.
14349
14350 Lines are prefixed with the position of their first character, starting at 0
14351 for the beginning of the buffer. At most one input line is printed per line,
14352 and large lines will be broken into multiple consecutive output lines so that
14353 the output never goes beyond 79 characters wide. It is easy to detect if a
14354 line was broken, because it will not end with '\n' and the next line's offset
14355 will be followed by a '+' sign, indicating it is a continuation of previous
14356 line.
14357
14358 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014359 $ echo "show errors" | socat stdio /tmp/sock1
14360 >>> [04/Mar/2009:15:46:56.081] backend http-in (#2) : invalid response
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014361 src 127.0.0.1, session #54, frontend fe-eth0 (#1), server s2 (#1)
14362 response length 213 bytes, error at position 23:
14363
14364 00000 HTTP/1.0 200 OK\r\n
14365 00017 header/bizarre:blah\r\n
14366 00038 Location: blah\r\n
14367 00054 Long-line: this is a very long line which should b
14368 00104+ e broken into multiple lines on the output buffer,
14369 00154+ otherwise it would be too large to print in a ter
14370 00204+ minal\r\n
14371 00211 \r\n
14372
Willy Tarreauc57f0e22009-05-10 13:12:33 +020014373 In the example above, we see that the backend "http-in" which has internal
Willy Tarreaue0c8a1a2009-03-04 16:33:10 +010014374 ID 2 has blocked an invalid response from its server s2 which has internal
14375 ID 1. The request was on session 54 initiated by source 127.0.0.1 and
14376 received by frontend fe-eth0 whose ID is 1. The total response length was
14377 213 bytes when the error was detected, and the error was at byte 23. This
14378 is the slash ('/') in header name "header/bizarre", which is not a valid
14379 HTTP character for a header name.
Krzysztof Piotr Oledzki2c6962c2008-03-02 02:42:14 +010014380
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014381show info
14382 Dump info about haproxy status on current process.
14383
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014384show map [<map>]
14385 Dump info about map converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014386 maps is returned. If a <map> is specified, its contents are dumped. <map> is
14387 the #<id> or <file>. The first column is a unique identifier. It can be used
14388 as reference for the operation "del map" and "set map". The second column is
14389 the pattern and the third column is the sample if available. The data returned
14390 are not directly a list of available maps, but are the list of all patterns
14391 composing any map. Many of these patterns can be shared with ACL.
Thierry FOURNIERd32079e2014-01-29 20:02:04 +010014392
14393show acl [<acl>]
14394 Dump info about acl converters. Without argument, the list of all available
Thierry FOURNIER65ce6132014-03-20 11:42:45 +010014395 acls is returned. If a <acl> is specified, its contents are dumped. <acl> if
14396 the #<id> or <file>. The dump format is the same than the map even for the
14397 sample value. The data returned are not a list of available ACL, but are the
14398 list of all patterns composing any ACL. Many of these patterns can be shared
14399 with maps.
Thierry FOURNIERc0e0d7b2013-12-11 16:55:52 +010014400
Willy Tarreau12833bb2014-01-28 16:49:56 +010014401show pools
14402 Dump the status of internal memory pools. This is useful to track memory
14403 usage when suspecting a memory leak for example. It does exactly the same
14404 as the SIGQUIT when running in foreground except that it does not flush
14405 the pools.
14406
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014407show sess
14408 Dump all known sessions. Avoid doing this on slow connections as this can
Willy Tarreau6162db22009-10-10 17:13:00 +020014409 be huge. This command is restricted and can only be issued on sockets
14410 configured for levels "operator" or "admin".
14411
Willy Tarreau66dc20a2010-03-05 17:53:32 +010014412show sess <id>
14413 Display a lot of internal information about the specified session identifier.
14414 This identifier is the first field at the beginning of the lines in the dumps
14415 of "show sess" (it corresponds to the session pointer). Those information are
14416 useless to most users but may be used by haproxy developers to troubleshoot a
14417 complex bug. The output format is intentionally not documented so that it can
Olivierce31e6e2014-09-05 18:49:10 +020014418 freely evolve depending on demands. You may find a description of all fields
14419 returned in src/dumpstats.c
14420
14421 The special id "all" dumps the states of all sessions, which must be avoided
14422 as much as possible as it is highly CPU intensive and can take a lot of time.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014423
14424show stat [<iid> <type> <sid>]
14425 Dump statistics in the CSV format. By passing <id>, <type> and <sid>, it is
14426 possible to dump only selected items :
14427 - <iid> is a proxy ID, -1 to dump everything
14428 - <type> selects the type of dumpable objects : 1 for frontends, 2 for
14429 backends, 4 for servers, -1 for everything. These values can be ORed,
14430 for example:
14431 1 + 2 = 3 -> frontend + backend.
14432 1 + 2 + 4 = 7 -> frontend + backend + server.
14433 - <sid> is a server ID, -1 to dump everything from the selected proxy.
14434
14435 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014436 $ echo "show info;show stat" | socat stdio unix-connect:/tmp/sock1
14437 >>> Name: HAProxy
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014438 Version: 1.4-dev2-49
14439 Release_date: 2009/09/23
14440 Nbproc: 1
14441 Process_num: 1
14442 (...)
14443
14444 # pxname,svname,qcur,qmax,scur,smax,slim,stot,bin,bout,dreq, (...)
14445 stats,FRONTEND,,,0,0,1000,0,0,0,0,0,0,,,,,OPEN,,,,,,,,,1,1,0, (...)
14446 stats,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,0,0,0,,0,250,(...)
14447 (...)
14448 www1,BACKEND,0,0,0,0,1000,0,0,0,0,0,,0,0,0,0,UP,1,1,0,,0,250, (...)
14449
14450 $
14451
14452 Here, two commands have been issued at once. That way it's easy to find
14453 which process the stats apply to in multi-process mode. Notice the empty
14454 line after the information output which marks the end of the first block.
14455 A similar empty line appears at the end of the second block (stats) so that
Krzysztof Piotr Oledzkif8645332009-12-13 21:55:50 +010014456 the reader knows the output has not been truncated.
Willy Tarreau9a42c0d2009-09-22 19:31:03 +020014457
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014458show table
14459 Dump general information on all known stick-tables. Their name is returned
14460 (the name of the proxy which holds them), their type (currently zero, always
14461 IP), their size in maximum possible number of entries, and the number of
14462 entries currently in use.
14463
14464 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014465 $ echo "show table" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014466 >>> # table: front_pub, type: ip, size:204800, used:171454
14467 >>> # table: back_rdp, type: ip, size:204800, used:0
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014468
Simon Horman17bce342011-06-15 15:18:47 +090014469show table <name> [ data.<type> <operator> <value> ] | [ key <key> ]
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014470 Dump contents of stick-table <name>. In this mode, a first line of generic
14471 information about the table is reported as with "show table", then all
14472 entries are dumped. Since this can be quite heavy, it is possible to specify
Simon Horman17bce342011-06-15 15:18:47 +090014473 a filter in order to specify what entries to display.
14474
14475 When the "data." form is used the filter applies to the stored data (see
14476 "stick-table" in section 4.2). A stored data type must be specified
14477 in <type>, and this data type must be stored in the table otherwise an
14478 error is reported. The data is compared according to <operator> with the
14479 64-bit integer <value>. Operators are the same as with the ACLs :
14480
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014481 - eq : match entries whose data is equal to this value
14482 - ne : match entries whose data is not equal to this value
14483 - le : match entries whose data is less than or equal to this value
14484 - ge : match entries whose data is greater than or equal to this value
14485 - lt : match entries whose data is less than this value
14486 - gt : match entries whose data is greater than this value
14487
Simon Hormanc88b8872011-06-15 15:18:49 +090014488
14489 When the key form is used the entry <key> is shown. The key must be of the
Simon Horman619e3cc2011-06-15 15:18:52 +090014490 same type as the table, which currently is limited to IPv4, IPv6, integer,
14491 and string.
Simon Horman17bce342011-06-15 15:18:47 +090014492
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014493 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014494 $ echo "show table http_proxy" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014495 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020014496 >>> 0x80e6a4c: key=127.0.0.1 use=0 exp=3594729 gpc0=0 conn_rate(30000)=1 \
14497 bytes_out_rate(60000)=187
14498 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14499 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014500
Willy Tarreau62a36c42010-08-17 15:53:10 +020014501 $ echo "show table http_proxy data.gpc0 gt 0" | socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014502 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020014503 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14504 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014505
Willy Tarreau62a36c42010-08-17 15:53:10 +020014506 $ echo "show table http_proxy data.conn_rate gt 5" | \
14507 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014508 >>> # table: http_proxy, type: ip, size:204800, used:2
Willy Tarreau62a36c42010-08-17 15:53:10 +020014509 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14510 bytes_out_rate(60000)=191
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014511
Simon Horman17bce342011-06-15 15:18:47 +090014512 $ echo "show table http_proxy key 127.0.0.2" | \
14513 socat stdio /tmp/sock1
Simon Horman64b28d02011-08-13 08:03:50 +090014514 >>> # table: http_proxy, type: ip, size:204800, used:2
Simon Horman17bce342011-06-15 15:18:47 +090014515 >>> 0x80e6a80: key=127.0.0.2 use=0 exp=3594740 gpc0=1 conn_rate(30000)=10 \
14516 bytes_out_rate(60000)=191
14517
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014518 When the data criterion applies to a dynamic value dependent on time such as
14519 a bytes rate, the value is dynamically computed during the evaluation of the
14520 entry in order to decide whether it has to be dumped or not. This means that
14521 such a filter could match for some time then not match anymore because as
14522 time goes, the average event rate drops.
14523
14524 It is possible to use this to extract lists of IP addresses abusing the
14525 service, in order to monitor them or even blacklist them in a firewall.
14526 Example :
Willy Tarreau62a36c42010-08-17 15:53:10 +020014527 $ echo "show table http_proxy data.gpc0 gt 0" \
14528 | socat stdio /tmp/sock1 \
Willy Tarreau88bc4ec2010-08-01 07:58:48 +020014529 | fgrep 'key=' | cut -d' ' -f2 | cut -d= -f2 > abusers-ip.txt
14530 ( or | awk '/key/{ print a[split($2,a,"=")]; }' )
Krzysztof Piotr Oledzki719e7262009-10-04 15:02:46 +020014531
Willy Tarreau532a4502011-09-07 22:37:44 +020014532shutdown frontend <frontend>
14533 Completely delete the specified frontend. All the ports it was bound to will
14534 be released. It will not be possible to enable the frontend anymore after
14535 this operation. This is intended to be used in environments where stopping a
14536 proxy is not even imaginable but a misconfigured proxy must be fixed. That
14537 way it's possible to release the port and bind it into another process to
14538 restore operations. The frontend will not appear at all on the stats page
14539 once it is terminated.
14540
14541 The frontend may be specified either by its name or by its numeric ID,
14542 prefixed with a sharp ('#').
14543
14544 This command is restricted and can only be issued on sockets configured for
14545 level "admin".
14546
Willy Tarreaua295edc2011-09-07 23:21:03 +020014547shutdown session <id>
14548 Immediately terminate the session matching the specified session identifier.
14549 This identifier is the first field at the beginning of the lines in the dumps
14550 of "show sess" (it corresponds to the session pointer). This can be used to
14551 terminate a long-running session without waiting for a timeout or when an
14552 endless transfer is ongoing. Such terminated sessions are reported with a 'K'
14553 flag in the logs.
14554
Cyril Bontée63a1eb2014-07-12 18:22:42 +020014555shutdown sessions server <backend>/<server>
Willy Tarreau52b2d222011-09-07 23:48:48 +020014556 Immediately terminate all the sessions attached to the specified server. This
14557 can be used to terminate long-running sessions after a server is put into
14558 maintenance mode, for instance. Such terminated sessions are reported with a
14559 'K' flag in the logs.
14560
Willy Tarreau0ba27502007-12-24 16:55:16 +010014561/*
14562 * Local variables:
14563 * fill-column: 79
14564 * End:
14565 */